An Efficient Route to Selective Bio-oxidation Catalysts: an Iterative Approach Comprising Modeling, Diversification, and Screening, Based on CYP102A1

Article abstract of DOI:10.1002/cbic.201100067

Perillyl alcohol is the terminal hydroxylation product of the cheap and readily available terpene, limonene. It has high potential as an anti-tumor substance, but is of limited availability. In principle, cytochrome P450 monooxygenases, such as the self-sufficient CYP102A1, are promising catalysts for the oxidation of limonene or other inert hydrocarbons. The wild-type enzyme converts (4R)-limonene to four different oxidation products; however, terminal hydroxylation at the allylic C7 is not observed. Here we describe a generic strategy to engineer this widely used enzyme to hydroxylate exclusively the exposed, but chemically less reactive, primary C7 in the presence of other reactive positions. The approach presented here turns CYP102A1 into a highly selective catalyst with a shifted product spectra by successive rounds of modeling, the design of small focused libraries, and screening. In the first round a minimal CYP102A1 mutant library was rationally designed. It contained variants with improved or strongly shifted regio-, stereo- and chemoselectivity, compared to wild-type. From this library the variant with the highest perillyl alcohol ratio was fine-tuned by two additional rounds of molecular modeling, diversification, and screening. In total only 29 variants needed to be screened to identify the triple mutant A264V/A238V/L437F that converts (4R)-limonene to perillyl alcohol with a selectivity of 97%. Focusing mutagenesis on a small number of relevant positions identified by computational approaches is the key for efficient screening for enzyme selectivity. Successive rounds of modeling (MD simulations of enzyme-substrate complexes to identify hotspots for selectivity), diversification (design of minimal library) and screening of a minimal library is shown to be an efficient approach to shift and maximize regioselectivity of CYP102A1, and thus to generate the valuable oxidation product perillyl alcohol from cheap and readily available limonene.

Full text of DOI:10.1002/cbic.201100067

DOI: 10.1002/cbic.201100067
An Efficient Route to Selective Bio-oxidation Catalysts: an
Iterative Approach Comprising Modeling, Diversification,
and Screening, Based on CYP102A1
Alexander Seifert, Mihaela Antonovici, Bernhard Hauer, and Jꢀrgen Pleiss*[a]
Perillyl alcohol is the terminal hydroxylation product of the
cheap and readily available terpene, limonene. It has high po-
tential as an anti-tumor substance, but is of limited availability.
In principle, cytochrome P450 monooxygenases, such as the
self-sufficient CYP102A1, are promising catalysts for the oxida-
tion of limonene or other inert hydrocarbons. The wild-type
enzyme converts (4R)-limonene to four different oxidation
products; however, terminal hydroxylation at the allylic C7 is
not observed. Here we describe a generic strategy to engineer
this widely used enzyme to hydroxylate exclusively the ex-
posed, but chemically less reactive, primary C7 in the presence
of other reactive positions. The approach presented here turns
CYP102A1 into a highly selective catalyst with a shifted prod-
uct spectra by successive rounds of modeling, the design of
small focused libraries, and screening. In the first round a mini-
mal CYP102A1 mutant library was rationally designed. It con-
tained variants with improved or strongly shifted regio-,
stereo- and chemoselectivity, compared to wild-type. From this
library the variant with the highest perillyl alcohol ratio was
fine-tuned by two additional rounds of molecular modeling, di-
versification, and screening. In total only 29 variants needed to
be screened to identify the triple mutant A264V/A238V/L437F
that converts (4R)-limonene to perillyl alcohol with a selectivity
of 97%. Focusing mutagenesis on a small number of relevant
positions identified by computational approaches is the key
for efficient screening for enzyme selectivity.
