An Enzymatic Route to α-Tocopherol Synthons: Aromatic Hydroxylation of Pseudocumene and Mesitylene with P450 BM3

Article abstract of DOI:10.1002/chem.201703647

Aromatic hydroxylation of pseudocumene (1 a) and mesitylene (1 b) with P450 BM3 yields key phenolic building blocks for α-tocopherol synthesis. The P450 BM3 wild-type (WT) catalyzed selective aromatic hydroxylation of 1 b (94 %), whereas 1 a was hydroxylated to a large extent on benzylic positions (46–64 %). Site-saturation mutagenesis generated a new P450 BM3 mutant, herein named “variant M3” (R47S, Y51W, A330F, I401M), with significantly increased coupling efficiency (3- to 8-fold) and activity (75- to 230-fold) for the conversion of 1 a and 1 b. Additional π–π interactions introduced by mutation A330F improved not only productivity and coupling efficiency, but also selectivity toward aromatic hydroxylation of 1 a (61 to 75 %). Under continuous nicotinamide adenine dinucleotide phosphate recycling, the novel P450 BM3 variant M3 was able to produce the key tocopherol precursor trimethylhydroquinone (3 a; 35 % selectivity; 0.18 mg mL^?1) directly from 1 a. In the case of 1 b, overoxidation leads to dearomatization and the formation of a valuable p-quinol synthon that can directly serve as an educt for the synthesis of 3 a. Detailed product pattern analysis, substrate docking, and mechanistic considerations support the hypothesis that 1 a binds in an inverted orientation in the active site of P450 BM3 WT, relative to P450 BM3 variant M3, to allow this change in chemoselectivity. This study provides an enzymatic route to key phenolic synthons for α-tocopherols and the first catalytic and mechanistic insights into direct aromatic hydroxylation and dearomatization of trimethylbenzenes with O₂.

Full text of DOI:10.1002/chem.201703647

A Journal of
Accepted Article
Title: An enzymatic route to α-tocopherol synthons: Aromatic
hydroxylation of pseudocumene and mesitylene with P450 BM3
Authors: Alexander Dennig, Alexandra Maria Weingartner, Tsvetan
Kardashliev, Christina Andrea Müller, Erika Tassano, Martin
Schürmann, Anna Joëlle Ruff, and Ulrich Schwaneberg
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: Chem. Eur. J. 10.1002/chem.201703647
Link to VoR: http://dx.doi.org/10.1002/chem.201703647
Supported by

10.1002/chem.201703647
Chemistry - A European Journal
Submission to: Chemistry – A European Journal
Manuscript type: Full Paper
An enzymatic route to α-tocopherol synthons:
Aromatic hydroxylation of pseudocumene and
mesitylene with P450 BM3
Alexander Dennig[a]‡*, Alexandra Maria Weingartner[a]‡, Tsvetan Kardashliev[a], Christina
Andrea Müller[a]†, Erika Tassano[c], Martin Schürmann[d], Anna Joëlle Ruff[a]*and Ulrich
Schwaneberg[a,b]*
[a] RWTH Aachen University, Institute of Biotechnology, Worringerweg 3, 52074 Aachen,
Germany
[b] DWI –Leibniz Institut für Interaktive Materialien, Forckenbeckstraße 50, 52074 Aachen,
Germany
[c] Department of Chemistry, University of Graz, Heinrichstrasse 28, 8010 Graz, Austria
[d] DSM Ahead R&D BV | DSM Innovative Synthesis, Post address: P.O. Box 1066, 6160
BB Geleen, The Netherlands
‡ Alexander Dennig and Alexandra M. Weingartner are joint first authors.
