P450 fingerprinting method for rapid discovery of terpene hydroxylating P450 catalysts with diversified regioselectivity

Article abstract of DOI:10.1021/ja109590h

Engineered P450 enzymes constitute attractive catalysts for the selective oxidation of unactivated C-H bonds in complex molecules. A current bottleneck in the use of P450 catalysis for chemical synthesis is the time and effort required to identify the P45

Full text of DOI:10.1021/ja109590h

COMMUNICATION
pubs.acs.org/JACS
P450 Fingerprinting Method for Rapid Discovery of Terpene
Hydroxylating P450 Catalysts with Diversified Regioselectivity
Kaidong Zhang, Shady El Damaty, and Rudi Fasan*
Department of Chemistry, University of Rochester, Rochester, New York 14627, United States
S Supporting Information
b
ABSTRACT: Engineered P450 enzymes constitute attrac-
tive catalysts for the selective oxidation of unactivated C-H
bonds in complex molecules. A current bottleneck in the use
of P450 catalysis for chemical synthesis is the time and effort
required to identify the P450 variant(s) with the desired
level of activity and selectivity. In this report, we describe a
method to map the active site configuration of engineered
P450 variants in high throughput using a set of semisyn-
Figure 1. (A) Surface representation of the active site shape and volume
thetic chromogenic probes. Through analysis of the result-
ing ‘fingerprints’, reliable predictions can be made regarding
the reactivity of these enzymes toward complex substrates
structurally related to the fingerprint probes. In addition,
fingerprint analysis offers a convenient and time-effective
means to assess the regioselectivity properties of the finger-
printed P450s. The described approach can represent a
valuable tool to expedite the discovery of P450 oxidation
catalysts for the functionalization of relevant natural pro-
ducts such as members of the terpene family.
of P450BM3 (Reprinted with permission from ref 8. Copyright 2008
Elsevier. PDB 1BVY). (B) Chromogenic probes used to map the active
site geometry of the engineered P450 variants.
Enzyme activity profiling across multiple substrates has been
used in the context of proteases, lipases, and kinases to generate
‘fingerprints’ for enzyme identification and classification.7 Our
goal was to develop a fingerprinting method for P450 enzymes
which could provide an indirect map of the size and geometry of
their active site (Figure 1A). We envisioned the resulting
fingerprint could (a) relay information regarding the accessibility
of the active site in the enzyme and thus its substrate scope, and
(b) enable a qualitative assessment of its regio- and stereoselec-
tivity properties, as these are in large part dictated by the
configuration of the enzyme active site.8,9 To this end, we
prepared a set of semisynthetic probes based on hydrocarbon
scaffolds with marked differences in structure, size, and bulkiness
(1-5, Figure 1B). ‘Reporter’ methoxy groups (-OCH3) were
installed on these structures to enable rapid profiling of P450
function on the probes using a Purpald-based colorimetric assay,
which detects the product (formaldehyde) of P450-dependent
demethylation of this functional group.4,5 These designs were
also intended to couple the screen readout with the C-H
oxidation activity of the P450, as this has proven important
toward the isolation of P450 catalysts which exhibit more
coupled catalytic cycles and can therefore support higher sub-
strate turnover numbers.2,4
ethods for selective oxidation of aliphatic (sp3) C-H
M
bonds are of huge synthetic relevance, in particular when
viable for the late-stage oxidation of complex molecules and
natural products.1 For this purpose, the use of cytochrome P450
monooxygenases offers the advantage that the regio- and stereo-
selectivity of these catalysts can be modulated by protein
engineering and potentially directed also toward energetically
and/or stereoelectronically ‘unbiased’ sp3 C-H sites. Cyto-
chrome P450 enzymes have been isolated and engineered for a
variety of relevant applications.2-4 Recently, the systematic
utilization of P450 variants with diversified substrate profile
and regioselectivity has constituted a powerful strategy toward
the late-stage transformation of single and multiple unactivated
sp3 C-H bonds in small-molecule substrates through P450-
mediated chemoenzymatic synthesis.5 The lack of time-effective
methods to gain access to P450 catalysts with varying regio- and
stereoselectivity properties, however, currently limits the scope
of this approach. While various methods are available for high-
throughput screening of P450 activity,6 these do not provide
information regarding the regio/stereoselectivity of the screened
P450s, which has to be established on a case-by-case basis
through laborious and time-consuming HPLC or GC analyses.