Animal studies suggest that perillyl alcohol might slow the
growth of pancreatic,[1] mammary,[2] colon,[3] and liver tumors.[4]
Therefore, perillyl alcohol research is sponsored by the Nation-
al Cancer Institute (NCI), and the compound is undergoing
phase II clinical trials. It can be obtained by the selective 7-hy-
droxylation of limonene, the cheap and readily available ter-
pene that is the product of the condensation of two molecules
of isoprene. The control of site-selectivity among similar CÀH
bonds is one of the key challenges in developing CH oxidation
catalysts. The most difficult application in site-selective CÀH
bond functionalization is the selective oxidation of a CÀH
bond that is neither positioned adjacent to a directing group
nor is inherently more reactive than alternate sites.[5] The selec-
tive 7-hydroxylation of limonene represents such a case, as
both C7 and C10 are at terminal allylic positions, and there are
five more-reactive positions (three allylic ring positions, and
two double bonds for epoxidation). Cytochrome P450 mono-
oxygenases (CYPs) are known to catalyze the oxidation of CÀH
bonds under environmentally friendly conditions by utilizing
molecular oxygen.[6,7] The screening of microbial sources is
widely applied to find selective enzyme catalysts. Previously,
the extensive screening of 1800 isolates led to the identifica-
tion of CYP153A11, an enzyme that is capable of the selective
conversion of limonene to perillyl alcohol.[8]
tase,[9] and is relatively stable under process conditions.[10]
Therefore, it is one of the most promising CYPs for application
in preparative synthesis. The wild-type enzyme is a highly
active fatty acid hydroxylase,[11] and has been shown to convert
(4R)-limonene into a variety of oxidation products. However,
perillyl alcohol is not among them.[12] Improving or changing
enzyme properties (like selectivity) with the help of random
mutant libraries typically also demands a large screening
effort, because of the huge size of the combinatorial libraries
and the lack of simple high throughput screening or selection
assays. Different strategies to enrich mutant libraries and there-
fore reduce the screening effort have been reported, for exam-
ple saturation mutagenesis at substrate-interacting residues
deduced from the crystal structure of an enzyme–substrate
complex.[13] However, the restricted availability of enzyme–sub-
strate complexes clearly limits the applicability of this strategy.
Another method is to focus libraries around the complete sub-
strate-binding pocket;[14] this method benefits from relatively
small and well defined substrate-binding pockets, as observed
in lipases, esterases, and transaminases. However, applying
these methods to CYP enzymes with their large, diverse, and
flexible substrate binding cavities[15–17] involves again a huge
screening effort. Hence, we developed a semi-rational ap-
proach that drastically decreases the screening effort. We took
Because of the high effort required for screening for selec-
tive enzymes from microbial sources, we have been working
on a generic approach that turns a well characterized enzyme,
the self-sufficient CYP102A1 from Bacillus megaterium (also
known as P450-BM3), into a selective catalyst for the conver-
sion of limonene to perillyl alcohol. The enzyme is a soluble
fusion protein of the monooxygenase and a diflavin reduc-
[a] A. Seifert, M. Antonovici, Prof. Dr. B. Hauer, Prof. Dr. J. Pleiss
Institute of Technical Biochemistry, University of Stuttgart
Allmandring 31, 70569 Stuttgart (Germany)
Fax: (+49)711-685-3196
E-mail: juergen.pleiss@itb.uni-stuttgart.de
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cbic.201100067.
1346
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2011, 12, 1346 – 1351

Selective Bio-oxidation Catalysts
advantage of the fact that enzymes control selectivity by stabi-
lizing the substrate in defined orientations that expose only se-
lected positions of the substrate to the active site, and we fo-
cused mutagenesis on residues involved in selectivity control.
We previously studied extensively enzyme–substrate interac-
tions by molecular dynamics simulations in order to elucidate
the molecular basis of selectivity control in CYP enzymes and
mutants.[18,19] The results suggested that structural elements
close to the active heme oxygen harbor residues that have a
major influence on selectivity. The systematic comparison of
this region in 31 CYP crystal structures and over 6300 sequen-
ces further showed that common selectivity-determining resi-
dues (hotspots) can be identified from the structures and se-
quences.[20] This study showed that hydrophobic residues are
preferentially found at these substrate-interacting hotspot po-
sitions, while others (e.g., charged amino acids) are not found
at all. Thus we constructed a minimal and highly enriched
CYP102A1 mutant library by introducing five hydrophobic
amino acids (alanine, valine, phenylalanine, leucine, and isoleu-
cine) at two hotspot positions (87 and 328).[12] The library con-
tained almost exclusively active mutants. Against a selection of
four cyclic and acyclic terpenoid substrates including (4R)-limo-
nene, eleven variants showed either a strong shift or improved
regio- or stereoselectivity during oxidation of at least one sub-
strate, as compared to wild-type CYP102A1. In a further study,
the screening of this library with inert cyclic and acyclic alkanes
showed variants with high selectivity for the 2-hydroxylation of
octane.[21]
Results
Round 1: Identification of CYP102A1 variants with shifted
product spectra
The minimal and highly enriched CYP102A1 mutant library (24
single and double mutants at positions 87 and 328) was
screened for enzyme variants that had the ability to convert
(4R)-limonene to perillyl alcohol. While conversion by the wild-
type enzyme did not result in perillyl alcohol, variant A328V
formed 27% of the desired product. In order to further im-
prove the selectivity of this variant we studied the impact of
the residues at hotspot positions 87 and 328 on substrate ori-
entation. Two variants, F87I/A328I and F87I/A328V, showed
similar selectivity; two further variants (F87L/A328L and A328L)
resulted in 10–15% perillyl alcohol production, and were not
considered for the next round of protein engineering.
