Experimental and theoretical study of the effect of active-site constrained substrate motion on the magnitude of the observed intramolecular isotope effect for the P450 101 catalyzed benzylic hydroxylation of isomeric xylenes and 4,4'-dimethylbiphenyl

Article abstract of DOI:10.1021/ja983000w

The validity of a cytochrome P450 (P450) 101 force field developed previously was tested by comparing to published results from other laboratories the predicted regioselectivity and stereoselectivity of both (R)- and (S)-norcamphor oxidation when the force field was used. Once validated, the force field was used to test the hypothesis that the magnitude of an observed intramolecular isotope effect is a function of the distance between equivalent but isotopically distinct intramolecular sites of oxidative attack. Molecular dynamics simulations and kinetic deuterium isotope effect experiments on benzylic hydroxylation were then conducted for a series of selectively deuterated isomeric xylenes and 4,4'-dimethylbiphenyl with P450 101. The molecular dynamics simulations predicted that the rank order of substrate mobility in the active site of P450 101 was o-xylene > p- xylene > dimethylbiphenyl. The observed isotope effects for the trideutero analogues were 10.6, 7.4, and 2.7, for the o-xylene, p-xylene, and 4,4'- dimethylbiphenyl, respectively. Thus, as the theoretically predicted rates of interchange between the isotopically distinct methyl groups decrease, the observed isotope effect decreases. The agreement between the theoretical predictions and experimental results provides strong support for the distance hypothesis stated above and for the potential of computational analysis to enhance our understanding of protein/small molecule interactions.

Full text of DOI:10.1021/ja983000w

J. Am. Chem. Soc. 1999, 121, 41-47
41
Experimental and Theoretical Study of the Effect of Active-Site
Constrained Substrate Motion on the Magnitude of the Observed
Intramolecular Isotope Effect for the P450 101 Catalyzed Benzylic
Hydroxylation of Isomeric Xylenes and 4,4′-Dimethylbiphenyl
Christian Audergon,† Krishna R. Iyer,† Jeffrey P. Jones,*,‡ John F. Darbyshire,† and
William F. Trager*,†
Contribution from the Department of Medicinal Chemistry, School of Pharmacy, UniVersity of Washington,
Seattle, Washington 98195, and Department of Chemistry, Washington State UniVersity,
Pullman, Washington 99164
ReceiVed August 20, 1998
Abstract: The validity of a cytochrome P450 (P450) 101 force field developed previously was tested by
comparing to published results from other laboratories the predicted regioselectivity and stereoselectivity of
both (R)- and (S)-norcamphor oxidation when the force field was used. Once validated, the force field was
used to test the hypothesis that the magnitude of an observed intramolecular isotope effect is a function of the
distance between equivalent but isotopically distinct intramolecular sites of oxidative attack. Molecular dynamics
simulations and kinetic deuterium isotope effect experiments on benzylic hydroxylation were then conducted
for a series of selectively deuterated isomeric xylenes and 4,4′-dimethylbiphenyl with P450 101. The molecular
dynamics simulations predicted that the rank order of substrate mobility in the active site of P450 101 was
o-xylene > p-xylene > dimethylbiphenyl. The observed isotope effects for the trideutero analogues were 10.6,
7.4, and 2.7, for the o-xylene, p-xylene, and 4,4′-dimethylbiphenyl, respectively. Thus, as the theoretically
predicted rates of interchange between the isotopically distinct methyl groups decrease, the observed isotope
effect decreases. The agreement between the theoretical predictions and experimental results provides strong
support for the distance hypothesis stated above and for the potential of computational analysis to enhance our
understanding of protein/small molecule interactions.
Introduction
be enormously valuable both in the design of new drugs
(agonists or antagonists) and the establishment of predictive
models for metabolism and toxicity. Similar progress for the
mammalian forms of P450 has been impeded because X-ray
crystallographic structures for these membrane-bound enzymes
are not available.
In the absence of crystal structures, insight into structural
aspects of this class of enzymes has necessitated the use of
indirect approaches. Such approaches have included homology
modeling of unknown structures based on the few soluble
prokaryote forms of the enzyme for which X-ray structures are
available, site directed mutagenesis, attempts at active-site
mapping with mechanism-based inhibitors, and kinetic isotope
effects. Previous studies have established that the magnitude
of an observed intramolecular deuterium isotope effect, (kH/
kD)obs, is highly dependent upon a rapid rate of equilibration,
relative to the actual bond-breaking event, between the chemi-
cally equivalent but isotopically distinct protio and deutero
metabolic sites in the substrate, with respect to a catalytically
susceptible orientation.10-12 If equilibration is rapid, the observed
The P450s are a superfamily of enzymes that utilize molecular
oxygen to catalyze the oxidation of organic compounds of
diverse structure. The importance of understanding their catalytic
and structural properties is underscored by their presence in
virtually all living organisms and their ability to metabolize
essentially any organic compound to which they are exposed.
An X-ray structure can be particularly valuable in elucidating
the properties of an enzyme by providing a structural foundation
for understanding its catalytic behavior. Once the atomic
coordinates are available, computer-driven modeling studies can
be used to probe enzyme properties that are not readily
accessible experimentally. For many enzyme and receptor
systems, computational approaches1-4 including pattern recogni-
tion,5 pharmacophore modeling,6,7 comparative molecular field
analysis,8 and molecular dynamics simulations9 have proven to
† University of Washington.
‡ Washington State University.
(1) Nemethy, G.; Scheraga, H. A. FASEB J. 1990, 4, 3189-3197.
(2) Sun, E.; Cohen, F. E. Gene 1993, 137, 127-132.
(3) Sezerman, U.; Vajda, S.; Cornette, J.; DeLisi, C. Protein Sci. 1993,
2, 1827-1843.
(4) Loew, G. H.; Villar, H. O.; Alkorta, I. Pharm. Res. 1993, 10, 475-
486.
(5) Kirschner, G.; Kowalski, B. In Drug Design; Ariens, E. J., Ed.;
Academic Press: New York, 1979; pp 73-131.
