Hydroxylation by the Hydroperoxy-Iron Species in Cytochrome P450 Enzymes

Article abstract of DOI:10.1021/ja038237t

Intramolecular and intermolecular kinetic isotope effects (KIEs) were determined for hydroxylation of the enantiomers of trans-2-(p-trifluoromethylphenyl)cyclopropylmethane (1) by hepatic cytochrome P450 enzymes, P450s 2B1, Δ2B4, Δ2B4 T302A, Δ2E1, and Δ2E1 T303A. Two products from oxidation of the methyl group were obtained, unrearranged trans-2-(p-trifluoromethylphenyl)cyclopropylmethanol (2) and rearranged 1-(p-trifluoromethylphenyl)but-3-en-1-ol (3). In intramolecular KIE studies with dideuteriomethyl substrates (1-d₂) and in intermolecular KIE studies with mixtures of undeuterated (1-d₀) and trideuteriomethyl (1-d₃) substrates, the apparent KIE for product 2 was consistently larger than the apparent KIE for product 3 by a factor of ca. 1.2. Large intramolecular KIEs found with 1-d₂ (k_H/k _D = 9-11 at 10 °C) were shown not to be complicated by tunneling effects by variable temperature studies with two P450 enzymes. The results require two independent isotope-sensitive processes in the overall hydroxylation reactions that are either competitive or sequential. Intermolecular KIEs were partially masked in all cases and largely masked for some P450s. The intra- and intermolecular KIE results were combined to determine the relative rate constants for the unmasking and hydroxylation reactions, and a qualitative correlation was found for the unmasking reaction and release of hydrogen peroxide from four of the P450 enzymes in the absence of substrate. The results are consistent with the two-oxidants model for P450 (Vaz, A. D. N.; McGinnity, D. F.; Coon, M. J. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 3555), which postulates that a hydroperoxy-iron species (or a protonated analogue of this species) is a viable electrophilic oxidant in addition to the consensus oxidant, iron-oxo.

Full text of DOI:10.1021/ja038237t

Published on Web 12/12/2003
Hydroxylation by the Hydroperoxy-Iron Species in
Cytochrome P450 Enzymes
R. Esala P. Chandrasena,† Kostas P. Vatsis,‡ Minor J. Coon,*,‡
Paul F. Hollenberg,*,§ and Martin Newcomb*,†
Contribution from the Department of Chemistry, UniVersity of Illinois at Chicago,
845 West Taylor Street, Chicago, Illinois 60607, and Departments of Biological Chemistry
and Pharmacology, UniVersity of Michigan, Medical School, Ann Arbor, Michigan 48109
Received August 30, 2003; E-mail: men@uic.edu
Abstract: Intramolecular and intermolecular kinetic isotope effects (KIEs) were determined for hydroxylation
of the enantiomers of trans-2-(p-trifluoromethylphenyl)cyclopropylmethane (1) by hepatic cytochrome P450
enzymes, P450s 2B1, ∆2B4, ∆2B4 T302A, ∆2E1, and ∆2E1 T303A. Two products from oxidation of the
methyl group were obtained, unrearranged trans-2-(p-trifluoromethylphenyl)cyclopropylmethanol (2) and
rearranged 1-(p-trifluoromethylphenyl)but-3-en-1-ol (3). In intramolecular KIE studies with dideuteriomethyl
substrates (1-d2) and in intermolecular KIE studies with mixtures of undeuterated (1-d0) and trideuteriomethyl
(1-d3) substrates, the apparent KIE for product 2 was consistently larger than the apparent KIE for product
3 by a factor of ca. 1.2. Large intramolecular KIEs found with 1-d2 (kH/kD ) 9-11 at 10 °C) were shown not
to be complicated by tunneling effects by variable temperature studies with two P450 enzymes. The results
require two independent isotope-sensitive processes in the overall hydroxylation reactions that are either
competitive or sequential. Intermolecular KIEs were partially masked in all cases and largely masked for
some P450s. The intra- and intermolecular KIE results were combined to determine the relative rate
constants for the unmasking and hydroxylation reactions, and a qualitative correlation was found for the
unmasking reaction and release of hydrogen peroxide from four of the P450 enzymes in the absence of
substrate. The results are consistent with the two-oxidants model for P450 (Vaz, A. D. N.; McGinnity, D.
F.; Coon, M. J. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 3555), which postulates that a hydroperoxy-iron
species (or a protonated analogue of this species) is a viable electrophilic oxidant in addition to the consensus
oxidant, iron-oxo.
The ubiquitous cytochrome P450 enzymes (P450s) catalyze
a wide range of oxidations in nature including the energetically
difficult hydroxylations of unactivated C-H positions. As the
catalysts for oxidations of drugs, pro-drugs, and xenobiotics,
as well as numerous natural substrates, P450s have attracted
considerable interest from the pharmacological community.
Their mechanisms of action are also of interest to the chemical
community because of the high-energy intermediates that must
be formed to hydroxylate a C-H bond at ambient temperature.
The result is that P450s, their models, and related oxidants are
Figure 1. The iron-oxygen intermediates formed in cytochromes P450.
the topics of thousands of publications each year.
The parallelogram represents heme.
The active sites of P450 enzymes contain iron in heme with
thiolate from protein cysteine as the fifth ligand to iron, which
reactions, creation of a molecule of water, and oxidation of
substrate. The iron-oxygen complexes produced in the latter
reaction steps are shown in Figure 1. The second reduction step
is in the Fe(III) oxidation state in the resting enzyme.1 The well-
characterized portion of the catalytic cycle1,2 of P450s involves
complexation of substrate, reduction of Fe(III) to Fe(II), and
gives a peroxo-iron species. Protonation of this intermediate
reversible binding of molecular oxygen. Subsequent rapid steps
on the distal oxygen gives a hydroperoxy-iron species. A second
in the cycle involve a second reduction, two protonation
protonation on the distal oxygen with subsequent or concomitant
loss of water gives the iron-oxo species. If the second proto-
† University of Illinois at Chicago.
nation occurs on the proximal oxygen atom, iron-complexed
hydrogen peroxide is formed, and hydrogen peroxide can be
released as product in an “uncoupled” process. Recent advances
in “cryo-reduction” methodology (radiolytic reductions at low
‡ Department of Biological Chemistry, University of Michigan.
§ Department of Pharmacology, University of Michigan.
(1) Ortiz de Montellano, P. R., Ed. Cytochrome P450 Structure, Mechanism,
and Biochemistry, 2nd ed.; Plenum: New York, 1995.
(2) White, R. E.; Coon, M. J. Annu. ReV. Biochem. 1980, 49, 315-356.
9
10.1021/ja038237t CCC: $27.50 © 2004 American Chemical Society
J. AM. CHEM. SOC. 2004, 126, 115-126
115