*Corresponding authors: alexander.dennig@tugraz.at, aj.ruff@biotec.rwth-aachen.de
u.schwaneberg@biotec.rwth-aachen.de
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/chem.2017xxxxx
1
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
Abstract
The aromatic hydroxylation of pseudocumene (1a) and mesitylene (1b) with P450 BM3 yields
key phenolic building blocks for α-tocopherol synthesis. The P450 BM3 wild-type (WT)
catalyzed selective aromatic hydroxylation of 1b (94%), whereas 1a was hydroxylated to a
large extent on benzylic positions (46-64%). Site-saturation mutagenesis generated a new P450
BM3 mutant, herein named “variant M3” (R47S, Y51W, A330F, I401M), with significantly
increased coupling efficiency (3 to 8-fold) and activity (75 to 230-fold) for 1a and 1b
conversion. Additional π-π interactions introduced by mutation A330F improved not only
productivity and coupling efficiency, but also selectivity toward aromatic hydroxylation of 1a
(61 to 75%). Under continuous NADPH-recycling the novel P450 BM3 variant M3 was able to
produce the key tocopherol precursor trimethylhydroquinone (TMHQ, 3a) (35% selectivity;
0.18 mg ml-1) directly from 1a. In case of 1b over-oxidation leads to dearomatization and
formation of a valuable p-quinol synthon that can directly serve as educt for synthesis of 3a.
Detailed product pattern analysis, substrate docking and mechanistic considerations support the
hypothesis that 1a binds in an inverted orientation in the active site of P450 BM3 WT as
compared to P450 BM3 variant M3 to allow this change in chemo-selectivity. This study
provides an enzymatic route to key phenolic synthons for α-tocopherols and the first catalytic
and mechanistic insights into direct aromatic hydroxylation and dearomatization of
trimethylbenzenes with O2.
KEYWORDS: monooxygenase, protein engineering, phenol, tocopherol, aromatic
hydroxylation, directed evolution.
2
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
Introduction
The selective hydroxylation of non-activated C-atoms with O2 remains a synthetic challenge
in organic chemistry.[1] Industrial synthesis of phenol from benzene is inefficient and mainly
proceeded via the multi-step cumene process (Hock synthesis, ~5% yield on benzene).[2]
A
vast number of pharmaceuticals, plastics, resins and vitamins[2b, 3] contain phenolic moieties
making this class of compounds indispensable in everyday life. Industrially important alkyl-
phenols can be obtained via phenol methylation[4] or chlorination followed by hydrolysis[5]
requiring at least three synthetic steps starting from benzene. Trimethylphenols (TMPs)
represent key synthons for α-tocopherols and α-tocotrienols (vitamin E), which find broad
applications as anti-oxidants, dietary supplements, UV-protectors, anti-tumor or anti-
inflammatory agents.[3b, 3c, 6] Natural sources of vitamin E do not cover the annual demand
(>30000 tons per year).[3b] In particular, the most potent form of vitamin E, namely
(2’R,4’R,8’R)-α-tocopherol, can only be extracted in racemic form from natural resources,
which makes innovative synthesis inevitable.[3c, 6c] Chemical synthesis of α-tocopherol
precursors such as 2,3,5- or 2,3,6-TMP (2a and 2b, Scheme 1) was summarized in an
excellent and detailed review.[3b] Synthesis of 2a and 2b requires typically three or more steps
including purification of intermediates. Alternatively, Pd-catalysts and high temperatures (350
°C) allow electrophilic acylation of 2,3-xylenol to yield 2a.[7] Only a few chemical routes
were reported for the direct synthesis of 2a or 2b from pseudocumene (1a).[6b, 8] These routes
often require either transition metal catalysts and/or strong oxidizing agents such as
trifluoroperacetic acid. Due to low selectivity of the reaction, formation of unwanted side
products such as quinones, mesitol or 2,4,5-TMP occurs. 2a and 2b can further be oxidized to