In this report, we introduce a method which can address this
limitation and streamline the discovery of P450s with the desired
regioselectivity toward complex terpene substrates.
To test the viability of the fingerprinting method, we tested 10
previously described variants of the fatty acid hydroxylase
P450BM3 from Bacillus megaterium.10 These P450 variants exhibit
varying activity and selectivity in the oxidation of non-native
small-molecule substrates.5 Parallel reactions with 1-5 were
carried out using these P450s in purified form and a NADPH
cofactor regeneration system containing a thermostable phos-
phite dehydrogenase (PTDH).11 Notably, each variant was
found to be associated to a unique fingerprint, supporting the
Received: October 25, 2010
Published: February 22, 2011
r
2011 American Chemical Society
3242
dx.doi.org/10.1021/ja109590h J. Am. Chem. Soc. 2011, 133, 3242–3245
|

Journal of the American Chemical Society
COMMUNICATION
ability of these profiles to capture the functional differences
among these P450s (Figure S1). In addition, the relative activity
of each P450 variant on 1, 2, and 3 varied considerably in spite of
the comparable size of these molecules ((20 Da), indicating that
these probes can effectively report on the different geometric
constraints within the active site of these enzymes.
Encouraged by these results, we investigated the utility of this
method for high-throughput analysis of engineered P450BM3
libraries in search of variants with diversified reactivity and
selectivity properties toward terpene substrates. To this end,
triple and quadruple mutant libraries were constructed by site-
saturation mutagenesis of positions V78, S81, V82, A87, L181,
and V184 in P450BM3 variant FL#62, which exhibited high
activity on all the probes (Figure S1, Table S1). Targeting these
sites was expected to be most effective in altering the active site
configuration of FL#62 variant as the corresponding amino acid
residues project their side chains toward the heme pocket and
substrate channel of the enzyme based on the available crystal
Figure 2. Fingerprints of P450BM3 (= WT), FL#62 and 25 FL#62-
derived variants from the collection of engineered P450 catalysts. Probe
activities are normalized to those of the reference enzyme P450BM3
-
12
(F87A). Mean values and standard deviations were calculated from three
replicates.
structure of P450BM3 (Figure 1A). The P450 libraries were
expressed in 96-well plates (DH5R cells) and screened against
probes 1-5 in parallel to acquire a fingerprint for each functional
P450 variant occurring in these libraries. Reactions with the
probes (1 mM, 60 min) were carried out using cell lysate in the
presence of the phosphite/PTDH cofactor regeneration system,
and probe activity was quantified based on absorbance at 550 nm
after 60-min incubation with Purpald (30 mM). To enable
comparisons among the fingerprints, the probe activities were
normalized to that of a reference P450, P450BM3(F87A), which
has minimal yet detectable activity on all five probes (Figure S1).
Wild-type P450BM3 was unsuitable for this purpose as it shows no
activity on three out of the five fingerprint probes (Figure S1).
From the screening of ∼10 000 recombinants, a total of 1220
catalytically active P450 variants were identified (threshold:
>20% parental activity on at least one probe). Comparative
analysis of the normalized P450 fingerprints revealed the occur-
rence of 261 variants (21%) featuring a unique profile, a
representative sample (25) of which is provided in Figure 2.
Since the variation coefficient of the assay is 15%, a variation
larger than (20% in at least one of the five fingerprint compo-
nents served as criterion to define two fingerprints as distinct.