A R T I C L E S
Chandrasena et al.
temperatures) permitted ESR characterization of the peroxo-
iron and hydroperoxy-iron species.3,4 Iron-oxo, although widely
believed to be an oxidant, could not be detected even at ultralow
temperatures,3,4 but evidence for transient iron-oxo production
in reactions of P450 enzymes with m-chloroperoxybenzoic acid
in the absence of substrates has been reported.5,6 Iron-complexed
hydrogen peroxide has not been detected in P450 studies, but
this species is formed on the pathway to the release of hydrogen
peroxide in an uncoupled reaction.1 It is also a requisite
intermediate when P450s are “shunted” with hydrogen perox-
ide;7 in shunt reactions, oxidations are effected by P450 and
hydrogen peroxide in the absence of the reductase enzymes that
transfer electrons to the P450s.
Iron-oxo has long been thought to be the active electrophilic
oxidant in P450,1 but results from our laboratories8-12 and
others13,14 suggested that two electrophilic oxidants are formed
in P450 reactions. One oxidant would be iron-oxo, and the other
would be hydroperoxy-iron, possibly hydrogen bonded, or the
protonated analogue, iron-complexed hydrogen peroxide. Much
of the evidence for two oxidants in P450 involved studies where
multiple oxidation pathways were implicated,15 but the two-
oxidants model is not the only possible explanation for multiple
reaction pathways. Shaik, Schwartz, and co-workers reported
computational results that suggested that iron-oxo might have
two accessible spin states that, in principle, could react
differently, providing a “two-states” alternative for multiple
reaction pathways,16 although that conclusion was recently
questioned.17,18
Figure 2. Portions of HPLC traces of amides from trans-2-(p-trifluoro-
methylphenyl)cyclopropanecarboxylic acid and (S)-1-phenylethylamine:
racemic acid (left), (+)-acid (center), (-)-acid (right).
substrate of the enzyme; the conclusion was that the hydro-
peroxy-iron intermediate effects epoxidation reactions.14 In
another study, highly variable kinetic isotope effects (KIEs) were
obtained for two products formed by P450-catalyzed oxidations
of the methyl group in a hypersensitive probe substrate, and
the conclusion was that both the iron-oxo and the hydroperoxy-
iron were effecting hydroxylations with different KIEs.19 We
report here KIE studies related to the latter work involving
hydroxylation reactions of both enantiomers of a hypersensitive
cyclopropane probe with five P450 enzymes that provide further
support for the two-oxidants model.
Results
Substrates and Products. The reactions studied were P450-
catalyzed oxidations of both enantiomers of trans-2-(p-trifluo-
romethylphenyl)cyclopropylmethane (1), which was previously
shown to be a substrate for hepatic P450s.9 Oxidations occur at
the methyl position to give unrearranged alcohol 2 and rear-
ranged alcohol 3.9 The trifluoromethyl group is electron
withdrawing, and it suppresses aromatic oxidation reactions that
ultimately give phenol products; phenol products are obtained
in high relative yield with the analogous substrate lacking the
trifluoromethyl group.20
Both the two-oxidants and the two-states models can rational-
ize selected experimental results, but recent studies designed
specifically to test for the two-oxidants model provided evidence
that hydroperoxy-iron (or a hydrogen-bonded or protonated
version of this species) is an active oxidant in P450-catalyzed
reactions. In one approach, Jin et al. showed that mutant
cytochrome P450cam T252A was capable of epoxidizing alkenes
even though it does not hydroxylate camphor, the natural
(3) Davydov, R.; Macdonald, I. D. G.; Makris, T. M.; Sligar, S. G.; Hoffman,
B. M. J. Am. Chem. Soc. 1999, 121, 10654-10655.
(4) Davydov, R.; Makris, T. M.; Kofman, V.; Werst, D. E.; Sligar, S. G.;
Hoffman, B. M. J. Am. Chem. Soc. 2001, 123, 1403-1415.
(5) Egawa, T.; Shimada, H.; Ishimura, Y. Biochem. Biophys. Res. Commun.
1994, 201, 1464-1469.
(6) Kellner, D. G.; Hung, S. C.; Weiss, K. E.; Sligar, S. G. J. Biol. Chem.
2002, 277, 9641-9644.
(7) Nordblom, G. D.; White, R. E.; Coon, M. J. Arch. Biochem. Biophys. 1976,
175, 524-533.
(8) Newcomb, M.; Le Tadic-Biadatti, M. H.; Chestney, D. L.; Roberts, E. S.;
Hollenberg, P. F. J. Am. Chem. Soc. 1995, 117, 12085-12091.
(9) Toy, P. H.; Dhanabalasingam, B.; Newcomb, M.; Hanna, I. H.; Hollenberg,
P. F. J. Org. Chem. 1997, 62, 9114-9122.
The enantiomers of 1 were obtained by resolution of trans-
2-(p-trifluoromethylphenyl)cyclopropanecarboxylic acid fol-
lowed by reduction reactions.9 The analysis of the % ee of the
enantiomers of purified acid was accomplished by conversion
of the acid to an amide with (S)-1-phenylethylamine and HPLC
analysis of the mixture. Figure 2 shows the results for a
diastereomeric mixture of amides from racemic acid and amides
from the enantiomerically enriched acids. Samples used for
preparation of the substrates were 99.3% ee for the (+)-
enantiomer of the acid and 98.4% ee for the (-)-enantiomer.
(10) Vaz, A. D. N.; McGinnity, D. F.; Coon, M. J. Proc. Natl. Acad. Sci. U.S.A.
1998, 95, 3555-3560.
(11) Toy, P. H.; Newcomb, M.; Coon, M. J.; Vaz, A. D. N. J. Am. Chem. Soc.
1998, 120, 9718-9719.
(12) Newcomb, M.; Shen, R.; Choi, S. Y.; Toy, P. H.; Hollenberg, P. F.; Vaz,
A. D. N.; Coon, M. J. J. Am. Chem. Soc. 2000, 122, 2677-2686.
(13) Pratt, J. M.; Ridd, T. I.; King, L. J. J. Chem. Soc., Chem. Commun. 1995,
2297-2298.
(14) Jin, S. X.; Makris, T. M.; Bryson, T. A.; Sligar, S. G.; Dawson, J. H. J.
Am. Chem. Soc. 2003, 125, 3406-3407.
(15) Newcomb, M.; Coon, M. J.; Hollenberg, P. F. Arch. Biochem. Biophys.
2003, 409, 72-79.
Enantiomers of substrate 1 had been prepared previously,9
but the absolute configuration of the enantiomers was not
determined. The structure of the amide from the (-)-enantiomer
(16) Shaik, S.; Filatov, M.; Schroder, D.; Schwarz, H. Chem.-Eur. J. 1998, 4,
193-199.
(17) Guallar, V.; Baik, M.-H.; Lippard, S. J.; Friesner, R. A. Proc. Natl. Acad.
Sci. U.S.A. 2003, 100, 6998-7002.
(18) QM/MM computations on a large model for cytochrome P450cam (ca. 6000
atoms MM, ca. 130 atoms QM) found considerable electronic differences
for spin states when the P450 model included hydrogen-bonded carboxylates
on the macrocycle. The spin shifts, which bias the energies of the spin
states, are a fundamental consequence of including a larger portion of the
enzyme in the QM model.17
(19) Newcomb, M.; Aebisher, D.; Shen, R.; Chandrasena, R. E. P.; Hollenberg,
P. F.; Coon, M. J. J. Am. Chem. Soc. 2003, 125, 6064-6065.
(20) Atkinson, J. K.; Hollenberg, P. F.; Ingold, K. U.; Johnson, C. C.; Le Tadic,
M. H.; Newcomb, M.; Putt, D. A. Biochemistry 1994, 33, 10630-10637.
9
116 J. AM. CHEM. SOC. VOL. 126, NO. 1, 2004

Hydroxylation in Cytochrome P450 Enzymes
A R T I C L E S
Figure 4. Representative GC-MS results from oxidation of a mixture of
1-d0 and 1-d3 where the two ion channels were monitored simultaneously.
The acetate derivative of 3 elutes at 16.8 min, and that of 2 elutes at 21.1
min. The inset shows the m/e ) 239 trace from 10 to 25 min.
Figure 3. Mass spectral fragmentation patterns (CI, methane) for the acetate
derivatives of alcohols 2 and 3.
diluted mixtures were analyzed by GC-MS (Table S1 in
Supporting Information). The ratios of d0 to d2 products found
were the same in these samples. With a sample diluted by a
factor of 10 000, the signal-to-noise ratio was so poor that
accurate product ratios were not obtained. The dilution study
shows that GC-MS measurements were made within the linear
range of the instrument.
of the acid and (S)-1-phenylethylamine was determined by X-ray
crystallography.21 The absolute configuration of the (-)-acid
is (R,R).
The resolved carboxylic acids were converted to substrates
1 by the method previously reported.9 Thus, LiAlH4 reduction
of the resolved acids gave alcohols 2. Conversion of alcohols 2
to the corresponding mesylates at low temperature followed by
LiBHEt3 reduction afforded substrates 1-d0. The dideuterio-
methyl substrates, 1-d2, were prepared in the same manner but
with LiAlD4 in the initial reduction. For the preparation of
trideuteriomethyl substrates, 1-d3, LiAlD4 and LiBDEt3 were
P450-Catalyzed Oxidations. Oxidations were conducted with
P450s 2B1, ∆2B4, ∆2B4 T302A, ∆2E1, and ∆2E1 T303A.10,22,23
These are mammalian hepatic enzymes that were expressed in
E. coli and purified. P450 reductase also was expressed in E.
coli and purified.22 P450 ∆2B4 and P450 ∆2E1 have short
deletions at the N-terminal ends of the proteins that have no
apparent effect on the catalytic activity. In the mutants, a highly
conserved active-site threonine was replaced with alanine.10,23
Active-site threonine is thought to be involved in the protonation
reactions of the iron-oxygen intermediates, and the two wild-
type-mutant pairs oxidize multiple reactive site substrates with
different regioselectivities.10,24
1
employed. The H NMR spectra of both enantiomers of 1-d3
showed <1% hydrogen content at the methyl positions,
confirming the high deuterium incorporation.
Analytical Methods. The products from P450-catalyzed
oxidations were analyzed by GC to determine yields, and GC-
MS was employed to determine isotopic composition. With
electron impact ionization, the ion produced from ring-opened
alcohol 3 efficiently cleaves the allyl group that contains the
isotopic label, and a molecular ion is not observed. Thus, the
alcohol products were converted to acetate derivatives and
analyzed with chemical ionization (CI) techniques. In the CI
mass spectra of the acetates from both 2 and 3, a significant
peak at (M - 19)+ (m/e ) 239) from loss of one fluorine atom
is observed (Figure 3). This convenient result ensures that the
isotopic compositions of the fragments we analyzed were the
same as those of the alcohol products.
In GC-MS analyses of the acetate derivatives of 2 and 3, we
continuously monitored selected ion channels and integrated the
peaks. This method avoids problems in determining isotope
ratios when the isotopomers are partially resolved by GC
separation. For studies with 1-d2, the mono- and dideuterio
products were formed, and we monitored the mass 240 and 241
ion channels. For studies with mixtures of 1-d0 and 1-d3, we
monitored the mass 239 and 241 ion channels. Typical results
are shown in Figure 4.
Oxidation reactions were conducted at 10 °C for 30 min.
Under the conditions we employed, the P450 enzymes remained
active over the course of the oxidations (Figure S1 in Supporting
Information). A series of trial runs wherein (R,R)-1-d0 at varying
concentrations was oxidized by P450 2B1 in a crude Michaelis-
Menten study indicated that saturation conditions were obtained
when the substrate was ca. 100 µM (Figure S2 in Supporting
Information). To ensure that, for intermolecular KIE studies,
the KIEs were in the kcat terms and not in KM, reactions were
conducted with substrate concentrations of ca. 500 µM. Control
reactions wherein substrate 1 in reaction mixtures containing
buffer but lacking enzymes was treated with 2 and 4 µmol of
H2O2 gave no oxidized products, ensuring that H2O2 released
by the enzymes (see below) did not complicate the results.
Two types of oxidation studies were performed. In one set
of studies, 1-d2 was oxidized to obtain intramolecular KIEs.
Duplicate experiments were conducted for each enzyme-
substrate combination (Supporting Information). The weighted
averages were determined to give the results in Table 1 where
we have multiplied the measured ratios of d2 to d1 products by
2 to obtain the apparent KIE values. Also listed in Table 1 are
the ratios of the apparent KIEs for products 2 and 3. The
The precision of the method over a range of concentrations
was tested with one of the product mixtures from P450 2B1
oxidation of a mixture of 1-d0 and 1-d3. The yields of oxidized
products were largest with this enzyme and were up to 10 times
greater than the amounts found with other P450 enzymes.
Following the standard GC-MS analysis, samples of the product
mixture were diluted by factors up to 1000, and the resulting
(22) Hanna, I. H.; Teiber, J. F.; Kokones, K. L.; Hollenberg, P. F. Arch. Biochem.
Biophys. 1998, 350, 324-332.
(23) Vaz, A. D.; Pernecky, S. J.; Raner, G. M.; Coon, M. J. Proc. Natl. Acad.
Sci. U.S.A. 1996, 93, 4644-4648.
(21) Chandrasena, R. E. P.; Newcomb, M.; Surve, B. C.; Wink, D. J.,
unpublished results.
(24) Toy, P. H.; Newcomb, M.; Hollenberg, P. F. J. Am. Chem. Soc. 1998,
120, 7719-7729.
9
J. AM. CHEM. SOC. VOL. 126, NO. 1, 2004 117