yield trimethylhydroquinone (TMHQ, 3a), the central phenol building block for vitamin E
(Scheme 1B).[3b, 6b, 9] Another route starts from mesitol (2,4,6-TMP; 2h, Scheme 1A), which
after oxidation and rearrangement yields 3a.[3b] 2h can also be obtained in good yields from
3
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
mesitylene (1b) using a Re-catalyst and explosive concentrations of H2O2 (85%)[10] or via
application of strong acids and expensive bis(trimethylsilyl) peroxide.[11] Just recently,
Lindhorst et al. could show oxidation of 1a with an iron–heterocyclic carbene complex and
H2O2 as the oxidant, thus yielding a mixture of various phenols and benzoquinones.[6b] The
increasing demand for phenolics and the usually harsh reaction conditions and low selectivity
during chemical hydroxylation underlines the importance to explore alternative synthetic
routes.[2b, 9] Ideally, a novel synthetic route should provide high efficiency, selectivity, and
also sustainability.[3c] The selective aromatic hydroxylation under mild reaction conditions
with either O2 or H2O2 as oxidant can be catalyzed by iron- or copper-containing proteins
such as mono- and dioxygenases[2a, 12], peroxygenases[13] or tyrosinases.[12b] However, most
enzymes lack the required operational performance for industrial applications, including poor
activities or low stability, jointly leading to low product yields.[1, 14] In two previous studies,
we reported the engineering of the P450 monooxygenase BM3[12e] toward efficient and
selective o-hydroxylation of mono-substituted benzenes[12a] and p-xylene.[2a] The catalyst was
further applied in combination with a tyrosine phenol lyase for preparative synthesis of
enantiopure L-tyrosine derivatives underlining its synthetic potential.[15] Site saturation
mutagenesis of residue F87 proved that the phenylalanine in the active site provides
indispensable interactions for efficient aromatic hydroxylation of benzenes.[12a] Based on
experimental results and substrate docking, a tight π-π interaction of residue F87 and anisole
as ligand was postulated.[12a, 12c] Despite that α-hydroxylation (hydride-abstraction followed
by hydride-rebound)[12e, 16] and aromatic hydroxylation (aryl epoxidation followed by NIH-
shift and re-aromatization)[12c, 17] require different steps in the P450 reaction mechanism, the
active site of P450 BM3 precisely directs the substrate to achieve highly regio- and chemo-
selective conversion of a broad range of benzenes.[2a, 12a, 12c, 12d, 15, 18] Encouraged by these
results, we decided to extend our investigations to the aromatic hydroxylation of sterically
more demanding benzenes such as 1a and 1b, which to date were only hydroxylated
4
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
(enzymatically) on benzylic positions e.g. employing xylene monooxygenases.[19] We
emphasized in particular on the synthesis of 2a, 2b, and 2h, which can serve as starting
materials for the production of antioxidants or antitumor agents such as the water soluble
vitamin E analogue Trolox (Scheme 1B).[20]
Scheme 1. A: Hydroxylation of pseudocumene (1a) with P450 BM3 yields 2,3,5- and 2,3,6-
trimethylphenol (2a, 2b). Additional side products (2c-2f) formed during conversion of 1a are
highlighted in brackets. Further hydroxylation enables synthesis of trimethylhydroquinone
(TMHQ, 3a) in one-pot mode. Hydroxylation of mesitylene (1b) with P450 BM3 yields 2,4,6-
TMP (2h). Further hydroxylation of 2h yields 3b (4-hydroxy-2,4,6-trimethylcyclohexa-2,5-
dien-1-one).[21] B: 3a represents the central aromatic precursor for synthesis of vitamin E or
the water-soluble anti-oxidant Trolox.[3b] NADP(H) = Nicotinamide adenine dinucleotide
phosphate.
Direct enzymatic hydroxylation of 1a or 1b would be advantageous and open novel and
greener routes to industrially relevant alkyl-phenols.