(6) Mayer, D.; Naylor, C. B.; Motoc, I.; Marshall, G. R. J. Comput.-
Aided Mol. Des. 1987, 1, 3-16.
(7) Wang, S.; Milne, G. W. A.; Yan, X.; Posey, I. J.; Nicklaus, M. C.;
Graham, L.; Rice, W. G. J. Med. Chem. 1996, 39, 2047-2054.
(8) Cramer, R. D., III; Patterson, D. E.; Bunce, J. D. Prog. Clin. Biol.
Res. 1988, 291, 161-165.
(9) Karplus, M.; Petsko, G. A. Nature 1990, 347, 631-639.
(10) Jones, J. P.; Korzekwa, K. R.; Rettie, A. E.; Trager, W. F. J. Am.
Chem. Soc. 1986, 108, 7074-7078.
(11) Jones, J. P.; Trager, W. F. J. Am. Chem. Soc. 1987, 109, 2171-
2173.
(12) Jones, J. P.; Korzekwa, K. R.; Rettie, A. E.; Trager, W. F. J. Am.
Chem. Soc. 1988, 110, 2018. (Additions and Corrections.)
10.1021/ja983000w CCC: $18.00 © 1999 American Chemical Society
Published on Web 12/16/1998

42 J. Am. Chem. Soc., Vol. 121, No. 1, 1999
Audergon et al.
isotope effect (kH/kD)obs is equal to the intrinsic isotope effect
(kH/kD) for the reaction.13-17 If the equilibration rate is slow, or
of the same order of magnitude as bond breaking, (kH/kD)obs
will vary between a value of 1 (no isotope effect) and kH/kD
(maximum (intrinsic) isotope effect). Factors that can affect
equilibration rate can affect the magnitude of (kH/kD)obs.
18 These
factors are termed masking factors, and include slow dissociation
of the enzyme-substrate complex and restrictions on reorienta-
tion of substrate in the active site. Thus, the value of (kH/kD)obs
and the extent of its departure from kH/kD has the capacity to
provide unique information on substrate motion in the active
site.
Figure 1. Intramolecular distances between methyl group carbon atoms
for o- and p-xylene and 4,4′-dimethylbiphenyl calculated using Spartan
Molecular Mechanics software.
Previous isotope effect studies indicate that the intrinsic
isotope effects for cytochrome P450 catalyzed hydroxylations
of methyl groups in different electronic environments are
relatively insensitive to changes in enzyme active site archi-
tecture.19,20 While differences in substrate electronic character
may dictate different intrinsic isotope effects for the methyl
group hydroxylation of a saturated hydrocarbon versus that of
the N-methyl of an aromatic amine, the intrinsic isotope effect
for the methyl group hydroxylation of a given substrate is
essentially invariant across different P450 isoforms that are
capable of catalyzing the reaction. This insensitivity further
suggests that any significant departure of (kH/kD)obs from kH/
kD, for the same or similar substrates, is not due to changes in
energetics or mechanism but rather to factors that are capable
of masking kH/kD. Likely factors are the distance between
chemically equivalent but isotopically distinct catalytic sites in
the substrate and/or active site constraint of substrate motion.
To test the hypothesis that distance is the factor that can affect
the magnitude of (kH/kD)obs, a set of substrates was chosen in
which this distance between the catalytic sites was systematically
varied. The chosen substrates included selectivity deuterated
o-xylene, p-xylene, and 4,4′-dimethylbiphenyl where the dis-
tance between carbon atoms of equivalent protio and deutero
sites is geometrically fixed for each substrate but varies from a
minimum of 2.48 Å (o-xylene) to a maximum of 11.05 Å (4,4′-
dimethylbiphenyl)21 between substrates,18 Figure 1. The fixed
distances and the lack of orienting elements such as polar- or
hydrogen-bonding sites allowed the combination of the effects
of distance and active-site constraints to be explored without
the complication of additional contributing features. The results
of this study strongly suggested that the near complete sup-
pression of kH/kD for the benzylic hydroxylation of 4-2H3,4′-
dimethylbiphenyl, whether catalyzed by P450 2B1 or various
microsomal preparations, was a direct consequence of the 11.05
Å distance separating the carbon atoms of the 4 and 4′ methyl
groups.
Although the earlier study18 suggested that intrinsic isotope
effect masking and distance are related, it did not have the
capacity to reveal the exact cause of masking or the relative
rates of dissociation and active site reorientation for each
substrate. In contrast, molecular dynamics simulations could
provide validation of the distance/constrained motion hypothesis
since such studies would allow direct assessment of the mobility
of a substrate in the active site of a P450 of known crystal
structure and one for which the atomic coordinates are readily
available. Indeed, molecular dynamics computational studies
have been used to determine active-site motions in enzymes
such as triosephosphate isomerase,22 ribonuclease T1,23 and
carboxypeptidase A.24 In the case of P450, the bacterial isoform
P450 101 has been studied extensively by using substrate-bound
models originating from known X-ray crystal data, e.g.,
camphor, norcamphor, and thiocamphor,25-28 whereas other
studies have adapted these models to study valproic acid, benzo-
[a]pyrene, and nicotine in the enzyme-substrate complex.29-31
Overall, these studies have led to a greater understanding of
active-site architecture, substrate-binding phenomena, and cata-
lytically related motion of both substrate and enzyme.
In this paper, we present results from parallel experimental
and computational studies that provide strong support for the
hypothesis that the degree of masking found in intramolecular
isotope effects is directly related to the rates of interchange
between chemically equivalent but isotopically distinct metabolic
sites in the active site. This methodology provides a means of
assessing the constraints of an enzyme active site toward
substrate motion. Furthermore, these experiments provide insight
into how to approach synchronization of experimental and
molecular dynamics clocks.
Results
Validation of the Molecular Dynamics Protocols with
D-(1R)- and L-(1S)-Norcamphor. The relative carbon radical
(13) Northrop, D. B. Biochemistry 1975, 14, 2644-2651.
(14) Northrop, D. B. In Isotope Effects On Enzyme Catalyzed Reactions;
Cleland, W. W., O’Leary, M. H., Northrop, D. B., Eds.; University Park
Press: Baltimore, 1977; pp 122-148.