The 261 variants with unique profile were pooled to form a
collection of fingerprinted P450 catalysts for the subsequent
proof-of-principle studies.
Probes 1, 2, 4, and 5 incorporate carbon skeletons
(cyclohexane, bicyclic norbornane, bicyclic decalin, and steroid)
occurring in numerous terpenes with relevant pharmacological
activity13 and of practical value for asymmetric catalysis.14 By
reflecting the geometric compatibility between the P450 active
site and these molecular scaffolds, we expected the fingerprints to
be useful for predicting the reactivity of the P450s toward
substrates sharing the same core structure. This hypothesis was
based on the observation that in most P450s substrate recogni-
tion is primarily mediated by hydrophobic and van der Waals
interactions rather than directional H-bonds or ionic interac-
tions.15,16 In particular, the recent engineering of P450BM3 into a
propane monooxygenase illustrates how substrate-active site
complementarity is sufficient to grant catalytic proficiency on a
substrate (propane) which cannot be recognized through
H-bonding or electrostatic interactions.8
Figure 3. Fraction of catalytically active P450 variants (TTN g 100)
among the predicted active (dark blue) and predicted inactive (light
blue) members of the P450 collection based on single fingerprint
component analysis. The core structure shared by the probes and the
target substrates is highlighted in blue.
(7)), the norbornane-based probe 2 (borneol (8), camphorsultam
(9)), and the decaline-based probe 4 (11,12-dihydronootkatone
(10), sclareolide (11)) (Figure 3). The P450s in the 261-member
collection were ranked according to their predicted activity on 6-
11 using single fingerprint component analysis, where probe 1-
activity was used as predictor for6- and 7-activity, probe 2-activity as
predictor for 8- and 9-activity, and probe 4-activity as predictor for
10-and11-activity. The 40 top-ranking P450s for each substrate pair
were extracted from the collection and tested for hydroxylation
activity on the target substrates in reactions in 96-well plates (∼0.1
mol % P450, KPi pH 8.0, 16 h) followed by GC analysis. For 6, 8, 10,
which are most closely related to the probes, activity predictions
were confirmed in 87-97% of the cases (activity threshold value:
100 total turnovers, TTN), as summarized in Figure 3. Importantly,
excellent rates of correct activity predictions (90-100%) were
found also in the context of 7, 9, and 11, which are more distantly
related to the probe structures. On average, 78% of the identified
P450 variants were found to support more than 400 total turnovers
(57%, >750 TTN; 30%, >1000 TTN), indicating that the large
majority of the identified P450 catalysts could be already useful for
synthesis at preparative scale.5 The fingerprint-based activity pre-
dictions were further tested by characterizing 10 bottom-ranking
P450 variants for each pair of target substrates. These enzymes were
To test this hypothesis, we investigated a panel of target
compounds structurally related (i.e., core structure related) to
the cyclohexane-based probe 1 (pentylcyclohexanol (6), menthol
3243
dx.doi.org/10.1021/ja109590h |J. Am. Chem. Soc. 2011, 133, 3242–3245

Journal of the American Chemical Society
COMMUNICATION
Figure 4. Product distribution (GC), total turnovers (TTN), fingerprints, and amino acid mutations (vs parent enzyme FL#62) of 8-active (A), 9-active
(B), and 10-active (C) variants extracted from the P450 collection. GC peak integrations are provided in Tables S2-S4. Amino acid mutations in FL#62
(vs P450BM3) are provided in Table S1. Oxidation products a and b were not characterized. (D) Hydroxylation products isolated from the reactions with
8, 9, 10, and 7.
predicted to be inactive as their fingerprints indicated no activity on
1, 2 or 4, respectively. On average, 88% of these variants showed no
detectable oxidation activity on the target substrate (Figure 3),
further supporting the reliability of the method and indicating that
the rate of ‘missed predictions’ is, on average, less than 15%.