Round 2: Improving selectivity of CYP102A1-A328V for
perillyl alcohol formation
To reveal the molecular basis of the observed selectivity of var-
iant A328V, the binding of (4R)-limonene to its active site was
studied by molecular dynamics simulations. As oxidation was
observed at both ends of the Y-shaped (4R)-limonene (the C8=
C9 double bond, and C7; Scheme 1), molecular dynamics (MD)
The screening of the minimal library with (4R)-limonene re-
vealed a total of six different products, and (4R)-limonene-8,9-
epoxide was identified as the major product.[12] However, the li-
brary also contained unselective variants that, unlike the wild-
type enzyme, produced perillyl alcohol as a minor product.
The work presented here constitutes an extension of the con-
cept of the minimal library, and focuses on those variants that
show perillyl alcohol formation. In order to increase selectivity
for perillyl alcohol formation, the variant with the highest per-
illyl alcohol ratio was selected, and the molecular basis of se-
lectivity of this variant was investigated by molecular dynamics
simulations. Based on observed binding modes, additional
amino acid positions involved in substrate binding in this com-
plex were identified. These positions were addressed by site-
directed mutagenesis with the aim of shaping the substrate
binding cavity of CYP102A1 to promote selective exposure of
C7 of (4R)-limonene to the activated heme oxygen, again by al-
lowing substitutions with solely the hydrophobic amino acids
alanine, valine, phenylalanine, leucine, and isoleucine. In sum-
mary, our approach combines the screening of our previously
introduced minimal CYP102A1 mutant library, and subsequent
fine tuning of the most successful variant by successive rounds
of molecular modeling, mutant design, and screening, while
keeping the numbers of generated and screened variants to a
minimum.
Scheme 1. Limonene, p-cymene, TIPMC and their derivatives.
ChemBioChem 2011, 12, 1346 – 1351
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org
1347

J. Pleiss et al.
simulations were performed by
starting with two initial orienta-
tions. In orientation I the methyl
group harboring C7 was pointed
towards the heme center, while
in orientation II the isopropene
group was pointed towards the
heme center. In variant A328V,
access to the activated oxygen is
restricted by the bulkier amino
acids phenylalanine and valine
at hotspot positions 87 and
328, respectively. To allow flexi-
bility, five MD simulations for
each initial orientation were car-
ried out, and the approach of
the substrate to the activated
heme oxygen was monitored.
During the 80 ns simulation
only C6, C7, C9, and C10 of the
substrate were at a distance
ꢀ4.0 ꢂ from the heme (for
0.08 ns, 0.59 ns, 0.3 ns, and
3.14 ns, respectively; Figure S1
in the Supporting Information).
In two simulations that started
from orientation I, and one sim-
ulation that started from orien-
tation II, the substrate ended up
in a stable conformation close
to the activated heme oxygen
(Figure S2). The two stable con-
formations after starting from
orientation I were very similar,
as indicated by an all-atom
Table 1. Selectivity of (4R)-limonene-converting mutants. Errors are <10%.
CYP102A1
variant
(4R)-limonene-1,2-
epoxide [%]
(4R)-limonene-8,9-
epoxide [%]
isopiperi-
tenol [%]
carveol
[%]
perillyl
alcohol [%]
A328V
5
10
7
10
3
54
78
57
55
27
1
5
0
3
2
0
0
3
1
4
3
0
0
27
10
25
24
60
97
A328VL/437A
A328VL/437I
A328VL/437V
A328VL/437F
A328VL/437F/A264V
2
Figure 1. Snapshots of stable binding orientations of (4R)-limonene (blue/white) after 10 ns of simulation in var-
iant A328V starting from A) orientation I and B) orientation II. The side chains of amino acids in hot spot positions
87 and 328 (orange) strongly restrict access to the activated heme oxygen (red sphere). In both simulations the
substrate is sandwiched between F87, V328 and L437. The size of the amino acid in position 264 controls the
available space in the immediate vicinity of the activated heme oxygen (magenta ellipse). White arrows indicate
the entrance route of the substrate to the active site.