(15) Miwa, G. T.; Garland, W. A.; Hodshon, B. J.; Lu, A. Y.; Northrop,
D. B. J. Biol. Chem. 1980, 255, 6049-6054.
(22) Brown, F. K.; Kollman, P. A. J. Mol. Biol. 1987, 198, 533-546.
(23) MacKerell, A. D.; Nilsson, L.; Rigler, R.; Heinemann, U.; Saenger,
W. Proteins 1989, 6, 20-31.
(16) Northrop, D. B. Ann. ReV. Biochem. 1981, 50, 103-131.
(17) Northrop, D. B. Biochemistry 1981, 20 4056-4061.
(18) Iyer, K. R.; Jones, J. P.; Darbyshire, J. F.; Trager, W. F. Biochemistry
1997, 35, 7136-7143.
(24) Makinen, M. W.; Troyer, J. M.; van der Werff, H.; Bernedsen, H.
J. C.; van Gunsteren, W. F. J. Mol. Biol. 1989, 207, 201-216.
(25) Collins, J. R.; Loew, G. H. J. Biol. Chem. 1987, 263, 3164-3170.
(26) Paulsen, M. D.; Bass, M. B.; Ornstein, R. L. J. Biomol. Struct. Dyn.
1991, 9, 187-203.
(19) Jones, J. P.; Rettie, A. E.; Trager, W. F. J. Med. Chem. 1990, 33,
1242-1246.
(27) Paulsen, M. D.; Ornstein, R. L. J. Comp.-Aided Mol. Des. 1992, 6,
449-460.
(20) Kharki, S. B.; Dinnocenzo, J. P.; Jones, J. P.; Korzekwa, K. R. J.
Am. Chem. Soc. 1995, 117, 3657-3664.
(28) Bass, B. B.; Paulsen, M. D.; Ornstein, R. L. Proteins 1992, 13,
26-37.
(21) The carbon atoms of the methyl groups are taken as the frame of
reference for defining inter-methyl group distance (calculated using Spartan
Molecular Mechanics software) since it is fixed, unlike the distances between
the rotationally mobile inter-methyl group hydrogens. The distance of closest
approach that can be achieved between two hydrogens that are bonded to
different methyl groups in the same substrate is approximately 1.4 Å for
o-xylene, 6.7 Å for p-xylene, and 11.1 Å for 4,4′-dimethylbiphenyl.
(29) Collins, J. R.; Camper, D. L.; Loew, G. H. J. Am. Chem. Soc. 1991,
113, 2736-2743.
(30) Jones, J. P.; Trager, W. F.; Carlson, T. J. J. Am. Chem. Soc. 1993,
115, 381-387.
(31) Jones, J. P.; Shou, M.; Korzekwa, K. R. Biochemistry 1995, 34,
6956-6961.

P450 101 Catalyzed Benzylic Hydroxylation
J. Am. Chem. Soc., Vol. 121, No. 1, 1999 43
Table 1. Heat of Formation (∆Hf), Relative Energiesa of Radical
from UHF AM1 Calculations and Relative Probability of Radical
Formation
Table 3. Predicted Percentage of Regio- and Stereoselectivity of
Hydroxylation of D-(1R)- and L-(1S)-Norcamphor by P450 101
based on Analysis of Steric Factors at Those Sites not Excluded by
Electronic Effects
position and
radical typeb
relative stability
(Kcal/mol)
relative probability
of radical formationc
position and
stereochemistry
D-(1R)-norcamphora
L-(1S)-norcamphor
C3 (20)
C4 (30)
C5 (20)
C6 (20)
C1 (30)
C7 (20)
1.65
16.26
0.04
0.00
18.03
4.49
0.063
0.000
0.932
1.000
0.000
0.001
C3-exo
C3-endo
C5-exo
C5-endo
C6-exo
C6-endo
0.2
0.0
28.6
33.9
37.0
0.3
0.3
0.0
49.4
2.0
48.3
0.0
a Energies relative to the most stable radical. b The term (20) indicates
a secondary radical center, and the term (30) indicates a tertiary radical
center. c Calculated by eq 2.
a Regioselectivities were calculated from the distance and angle of
a given hydrogen from the perferryl oxygen as described in Experi-
mental Section.
Table 2. Predicted Percentage of Regio- and Stereoselectivity of
Hydroxylation of D-(1R)- and L-(1S)-Norcamphor by P450 101 from
Analysis of Steric Factors
Table 4. Experimental and Theoretical Regioselectivity (as
percent) for the P450 101 Catalyzed Hydroxylation of D-(1R)- and
L-(1S)-Norcamphor
position and
stereochemistry
D-(1R)- norcamphora
L-(1S)- norcamphor
experimenta
Audergon
Loidaa
Harrisb
C3-exo
C3-endo
C4
C5-exo
C5-endo
C6-exo
C6-endo
C1
1.2
0.2
53.7
12.1
14.2
14.5
0.1
0.0
2.7
1.4
1.6
0.0
45.1
15.3
0.6
13.9
0.0
0.0
position D-(R) L-(S) D-(R) L-(S) D-(R) L-(S) D-(R) L-(S)
C5
C6
C3
65
30
5
28
62
10
62.5 51.4
37.3 48.3
54
41
4
39
57
3
75.0 43.3
18.7 38.9
6.3 17.7
0.2
0.3
a Taken from Loida et al.33 b Taken from Harris and Loew.32
Table 5. Statistically Corrected Kinetic Isotope Effects for the
Hydroxylation of the R-Methyl Group of o-Xylene, p-Xylene and
4,4′-Dimethylbiphenyl
C7-exo
C7-endo
22.2
1.3
substrate
isotope effecta
a Regioselectivities were calculated from the distance and angle of
a given hydrogen from the perferryl oxygen as described in Experi-
mental Section.
o-xylene-R-2H1-R′-2H1
o-xylene-R-2H2-R′-2H2
o-xylene-R-2H3
8.4 ( 0.42 (3)b
8.1 ( 0.29 (3)
10.6 ( 0.41 (3)
7.5 ( 0.38 (6)
7.4 ( 0.07 (6)
7.4 ( 0.37 (6)
5.5 ( 0.53 (6)
6.2 ( 0.15 (6)
2.7 ( 0.11 (6)
energies for each carbon site were calculated from the heat of
radical formation at each site. The results indicated that the rank
order of energetically preferred radicals was C5 > C6 > C3 >
C7 > C4 > C1. The relative stability values were then used to
calculate the relative probability of specific radical formation.