5
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
Results and Discussion
Conversion of 1a and 1b with P450 BM3 WT and variants M1 and M2
Catalytic characterization of P450 BM3 for conversion of 1a and 1b was done measuring
NADPH oxidation rates and product formation by GC-FID/-MS with authentic standards (see
SI). The P450 BM3 WT hydroxylates 1a and 1b at very low turnover frequencies (TOFs) of
3.2 and 0.3 min-1, respectively (Table 1, entry 1 and 4). Measured productivities correlate with
the low NADPH oxidation rates of P450 BM3 WT (1a = 22 min-1; 1b = 13 min-1), indicating
that both substrates do not enter efficiently the active site of the monooxygenase to initiate the
catalytic cycle.[12e, 16-17] The two polar amino acid residues R47 and Y51 (‘gate keepers’)[12e]
often hinder sterically demanding substrates to efficiently enter the substrate channel of P450
BM3.[2a, 12a, 18b, 18c] In addition, a very low coupling efficiency of only 3% for 1b indicates a
loose-fitting in the active site[22], possibly due to higher sterical demands as compared to p-
xylene (45%)[2a] or toluene (10%).[12a] Unexpectedly and despite of the low catalytic activity,
a total turnover number (TTN) of 1602 (Table 2, entry 2) could be measured for
hydroxylation of 1a using cell free extracts (CFE) of P450 BM3 WT and a glucose
dehydrogenase (GDH) for NADPH recycling.[15, 23] A TTN >1500 at only 15% coupling
efficiency (Table 1, entry 1) is a good value and still one order of magnitude higher than
TTNs of most chemical catalysts used for oxyfunctionalization.[6b, 24] A comparable chemical
catalyst (1% loading) showed after reaction optimization 54 turnover cycles for conversion of
1a using H2O2 as oxidant.[6b] For conversion of 1b a TTN of only 393 was determined, which
is 4-fold lower than for 1a and can be explained by the significant lower coupling efficiency
(3%; 5-fold lower; Table 1, entry 4). A low coupling efficiency often results in formation of
-
reactive oxygen species such as H2O2 or O2 that can lead to rapid catalyst inactivation.[22b, 25]
Chemoselectivity of P450 BM3 WT for 1b is preferentially toward the aromatic ring leading
to formation of 2h (83%; Table 2, entry 11), which is in good agreement with other reports on
6
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
structural related benzene substrates (e.g. m-xylene).[2a, 12a, 15, 18c] 3,5-dimethylbenzylalcohol
(2g) was detected only in traces (6%). In contrast, P450 BM3 WT converted 1a to all
theoretically possible mono-hydroxylated products (2a-2f; Table 2, entry 1 and 2). A low
chemoselectivity of P450 BM3 WT for aromatic hydroxylation of o-xylene (45%) as
compared to p-xylene (>95% o-hydroxylation) and m-xylene (98%) has been shown before.[2a,
18b] Major products during conversion of 1a with P450 BM3 WT are 2,4-DMBA (2e; 40%),
2,4,5-TMP (2c; 23%) and 2,5-DMBA (2f; 21%), in summary 84% (Table 2, entry 1).
Secondary hydroxylation products of 1a were obtained in traces (3%) and only in presence of
a cofactor regeneration system (Table 2, entry 2). To overcome the drawbacks of P450 BM3
WT, we decided to employ P450 BM3 variant M2 (R47S, Y51W, I401M) as catalyst, which
was engineered for aromatic hydroxylation of p-xylene.[2a] Compared to the P450 BM3 WT,
variant M2 displayed improved hydroxylation of 1a and 1b (Table 1). P450 BM3 variant M2
displayed a 2.6-fold higher NADPH oxidation rate during conversion of 1b, resulting overall
in a 4.7-fold higher TOF (1.4 min-1; Table 1, entry 5) as compared to the P450 BM3 WT.