Altogether, these studies demonstrated the efficiency of the finger-
print analysis strategy in accelerating the discovery of synthetically
useful P450 catalysts for a variety of target compounds (6-11)
through simple inspection of their fingerprints.
targeted by hydroxylation in 6-10, the major oxidation products
were isolated from larger scale reactions (50 mg substrate, 0.2 mol %
P450, 24 h) and their identity elucidated by 1D and 2D-NMR
(1H-1H COSY, HMBC, HSQC, NOESY). The data correspond-
ing to 8 (borneol) and 10 (11,12-dihydronootkatone) are of
particular interest because they highlight two key aspects of the
overall approach. First, the isolated P450 variants were found to
target, collectively, 40% of the sp3 C-H sites occurring in these
compounds (3/7 and 4/10, respectively), including tertiary, sec-
ondary, and even primary positions as indicated in Figure 4D. While
the selectivity of these variants for some of these positions is
moderate (<40%, Tables S2-S4), these results are remarkable
considering the small number of variants tested (40) and their
straightforward identification through fingerprinting and fingerprint
analysis. Another important finding was that P450-catalyzed hydro-
xylation in these substrates occurred also at positions which are
remote with respect to the reporter functional group in the
corresponding probe (e.g., products 13, 14, 19 in Figure 4).
Altogether, these studies demonstrate the ability of the described
fingerprint-based method not only to expedite the search of P450
oxidation catalysts with diversified regioselectivity but also to enable
the discovery of P450s useful for targeting aliphatic positions across
the whole carbon skeleton of terpenes structurally related to the
fingerprint probes.
While the individual fingerprint components were useful to
predict substrate acceptance in the P450 variants, we envisioned
that inspection of the whole fingerprint could provide a means to
anticipate differences in the regioselectivity properties of these
enzymes. In P450s, the regio/stereoselectivity of the oxidation
reaction depends upon the orientation of the substrate above the
distal side of the heme prior to oxidative attack by the oxo-ferryl
porphyrin π-cation radical species.16,17 This is influenced by the
active site configuration of the P450 enzyme, which can be
mapped through the described fingerprinting approach. To test
this idea, we isolated P450 variants featuring divergent finger-
prints and compared the product distribution after reaction with
6-11. With the majority (5/6) of these compounds, P450s with
different fingerprints exhibited also important differences in
regioselectivity, as illustrated by the examples in Figure 4 (see
also Tables S2-S4). This occurred with a frequency of 43% (6),
32% (7), 42% (8), 41% (9), and 45% (10), indicating that a large
fraction (∼40% on average) of the active site changes captured
by fingerprinting affected the binding mode of these substrates
during catalysis. With 11, a sole oxidation product ((S)-3-hydroxy
sclareolide) was obtained with various P450 variants, suggesting that
the ability of this compound to adopt alternate orientations within
the enzyme active site may be limited. To shed light on the sites
Given the observed relationship between fingerprint and
regioselectivity, we anticipated that two variants sharing a similar
fingerprint would display similar selectivity in substrate oxida-
tion. To test this hypothesis, we isolated three variants from the
78/81/87 library (5-G9, 5-C4, 5-C2) possessing an identical finger-
print. Characterization of these variants revealed that these enzymes
exhibit remarkably similar product profiles across multiple substrates
(Figure 5, Table S5). While these P450s differ from each other by up
3244
dx.doi.org/10.1021/ja109590h |J. Am. Chem. Soc. 2011, 133, 3242–3245

Journal of the American Chemical Society
COMMUNICATION
’ ACKNOWLEDGMENT
We thank Frances H. Arnold for kindly providing the P450BM3
variants used in the preliminary experiments (Figure S1) and
Huimin Zhao for kindly providing vectors for the expression of
the thermostable PTDH. This work was supported by startup
funds from the University of Rochester.
’ REFERENCES
(1) (a) Cook, B. R.; Reinert, T. J.; Suslick, K. S. J. Am. Chem. Soc.