The results of this calculation are presented in Table 1, and are
used as electronic criteria in predicting the regioselectivity of
norcamphor hydroxylation. Because of the lower energies
associated with radical formation at C3, C5, and C6, oxidation
at these sites was deemed to be favored, whereas radical
formation at C1, C4, or C7 was excluded because of the higher
energies associated with these sites. Although the assumption
that only these three carbons can form radicals is arbitrary, it is
similar to the assumptions made by others.32,33
The Cartesian coordinates for the perferryl oxygen-substrate
complex that were stored during the molecular dynamic runs
were analyzed, using the steric criteria defined above, to
establish which hydrogens on D-(1R)- or L-(1S)-norcamphor
were oriented for catalysis. The results of this analysis are
summarized in Table 2, and are expressed as a percentage.
Hydroxylation is predicted to preferentially occur at carbon C4,
(>45%), but with substantial additional oxidation at C5 (>15%)
and C6 (>13%), for both enantiomers. Considerable hydroxy-
lation was also predicted at C7 (>22%), but only for L-(1S)-
norcamphor. Stereoselectivity of hydroxylation was confined
to the exo position with one notable exception, the C5-endo
position of D-(1R)-norcamphor. The C3-endo, C6-endo, and C1
hydrogen do not meet the defined steric criteria necessary for
catalysis.
p-xylene-R-2H1-R′-2H1
p-xylene-R-2H2-R′-2H2
p-xylene-R-2H3
4-2H1,4′-2H1-dimethylbiphenyl
4-2H2,4′-2H2-dimethylbiphenyl
4-2H3,4′-dimethylbiphenyl
a Numerical value of the isotope effect plus or minus the standard
deviation. b Number in parentheses is the number of determinations.
instability of the C4 radical signaled by the electronic component
precludes its formation. Thus, hydroxylation is predicted to be
confined to the C3, C5, and C6 sites for both stereoisomers.
Stereoselectivity of hydroxylation is essentially exo for all
L-(1S)-norcamphor carbon positions. However, in the case of
D-(1R)-norcamphor, although reaction at the C6 and C3 positions
is predicted to be exo, hydroxylation of C5 is predicted to be
nonstereoselective. The experimentally determined percent
regioselectivity values of the P450 101 catalyzed hydroxylation
of D-(R)- and L-(S)-norcamphor33 are tabulated together with
the computationally derived values from the present study and
those of Loida et al.33 and Harris and Lowe32 in Table 4.
Isotope Effects for the Isomer Xylenes and 4,4′-Dimeth-
ylbiphenyl. The observed deuterium isotope effects for selec-
tively deuterated xylene isomers and 4,4′-dimethylbiphenyl are
listed in Table 5. With o-xylene and p-xylene, the observed
isotope effect for the R-2H3-substrates were 10.6 and 7.4,
respectively, and are similar to the observed isotope effect values
for R-2H1-R′-2H1 and R-2H2-R′-2H2-deuterated substrates (8.4
and 8.1; 7.4 and 7.37, respectively). In the case of 4-2H3,4′-
dimethylbiphenyl the observed isotope effect (2.7) is signifi-
cantly smaller than the observed isotope effect (5.5 and 6.2)
for the di- and tetradeuterated substrates, respectively.
The steric predictions minus the potential sites of oxidation
excluded by electronic effects are summarized in Table 3. The
(32) Harris, D.; Loew, G. J. Am. Chem. Soc. 1995, 117, 2738-2746.
(33) Loida, P. J.; Sligar, S. G.; Paulsen, M. D.; Arnold, G. E.; Ornstein,
R. L. J. Biol. Chem. 1995, 270, 5325-5330.
Molecular Dynamics of Isomeric Xylenes and 4,4′-Dimeth-
ylbiphenyl. The results of the molecular dynamics simulations

44 J. Am. Chem. Soc., Vol. 121, No. 1, 1999
Audergon et al.
Table 6. Number of Switches between Methyl Group and a
Reactive Orientation during the Course of a Molecular Dynamics
Run
101 catalyzed hydroxylation of these substrates by computer
simulation is particularly challenging because multiple products
are formed. For example, an enantiomeric mixture of norcam-
phor metabolized by P40 101 results in the formation of 45%
5-exo-hydroxy-, 47% 6-exo-hydroxy-, and 8% 3-exo-hydrox-
ynorcamphor,34 while the 5-, 6-, and 3-hydroxy products are
obtained in a ratio 65:30:5, respectively, from enantiomerically
pure D-(1R)-norcamphor and in a ratio of 28:62:10, respectively,
from enantiomerically pure L-(1S)-norcamphor.33
distance between
methyl group
carbon atoms (Å)
number of
switches in 1 ns
molecular dynamics
substrate
o-xylene
m-xylene
2.48
5.0
6.62
11.05
1172a
779b
3
p-xylene
4,4′-dimethylbiphenyl
0
In this study the predicted regioselective results obtained for
the formation of the 5-, 6-, and 3-hydroxy products from D-(1R)-
norcamphor gave a ratio of 62.5:37.3:0.2, whereas those from
L-(1S)-norcamphor gave a ratio of 51.4:48.3:0.3 for the corre-
sponding products. Although not exact, these values are in
reasonable accord with the experimental results for the individual
enantiomers,33 Table 4. However, evaluation of the predicted
product stereoselectivities (exo versus endo, Table 3) are
complicated by the fact that experimental product stereoselec-
tivities have only been reported for racemic norcamphor34 and
not for the pure enantiomers.33 The predicted product stereo-
selectivity for L-(1S)-norcamphor hydroxylation was predomi-
nantly exo at all carbon sites. This is also the major stereo-
chemistry predicted for product formation at the C6 and C3
positions of D-(1R)-norcamphor. All of these results are in
agreement with experiment and previously reported theoretical
results,32,33 Table 4. However, a major difference arises for the
5-hydroxylation of D-(1R)-norcamphor where our method
predicts an exo/endo ratio of 0.85:1, Table 3. Although no
experimental studies have appeared reporting the exo to endo
product distribution for this enantiomer, these results would
appear not to agree with the experimental results reported for
the racemic mixture.34
a Calculated from 500 ps of molecular dynamics. b Although no
experimental results were obtained for m-xylene due to analytical
complications, the theoretical data is worthy of consideration since it
shows good correlation with the other isomers.