Despite this improvement, the productivity remained very low and more than three orders of
magnitude lower as compared to p-xylene (1953 min-1)[2a] or anisole (2427 min-1).[12a]
Differences in substrate solubility could play an important role; in particular in case of
anisole, which is 3900-fold more soluble in water (1.85 mol L-1)[26] than 1a (0.4 µmol L-1).[27]
Regioselectivity of P450 BM3 variant M2 remained high for aromatic hydroxylation of 1b
yielding 89% of 2h (Table 2, entry 11). Minor changes in chemoselectivity and coupling
efficiency indicate a similar orientation of 1b in the active site of P450 BM3 WT and variant
M2 (no active site residue with direct substrate interaction exchanged).[2a] Conversion of 1a
with P450 BM3 variant M2 displayed a 4-fold improved NADPH depletion rate (87 min-1) as
well as an increased coupling efficiency by 4% (Table 1, entry 2). Still, a TOF of only 17 min-
1 is a comparably low value for benzene conversion with P450 BM3.[2a, 12a, 18b, 18c]
Chemoselectivity during conversion of 1a by P450 BM3 variant M2 was shifted toward
7
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
increased aromatic hydroxylation and accumulation of an additional product was detected
when a cofactor regeneration system was applied. The product was after preparative isolation
and comparison with a commercial standard identified as TMHQ (3a, 32%, for 1H-NMR
analysis see SI). Major products after conversion of 1a with purified P450 BM3 variant M2
are 2,4-DMBA (2e, 25%) and 2,4,5-TMP (2c, 39%) (Table 2, entry 5). To investigate the
influence of the two mutations at the entrance of the substrate channel (residues 47 and 51),
we decided to use also the P450 BM3 variant M1 (R47S, Y51W)[2a] for conversion of 1a.
P450 BM3 variant M1 displayed similar regioselectivity as compared to P450 BM3 WT,
while TTN values could be improved significantly from 1602 to 4910 (Table 2, entry 2 vs. 4).
Table 1. Catalytic data for conversion of 1a and 1b with purified P450 BM3 enzymes.
Entry
1
Substrate
Catalyst
WT
NADPH ox. rate [min-1][a]
22 ± 4
Coupling [%][b]
15 ± 2
TOF [min-1][c]
3.2
1a
87 ± 3
499 ± 91
13 ± 2
19 ± 2
45 ± 6
3 ± 1
17
226
0.3
1.4
69
2
3
4
5
6
1a
1a
1b
1b
1b
M2
M3
WT
M2
M3
34 ± 14
285 ± 20
4 ± 2
24 ± 2
[a] NADPH oxidation rate (molcofactor molP450-1 min-1) was determined spectrophotometrically at 340 nm absorbance;
[b] Coupling efficiency (%) = ratio between product formation [µM] and oxidized cofactor [µM]; [c] Turnover
-1
frequency (molproduct molP450 min-1). WT = P450 BM3 wild type; M2 = P450 BM3 mutant (R47S, Y51W,
I401M)[2a]; M3 = P450 BM3 mutant (R47S, Y51W, I401M, A330F) (obtained in this study). Conversions were done
using purified P450 BM3 protein (0.3 μM WT enzyme or 0.1 μM for the P450 BM3 variants), 10 mM substrate (1a
or 1b), 1.5% DMSO as co-solvent in a final volume of 5 mL. 200 µM NADPH were added and the activity of P450
BM3 was measured as initial NADPH oxidation rates at 340 nm. Products were quantified using GC-FID and
commercial standards for 2a-h.
Under continuous NADPH regeneration, P450 BM3 variant M2 displayed even a higher TTN
of 5268 (3.3-fold increase compared to P450 BM3 WT) for conversion of 1a (Table 2, entry
6). For conversion of 1b with CFE of P450 BM3 variant M2 a TTN of 1243 was measured
which is a 3.2-fold improvement compared to P450 BM3 WT (Table 2). A higher coupling
8
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
efficiency (Table 1), the improved electron transfer/O2 activation (substitution I401M)[2a, 12a]
and a better access of the substrate to the active site (substitutions R47S and Y51W)[2a, 12a, 12e,
18b, 18c] jointly improve the catalytic performance of P450 BM3 variant M2 for conversion of
1a and 1b.