RMSD of 0.7 ꢂ for limonene after fitting both structures. 81%
of the oxidation products observed for variant A328V derived
from oxidation at the two ends of the (4R)-limonene molecule
(Table 1; Scheme 1). These products can be explained by the
two productive binding modes observed in the simulations. In
both stable complexes the substrate was sandwiched between
the two hotspot amino acids (87, 328) and a leucine residue
(L437, Figure 1). This result indicated that, in addition to the
amino acids in positions 87 and 328, the amino acid at posi-
tion 437 strongly influences the orientation of the substrate in
variant A328V, and therefore the variant’s selectivity.
alanine at position 437 increased the selectivity to 60%
(Table 1). In addition, double mutant A328V/L437F showed,
almost exclusively, oxidation at the two ends of the Y-shaped
(4R)-limonene molecule.
Round 3: Improving CYP102A1-A328V/L437F selectivity for
perillyl alcohol formation
The double mutant in complex with (4R)-limonene was mod-
eled by starting from orientations similar to those for A328V.
The MD simulations revealed binding orientations of the sub-
strate very similar to those observed for variant A328V (data
not shown). The observed stable conformations identified an
additional position with the potential to affect selectivity for
perillyl alcohol formation. Position 264 confined the substrate
binding cavity in the I-helix region, and restricted the available
space in the immediate vicinity of the activated heme oxygen
(Figure 1). As the two ends of the (4R)-limonene molecule
clearly differ in their bulkiness, the size of the amino acid side
chain in position 264 determines the size of the substrate
moiety of the Y-shaped (4R)-limonene that can reach to the
activated heme oxygen. The methyl group harboring C7 occu-
As it is rather difficult to predict the impact on substrate ori-
entation of a particular amino acid substitution at position
437, a second minimal library was designed to identify the op-
timal amino acid at position 437. As substrate-interacting
amino acids in cytochrome P450 monooxygenases are prefer-
entially hydrophobic,[20] isoleucine, phenylalanine, alanine, and
valine were chosen as substitutes for leucine at position 437,
and these were introduced into variant A328V. The small
double-mutant library was screened for perillyl alcohol forma-
tion. While alanine, valine, and isoleucine did not improve the
selectivity of the enzyme for perillyl alcohol formation, phenyl-
1348
www.chembiochem.org
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2011, 12, 1346 – 1351

Selective Bio-oxidation Catalysts
pies considerably less space than the branched isopropene
moiety that contains the double-bond forming carbons (C8
and C9). This feature of the substrate molecule, and the bind-
ing orientation observed for variants A328V and A328V/L437F
reveal a strategy for how to prevent 8,9-epoxide formation
and to increase the selectivity for 7-hydroxylation: the size of
the pocket in the direct vicinity of the heme centre can be
reduced by substituting A264 with the slightly more bulky
valine. While a smaller pocket would be expected to hinder
the approach of the more bulky isopropene moiety to the ac-
tivated heme oxygen, the smaller methyl group would be
expected to remain able to reach the activated heme oxygen
and therefore to be selectively oxidized. As the size of the
side-chain at position 264 is limited (because of the proximity
to the heme center), larger hydrophobic residues were not
tested at this position. As expected, the substitution of A264
with valine in variant A328V/L437F strongly increased selectivi-
ty for the 7-hydroxylation of (4R)-limonene to 97% (Figure S3).
The product formation rate for perillyl alcohol with purified
variant A264V/A328V/L437F within the first minute was 72Æ
8 mmolminÀ1 mmolÀ1 P450, and 37Æ6% of the substrate was
converted to perillyl alcohol after two hours.
mutants, and that the positions that are controlling selectivity
for (4R)-limonene are also driving selectivity towards TIPMC.
The lower selectivity for the terminal hydroxylation for TIPMC
in comparison to (4R)-limonene and p-cymene is probably at-
tributable to absence of the planar character of the hexane
ring.
Discussion
Cytochrome P450 monooxygenases catalyze oxidation reac-
tions at physiological conditions by using molecular oxygen. In
contrast to conventional oxidation chemistry, this constitutes
an environmentally friendly alternative. In addition, there is an
increased demand for biologically produced flavorings and
aromatic compounds. The latter are often selective oxidation
products of terpenes and terpenoids—both common
CYP102A1 substrates. The fatty acid hydroxylase CYP102A1 is
one of the most promising monooxygenases for application in
preparative synthesis.[22] Various mutations (often in the sub-
strate entrance region) have been described that widen the
substrate scope toward more bulky molecules[23,24] However,
for many of the substrates the wild-type enzyme (or variants
with extended substrate spectra) show low selectivity, or the
desired product is not obtained. Thus, an efficient strategy to
design an enzyme with desired selectivity is needed.