are shown in Table 6. As indicated in the methods section the
simulations for o-xylene were conducted for 500 ps, whereas
those for the other substrates were run for 1 ns. The predicted
number of equilibration events, or switches, between the methyl
groups were 1172, 779, 3, and 0 for o-xylene, m-xylene,
p-xylene, and 4,4′-dimethylbiphenyl, respectively, when normal-
ized to a 1-ns time span.
Discussion
In the earlier experimental paper on the isotope effects
associated with the P450 catalyzed methyl group hydroxylation
of o- and p-xylene and 4,4′-dimethylbiphenyl, the magnitudes
of the observed isotope effects were found to decrease as the
distance between reaction sites increased.18 The results of these
experiments support the hypothesis that the degree of masking
of an intrinsic isotope effect in an intramolecular isotope effect
experiment is directly related to the distance between equivalent
but isotopically distinct methyl groups. Intuitively the relation-
ship is appealing. One would certainly expect that the rate of
equilibration of two methyl groups, with respect to a defined
point in space, would decrease as the distance between the two
methyl groups increased, particularly if any steric impediments
(active-site topography) began to assume importance. A primary
goal of this study was to carry out both computational and
additional experimental studies that would further test this
hypothesis, using the same set of substrates (isomeric xylenes
and 4,4′-dimethylbiphenyl) but with an enzyme, P450 101, with
known active site dimensions. If successful, these studies would
also reinforce the notion that P450 101 can serve as a useful
model for the mammalian P450s. Since the active-site structures
of mammalian P450 are unknown, in computero studies based
on the known crystal structures of several soluble cytosolic
bacterial P450s might be used to predict and gain an under-
standing of mammalian P450s, provided they can be validated.
Indeed, this is often the only way such information can be
obtained. Thus, the validation of computational methods is
extremely important toward evaluating the potential of the
predictive power of this method.
Overall, the initial molecular dynamics simulations of the
enzyme-substrate complex of P450 101 with D-(1R)- and
L-(1S)-norcamphor incorporating only the motion of residues
within 13 Å of the heme iron, and using a combination of steric
and electronic criteria for analysis, gave good correlation with
previously published values for the regio- and stereoselective
hydroxylation of the norcamphor enantiomers by P450 101. It
was therefore concluded that our methodology could serve as
an appropriate model for testing the distance hypothesis.
The low energy barrier associated with rotation ensures that
the rate of methyl group rotation is much faster than the rate of
methyl group oxidation. If a substrate has a methyl group that
is oxidized and the methyl group contains deuterium and
protium, the enzyme has the choice of oxidizing either a C-H
or a C-D bond. Since the rate of methyl group rotation is fast
relative to oxidation, (kH/kD)obs will closely approach kH/kD.
Thus, the observed intramolecular isotope effects for P450 101
catalyzed benzylic hydroxylation of the di- and tetradeuterated
substrates studied here can be expected to closely approach the
intrinsic isotope effects.18 All values of (kH/kD)obs for these
substrates are large and fall within a range of 5.5 to 8.4, (Table
5). The values of (kH/kD)obs for the trideuteromethyl xylenes are
equally large, suggesting that within the active site of P450 101
even methyl group interchange is fast enough to effectively
unmask the intrinsic isotope effect. In contrast, (kH/kD)obs for
4-2H3,4′-dimethylbiphenyl is only 2.7, suggesting that kH/kD for
this substrate is highly masked. These results mirror the trends
that were observed with these same substrates when either rat
liver microsomes or purified P450 2B1 was used as the source
of P45018 and are consistent with the hypothesis that the extent
of kH/kD masking increases with increasing distance between
the catalytically susceptible methyl groups. For P450 101, just
Prior to determining whether molecular dynamic simulations
of the motion of the isomeric xylenes and 4,4′-dimethylbiphenyl
substrates in the active site of P450 101 were predictive of
distance-dependent masking effects, our working model of the
active enzyme needed to be validated. To this end D-(1R)- and
L-(1S)-norcamphor were chosen as probe substrates. An X-ray
crystal structure of D-norcamphor-P450 101 substrate-enzyme
complex (7cpp.pdb) was available, and both computational and
experimental results describing the P450 101 catalyzed hy-
droxylation of the racemic and enantiomeric forms of norcam-
phor were also available.32-34 Moreover, predicting the P450
(34) Atkins, W. M.; Silgar, S. G. Biochemistry 1988, 27, 1610-1616.

P450 101 Catalyzed Benzylic Hydroxylation
J. Am. Chem. Soc., Vol. 121, No. 1, 1999 45
Scheme 1. Kinetic Scheme for Binding Substrate to Enzyme
for an Intramolecular Isotope Effect Experiment with the
Representative ES Complexes for p-xylene.
the substrate remaining in the binding site of the enzyme.
Steady-state analysis of the product ratio leads to the following
equation for the observed isotope effect for Scheme 1.
k5H
(k2 + 2kr) + k5H
(
)
kH
k5D
d[P1]
d[P2]
)
)
(1)
( )
kD
(k2 + 2kr) + k5H
obs
Evaluation of the limits of this equation indicates that fast
rotation, kr . k5H, leads to the intrinsic isotope effect being
observed. For compounds with slow rotation an intrinsic isotope
effect can still be observed if k2 . k5H. However, as k2 and kr
become slower than k5H the observed isotope effect will tend
towards 1. Thus, isotope effects that are less than the intrinsic
isotope effect require that both of the rate constants kr and k2
as for P450 2B1, the 11.05 Å separating the 4 and 4′ methyl
groups of dimethylbiphenyl substrate is sufficiently large to
prevent equilibration of these two groups with respect to the
catalytically active oxygen in the time frame of the hydrogen
atom abstraction.
be slow relative to, or of the same order of magnitude as, k5H
.