A R T I C L E S
Chandrasena et al.
Table 1. Apparent Kinetic Isotope Effects for Oxidations of
Substrates 1-d2 at 10 °Ca
c
enzymeb
substrate
KIEapp 2
KIEapp 3
KIE ratio
2B1
(R,R)-1
(S,S)-1
(R,R)-1
(S,S)-1
(R,R)-1
(S,S)-1
(R,R)-1
(S,S)-1
(R,R)-1
(S,S)-1
10.72 ( 0.02
11.16 ( 0.03
9.19 ( 0.01
9.52 ( 0.01
9.83 ( 0.02
9.76 ( 0.02
9.96 ( 0.02
8.48 ( 0.01
9.85 ( 0.03
9.71 ( 0.04
8.64 ( 0.04
9.54 ( 0.01
7.53 ( 0.04
8.15 ( 0.04
7.67 ( 0.02
8.24 ( 0.05
8.72 ( 0.05
7.0 ( 0.2
1.24
1.17
1.22
1.17
1.28
1.18
1.14
1.21
1.22
1.19
∆2B4
∆2B4 T302A
∆2E1
Figure 5. Arrhenius plots for variable temperature KIE studies. Product 2
results are solid circles and solid lines, product 3 results are open circles
and dotted lines, P450 2B1 results are black, P450 ∆2B4 T302A results
are red, and the 95% confidence interval for product 2 oxidations by P450
2B1 is gray.
∆2E1 T303A
8.08 ( 0.02
8.14 ( 0.06
a Apparent KIEs are the ratios of d2 to d1 products multiplied by 2. b P450
enzymes. c Ratio of KIE for product 2 to that for product 3.
Table 3. Hydrogen Peroxide Release from P450 Enzymes at
30 °Ca
enzymeb
pH 6.8
pH 7.4
Table 2. Apparent Kinetic Isotope Effects in Competitive
Oxidations of 1-d0 and 1-d3 at 10 °C
∆2B4
∆2B4 T302A
∆2E1
17.9
34.6
76.1
115
10.6c
19.4c
28.3
58.5
enzymea
substrate
KIEapp 2
KIEapp 3
KIE ratiob
2B1
(R,R)-1
(S,S)-1
(R,R)-1
(S,S)-1
(R,R)-1
(S,S)-1
(R,R)-1
(S,S)-1
(R,R)-1
(S,S)-1
2.29 ( 0.04
1.71 ( 0.01
1.99 ( 0.04
2.32 ( 0.01
3.70 ( 0.01
2.62 ( 0.01
8.5
8.3
8.03 ( 0.16
8.18 ( 0.01
1.80 ( 0.04
1.28 ( 0.01
1.68 ( 0.01
1.63 ( 0.01
3.18 ( 0.01
2.31 ( 0.01
8.2
6.2
7.32 ( 0.01
7.69 ( 0.01
1.3
1.3
1.2
1.4
1.2
1.1
∆2E1 T303A
∆2B4
a Turnover in nmol of H2O2/nmol of P450/min. Average of six
determinations from two separate experiments. b P450 enzymes. c The results
for P450s ∆2B4 and ∆2B4 T302A at pH 7.4 were previously reported; see
ref 28.
∆2B4 T302A
∆2E1c
Reactions were conducted between 10 and 40 °C, and the results
are given in Table S2 (Supporting Information). The (AH/AD)
values for these data were in the range expected for cases where
tunneling effects are not biasing the results.27 Values of log-
(AH/AD) were within the 95% confidence interval of zero (Figure
5), and the largest deviation from zero was positive; in the region
where tunneling is found for H and not D, log(AH/AD) has a
negative value.27
∆2E1 T303A
1.1
1.1
a P450 enzymes. b Ratio of KIE for product 2 divided by that for product
3. c Low yields of d2 products were obtained in these reactions, and the
values should be considered approximate.
consistency of the ratios of KIEs is an important point that will
be discussed later.
The second type of experiment was conducted to determine
intermolecular KIEs in the hydroxylation reactions, which are
not necessarily the same as intramolecular KIEs due to masking
effects. Intermolecular KIEs could be determined by comparing
the results of independent experiments with 1-d0 and 1-d3, but
we used a design where oxidations of the two substrates were
conducted in one experiment. In this manner, there can be no
differences in the activity of the P450 enzymes, and (unlikely)
differences in product release rates do not affect the results.
Again, the results (Table 2) are weighted averages. Reactions
were conducted with all P450s. The yields of product 3 from
oxidations of 1-d3 catalyzed by P450 ∆2E1 were too small to
permit accurate quantitation, however, and approximate values
for the KIEs with this enzyme are listed.
The intermolecular KIE results appear to be more random
than the intramolecular KIE results because extensive mask-
ing25,26 occurred with P450 2B1, ∆2B4, and ∆2B4 T302A.
Nonetheless, the ratio of KIEs for products 2 and 3 converged
at 1.2, reflecting an underlying uniformity in the results that
was present in the intramolecular KIE data.
In discussion, we will highlight the uncoupling reactions of
P450 enzymes where NADPH is consumed and H2O2 is
released; such uncoupling reactions can “unmask” a masked
intermolecular KIE. Hydrogen peroxide release was determined
for reactions of P450s ∆2B4, ∆2B4 T302A, ∆2E1, and ∆2E1
T303A in the absence of substrate, and the results are given in
Table 3.28 In qualitative terms, ∆2E1 released more H2O2 than
∆2B4, and both mutants were more active than their wild-type
parents. In addition, the ∆2B4 T302A mutant has been found
to generate more H2O2 than wild-type ∆2B4 in the presence of
typical substrates,28 and preliminary results with p-substituted
phenols also reveal this trend with the ∆2E1 pair of cyto-
chromes.29 Hydrogen peroxide release from P450 2B1 in the
absence of substrate at pH 7.5 was reported to be 33 nmol/
nmol of enzyme/min,30 but we caution it is not appropriate to
compare the results from different laboratories without stan-
dardization experiments.
Discussion
A key objective of the present study was to attempt to
distinguish between the “two-oxidants” model and the “two-
states” model for P450-catalyzed oxidations. Evidence that
multiple reaction pathways are available for P450-catalyzed
oxidations has been accumulating for years,15 and these two
The intramolecular KIE values in Table 1 are large, suggest-
ing the possibility that H-atom tunneling effects were complicat-
ing the results. We investigated the temperature dependence of
the KIEs in oxidations of (S,S)-1-d2 by two P450 enzymes to
evaluate whether changes in tunneling effects were important.
(27) Kwart, H. Acc. Chem. Res. 1982, 15, 401-408.
(28) Vatsis, K. P.; Peng, H. M.; Coon, M. J. J. Inorg. Biochem. 2002, 91, 542-
553.
(29) Vatsis, K. P.; Coon, M. J., unpublished results.
(30) Roberts, E. S.; Ballou, D. P.; Hopkins, N. E.; Alworth, W. L.; Hollenberg,
P. F. Arch. Biochem. Biophys. 1995, 323, 303-312.
(25) Masking and unmasking are briefly discussed in the Appendix in the
Supporting Information.
(26) For an overview of masking effects in P450-catalyzed reactions, see:
Higgins, L.; Bennett, G. A.; Shimoji, M.; Jones, J. P. Biochemistry 1998,
37, 7039-7046.
9
118 J. AM. CHEM. SOC. VOL. 126, NO. 1, 2004