Selectivity towards molecules of similar shape
To determine the impact of the three mutations on regioselec-
tivity towards molecules of similar shape but with different in-
trinsic reactivity, (4S)-limonene, p-cymene, and trans-4-isopro-
pyl-1-methylcyclohexane (Scheme 1) were converted by variant
A264V/A328V/L437F.
Directed evolution is a powerful method to change and im-
prove enzyme properties such as activity, stability and selectivi-
ty.[25] However, this approach relies on effective screening
assays, because a large number of variants have to be generat-
ed and screened. This is straightforward with assays for activity
or stability. However, if directed evolution is applied to im-
prove regio-, chemo-, or stereoselectivity, the reaction products
have to be quantitatively analyzed. Even though high-through-
put screening methods for selectivity on the basis of GC analy-
sis have been developed,[26] increasing efforts have been un-
dertaken to limit the size of mutant libraries by focusing muta-
genesis on potential substrate-interacting residues, selected by
examination of the crystal structure information.[13,14] However,
because of the large and diverse substrate binding cavity of
CYP enzymes,[15–17] the number of potential substrate-interact-
ing residues is of the order of ten; thus a combinatorial library
would be too large to be screened. However, by an integrated
approach of sequence analysis, molecular modeling, and
screening of focused libraries, we drastically reduced the size
of mutant library to be screened. After only three rounds of
modeling, design, and screening we were able to extend the
product spectra of CYP102A1 to include perillyl alcohol
(round 1: screening of the previously designed minimal
CYP102A1 library). Subsequently, we identified two additional
mutations at position 437 (round 2) and position 264
(round 3), where amino acid substitutions increased the selec-
tivity for the new product up to 97%. This is among the high-
est regioselectivities for terminal hydroxylation ever reported
for the comprehensively investigated CYP102A1 enzyme.[27]
With our approach, the screening of only 29 variants was suffi-
cient to turn a generally unselective enzyme (which does not
show terminal hydroxylation activity) into a catalyst that
The (S)-enantiomer of limonene was preferentially hydroxy-
lated at C7 by the triple mutant, though with a slightly lower
selectivity (87%), as compared to (4R)-limonene.
In contrast to (4R)-limonene, p-cymene contains a benzyl
ring and is therefore entirely planar. The reactivity of C7 in p-
cymene was comparable to that of C7 in (4R)-limonene. In con-
trast to the wild-type enzyme, which forms preferentially p-
cymene-8-ol (75%) and only small amounts of the 7-hydroxy
product (<10%), mutant A264V/A328V/L437F converted p-
cymene-like (4R)-limonene with high regioselectivity for C7
(99%) to p-cymene-7-ol.
In contrast to (4R)-limonene and p-cymene, trans-4-isopro-
pyl-1-methylcyclohexane (TIPMC) is not planar and contains no
double bonds. Therefore the molecule is more bulky and less
reactive, especially at C7. The highly unselective wild type
enzyme hydroxylates TIPMC preferentially at ring positions
(and a minor product derived from hydroxylation at C8), how-
ever, formation of the 7-hydroxy product was not observed. In
variant A328V a strong elevation of the 8-hydroxy product at
the expense of the ring-hydroxylation was observed, and this
variant also showed hydroxylation at the non-activated posi-
tion C7 to form TIPMC-7-ol (7%). The substitution of leucine
by phenylalanine at position 437 increased the ratio for the 7-
hydroxy product to 27%. However, conversion of this substrate
by the triple mutant A264V/A328V/L437F was not observed, as
access to the heme oxygen was too narrow. This clearly shows
that even non-activated terminal carbons are oxidized by our
ChemBioChem 2011, 12, 1346 – 1351
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org
1349

J. Pleiss et al.
The final reaction mixture (1 mL) contained Tris-HCl (50 mm,
pH 7.5), <2% (v/v) DMSO, substrate (0.5 mm), CYP enzyme
(0.5 mm), a cofactor (NADPH) recycling system (5 U glucose-6-phos-
phate dehydrogenase) and b-d-glucose-6-phosphate Na salt
(100 mm). The reaction was started by adding NADPH (100 mL,
2 mm). The mixture was left for 1, 5, 10, 20, 30, 60, and 120 min at
308C with gentle mixing, and the reaction was terminated by
adding HCl (20 mL, 37%) prior to extraction.