This appears to be the case for 4,4′-dimethylbiphenyl which
gives a small isotope effect relative to the intrinsic isotope effect
for the trideutero substrate. Thus, the rate constant k2 is likely
to be slow or of the same order of magnitude as k5H for these
reactions.
The results of the molecular dynamics simulations are shown
in Table 6. As described in the Experimental Section the
simulations for o-xylene were conducted for 500 ps whereas
those for the other substrates were run for 1 ns. In general, the
aromatic rings remain perpendicular to the iron-oxygen bond
during the course of the dynamics simulations for each substrate.
The main motion of the molecules is a translational motion
which allows each methyl group to approach the active-oxygen
species. Thus, the molecule swings back and forth, placing first
one than the other methyl group in position for oxidation. In
the initial analysis of the run, the distance of every hydrogen
atom of the substrate from the iron-oxygen bond was deter-
mined. Since the aromatic ring adopted an orientation coplanar
to the heme, the average calculated distances from the oxygen
atom were the smallest for the aromatic hydrogen. However,
previously published experimental studies with the substrates
used in this investigation together with several additional
metabolic studies35-37 indicate that the major metabolic pathway
for substituted toluene is benzylic hydroxylation. Benzylic
hydroxylation accounts for greater than 90% of the metabolic
turnover of these substrates. We therefore did not include
aromatic hydrogen in our molecular dynamics simulations and
only considered the distances of the iron-oxygen from the
methyl hydrogen.
The number of interchanges between the methyl groups were
1172, 779, 3, and 0 for o-xylene, m-xylene, p-xylene, and 4,4′-
dimethylbiphenyl, respectively, when normalized to 1 ns. The
trend in the data is exactly what would be expected. As the
distance between intramolecular methyl groups increases, the
number of methyl group interchanges decreases, and the
intramolecular isotope effect gets smaller. The relationship
between the number of interchanges and the degree of masking
is hyperbolic and is given by the equation for the expression of
an intramolecular isotope effect,10 eq 1. The kinetic scheme
outlined in Scheme 1 describes the simplest scheme that
accounts for our observations. The rate constants k1 and k2 are
for the rates of binding and debinding of substrate to the enzyme
to form either an enzyme-substrate complex that can lead to
hydrogen atom abstraction (ESH) or to deuterium abstraction
(ESD). The rate constants for hydrogen or deuterium atom
abstraction are k5H for hydrogen atom abstraction, and k5D for
deuterium abstraction. The rate constant kr determines the rate
of interchange of the two ES complexes, ESH and ESD, with
In general, these observations indicate that for o-xylene, the
rate of equilibration between the methyl and perdeuteromethyl
groups in the P450 101 active site is very fast and thus should
not contribute to any observable masking of the intrinsic isotope
effect. The rate of equilibration between the methyl groups drops
dramatically for p-xylene but is apparently still rapid enough,
relative to bond breaking, to allow the observation of an isotope
effect that is close in magnitude to the intrinsic isotope effect.
In contrast, no switching and a masked isotope effect of 2.7 is
observed for 4,4′-dimethylbiphenyl during the time course of
the dynamics simulations. These results suggest that an in-
tramolecular distance of 11.05 Å is large enough to slow methyl
group interchange to an extent that is sufficient to extensively
mask the intrinsic isotope effect. In fact, eq 1 allows the rotation
rate to be zero and the isotope effect to still be slightly unmasked
by k2 being approximately equal in rate to k5h.
Conclusions
Both the computational and experimental results obtained with
P450 101 in this study are entirely consistent with the
experimental results found in the previous study.18 Together the
two studies provide strong evidence in support of the hypothesis
that the degree of intrinsic isotope effect masking is directly
related to the distance between chemically equivalent but
isotopically distinct intramolecular catalytic sites, i.e., the
distance hypothesis. The results of the computational studies
provide additional support for establishing computational analy-
sis as a valid methodology for understanding and predicting
protein/small molecule interactions. Several other investigators
have also conducted similar parallel experimental and theoretical
studies with P450 101 and successfully predicted metabolite
profiles, the salient examples being norcamphor,33 cis-â-
methylstyrene,38 and nicotine.30 However, to our knowledge this
is he first example where the trend in masking an intrinsic
isotope effect has been qualitatively predicted from molecular
dynamics simulations.
Finally, an additional interesting feature to emerge from this
study is the indication that it might be possible to synchronize
the clock associated with defined conditions of molecular
dynamics runs with relative methyl group interchange rates and
the magnitude of the observed isotope effect. The most telling
(35) Hanzlik, R. P.; Hogberg, K.; Moon, J. B.; Judson, C. M. J. Am.
Chem. Soc. 1985, 107, 7164-7167.
(36) Riley, P.; Hanzlik, R. P. Xenobiotica 1994, 24, 1-16.
(37) Higgins, L.; Bennett, G. A.; Shimoji, M.; Jones, J. P. Biochemistry
1998, 37, 7039-7046.
(38) Fruetel, J. A.; Collins, J. R.; Camper, D. L.; Loew, G. H.; Ortiz de
Montellano, P. R. J. Am. Chem. Soc. 1992, 114, 6987-6993.

46 J. Am. Chem. Soc., Vol. 121, No. 1, 1999
Audergon et al.
data in this regard are that collected for p-xylene. The observa-
tion of an intrinsic isotope effect for hydroxylation of this
substrate requires that a number of interchanges must occur
between the enzyme-substrate complexes prior to product
formation. Three such interchanges are predicted to occur over
a 1 ns molecular dynamics simulation. The fact that only three
interchanges occur over the 1 ns of simulation, even though
experimental observation indicates that the intrinsic isotope
effect is completely unmasked, strongly suggests that shorter
simulation times would be inadequate. On the other hand, the
zero interchanges observed for 4,4′-dimethylbiphenyl over 1 ns
of simulation, coupled to its extensively masked intrinsic isotope
effect suggests that the rate of interchange will be sufficiently
slow so that the effects of masking begin to become observable.