Hydroxylation in Cytochrome P450 Enzymes
A R T I C L E S
Scheme 1
Figure 6. Ratios of product 2 to product 3 obtained in oxidations of
substrates 1-d2. The d2 products are gray bars, and the d1 products are black
bars. TA in the enzyme name indicates the active-site Thr to Ala mutant.
models evolved to rationalize the results. In the two-oxidants
model, both iron-oxo and another electrophilic oxidant effect
oxidation reactions. We refer to the other oxidant as hydro-
peroxy-iron (FeOOH), but there is no evidence to distinguish
between hydroperoxy-iron, a hydrogen-bonded analogue of this
species, or iron-complexed hydrogen peroxide. In the two-states
model, low-spin and high-spin states of iron-oxo are proposed
to be energetically accessible.16-18
Figure 7. Ratios of cyclic to acyclic products from reactions of d0 substrates
(gray bars) and d3 substrates (black bars). TA in the enzyme name indicates
the active-site Thr to Ala mutant.
The two-oxidants and two-states models are compared in
Scheme 1 in the context of substrate 1 studied in this work. In
the two-oxidants model, iron-oxo reacts by an insertion reaction
that produces cyclic alcohol 2 directly. The hydroperoxy-iron
species effects oxidations by insertion of the elements of OH+
to give, initially, the protonated alcohol intermediate 4, that can
be deprotonated, producing 2, or can react by a solvolytic-type
heterolysis to give the ring-opened product 3 via the cationic
intermediate 5. In the two-states model,16 a low-spin state of
iron-oxo reacts by an insertion reaction to produce 2. The high-
spin state of iron-oxo reacts by a hydrogen atom abstraction
reaction that gives the radical intermediate 6, which can be
trapped in a rebound step (displacement of OH from iron) or
ring open to give 3 via radical 7. Both models accommodate
iron-oxo insertion reactions, and the difference involves the
identity of the “other” oxidant; is it hydroperoxy-iron or a second
spin state of iron-oxo?
cyclopropylcarbinyl radicals or cations,31-33 but, irrespective of
that, they can be excluded from first principles. In either model
that explains P450-catalyzed oxidations, a secondary KIE in
the corresponding putative intermediate would result in the
correct prediction of apparent KIEs for products 2 and 3 in one
set of experiments (intra- or intermolecular studies) and an
incorrect prediction of apparent KIEs for these products in the
other set of experiments. Thus, the small apparent KIE of ca.
1.2 must be due either to a small primary KIE in addition to a
large KIE or to a constant ratio of KIEs for the processes leading
to the two products.
Two-Oxidants Model. The results of the present study
support the “two-oxidants” model. This model contains three
possible isotope-sensitive reactions. In addition to KIEs for
reactions of the two oxidants, a small primary KIE will exist in
the deprotonation reactions of protonated alcohol intermediate
4. In the intermolecular KIE studies, 4a, produced by insertion
of OH+ into a C-H bond in 1-d0, will be deprotonated
somewhat faster than 4b, produced by insertion into a C-D
bond in 1-d3. Because the deprotonation reaction competes with
the heterolytic fragmentation reaction, intermediate 4a will give
more of the cyclic product 2 than intermediate 4b. Therefore,
one expects that the ratio of cyclic product 2 to rearranged
product 3 will decrease in the oxidation of the d3 substrate, as
found experimentally. In the intramolecular KIE studies, the
intermediates formed are 4c from insertion of OH+ into the
C-H bond and 4d from insertion into a C-D bond. Intermediate
4c will be deprotonated faster than intermediate 4d due to a
KIE in the deprotonation reaction, and this will result in greater
yields of the cyclic product containing two deuterium atoms,
again as observed.
Figures 6 and 7 show the ratios of products 2 to 3 obtained
in the intra- and intermolecular KIE studies, respectively. In
experiments with 1-d2, the ratio of [2]/[3] was larger for the d2
products than for the d1 products, and, in experiments with the
1-d0 and 1-d3 substrates, the [2]/[3] ratio was larger for the d0
products than for the d2 products. The changes in product ratios
are graphic representations of the larger apparent KIEs for
product 2 than for product 3. The figures also show the absolute
values of the product ratios for the various enzymes. In general,
the results in the two figures show high correspondence.
Especially noteworthy is the large change in product ratios for
the mutant enzymes in comparison to their wild-type parents.
Both the intramolecular and the intermolecular studies indicate
that two KIEs are involved in the formation of products 2 and
3. They could be two relatively large KIEs, one for formation
of each product, or one large KIE for formation of both products
with a small KIE superimposed. If the latter holds, then the
small KIE is approximately 1.2, which is on the order of a
secondary KIE, but it is important to note at the outset that the
ratio of 1.2 for the apparent KIEs for the two products cannot
possibly result from a secondary KIE. Secondary KIEs would
be expected to be quite small for reactions of aryl-substituted
(31) Newcomb, M.; Johnson, C. C.; Manek, M. B.; Varick, T. R. J. Am. Chem.
Soc. 1992, 114, 10915-10921.
(32) Newcomb, M.; Choi, S. Y.; Toy, P. H. Can. J. Chem. 1999, 77, 1123-
1135.
(33) Wiberg, K. B.; Shobe, D.; Nelson, G. L. J. Am. Chem. Soc. 1993, 115,
10645-10652.
9
J. AM. CHEM. SOC. VOL. 126, NO. 1, 2004 119

A R T I C L E S
Chandrasena et al.
Scheme 2
The qualitative results, that is, the product ratios in Figures
6 and 7, implicate formation of intermediates 4 and hydroxy-
lation by the hydroperoxy-iron species in P450. A quantitative
evaluation of the results indicates that hydroxylation of substrate
1 involved predominantly oxidation by hydroperoxy-iron and
that hydroxylation by iron-oxo was a minor process that resulted
in “noise” in the KIE values. This conclusion is based on the
consistent ratios of the apparent KIEs for products 2 and 3, the
consistent values for the total KIEs for each enzyme found in
the intramolecular studies, and the high correspondence for
deprotonation KIEs and fragmentation rate constants in reactions
of intermediate 4 obtained from the intra- and intermolecular
data.
The KIEs for product 2 were ca. 1.2 times as large as the
KIEs for product 3 for any enzyme-substrate combination,
irrespective of the absolute values. If two oxidants (iron-oxo
and hydroperoxy-iron) with the same inherent KIEs in the
oxidation steps were involved in formation of the products, the
difference in KIEs for 2 and 3 should be attenuated depending
on the amounts of products arising from each oxidant. Moreover,
if the KIEs for iron-oxo and hydroperoxy-iron differed consider-
ably, then the KIE for product 2, formed in reactions of both
oxidants, would be randomly different from that for 3, formed
only in the hydroperoxy-iron reaction. The consistent ratio of
KIEs for 2 and 3 indicates either (1) that most of the oxidation
of substrate 1 was effected by hydroperoxy-iron or (2) that
approximately equal relative percentages of intermediate 4 were
produced in all 10 enzyme-substrate combinations we studied
and the inherent KIEs for reactions of iron-oxo and hydroper-
oxy-iron are similar. The former explanation is simpler; the latter
requires an extraordinary number of coincidences.
Table 4. Kinetic Values from Intramolecular KIE Studies Analyzed
According to Scheme 2a
enzyme
2B1
substrate
kH/kD
2kh/(kh + kd)c
2kh/kF
b
d
(R,R)-1
(S,S)-1
(R,R)-1
(S,S)-1
(R,R)-1
(S,S)-1
(R,R)-1
(S,S)-1
(R,R)-1
(S,S)-1
10.0
10.8
9.0
9.4
9.2
9.3
9.7
8.3
9.2
9.0
1.24
1.17
1.22
1.17
1.28
1.18
1.14
1.21
1.22
1.19
2.5
4.7
11.2
14.1
3.4
3.0
4.5
9.4
2.3
∆2B4
∆2B4 T302A
∆2E1
∆2E1 T303A
1.5
a For reactions at 10 °C. b Intramolecular KIE from eq 1. c From eq 3.
d From eq 2.
d1 products (eq 1). The ratio of the deprotonation to heterolytic
fragmentation rate constants is given by the ratio of the two d2
products (eq 2). The ratio of deprotonation rate constants for
4-d2 and 4-d1, given by eq 3, is also equal to the observed ratios
of apparent KIEs for 2 and 3 listed in Table 1.
kH/kD ) 2(2-d2 + 3-d2)/(2-d1 + 3-d1)
2kh/kF ) 2-d2/3-d2
(1)
(2)
(3)
A comparison of the present results for substrate 1 and
previous KIE studies19 of the related substrate 8 reinforces the
above conclusion. Unlike the situation with substrate 1, the ratios
of KIEs for products 9 and 10 from substrate 8 were not constant
with various P450s. In fact, the ratios of KIEs for the two
products were not the same in intramolecular and intermolecular
KIE studies with one enzyme. Three isotope-sensitive reactions
were needed to model the results, oxidation by iron-oxo,
oxidation by hydroperoxy-iron, and deprotonation of the pro-
tonated alcohol intermediates.19
2kh/(kh + kd) ) (2-d2/3-d2)/(2-d1/3-d1)
The results in Table 4 show quite good agreement. The KIEs
for hydroxylation of 1 by the hydroperoxy-iron species in each
enzyme differ only slightly. The agreement in the KIEs for the
hydroxylation reactions of each enantiomer of 1 catalyzed by
the P450 ∆2B4 and ∆2B4 T302A wild-type/mutant pair is
excellent, with an average difference from the mean for these
four reactions of 1%. One intramolecular KIE for hydroxylation,
that for oxidation of (S,S)-1 by P450 ∆2E1, appears out of line
given the good agreement for the other reactions; it is possible
that this is a case where hydroxylation by iron-oxo was
substantial enough to perturb the overall apparent KIE notice-
ably. If the (S,S)-1 results for ∆2E1 are excluded, then the
average difference from the mean for the remaining three KIEs
for ∆2E1 and its mutant is less than 3%. It is important to note
that the uniformity in the actual KIEs for oxidation by a P450
and its mutant is a consequence of consecutive isotope-sensitive
reactions and cannot be obtained in a model with competing
isotope-sensitive reactions.
The intramolecular KIE results are analyzed for the model
of predominant oxidation by hydroperoxy-iron according to
Scheme 2. The important rate constants are those for insertion
of OH+ into a C-H bond (kH) or a C-D bond (kD), those for
deprotonation of the isotopomers of 4 by loss of a proton (kh)
or deuteron (kd), and that for heterolytic fragmentation of 4 (kF).
We make the reasonable assumption that heterolytic fragmenta-
tions of the isotopomers of 4 would have negligible secondary
KIEs.
Table 4 contains kinetic values evaluated according to Scheme
2. The intramolecular KIE is given by twice the ratio of d2 to
The KIE for the deprotonation reactions of intermediates 4,
that is, 2kh/(kh + kd), is quite consistent, as we have noted earlier
9
120 J. AM. CHEM. SOC. VOL. 126, NO. 1, 2004