equals CYP153A11, a specialized enzyme for terminal hydroxyl-
ation.[28]
The initial activity measured for perillyl alcohol formation by
the most selective variant is in the same range as that ob-
served for CYP153A11. Nevertheless, there is room for improve-
ment of activity. As the substrate entrance to the active site
cavity has a strong influence on activity, we expect that muta-
tions in this region (far away from the heme) could improve
activity drastically without interfering with selectivity. As an al-
ternative, directed evolution could be applied to improve over-
all stability, solubility, and activity, as was shown previously for
CYP102A1 to improve its activity toward limonene and other
compounds.[29]
GC-MS analysis: Diethyl ether was used to extract the aqueous re-
action mixture twice. The organic phase was dried over magnesi-
um sulfate. Analysis of the reaction products was performed on a
Shimadzu QP2010 GC/MS with EI-ionization; the GC was equipped
with a FS-supreme-5 capillary column (length: 30 m, internal diam-
eter: 0.25 mm, film thickness: 0.25 mm). For the analysis of (4R)-li-
monene, (4S)-limonene, p-cymene, trans-4-isopropyl-1-methylcyclo-
hexane, and their oxidation products, the GC was programmed as
follows: 808C, 2 min. iso; 58CminÀ1 to 1808C; 308CminÀ1 to
2508C; injector temperature 2508C.
The conversion of molecules of similar shape but with differ-
ent intrinsic reactivity confirmed our observations from model-
ing: the orientation of the substrate close to the activated
heme oxygen is a major determinant of regio- and chemose-
lectivity. Because of this shape selectivity, nonactivated carbon
atoms are oxidized even in the presence of other reactive
atoms. This is in agreement with previous reports that showed
that CYP102A1 can be engineered for hydroxylation at the
nonactivated terminal positions of alkanes.[30]
The oxidation products, carveol, (4R)-limonene-8,9-epoxide, and
perillyl alcohol were identified by comparison to authentic sam-
ples. (4R)-limonene-1,2-epoxide, p-cymene-7-ol, p-cymene-8-ol,
trans-4-isopropyl-1-methylcyclohexane-7-ol ,and trans-4-isopropyl-
1-methylcyclohexane-8-ol were identified by comparison of their
characteristic mass fragmentation patterns in the NIST mass spec-
trometry database.[34] Isopiperitenol was identified by comparison
of the mass spectra to literature data.[35] Regioselectivity was deter-
mined from gas chromatograms by integrating and comparing the
product peaks.
As mutants of CYP102A1 have been shown to accept a wide
range of substrates,[31] and the minimal library was highly
enriched for selective catalysts towards cyclic and acyclic al-
kanes,[21] we are confident that our approach is generic and
can be successfully applied to develop highly selective oxida-
tion catalysts for a wide range of substrates.
For quantitative GC analysis the FID response was calibrated for
perillyl alcohol. Mixtures of potassium phosphate buffer [50 mm,
pH 7.5, containing perillyl alcohol final concentration: 30–500 mm)]
and nerol (100 mm) as an internal standard were extracted with di-
ethyl ether. The ratio of the area corresponding to the perillyl alco-
hol to that of the internal standard was plotted against perillyl al-
cohol concentration to give a straight line.
Experimental Section
Yeast extract and tryptone/peptone from caseine were purchased
from Roth (Karlsruhe, Germany). NADPH tetrasodium salt was ob-
tained from Codexis (Jꢀlich, Germany). Trans-4-isopropyl-1-methyl-
cyclohexane was purchased from Chemos (Regenstauf, Germany).
All other chemicals used in this work were purchased from Fluka
or Sigma and were of analytical grade or higher.
Molecular modeling
Molecular dynamics simulations: Molecular dynamics (MD) simula-
tions were carried out to study how (4R)-limonene is accommodat-
ed in the active site cavity of CYP102A1 variants A328V and
A328V/L437F. The mutations A328V and L437F were introduced
into the structure of CYP102A1 (PDB ID: 1BU7, chain A) with the
Pymol 0.99 program.[36] The (4R)-limonene molecule was manually
placed in the active site cavity. The complexes were simulated in
explicit water by using the Amber 9 molecular dynamics simulation
program.[37] The system was coupled to an external pressure and
temperature bath.[38] The Amber force field ff03 was used, and the
partial charges of (4R)-limonene and the heme group including the
activated oxygen were assigned by ab initio calculations and re-
strained electrostatic potential (RESP) fitting as described previous-
ly.[19] During the first 600 ps the protein was restrained by a har-
monic potential, while the substrate remained free. The restraints
were gradiently decreased, and the system was finally simulated
unrestrained for 10 ns. The systems were equilibrated after 1.4 ns,
and the last 8 ns of each simulation was used for analysis.