Thus, it appears that longer simulations would be required to
define a substrate that would have a partially masked isotope
effect. While it is clear that simulation time cannot be directly
related to wall clock time, these simulations and the conditions
under which they were run provide a lower limit for the
simulation time required to sample active site motion.
Figure 2. The structures and carbon framework numbering system of
L-(1S)- and D-(1R)-norcamphor.
and 4,4′-dimethylbiphenyl metabolites. Mass spectral conditions were
as follows: 50 ms dwell time, -70 eV ionizing voltage, and 200-205
°C source temperature.
The deuterium incorporation of each substrate was determined by
using the same GC parameters as those used for the respective
metabolites or by bleeding the compound through the reference inlet.
The mass spectral conditions were identical to those used for the
metabolites except that the ionizing voltage was -12.5 eV. The
measured ion intensity of each ion was corrected for the natural isotopic
2
abundance of H, 13C, 14C, 18O, and 29Si,41 and isotope effects were
determined as described previously.10,12
Experimental Section
Chemicals. o-Methyltoluate, p-methyltoluate, pentane, silica gel,
dimethylphthalate, terephthalic acid, sodium borohydride, sodium boro-
deuteride, pthalic dicarboxaldehyde, terephthaldehyde, 4,4′-dimethyl-
biphenyl chromium trioxide, and Diazald were obtained from Aldrich
Chemical Co., and lithium aluminum deuteride and lithium aluminum
hydride were obtained from Fluka. Organic solvents were purchased
from J.T. Baker and were of analytical grade. N-methyl-N-(tert-
butyldimethylsilyl) trifluoroacetamide (MTBSTFA) was purchased from
Pierce, and p-toluenesulfonyl chloride was obtained from Eastman
Kodak. Biochemicals were obtained from Sigma.
Synthesis of Substrates. Selectively deuterated xylene isomers and
4,4′-dimethylbiphenyl, o-xylene-R-2H3, p-xylene-R-2H3, 4-2H3,4′-di-
methylbiphenyl, o-xylene-R-2H2-R′-2H2, p-xylene-R-2H2-R′-2H2, 4-2H2,4′-
2H2-dimethylbiphenyl, o-R-2H1-R′-2H1, p-xylene-R-2H1-R′-2H1, 4-2H1,4′-
2H1-dimethylbiphenyl were synthesized as was described previously.18
Enzyme Preparation. Pseudomonas putida was grown according
to literature procedure.39 The cell free lysate that contained P450 101
was stored frozen in a saturated solution of camphor. The camphor
was removed just prior to use by eluting the solution through a Sephadex
G-15 column with 50 mM potassium phosphate buffer, pH 7.
Incubation Conditions. An aliquot of camphor-free phosphate buffer
from the Sephadex G-15 column eluant described above, sufficient to
contain 10 nmol of P450 101,40 was added to each incubation vial,
along with NADH, 12.3 µmol, and substrate (xylenes 4.08 mM and
4,4′-dimethylbiphenyl 0.67 mM) in a total volume of 2 mL of 50 mM
potassium phosphate buffer, pH 7. Incubations were run for 30 min at
27 °C then terminated by the addition of 5 mL of pentane and stored
at -78 °C until analysis.
Derivatization and Analysis by GC/MS. The frozen incubation
mixtures were thawed, and the aqueous layer was extracted with 2 ×
5 mL portions of pentane. The pentane layers were pooled, dried over
sodium sulfate, and concentrated under a stream of nitrogen gas to 30
µL. The samples were derivatized with 150 µL of 60% MTBSTFA in
acetonitrile at 65 °C for 3 h and analyzed by GC/MS. GC/MS analysis
was performed on a VG7070 mass spectrometer, interfaced to a HP-
5710A GC fitted with a 15 m J & W DB-5 fused silica capillary column
and operated in the selected ion-monitoring mode. The derivatized
xylenes were cold trapped at 40 °C, the temperature was raised linearly
at 15 °C/min to 135 °C followed by isothermal elution for 5 min. For
the derivatized biphenyls an initial temperature of 100 °C for 1 min
followed by a linear ramp at 15 °C/min to 250 °C followed by
isothermal elution for 5 min was used. The [M - 57]+ ion of the silyl
derivative was monitored for the various xylene isomer metabolites
Computational Methodology
A. Validation of the Model with D-(1R)- and L-(1S)-Norcamphor.
A detailed description of the methods is given by Jones and Korzekwa.42
The semi-empirical quantum chemical Austin method (AM1), as
provided in the MOPAC 6.0 programs was used to calculate the
minimized geometry, the charges, and the heat of formation of all the
molecules. The unrestricted Hartree-Fock (UHF) Hamiltonian was used
for all open-shell calculations (radicals). The protoporphyrin parameters
were developed, with an oxygen bound to the iron 1.7 Å above the
heme, as previously reported.30 Briefly, the coordinates of the heme-
iron-sulfur complex were obtained from the crystal structure of P450
101 with substrate bound (Brookhaven data bank, 2cpp.pdb). The full
porphyrin complex was used including all of the side chains and the
anionic forms of the propionate groups. The cysteine residue was
replaced with a thiomethane anion and the methyl group oriented in a
configuration similar to that of the methylene group of cysteine. The
charges were determined with the Gaussian 90 ab initio package, the
LANL1MB basis set. Natural population analysis was done by the
method of Reed and Weinhold.43
Molecular dynamic simulations were performed using the AMBER
4.0 suite of programs. The Cartesian coordinations for all of the atoms
except the ferryl oxygen and hydrogen atoms used for the initial D-(1R)-
norcamphor complex with P450 101 were obtained from the X-ray
crystal structure of D-(1R)-norcamphor bound to P450 101 (Brookhaven
data bank, 7cpp.pdb). In the case of L-(1S)-norcamphor the substrate
was docked in a configuration similar to that of D-(1R)-norcamphor,
by overlaying the oxygen atoms and the C2 carbons and orienting C1
close to C3, Figure 2.