Engineering P450 BM3 toward improved conversion of 1a
In order to improve the catalytic performance (activity and coupling efficiency) of P450 BM3
variant M2, the position A330 was, after structural inspection, mutated using P450 BM3
variant M2 as template followed by screening for improved hydroxylation of 1a (see SI).
Position 330 was previously reported to influence the activity of P450 BM3 for aromatic
hydroxylation of toluene and ethyl benzene.[18c] Further it could be shown that rational
introduction of a proline on position 330 (A330P) led to relocation of a neighboring proline
(P329) toward the substrate binding pocket resulting in higher coupling efficiency and activity
of P450 BM3.[18c] Here, position 330 was saturated to all 20 canonical amino acids and
screened for improved NADPH oxidation (1a as substrate). Using 1a as substrate allows
formation of the target products 2a and 2b, which are easily detectable in microtiter plate
(MTP) format by the colorimetric 4-AAP assay.[28] A high number of improved variants
(76%) were identified within the generated library (see SI; 180 variants screened; theoretical
calculated >95% diversity coverage).[29] This indicates that the P450 BM3 WT active site
configuration (and in particular residue A330) reduces the efficient binding and conversion of
1a (Table 1). The best three P450 BM3 muteins (based on phenol production; see SI)
contained each a phenylalanine at position 330 (A330F) which is, to the best of our
knowledge, an unreported amino acid exchange in P450 BM3.[12e] One variant (herein named
P450 BM3 variant M3 = R47S, Y51W, A330F, I401M) was selected for purification and
detailed catalytic characterization. P450 BM3 variant M3 displayed a 22-fold higher NADPH
oxidation rate during conversion of 1b than P450 BM3 WT and a 8.4-fold higher rate than
9
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
P450 BM3 variant M2, respectively (Table 1; Entry 4-6). Moreover, an improved productivity
(TOF = 226 min-1) was measured during conversion of 1a. For both substrates the coupling
efficiency was increased (24% for 1b; 45% for 1a) in the range of 2- to 6-fold as compared to
P450 BM3 variant M2 (Table 1). The latter is in good agreement with the results obtained
during activity screenings in MTP format. When comparing TOF values for 1b, P450 BM3
variant M3 is a 230-fold better hydroxylase than the P450 BM3 WT and 49-fold better than
P450 BM3 variant M2, respectively (Table 1). In case of 1a, the improvements range from
13-fold compared to P450 BM3 variant M2 and 71-fold compared to P450 BM3 WT.
Selectivity of P450 BM3 variant M3 remained high for aromatic hydroxylation of 1b (89-
94%, Table 2, entries 13 and 14). In case of 1a, minor changes in product distribution were
detected (Table 2) as compared to P450 BM3 variant M2. In presence of a NADPH
regeneration system, P450 BM3 variant M3 (applied as CFE) reached a product concentration
of up to 0.48 g L-1 (mixture of 2a-2f) for conversion of 1a, which corresponds to a TTN of
4569. Conversion of 1b with P450 BM3 variant M3 showed an increase in TTN (19% vs.
P450 BM3 variant M2) underlining the overall improved operational performance. Higher
TOF values were (as a general trend) obtained for conversion of 1a as compared to 1b (Table
1; entry 1-3 vs. entry 4-6). This indicates that only small structural differences can lead to
dramatic changes in selectivity, coupling efficiency and activity, which has been shown for
conversion of xylene isomers.[2a, 18b] In addition to 1a and 1b, we also investigated the
hydroxylation of anisole and toluene with P450 BM3 variant M3 (CFE as catalyst supported
by a GDH cofactor regeneration system). Selectivity of P450 BM3 variant M3 for
hydroxylation of o-positions remained very high for both substrates (95-99%)[12a], whereas the
TTN values could be further increased (anisole: 8536; toluene: 6262) as compared to P450
BM3 variant M2 (anisole: 6195; toluene 3740)[12a] making P450 BM3 variant M3 an overall
better catalyst for aromatic hydroxylation of benzenes. It is worth mentioning that for
10
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
anisole/toluene secondary hydroxylation to the respective 1,4-hydroquinones occurred as
reported before.[15, 30]
Table 2. Selectivity and productivity of P450 BM3 WT and variants using cell-free extracts
(CFE) or purified enzymes.