Hydroxylation in Cytochrome P450 Enzymes
A R T I C L E S
Scheme 3
Table 5. Kinetic Values from Intermolecular KIE Studies Analyzed
According to Scheme 3a
b
c
e
enzyme
2B1
substrate
KIE
k
unc/kH
2kh/(kh + kd)d
2kh/kF
obs
(R,R)-1
(S,S)-1
(R,R)-1
(S,S)-1
(R,R)-1
(S,S)-1
(R,R)-1
(S,S)-1
(R,R)-1
(S,S)-1
2.1
1.6
1.9
2.2
3.5
2.5
8
8
7.8
8.0
0.09
0.04
0.08
0.10
0.24
0.13
1.2
1.5
1.12
1.26
1.27
1.34
1.18
1.42
1.16
1.13
1.0
1.3
1.10
1.06
2.5
5.3
6.6
8.3
2.4
2.9
5
6
1.9
1.5
∆2B4
∆2B4 T302A
∆2E1f
∆2E1 T303A
a For reactions at 10 °C. b Masked intermolecular KIE from eq 4. c From
eq 7. d From eq 6. e From eq 5. f The values for P450 ∆2E1 are low
precision.
in the context of the ratios of KIEs. This value translates into
a rate constant for loss of D+ from 4-d1 that is about 70% as
fast as loss of H+ from 4-d1 or from 4-d2, indicating a highly
exothermic proton transfer reaction that competes with the
undoubtedly fast solvolytic ring opening of 4. The KIE in the
deprotonation reaction of the protonated alcohol is specific to
the substrate, and it is predicted to be a constant in our studies
when different enzymes are employed because the only sub-
strates were the enantiomers of 1. This consistency for different
enzymes provides strong support for the two-oxidants model.
The rate constants for deprotonation of 4 (kh) are expected
to be enzyme specific, and they should display small differences
for the two enantiomers of 1 due to diastereomeric interactions
in the enzymes’ active sites. The heterolytic fragmentation
reaction should be substrate specific and constant in our studies.
Thus, changes in the 2kh/kF ratios reflect mainly differences in
the deprotonation rate constants. In general, the trends in 2kh/
kF are in reasonable agreement in that, for example, deproto-
nations in P450 ∆2B4 are much faster than in its mutant for
both enantiomers of substrate, and both enantiomers of substrate
are deprotonated slowly in P450 ∆2E1 T303A in comparison
to the other enzymes.
It is noteworthy that the deprotonation reactions of intermedi-
ates 4 are considerably retarded for both mutants in comparison
to their wild-type analogues. A modeling study of P450 2E1
indicates that Thr303 is located adjacent to bound substrates,34
and this observation offers an attractive explanation of the results
for P450 ∆2E1 and its T303A mutant. Specifically, the
conversion of Thr303 to Ala removed a basic residue from a
position that might be adjacent to the protonated alcohol moiety
in 4. This speculation could be tested by modeling 1 and 4 in
the active site of P450 2E1.
can calculate the relative rates of the two processes from the
observed KIE in the intermolecular studies and the intrinsic KIE
determined from the intramolecular studies (see below). KIE
masking of the oxidation reaction will not have an effect on
the deprotonation and heterolytic fragmentation rate constants,
and the values of kh, kd, and kF obtained from the intermolecular
KIE results should match those from the intramolecular studies.
Table 5 contains rate constants from the intramolecular KIE
studies derived from the model in Scheme 3 using eqs 4-7.
The observed KIE (KIEobs), given by eq 4, is partially masked.
The rate constants for deprotonation and heterolytic fragmenta-
tion of intermediates 4 are unaltered by the intermolecular design
of the experiments, and eqs 5 and 6 are analogues of eqs 2 and
3. The ratio of rate constants for uncoupling and hydroxylation
(kunc/kH) can be calculated with eq 7 if the intrinsic KIE value
(KIEint) is known. To calculate KIEint for each enzyme-substrate
combination, we used the intramolecular KIEs from Table 4
(kH/kD) and multiplied those values by (1.15)3 ) 1.5. The
intramolecular experiment gives a primary KIE divided by a
secondary KIE, whereas the intermolecular experiment gives a
primary KIE times two secondary KIEs,35 and we assume that
the secondary KIE is normal and approximately equal to 1.15
as found for P450-catalyzed hydroxylation of octane.36
KIEobs ) (2-d0 + 3-d0)/(2-d2 + 3-d2)
2kh/kF ) 2-d0/3-d0
(4)
(5)
2kh/(kh + kd) ) (2-d0/3-d0)/(2-d2/3-d2)
unc/kH ) (1 - KIEobs)/(KIEobs - KIEint)
(6)
(7)
k
For the intermolecular KIE studies, the model is modified as
shown in Scheme 3, where the dotted line separates reactions
occurring in different enzyme-substrate complexes. The second
reduction step in the P450 reaction sequence is irreversible,1
and, assuming that binding of substrate is not isotope sensitive,
the intermolecular KIEs would be kH/kD ) 1 (a masked KIE)
unless some other reaction of the oxidant is possible that can
“unmask” the masked KIE.25,26 In Scheme 3, the “unmasking”
reaction is the uncoupling reaction, release of H2O2 in competi-
tion with the oxidation of substrate, which has the rate constant
For a given enzyme-substrate combination, the values in
Table 5 show excellent agreement with the corresponding values
in Table 4. The average deviation from the means for the 2kh/
(kh + kd) values is 5%, and the average deviation from the means
for the 2kh/kF values is 11%. Importantly, the qualitative ranking
of 2kh/kF for each enzyme-substrate combination is essentially
unchanged in the two data sets, and the larger 2kh/kF values for
P450 ∆2B4 and P450 ∆2E1 as compared to their mutants found
k
unc. Because the rate of the uncoupling reaction relative to the
(35) Hanzlik, R. P.; Hogberg, K.; Moon, J. B.; Judson, C. M. J. Am. Chem.
rate of hydroxylation determines the extent of unmasking, one
Soc. 1985, 107, 7164-7167.
(36) Jones, J. P.; Rettie, A. E.; Trager, W. F. J. Med. Chem. 1990, 33, 1242-
(34) Park, J. Y.; Harris, D. J. Med. Chem. 2003, 46, 1645-1660.
1246.
9
J. AM. CHEM. SOC. VOL. 126, NO. 1, 2004 121