The belly option was used to limit the dynamic motion to the
substrate and the 53 amino acids located within a radius of 13 Å above
the heme iron, forming an hemisphere representing the “belly”. All
calculations were performed with a nonbonded cut-off distance of 8
Å, and the nonbonded pair list was updated every 0.1 ps. A constant
dielectric function was used for electrostatic calculations. The energy
of the entire system was minimized with 1500 steps of steepest descent
followed by a conjugated gradient minimization until the convergence
gradient for either energy (2 cal/mol) or norm of gradient energy (2
cal/mol/Å) was reached. The system was warmed and stabilized in three
stages of 10 ps each at 100 K, 200 K, and 300 K, respectively.
(41) Korzekwa, K. R.; Howald, W. N.; Trager, W. F. Biomed. EnViron.
Mass Spectrom. 1990, 19, 211-217.
(42) Jones, J. P.; Korzekwa, K. R. Methods Enzymol. 1996, 272, 326-
(39) Gunsalus, I. C.; Wagner, G. C. Methods Enzymol. 1978, 52, 166-
188.
(40) Omura, T.; Sato, R. J. Biol. Chem. 1964, 239, 2370-2378.
335.
(43) Reed, A. E.; Weinstock, R. B.; Weinhold, F. J. Chem. Phys. 1985,
83, 735-746.

P450 101 Catalyzed Benzylic Hydroxylation
J. Am. Chem. Soc., Vol. 121, No. 1, 1999 47
Simulations were performed over 500 ps on each enantiomer at 300
K, and the Cartesian coordinates of each atom in the enzyme-substrate
complex were stored every 0.1 ps. A constant temperature for the
simulation was maintained by weakly coupling a thermal bath with a
0.1-ps strength constant. The rms changes in total energy and
temperature stabilized at less then 13.6 Kcal/mol and 6.0 K respectively,
indicating that the system was reasonably equilibrated.
Molecular dynamic simulations were analyzed using the Tripos
Associates SYBYL program. Previous publications have used the
substrate hydrogen-ferryl oxygen distance and the corresponding
carbon-hydrogen-ferryl oxygen angle as defining steric criteria for
establishing when a given hydrogen’s position has reactive geom-
etry.32,33 These two criteria were applied simultaneously to the
simulation results to define the reactivity of each of the various
hydrogens; (1) the hydrogen atom on a specific carbon atom had to be
close enough to the ferryl oxygen atom to be abstracted (less than 3.5
Å), and (2) the angle defined by the ferryl oxygen, hydrogen, and carbon
(O-H-C) had to be linear with a deviation less than (70°. Only one
reactive position is allowed at one time. If more than one site meets
the above steric criteria, an electronic criterion is invoked to select the
reactive position. The electronic criterion is based on the relative
stability of the radical intermediates (see below). Finally, if two
hydrogens from the same carbon are potentially reactive, the one nearest
to the ferryl oxygen atom is chosen.
B. Molecular Dynamics of Isomeric Xylenes and 4,4′-Dimethyl-
biphenyl. The molecular dynamics of the isomeric xylenes and 4,4′-
dimethylbiphenyl were studied by using the protocols established above.
Numerous initial docking positions were tested for p-xylene in the active
site of P450 cam. However, after 10 ps of molecular dynamics, all of
the initial docking positions converged to approximately the same loca-
tion within the active site. This location, in which p-xylene is oriented
for catalysis, was then used as a starting position for molecular dynamics
calculations. Subsequently, it was also used as a template to locate the
starting positions for o-xylene, m-xylene, and 4,4′-dimethylbiphenyl.
To obtain a reasonable sample of the possible conformational space
available to the various substrates, dynamic simulations were performed
for 500 ps for o-xylene but extended to 1 ns for m-xylene, p-xylene,
and 4,4′-dimethylbiphenyl. For the same reason, molecular dynamics
studies were conducted at 600 K instead of 300 K. The final run
temperature of 600 K was reached by warming each system in 4 stages
of 10 ps each (100 K, 200 K, 300 K, and 600 K). The rms changes in
total energy and temperature for all of the molecules were similarly
stabilized at less than 16.9 Kcal/mol and 11.6 K, respectively, indicating
that the systems were reasonably equilibrated.
The distance between each benzylic hydrogen and the perferryl
oxygen was measured. The interchange, or switching, between the two
methyl groups with respect to the perferryl oxygen was defined by the
following criteria: (1) how often a hydrogen atom from a given methyl
group that is poised for catalysis (within 3.5 Å of the perferryl oxygen)
is replaced by a hydrogen from the alternate methyl group within the
time allocated (up to 1 nsec) and (2) when the selected hydrogen forms
a carbon-hydrogen-oxygen angle that deviates less than (70° from
linearity.
The relative radical energies for each carbon site were calculated
from the heat of formation of the radicals. This method had previously
been used by Harris and Loew32 in the assessment of the relative
stability of primary, secondary, and tertiary radicals. The relative
probability of radical formation was calculated using eq 2, and
represents the electronic reactivity of a given carbon site.
Acknowledgements. This research was supported in part by
the National Institutes of Health (Grants ES 09122 (J.P.J.) and
GM 36922 (W.F.T.) and in part by the Swiss National Science
Foundation (C.A.). We also thank the reviewer who suggested
that we present eq 1 and Scheme 1 to describe our results and
Ken Korzekwa for his helpful comments.
n1/n2 ) exp(∆∆Hf/RT)
(2)
where ∆∆Hf is equal to the relative radical stability shown in Table 1
at sites n1 and n2, R the gas constant (0.001987 Kcal/deg mol]), and
T the absolute temperature in degrees kelvin (K). Finally, the contribu-
tion of the steric and electronic reactivity factors are used to predict
the regio- and stereospecificity of D-(1R)- and L-(1S)-norcamphor
hydroxylation.
JA983000W