1986, 108, 7281. (b) Das, S.; Incarvito, C. D.; Crabtree, R. H.; Brudvig,
G. W. Science 2006, 312, 1941. (c) Yang, J.; Gabriele, B.; Belvedere, S.;
Huang, Y.; Breslow, R. J. Org. Chem. 2002, 67, 5057. (d) Wender, P. A.;
Hilinski, M. K.; Mayweg, A. V. Org. Lett. 2005, 7, 79. Chen, M. S.; White,
M. C. Science 2007, 318, 783. (e) Lee, S.; Fuchs, P. L. J. Am. Chem. Soc.
2002, 124, 13978. (f) Brodsky, B. H.; Du Bois, J. J. Am. Chem. Soc. 2005,
127, 15391. (g) Chen, M. S.; White, M. C. Science 2010, 327, 566.
(2) Fasan, R.; Chen, M. M.; Crook, N. C.; Arnold, F. H. Angew.
Chem., Int. Ed. 2007, 46, 8414.
(3) (a) Lewis, J. C.; Bastian, S.; Bennett, C. S.; Fu, Y.; Mitsuda, Y.;
Chen, M. M.; Greenberg, W. A.; Wong, C. H.; Arnold, F. H. Proc. Natl.
Acad. Sci. U.S.A. 2009, 106, 16550. (b) Li, S.; Chaulagain, M. R.; Knauff,
A. R.; Podust, L. M.; Montgomery, J.; Sherman, D. H. Proc. Natl. Acad.
Sci. U.S.A. 2009, 106, 18463. (c) Whitehouse, C. J.; Bell, S. G.; Tufton,
H. G.; Kenny, R. J.; Ogilvie, L. C.; Wong, L. L. Chem. Commun.
(Cambridge, U.K.) 2008, 966. (d) Zehentgruber, D.; Hannemann, F.;
Bleif, S.; Bernhardt, R.; Lutz, S. ChemBioChem 2010, 11, 713. Sun, L.;
Chen, C. S.; Waxman, D. J.; Liu, H.; Halpert, J. R.; Kumar, S. Arch.
Biochem. Biophys. 2007, 458, 167. (e) Liu, L.; Schmid, R. D.; Urlacher,
V. B. Biotechnol. Lett. 2010, 32, 841. (f) Bottner, B.; Schrauber, H.;
Bernhardt, R. J. Biol. Chem. 1996, 271, 8028. (g) Tang, W. L.; Li, Z.;
Zhao, H. Chem. Commun. (Cambridge, U.K.) 2010, 46, 5461.
(4) Peters, M. W.; Meinhold, P.; Glieder, A.; Arnold, F. H. J. Am.
Chem. Soc. 2003, 125, 13442.
(5) Rentmeister, A.; Arnold, F. H.; Fasan, R. Nat. Chem. Biol. 2009,
5, 26.
(6) Rabe, K. S.; Gandubert, V. J.; Spengler, M.; Erkelenz, M.;
Niemeyer, C. M. Anal. Bioanal. Chem. 2008, 392, 1059.
(7) (a) Goddard, J. P.; Reymond, J. L. J. Am. Chem. Soc. 2004,
126, 11116. (b) Basile, F.; Ferrer, I.; Furlong, E. T.; Voorhees, K. J. Anal.
Chem. 2002, 74, 4290. (c) Maly, D. J.; Huang, L.; Ellman, J. A.
ChemBioChem 2002, 3, 16.
Figure 5. Product distribution (GC), TTNs, fingerprints, and amino
acid mutations (vs parent enzyme FL#62) of three P450 variants sharing
an identical fingerprint and one having a different fingerprint from the
same library. GC peak integrations are provided in Table S5. Oxidation
product c was not characterized.
to three amino acid substitutions in their active sites, these mutations
have apparently resulted in equivalent active site geometries. Notably,
this feature could be captured through inspection of their finger-
prints. Thus, we conclude that fingerprint analysis could also offer a
convenient strategy to identify, within a library of engineered P450s,
those that possess a desired type of regioselectivity, once the
corresponding fingerprint is known. The total probe activity would
provide then a way to recognize, among these variants, the one(s)
which are more catalytically efficient as this parameter is related to the
number of turnovers supported by the enzyme. For comparison,
another P450 variant was extracted from the same library (5-C12)
and found to differ from 5-C4 by a single amino acid (V78 vs F78).