Mutant expression: Wild type CYP102A1 and its mutants were
heterologously expressed from pET22b and pET28a+ vectors in
E. coli as reported previously.[32] For the introduction of site direct-
ed mutations, the QuikChange site-directed mutagenesis kit from
Stratagene was used.
Enzyme purification: Purification was done by metal-affinity chro-
matography by using Ni-NTA matrices from Qiagen. Protein lysates,
filtered through a 1–2 mm filter, were applied to a 5 mL bed
volume column that was pre-equilibrated with purification buffer
[potassium phosphate (50 mm, pH 7.5), KCl (300 mm), imidazole
(20 mm), and PMSF (0.1 mm)]. Nonspecifically bound proteins
were washed from the column with four column volumes of purifi-
cation buffer containing 80 mm imidazole. The bound protein was
eluted with purification buffer containing imidazole (200 mm). Puri-
fied sample was dialyzed overnight against potassium phosphate
buffer [2 L, 50 mm, pH 7.5, containing PMSF (0.1 mm)], and frozen
at À308C until use.
Acknowledgement
Enzyme activity measurements: CO-difference spectra measure-
ments with an extinction coefficient of 91 mmÀ1 cmÀ1 were used to
determine CYP concentrations as described elsewhere.[33] CYP activ-
ity was determined by measuring product formation over time.
Financial support by the Deutsche Forschungsgemeinschaft (Son-
derforschungsbereich 706) and by the Federal Ministry of Educa-
1350
www.chembiochem.org
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemBioChem 2011, 12, 1346 – 1351

Selective Bio-oxidation Catalysts
[19] A. Seifert, S. Tatzel, R. D. Schmid, J. Pleiss, Proteins Struct. Funct. Bioinf.
2006, 64, 147–155.
[20] A. Seifert, J. Pleiss, Proteins Struct. Funct. Bioinf. 2009, 74, 1028–1035.
[21] E. Weber, A. Seifert, M. Antonovici, C. Geinitz, J. Pleiss, V. B. Urlacher,
Chem. Commun. 2011, 47, 944–946.
[22] V. B. Urlacher, S. Eiben, Trends Biotechnol. 2006, 24, 324–330.
[23] K.-H. Kim, J.-Y. Kang, D.-H. Kim, S.-H. Park, S. H. Park, D. Kim, K. D. Park,
Y. J. Lee, H.-C. Jung, J.-G. Pan, T. Ahn, C.-H. Yun, Drug Metab. Dispos.
2011, 39, 140–150.
tion and Research (“Systems Biology in Pseudomonas for Industri-
al Biocatalysis”) is gratefully acknowledged.
Keywords: cytochrome
p450 monooxygenase
·
MD
simulations · minimal mutant library · oxidation · protein
engineering
[24] B. M. A. van Vugt-Lussenburg, E. Stjernschantz, J. Lastdrager, C. Oosten-
brink, N. P. E. Vermeulen, J. N. M. Commandeur, J. Med. Chem. 2007, 50,
455–461.
[25] O. Kuchner, F. H. Arnold, Trends Biotechnol. 1997, 15, 523–530.
[26] M. T. Reetz, K. M. Kꢀhling, S. Wilensek, H. Husmann, U. W. Hꢃusig, M.
Hermes, Catal. Today 2001, 67, 389–396.
[27] E. G. Funhoff, J. B. Van Beilen, Biocatal. Biotransform. 2007, 25, 186–193.
[28] E. G. Funhoff, J. Salzmann, U. Bauer, B. Witholt, J. B. van Beilen, Enzyme
Microb. Technol. 2007, 40, 806–812.
[29] R. Fasan, Y. T. Meharenna, C. D. Snow, T. L. Poulos, F. H. Arnold, J. Mol.
Biol. 2008, 383, 1069–1080.
[30] P. Meinhold, M. W. Peters, A. Hartwick, A. R. Hernandez, F. H. Arnold,
Adv. Synth. Catal. 2006, 348, 763–772.
[31] V. B. Urlacher, S. Lutz-Wahl, R. D. Schmid, Appl. Microbiol. Biotechnol.
2004, 64, 317–325.
[32] S. C. Maurer, H. Schulze, R. D. Schmid, V. Urlacher, Adv. Synth. Catal.
2003, 345, 802–810.
[33] T. Omura, R. Sato, J. Biol. Chem. 1964, 239, 2370–2378.
[34] In NIST 05 Mass Spectral Library, National Institute of Standards and
Technology, Gaithersburg, Maryland, USA.
[1] M. J. Stark, Y. D. Burke, J. H. McKinzie, A. S. Ayoubi, P. L. Crowell, Cancer
Lett. 1995, 96, 15–21.