A R T I C L E S
Chandrasena et al.
from analysis of the intramolecular results are reproduced in
the analysis of the intermolecular results, albeit somewhat
attenuated.
competitions were similar to the experimental values when the
computations included tunneling using the Bell correction. From
the Bell-corrected results, the KIE for product 9 is now predicted
to be 1.22 times greater than that for product 10 in intra-
molecular studies and 1.12 times greater in intermolecular
studies.38 For comparisons to experimental results, Shaik and
co-workers averaged four of the six sets of results from the
intramolecular KIE studies of 8 and did not consider the values
from the intermolecular studies. They stated that the new
computational results are in “perfect accord with experiment”
and provide a “proof” that a two-state mechanism is operative
in the P450-catalyzed oxidations of 8.38
For P450s 2B1, ∆2B4, and ∆2B4 T302A, the large degree
of KIE masking25,26 reflects the fact that the hydroxylation
reactions of 1-d0 are 4-25 times faster than the uncoupling
reactions for these enzymes. For P450s ∆2E1 and ∆2E1 T303A,
however, release of hydrogen peroxide is faster than the
hydroxylation reaction, which results in largely unmasked values
for KIEobs that approach KIEint. It is reasonable to assume that,
to a first approximation, the absolute values of the hydroxylation
rate constants kH are similar for the various enzymes, in which
case the major changes in kunc/kH are due to changes in kunc
.
The criteria applied for these conclusions are elusive. The
ratios of intramolecular KIEs for products 9 and 10 for the six
combinations of enzymes and substrate ranged from (0.93 (
0.01) to (1.24 ( 0.02) (deviations at 2σ), and the ratios of the
intermolecular KIEs for 9 and 10 (four combinations of enzymes
and substrate) ranged from 1.3 to 2.8 with an average of 1.9.19,39
The demonstrably significant variations in the experimental KIE
ratios were not addressed in the recent computational report,
and the remarkably large errors in the two-states model
predictions for the intramolecular KIE ratios were not noted.38
In our view, the new KIE predictions do not provide convincing
evidence that the two-state model is operative in oxidations of
8, and, in fact, they indicate that it is not. As noted above, we
concluded that the results with substrate 8 could not be modeled
with only two isotope-sensitive reactions; a third KIE was
required.19
Unlike the case with 8, the results with substrate 1 can be
modeled with two isotope-sensitive reactions. Thus, any “two-
reactions” model can fit the consistent ratio of product KIEs
because the only necessary conditions are that the products are
formed in different reactions and the KIEs for these reactions
are not the same. The two-states model is inadequate in
explaining other results for substrate 1, however, that include
(1) the variable apparent KIEs for the products from reactions
of each P450 with the enantiomeric substrates, (2) the unmasking
effects found in the intermolecular studies, and (3) the changes
in product ratios observed for the wild-type and mutant enzyme
pairs. The same shortcomings exist if one attempts to rationalize
the results for 8.
In the two-states model, the only route to product 3 is ring
opening of radical 6 produced in the high-spin reaction. As a
consequence of the constant ratio of KIEs for the two products,
the model then requires that the low-spin reaction gave only
product 2. One would expect that a particular P450 enzyme
would display the same KIE for the products formed from the
two enantiomeric substrates in the intramolecular studies, but
the apparent KIEs for 2 and for 3 formed from a given enzyme
vary by up to 25%.
The variations in apparent KIEs for products 2 and 3 from
oxidation of the two enantiomers by a given enzyme cannot be
explained adequately by any model that postulates two compet-
ing reactions with different KIEs. The substrates differ only in
Therefore, the lifetimes of the FeOOH species (τ ) 1/kunc) in
P450 2B1 are up to 40 times longer than those in ∆2E1.
Knowledge of these lifetimes is important for studies that follow
reactions of the hydroperoxy-iron species directly, such as cryo-
reduction experiments.3,4
The hydrogen peroxide turnover values in the absence of
substrate correlate with the values of kunc/kH at the qualitative
level. For the enzymes studied in this work, P450 ∆2B4 released
the smallest amount of H2O2 and had the smallest value of kunc
/
kH, and P450 ∆2B4 and its mutant had smaller values of kunc
/
kH and smaller amounts of H2O2 turnover than P450 ∆2E1 and
its mutant. Hydrogen peroxide turnover/release values in the
absence and presence of substrate are not necessarily identical,
but the trend is clear that P450 ∆2E1 and its mutant appear to
be “leaky” in comparison to P450 ∆2B4 and its mutant in regard
to both the H2O2 release rates and the extent of unmasking.
This correlation provides support for oxidation by hydroperoxy-
iron and, thus, the two-oxidants model, because hydrogen
peroxide release would not unmask a KIE for iron-oxo (see
below).
Two-States Model. The two-states model has two distinct
reactions (Scheme 1), and it is conceivable at the superficial
level that all of product 2 was formed in reaction of the low-
spin state with one KIE and all of product 3 was produced in
the high-spin state reaction with a different KIE. Although this
model can fit the ratio of apparent product KIEs, it is not in
accord with other aspects of the results as detailed below.
The KIEs expected for reactions in the two-states model were
computed by Shaik and co-workers who reported that the KIE
in reactions of the high-spin state will be greater than the KIE
in reactions of the low-spin state.37 Irrespective of the extent of
reaction by the two spin states or details of the partitioning of
putative radical 6, the prediction using those results37 is that
the KIE in product 3 will be greater than the KIE in product 2.
Therefore, the ratio of cyclic to acyclic product would be
predicted to increase in the oxidation of the d3 substrate, which
is the opposite of the experimental results in Figure 6. In a
similar manner, the ratio of products is predicted incorrectly
for the intramolecular results shown in Figure 7.
The predicted37 relative KIEs also were in the wrong direction
for the products formed in P450-catalyzed oxidations of
substrate 8.19 However, after the appearance of the communica-
tion reporting results for 8, Shaik and co-workers published a
new study for this substrate in which the relative order of KIEs
for the low- and high-spin states was reversed from the original
report.38 Computed absolute KIE values for intramolecular
(38) Kumar, D.; de Visser, S. P.; Shaik, S. J. Am. Chem. Soc. 2003, 125, 13024-
13025.
(39) The intramolecular KIE results in ref 19 were from comparisons of
independent reactions of 8-d0 and 8-d3. In more recent studies, we used
the same experimental design as in the present study (i.e., competition
between 8-d0 and 8-d3) and found a range of KIE ratios from 1.4 to 2.3
with an average of 1.75 for 10 examples of enzymes and substrate
(Newcomb, M.; Chandrasena, R. E. P., unpublished results).
(37) Ogliaro, F.; Filatov, M.; Shaik, S. Eur. J. Inorg. Chem. 2000, 2455-2458.
9
122 J. AM. CHEM. SOC. VOL. 126, NO. 1, 2004

Hydroxylation in Cytochrome P450 Enzymes
A R T I C L E S
their configurations, and the basic energetics for the competing
reactions in a given P450 are expected to be the same. As we
demonstrated above, however, a model with consecutive isotope-
sensitive steps not only explains the variability in apparent KIEs
for the products but also gives consistent actual KIEs for the
initial oxidation step by a given P450. The question thus
arises: can the two-states model be salvaged by reduction to a
“one-state” model, wherein all reactions occur via the high-
spin iron-oxo (with one KIE), and the radical intermediate 6 is
distributed to products 2 and 3 in a second isotope-sensitive
step? It cannot. An isotope-sensitive trapping reaction for 6
would involve secondary KIEs from the remaining H/D atoms
on the radical, and, as noted earlier, any secondary KIE effects
will have opposite results in the intra- and intermolecular studies.
That is, any secondary KIE effect will be larger for product 2
than for product 3 in one set of studies and smaller in the other.
The unmasking effects found in the intermolecular KIE
studies are difficult to explain in the context of the two-states
model. Because the formation of the peroxo-iron intermediate
is irreversible,1 an intramolecular KIE of kH/kD ) 1 will be found
unless some other pathway for reaction of the oxidant exists
that “unmasks” the masked KIE.25,26 If the hydroperoxy-iron
intermediate is an oxidant, then release of hydrogen peroxide
(the uncoupling reaction) serves as a KIE unmasking reaction.
The uncoupling reaction cannot unmask a KIE if the iron-oxo
species is the oxidant, however, because formation of iron-oxo
by loss of water from the (FeOOH2) species is computed to be
highly exothermic and irreversible.40 Thus, a different competing
reaction is needed to unmask KIEs for iron-oxo. In principle,
another unmasking reaction is possible, an oxidase-type reaction
wherein iron-oxo is further reduced to give water, but the extent
of oxidase activity (if any) in the reactions studied here is not
known.
In regard to the changes in ratios of products 2 and 3 for the
wild-type and mutant enzyme pairs, the two-states model offers
no satisfactory explanation for why the mutation should have a
profound effect. This problem for the model also surfaces in
wild type-mutant pair studies in the form of variable regio-
selectivities in oxidations of substrates with multiple reactive
sites (see below). We noted that the two-states model requires
that essentially all of product 2 was formed by the low-spin
reaction and all of product 3 was formed in the high-spin
reaction. Therefore, the ratios of product 2 to product 3 would
directly reflect the amounts of reaction through each channel.
For the results found here, the difference in activation energies
for oxidations by the two states must change by as much as 1.0
kcal/mol in the case of oxidations of (S,S)-1 by P450 ∆2E1
and its T303A mutant. How the Thr to Ala mutation could have
such a large differential effect on the activation barriers for the
two similar processes postulated in the two-states model is
unknown.
below). This deficiency cannot be addressed with an ad hoc
rationalization that radicals were initially produced and then
oxidized to cations because cationic products are found for
cyclopropylmethyl probes that are precursors to radicals with
lifetimes as short as 2 ps12,41 when no rearranged products are
found from related probes that are precursors to radicals with
lifetimes up to 10 ns.42 A recent challenge to the model arose
from computational studies that indicated that only one spin
state of P450 iron-oxo was energetically accessible,17,18 calling
into question the premise of the two-states model for P450.
Comparison with Other Mechanistic Studies. The results
of the present work and the related study with substrate 8
provide support for the two-oxidants model and indicate that
the species we have called hydroperoxy-iron is the predominant
oxidant of substrate 1. Formation of protonated alcohol 4 by
insertion of OH+ into a C-H bond from the hydroperoxy-iron
species (or iron-complexed hydrogen peroxide) leads to re-
arrangement product 3 by a cationic process, not a radical
pathway. Both conclusions, that hydroperoxy-iron is an oxidant
form in P450 and that rearrangements involve cationic inter-
mediates, are consistent with previous mechanistic studies.
The formation of cation-derived rearrangement products in
P450-catalyzed oxidations previously was observed when
hypersensitive probe substrates 11 were used with P450 2B1
(11a)8 and with the same P450 enzymes used in the present
study (11b).12 Probe substrates 11 were designed such that a
radical intermediate will open by breaking the cyclopropyl bond
to the benzylic carbon, whereas a cationic intermediate will open
by breaking the bond to the ether-substituted carbon.41 A cation-
derived product also was found from methyl group oxidation
of methylcubane (12), which gives a distinct cationic rearrange-
ment product.12 In recent works, a cation-derived rearrangement
product was found in oxidations of norcarane (13) by a variety
of P450 enzymes.43,44 Most of the mechanistic probes that have
been applied in P450 studies give the same products from radical
and cationic rearrangements, and no conclusions about the
pathway for rearrangements of these substrates can be made.
The result is that cationic rearrangement products were found
in all cases where rearranged products were obtained from a
substrate that provided a means for differentiation between
radical and cationic intermediates.42
Oxidations by the hydroperoxy-iron intermediate in P450
were implicated in the mechanistic probe studies discussed
above. One pathway to the cationic rearrangement products
involves solvolysis reactions of protonated alcohols such as that
shown in Scheme 1 for intermediate 4, and insertion of OH+
from hydroperoxy-iron or iron-complexed hydrogen peroxide
was postulated.8 In early work, Pratt et al. concluded that iron-
Although computational in origin,16 the two-states model is
largely qualitative. It successfully rationalizes selected P450
experimental results but is inconsistent with other results,
including some important trends. For example, it predicts that
only radical-derived rearrangement products will be formed in
mechanistic probe studies16,38 and does not explain the cationic
rearrangement products consistently found in studies with probes
that distinguish between cationic and radical intermediates (see
(41) Le Tadic-Biadatti, M. H.; Newcomb, M. J. Chem. Soc., Perkin Trans. 2
1996, 1467-1473.
(42) Newcomb, M.; Toy, P. H. Acc. Chem. Res. 2000, 33, 449-455.
(43) Newcomb, M.; Shen, R. N.; Lu, Y.; Coon, M. J.; Hollenberg, P. F.; Kopp,
D. A.; Lippard, S. J. J. Am. Chem. Soc. 2002, 124, 6879-6886.
(44) Auclair, K.; Hu, Z.; Little, D. M.; Ortiz de Montellano, P. R.; Groves, J.
T. J. Am. Chem. Soc. 2002, 124, 6020-6027.
(40) Harris, D. L.; Loew, G. H. J. Am. Chem. Soc. 1998, 120, 8941-8948.
9
J. AM. CHEM. SOC. VOL. 126, NO. 1, 2004 123