Interestingly, this single amino acid substitution causes a remarkable
change in the active site configuration of the enzyme as evinced from
the difference in the product distribution with 7, 8, and 10 and as it
could be anticipated from comparison of the respective fingerprints
(Figure 5).
In summary, we have reported an efficient and time-effective
method to gain insights regarding the reactivity and regioselectivity
properties of a P450 enzyme. This approach is amenable to the high-
throughput screening of engineered P450 libraries and offers the
unprecedented capability to allow for the rapid identification of P450
variants with diversified regioselectivity as well as variants with a
specific type of regioselectivity (via fingerprint comparison) within
such libraries. We expect this method to prove valuable in the
development of P450 catalysts for the functionalization of unacti-
vated C-H bonds in complex terpenes via P450-mediated synthesis.
(8) Fasan, R.; Meharenna, Y. T.; Snow, C. D.; Poulos, T. L.; Arnold,
F. H. J. Mol. Biol. 2008, 383, 1069.
(9) (a) Oliver, C. F.; Modi, S.; Primrose, W. U.; Lian, L.; Roberts,
G. C. K. Biochem. J. 1997, 327, 537. (b) Branco, R. J.; Seifert, A.; Budde,
M.; Urlacher, V. B.; Ramos, M. J.; Pleiss, J. Proteins 2008, 73, 597.
(10) Narhi, L. O.; Fulco, A. J. J. Biol. Chem. 1987, 262, 6683.
(11) McLachlan, M. J.; Johannes, T. W.; Zhao, H. Biotechnol. Bioeng.
2008, 99, 268.
(12) Sevrioukova, I. F.; Li, H.; Zhang, H.; Peterson, J. A.; Poulos,
T. L. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 1863.
(13) Fraga, B. M. Nat. Prod. Rep. 2003, 20, 392.
(14) (a) Gayet, A.; Bolea, C.; Andersson, P. G. Org. Biomol. Chem.
2004, 2, 1887. (b) Garcia, C.; LaRochelle, L. K.; Walsh, P. J. J. Am. Chem.
Soc. 2002, 124, 10970. (c) Kumara, S.; Sobhiab, M. E.; Ramachandran,
U. Tetrahedron: Asymmetry 2005, 16, 2599.
(15) (a) Meharenna, Y. T.; Li, H.; Hawkes, D. B.; Pearson, A. G.; De
Voss, J.; Poulos, T. L. Biochemistry 2004, 43, 9487. (b) Yao, H.;
McCullough, C. R.; Costache, A. D.; Pullela, P. K.; Sem, D. S. Proteins
2007, 69, 125. (c) Xu, L. H.; Fushinobu, S.; Takamatsu, S.; Wakagi, T.;
Ikeda, H.; Shoun, H. J. Biol. Chem. 2010, 285, 16844. (d) Denisov, I. G.;
Makris, T. M.; Sligar, S. G.; Schlichting, I. Chem. Rev. 2005, 105, 2253.
(16) Pylypenko, O.; Schlichting, I. Annu. Rev. Biochem. 2004,
73, 991.
’ ASSOCIATED CONTENT
S
Supporting Information. Experimental details and char-
b
acterization data. This material is available free of charge via the
Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION
(17) Sono, M.; Roach, M. P.; Coulter, E. D.; Dawson, J. H. Chem.
Rev. 1996, 96, 2841.
Corresponding Author
fasan@chem.rochester.edu
3245
dx.doi.org/10.1021/ja109590h |J. Am. Chem. Soc. 2011, 133, 3242–3245