[2] J. D. Haag, M. N. Gould, Cancer Chemother. Pharmacol. 1994, 34, 477–
483.
[3] S. A. Broitman, J. Wilkinson IV, S. Cerda, S. K. Branch, Adv. Exp. Med. Biol.
1996, 401, 111–130.
[4] J. J. Mills, R. S. Chari, I. J. Boyer, M. N. Gould, R. L. Jirtle, Cancer Res. 1995,
55, 979–983.
[5] S. Li, M. R. Chaulagain, A. R. Knauff, L. M. Podust, J. Montgomery, D. H.
Sherman, Proc. Natl. Acad. Sci. USA 2009, 106, 18463–18468.
[6] S. Panke, B. Witholt, A. Schmid, M. G. Wubbolts, Appl. Environ. Microbiol.
1998, 64, 2032–2043.
[7] E. W. van Hellemond, D. B. Janssen, M. W. Fraaije, Appl. Environ. Micro-
biol. 2007, 73, 5832–5839.
[8] J. B. van Beilen, R. Holtackers, D. Lꢀscher, U. Bauer, B. Witholt, W. A.
Duetz, Appl. Environ. Microbiol. 2005, 71, 1737–1744.
[9] L. O. Narhi, A. J. Fulco, J. Biol. Chem. 1987, 262, 6683–6690.
[10] K. Kꢀhnel, S. C. Maurer, Y. Galeyeva, W. Frey, S. Laschat, V. B. Urlacher,
Adv. Synth. Catal. 2007, 349, 1451–1461.
[11] A. W. Munro, D. G. Leys, K. J. McLean, K. R. Marshall, T. W. B. Ost, S. Daff,
C. S. Miles, S. K. Chapman, D. A. Lysek, C. C. Moser, C. C. Page, P. L.
Dutton, Trends Biochem. Sci. 2002, 27, 250–257.
[35] J. Lꢀcker, W. Schwab, M. C. Franssen, L. H. Van Der Plas, H. J. Bouw-
meester, H. A. Verhoeven, Plant J. 2004, 39, 135–145.
[36] W. L. DeLano in The PyMOL Molecular Graphics System, DeLano Scientif-
ic, San Carlos, CA, USA, 2002.
[12] A. Seifert, S. Vomund, K. Grohmann, S. Kriening, V. B. Urlacher, S. La-
schat, J. Pleiss, ChemBioChem 2009, 10, 853–861.
[37] D. A. Case, T. A. Darden, T. E. Cheatham III, C. L. Simmerling, J. Wang,
R. E. Duke, R. Luo, K. M. Merz, D. A. Pearlman, M. Crowley, R. C. Walker,
W. Zhang, B. Wang, S. Hayik, A. Roitberg, G. Seabra, K. F. Wong, F. Paesa-
ni, X. Wu, S. Brozell, V. Tsui, H. Gohlke, L. Yang, C. Tan, J. Mongan, V.
Hornak, G. Cui, P. Beroza, D. H. Mathews, C. Schafmeister, W. S. Ross,
P. A. Kollman, Amber 9, University of California, San Francisco, CA, 2006.
[38] H. J. C. Berendsen, J. P. M. Postma, W. F. van Gunsteren, A. DiNola, J. R.
Haak, J. Chem. Phys. 1984, 81, 3684–3690.
[13] T. Kubo, M. W. Peters, P. Meinhold, F. H. Arnold, Chem. Eur. J. 2006, 12,
1216–1220.
[14] M. T. Reetz, L.-W. Wang, M. Bocola, Angew. Chem. 2006, 118, 1258–
1263; Angew. Chem. Int. Ed. 2006, 45, 1236–1241.
[15] H. Li, T. L. Poulos, Nat. Struct. Biol. 1997, 4, 140–146.
´
[16] P. A. Williams, J. Cosme, A. Ward, H. C. Angove, D. Matak Vinkovic, H.
Jhoti, Nature 2003, 424, 464–468.
[17] J. K. Yano, M. R. Wester, G. A. Schoch, K. J. Griffin, C. D. Stout, E. F. John-
son, J. Biol. Chem. 2004, 279, 38091–38094.
[18] R. J. F. Branco, A. Seifert, M. Budde, V. B. Urlacher, M. J. Ramos, J. Pleiss,
Proteins Struct. Funct. Bioinf. 2008, 73, 597–607.
Received: January 28, 2011
Published online on May 17, 2011
ChemBioChem 2011, 12, 1346 – 1351
ꢁ 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.chembiochem.org
1351