A R T I C L E S
Chandrasena et al.
complexed hydrogen peroxide was an important oxidant in P450
on the basis of product differences found when H2O2 and
iodosobenzene were used in shunt reactions.13
work. The intramolecular KIE values indicate a relatively low
reactivity for the hydroperoxy-iron species. The intramolecular
KIEs we obtained are equal to a primary KIE divided by a
secondary KIE.35 Invariably, the secondary KIEs for P450-
catalyzed hydroxylations are normal, and the average value for
the secondary KIE in hydroxylation of the methyl group in
octane was kH/kD ≈ 1.15.36 If that value applies in the
functionalization of the methyl group in 1, then the primary
KIE at 40 °C for reaction of substrate (S,S)-1 with P450 2B1
would be kH/kD ) 10, which is larger than the primary KIE
found for oxidation of the methyl group in octane by the same
P450 enzyme at 37 °C (e.g., 9.1-9.2).36,48 Because the C-H
bond energy of the methyl group in substrate 1 is about 3 kcal/
mol smaller than that of a methyl group in an alkane,49 however,
hydroxylation of 1 should have a smaller KIE than hydroxylation
of octane. It appears possible, therefore, that the large KIEs
found for reaction of substrate 1 are those for the less reactive
hydroperoxy-iron species, whereas the smaller KIE found for
octane might be for reaction of a more highly reactive iron-
oxo species or for a mixture of reactions from hydroperoxy-
iron and iron-oxo. Consistent with this conclusion, the primary
KIEs determined for authentic iron-oxo intermediates produced
from porphyrin-iron complexes and sacrificial oxidants were
found to be kH/kD ) 7.5-8.7 for oxidation of the tertiary C-H
positions of adamantane at 20 °C,50 substantially smaller than
the primary KIEs found here for oxidation of substrate 1 at 10
°C (kH/kD ) 10.3-12.4).
Alternative electrophilic oxidants to iron-oxo have been
implicated in systems related to P450 enzymes, providing
additional support for the two-oxidants model. In heme oxy-
genase, the enzyme that degrades heme, a hydroperoxy-iron
species is thought to be the active species that oxidizes the heme
macrocycle.51 In several studies with iron-porphyrin complexes
that model P450 and other heme-containing enzymes, multiple
oxidants have been implicated from changes in product yields
as a function of sacrificial oxidant or experimental conditions.52-57
The conclusions in the model studies are that both iron-oxo and
a complex between iron and the sacrificial oxidant are active
oxidants, in direct analogy to the two electrophilic oxidants
postulated in the two-oxidants model.
A compelling case for hydroperoxy-iron as an oxidant is
provided by several studies of P450 wild-type enzymes and their
mutants in which the highly conserved active-site Thr was
replaced with Ala, including the P450s used in this work.15,45
From changes in the regioselectivity of oxidations of simple
alkenes (epoxide versus allyl alcohol products), Vaz et al.
concluded that P450s ∆2B4 and ∆2E1 and their threonine-to-
alanine mutants had two active electrophilic oxidants.10 In that
work, the authors suggested that the hydroperoxy-iron species
was a preferential epoxidizing agent, and iron-oxo was a
preferential hydroxylating agent. The regioselectivity changes
found in oxidations of probe 8, coupled with product ratio
changes in oxidations of probe substrate 1 by the same wild-
type and mutant pairs, further supported the two-oxidants model
with the added condition that the hydroperoxy-iron species must
be able to effect hydroxylation to some extent.11
Similar results were found with another wild-type and mutant
P450 enzyme pair.46 The relative amounts of N-dealkylation
and sulfoxidation in oxidations of p-(N,N-dimethylamino)-
thioanisole varied by a factor of 4 for P450 BM3 and its T268A
mutant (equivalent to the mutants used in the present study).
The absence of a KIE in competitive oxidations of undeuterated
substrate and its perdeuteriomethyl analogue, coupled with the
demonstration of an intrinsic KIE for N-demethylation and fast
tumbling of substrate in the active site, resulted in the conclusion
that two distinct oxidants existed that did not interchange.46 The
authors noted that, in principle, the two oxidants might be two
spin states of iron-oxo, but that can only be the case if
interconversion between the two spin states is slower than the
oxidation reaction, which is not expected.
Studies with P450cam, the best characterized P450 enzyme,
were recently reported that also indicate that the hydroperoxy-
iron species is an electrophilic oxidant.14 The P450cam T252A
mutant, which apparently does not produce iron-oxo, does not
oxidize camphor, the natural substrate of the enzyme. Jin et al.
reported that the T252A mutant was capable of epoxidizing
alkenes, however, and they concluded that epoxidation reactions
were effected by the hydroperoxy-iron intermediate produced
in the mutant.14
Conclusion
The present study adds to the growing evidence that hydro-
peroxy-iron is a second electrophilic oxidant species in cyto-
chromes P450, effecting oxidations by insertion of OH+ into a
C-H bond. The intermediate protonated alcohol thus formed
is rapidly deprotonated to give neutral alcohol products. Because
Multiple electrophilic oxidant forms also were implicated in
a recent study of the effect of anions on oxidations catalyzed
by P450 2D6 and other P450 enzymes.47 For P450 2D6,
carbonate was found to slow O-demethylation reactions but have
no effect on the kinetics of N-demethylation reactions. Most
interestingly, when the P450 was shunted with cumyl hydro-
peroxide, carbonate had no obvious effect on the rate of an
O-demethylation reaction. Because the shunt reaction is thought
to produce iron-oxo directly, the authors concluded that carbon-
ate differentially affected the production of iron-oxo in the
normal reaction sequence in comparison to production of
hydroperoxy-iron.47
(48) Jones, J. P.; Trager, W. F. J. Am. Chem. Soc. 1987, 109, 2171-2173.
Correction: 1988, 110, 2018.
(49) Halgren, T. A.; Roberts, J. D.; Horner, J. H.; Martinez, F. N.; Tronche, C.;
Newcomb, M. J. Am. Chem. Soc. 2000, 122, 2988-2994.
(50) Sorokin, A.; Robert, A.; Meunier, B. J. Am. Chem. Soc. 1993, 115, 7293-
7299.
(51) Ortiz de Montellano, P. R. Acc. Chem. Res. 1998, 31, 543-549.
(52) Kamaraj, K.; Bandyopadhyay, D. J. Am. Chem. Soc. 1997, 119, 8099-
8100.
(53) Nam, W.; Lim, M. H.; Moon, S. K.; Kim, C. J. Am. Chem. Soc. 2000,
122, 10805-10809.
Support for the two-oxidants model also is found in an
evaluation of the magnitude of the KIEs obtained in the present
(54) Nam, W.; Lim, M. H.; Lee, H. J.; Kim, C. J. Am. Chem. Soc. 2000, 122,
6641-6647.
(55) Collman, J. P.; Chien, A. S.; Eberspacher, T. A.; Brauman, J. I. J. Am.
Chem. Soc. 2000, 122, 11098-11100.
(45) Coon, M. J. Biochem. Biophys. Res. Commun. 2003, 312, 163-168.
(46) Volz, T. J.; Rock, D. A.; Jones, J. P. J. Am. Chem. Soc. 2002, 124, 9724-
9725.
(47) Hutzler, J. M.; Powers, F. J.; Wynalda, M. A.; Wienkers, L. C. Arch.
Biochem. Biophys. 2003, 417, 165-175.
(56) Wadhwani, P.; Mukherjee, M.; Bandyopadhyay, D. J. Am. Chem. Soc. 2001,
123, 12430-12431.
(57) Collman, J. P.; Zeng, L.; Decre´au, R. A. Chem. Commun. 2003, 2974-
2975.
9
124 J. AM. CHEM. SOC. VOL. 126, NO. 1, 2004