Entry
Substrate
Conv. [%]
Catalyst
WT
2a-2c [%]
9/4/23
8/9/34
3/5/34
4/5/34
6/17/39
7/4/31
6/14/41
7/4/29
8/5/28
---
2d-2f [%]
3/40/21
5/27/14
3/37/18
2/31/12
2/25/11
2/17/7
1/29/9
2/17/6
2/25/9
---
3a [%]
2a-c:2d-f:3a [%]
36:64:0[a]
51:46:3
TTN [-]
---
1
1a
1a
---
16
0
2
WTcfe
M1
3
1602
---
3
1a
---
0
42:58:0[a]
43:45:12
62:38:0[a]
42:26:32
61:39:0[a]
40:25:35
41:36:23
-
4
1a
44
M1cfe
M2
12
4910
---
5
1a
---
0
6
1a
40
M2cfe
M3
32
0
5268
---
7
1a
---
8
9 (10 mL)
10
1a
34
M3cfe
M3cfe
M3cfe
35
4569
3528
4225/4978
1a
29
23
2a/2b
42/49
>99/>99
Entry
11
Substrate
Conv. [%]
Catalyst
WT
2g [%]
2h [%]
83
3b [%]
11
2g:2h:3b [%]
6:83:11
TTN [-]
393
1
4
6
3
1b
1b
1b
1b
2h
2h
12
M2
89
8
3:89:8
1243
1526
n.d.
13
5
M3
4
85
11
4:85:11
14[b] (50 mL)
15[b] (1 mL)
16[c] (100 mL)
>99
45
89
M3cfe
M3cfe
M3cfe
11
---
---
89
n.d.
44
11:89:n.d.
0:45:44
45
n.d.
11
89
0:11:89
n.d.
WT = P450 BM3 WT. M1 = P450 BM3 (R47S, Y51W)[2a]; M2 = P450 BM3 (R47S, Y51W, I401M)[2a]; M3 = P450 BM3
(R47S, Y51W, A330F, I401M). Selectivity/productivity was determined with commercial standards of 2a-h and 3a.
Selectivity of purified enzymes and CFE are shown to report values under cofactor limited/unlimited conditions. Conversions
with purified P450 BM3 protein (entries 1,3,5,7 and 11-13; 0.3 μM P450 BM3 WT or and 0.1 μM for the P450 BM3 variants)
contained: 10 mM 1a/1b, 1.5% DMSO as co-solvent in a final volume of 5 mL. 200 µM NADPH (final concentration) was
added to start the reaction. Reaction conditions for entries 2,4,6,8 and 9: 10 mM substrate, 2% DMSO, CFE (1 µM P450 BM3
WT or variants), GDH cofactor regeneration system. All reactions (except for entry 13 and 15) were done in triplicate. [a]
Secondary hydroxylations did not occur in these reactions due to limited amount of NADPH. [b] Reaction contained: 4 mM
1b, 30 mg of freeze dried CFE of variant M3, GDH-cofactor regeneration system. [c] Reaction was done under continuous
pH control and with pressurized air constantly supplied from top into the glass flask, which leads to significant evaporation
of 2g (no substrate recovered after extracting with organic solvent). Further experimental details and analytical data of isolated
compounds can be found in SI. n.d. = not determined.