Hydroxylation in Cytochrome P450 Enzymes
A R T I C L E S
mmol) in 10 mL of CH2Cl2 containing 300 µL of Et3N at 0 °C. The
mixture was stirred at room temperature for 3 h and quenched by
addition of HCl solution. The mixture was washed with a 2 N HCl
solution (3 × 20 mL), dried over Na2SO4, and filtered. Solvent was
removed in vacuo to give crude product that was analyzed by HPLC.
Recrystallization of the sample prepared from the (-)-acid from ethyl
acetate gave a sample that was suitable for X-ray structure analysis.21
P450-Catalyzed Oxidations. Cytochrome P450 enzymes and P450
reductase were expressed in E. coli and purified as previously
described.10,22,23 A stock solution of substrate was prepared from
substrate 1 (2 mg in 100 µL of methanol) and DLPC (2 mg) in 1.0 mL
of 50 mM phosphate buffer, pH 7.4; the mixture was sonicated and
stored at 0 °C. A stock solution of NADPH was prepared by addition
of 15 mg of NADPH to 375 µL of the same buffer, and this solution
also was stored at 0 °C. Reaction mixtures were prepared by mixing 1
nmol of P450 enzyme, 2 nmol of P450 reductase, 60 µL of DLPC,
and the above buffer solution to bring the total volume to 1.9 mL at 0
°C. After standing at 0 °C for 10 min, 10 µL of the stock solution of
substrate was added. The mixture was stored for 15 min at 0 °C and
then placed in a 10 °C water bath. After 5 min at 10 °C, the reaction
was initiated by addition of 100 µL of the NADPH stock solution.
Reactions were maintained at 10 °C for 30 min and then quenched by
the addition of 2 mL of CH2Cl2. The phases were separated, and the
aqueous phase was extracted with CH2Cl2 (2 × 2 mL). The combined
organic phase was dried (MgSO4) and filtered. For quantitation studies,
an internal standard (1-phenylpropan-1-ol) was added, the mixture was
concentrated under a stream of nitrogen to ca. 0.2 mL, and the resulting
mixture was analyzed by GC (Carbowax) using flame ionization
detection.
hydroperoxy-iron is the predominant oxidant of substrate 1, we
were able to calculate the kH/kD values in both the oxidation
reaction and the deprotonation reactions of each P450 and
relative rate constants for oxidation and hydrogen peroxide
release (uncoupling reaction) in each enzyme. The relative rate
constants for fragmentation and deprotonation of intermediate
protonated alcohol 4 also were obtained, and these will be useful
for comparison to results with other P450 substrates.
We consider the complete characterization of the hydroper-
oxy-iron oxidant to be an important goal both for understanding
the biological oxidations and for chemical modeling of the
oxidants. One important question involves the detailed identity
of the oxidant: is it hydroperoxy-iron, hydrogen-bonded hy-
droperoxy-iron, or iron-complexed hydrogen peroxide? Another
question involves the reactive character of the species.
The present results implicate hydroxylation by hydroperoxy-
iron, and this is supported by previous results from our
laboratories.11,19 Other properties have been ascribed to this
species, however. Thus, Vaz et al.10 concluded that hydroperoxy-
iron was a preferential epoxidizing agent in comparison to
hydroxylation, Jin et al.14 concluded that it could epoxide alkenes
but could not hydroxylate the high-energy C-H bond in
camphor, Volz et al.46 concluded that it was a preferential
sulfoxidizing agent in comparison to oxidation of an N-methyl
group, and Hutzler et al.47 concluded that it preferentially
oxidizes an N-methyl group in comparison to an O-methyl
group. None of these conclusions are necessarily in conflict with
one another, and the collective message apparently is that
hydroperoxy-iron is a less reactive oxidant than iron-oxo.
Our results also present a caution regarding characterizations
of the reactivity of the iron-oxo species. Previous P450 reactivity
studies have been conducted with the assumption that iron-oxo
was the only oxidant, but it is quite possible that the results
reflect a mixture of the reactivities of the two electrophilic
oxidants. Methods for unambiguous production of one oxidant
and the exclusion of the other will undoubtedly be important
for fully understanding the complex oxidizing system in
cytochrome P450.
Isotopic Compositions of 2 and 3. These were determined by GC-
MS analysis of the corresponding acetates. Following the P450-
catalyzed oxidation reactions, the combined organic phases were
concentrated to ca. 0.5 mL under a nitrogen stream. Pyridine (50 µL)
and acetic anhydride (100 µL) were added, and the mixtures were
allowed to stand at room temperature overnight. The reaction mixtures
were diluted to 1.5 mL (CH2Cl2) and washed with a 1 N HCl solution
(3 × 1 mL) and a 1 N NaOH solution (3 × 1 mL). The solutions were
dried (MgSO4), filtered, and concentrated to ca. 0.1 mL under a nitrogen
stream. The mixtures were analyzed by GC-MS (Carbowax) with
methane gas chemical ionization. The mass spectra and representative
results are shown in Figures 3 and 4.
Intermolecular KIE Studies. Stock solutions of substrates contain-
ing 1-d0 and 1-d3 were prepared as described above. The ratio of
substrates in each was determined by GC-MS using chemical ionization
by monitoring the (M - 19)+ (loss of F) channels at m/e ) 181 and
184. For studies with (S,S)-1, the d0/d3 ratio was 1.25. For studies with
(R,R)-1, the d0/d3 ratio was 1.04. The ratios of d0 to d2 products were
corrected by the ratios of the substrates to give the apparent KIE values.
Determination of H2O2 Production. Reconstitution of the P450
enzymes with P450 reductase and DLPC was previously reported.28,59
Reaction mixtures (total volume 1.2 mL) were prepared in triplicate
and consisted of P450 (0.08 µM ∆ 2B4 or ∆2B4 T302A, 0.04 µM
∆2E1 or ∆2E1 T303A), reductase in an amount equal to that of the
P450, and DLPC in a 75:1 molar ratio with total protein (reductase +
P450). The protein-phospholipid mixtures were allowed to stand at
room temperature for 30 min, diluted with potassium phosphate buffer
(pH 6.8 or 7.4, final concentration 0.1 M), and equilibrated for 5 min
at 30 °C before initiation of the reactions with excess NADPH (0.7
M). Reactions were carried out at the same temperature with shaking
under air for 5 min with the ∆2E1 pair of proteins at pH 6.8 and for
8-15 min in all other cases. After termination of the reactions by
addition of perchloric acid (final concentration, 2%) and precipitation
of the denatured protein, H2O2 was quantified spectrophotometrically
as the ferrithiocyanate complex in 1.0-mL aliquots of the supernatant
Experimental Section
Substrates and Products. The preparation of enantiomerically
enriched substrates 1 and alcohols 2 and 3 followed the reported
methods.9 For dideuteriomethyl substrates 1-d2, LiAlD4 was used in
the initial reduction step. For tridueteriomethyl substrates, LiAlD4 and
1
LiBDEt3 were used for the reduction steps. The H NMR spectra of
samples of 1-d3 showed <1% protium content in the region for methyl
group absorbance. The enantiomeric purities of the samples of resolved
acids used for the preparation of 1 were determined by conversion of
the acid to an amide by reaction with (S)-(+)-2-phenylethylamine and
analysis by HPLC (C-18 column, water-acetonitrile, 1:1, v:v); Figure
2 shows the results.
N-(1-Phenylethyl)-trans-2-(p-trifluoromethylphenyl)cyclopro-
panecarboxamide. The procedure of Zhang et al. was used.58 The
above acid (0.41 g, 1.78 mmol) was added to 2 mL of SOCl2, and the
mixture was heated at reflux for 1 h. Excess SOCl2 was removed by
rotary evaporation, the residue was dissolved in 50 mL of CH2Cl2, and
the CH2Cl2 was removed by rotary evaporation. The crude acid chloride
was added to a solution of (S)-(+)-1-phenylethylamine (0.3 g, 2.5
(58) Zhang, X.; Hodgetts, K.; Rachwal, S.; H., Z.; Wasley, J. W. F.; Craven,
K.; Brodbeck, R.; Kieltyka, A.; Hoffman, D.; Bacolod, M. D.; Girard, B.;
Tran, J.; Thurkauf, A. J. Med. Chem. 2000, 43, 3923-3932.
(59) Vatsis, K. P.; Coon, M. J. Arch. Biochem. Biophys. 2002, 397, 119-129.
9
J. AM. CHEM. SOC. VOL. 126, NO. 1, 2004 125

A R T I C L E S
Chandrasena et al.
fraction.28 Zero-time controls contained all of the components of the
reconstituted enzyme system but were treated with perchloric acid
before addition of NADPH. Standard curves were constructed with
known amounts of H2O2 added to perchlorate-denatured, zero-time
mixtures. The turnover numbers listed in Table 3 are the averages of
values obtained in two separate experiments. No H2O2 was formed when
the flavoprotein or NADPH was omitted from the reaction mixtures.
16954 to P.F.H., and GM-48722 to M.N.). We thank Prof. M.-
H. Baik for providing details concerning ref 17.
Supporting Information Available: Appendix discussing
masking effects, tables of results from the dilution study,
variable temperature studies, intra- and intermolecular KIE
studies, and the results of the time-course and Michaelis-
Menten studies (PDF). This material is available free of charge
via the Internet at http://pubs.acs.org.
Acknowledgment. This work was supported by grants from
the National Institutes of Health (DK-10339 to M.J.C., CA-
JA038237T
9
126 J. AM. CHEM. SOC. VOL. 126, NO. 1, 2004