11
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
One-pot double-hydroxylation of 1a and 1b
In the presence of a co-factor regeneration system, secondary hydroxylation of 1a and 1b
were detected by GC-FID/MS. In case of 1b over-oxidation leads to formation of the p-quinol
4-hydroxy-2,4,6-trimethylcyclohexa-2,5-dien-1-one (3b), a versatile building block for
natural products, therapeutics and organic synthesis in general.[21] Beneficially, 3b can further
serve as intermediate for synthesis of 3a[3b] which was not detected during conversion of
1b/2h. For detailed compound analysis and identification, 3b was produced on 50 mL scale
using 2h as substrate (10 mM; P450 BM3 variant M3 as catalyst). After isolation and
purification (11 mg isolated) the product was analyzed using 1H-/13C-NMR and GC-MS and
compared to literature values (see SI).[21] Despite a huge industrial interest in 2h (~90 patent
applications)[31] there has been (to the best of our knowledge) no biocatalytic route described
showing conversion of 1b into 2h or 3b. In case of 1a, the over-oxidation of 2a and/or 2b
enables the direct/linear synthesis of 3a without the need for additional catalysts/reactants or
intermediate isolation (Scheme 1A). The formation of 3a in the reactions using 1a, (or 2a/2b)
as substrate revealed that all P450 BM3 enzymes are able to produce the desired vitamin E
synthon. Employing P450 BM3 WT for conversions of 1a generates 3a only in traces (<0.01
g L-1), whereas P450 BM3 variant M1 displayed an 8-fold increased product formation of up
to 0.08 g L-1 of 3a. Conversion of 1a with P450 BM3 variants M2 and M3 revealed further
increased selectivity (up to 32-35%; one-pot conversion of 1a) for 3a (P450 BM3 variant M2
= 0.19 g L-1; P450 BM3 variant M3 = 0.1 to 0.18 g L-1; Table 2 entry 8 and 9). Starting from
2a or 2b, P450 BM3 variant M3 generated (without additional reaction optimization) even
higher product concentrations for 3a (0.64 g L-1 and 0.76 g L-1; Table 2, entry 10; >99%
selectivity). To prevent rapid oxidation of 3a to the respective 1,4-benzoquinone, all
conversions were performed in presence of 10 mM ascorbic acid (weak/mild reducing agent),
12
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
which had no detrimental effect on the biocatalysts. In order to analyze the diverse product
patterns obtained from secondary hydroxylation of 1a, all six mono-hydroxylated products
(2a-f) were converted separately by P450 BM3 variant M3. Since the formation of 3a (major
double hydroxylated product) would be expected to proceed via 2a and 2b hydroxylation
(excluding methyl group shifts), we determined NADPH oxidation rates using both
compounds as substrates. For 2a/2b a significantly higher NADPH oxidation rate was
measured as compared to 2c-f and 1a (see SI); the latter indicates that hydroxylation of 2a and
2b is favored and high selectivity (>99%) drives the reaction efficiently toward formation of
3a. During conversions of 1a additional di-hydroxylated products could be detected, which
were most likely derived from conversion of 2c and 2e (see SI). Secondary hydroxylation of
substrates is usually unwanted and requires careful adjustment of the reaction system[15] or
further catalyst engineering. For accessing the vitamin E synthons 3a/3b, the observed over-
oxidations of phenolic intermediates are desirable since intermediate work-up or additional
catalysts/reagents are not required.[3b]
Substrate binding and chemo-selectivity of P450 BM3
Computational assisted docking studies (using YASARA software package)[32] were
performed to generate a molecular understanding of achieved improvements in activity,
coupling efficiency (Table 1), and altered chemoselectivity (Table 2). Docking of substrates
provides binding modes during catalysis, which can be described by the Gibbs free energy of
binding (ΔGbind).[33] Docking was performed using 1a and 1b as ligand and the crystal
structure of P450 BM3 WT was employed as receptor molecule (PDB: 1BU7).[34] In addition,
a homology model structure of P450 BM3 variant M3 was generated for docking of both
substrates. It is well accepted for P450-catalyzed reactions that a close interaction of the
substrate and the heme-bound water ligand is essential to initiate the catalytic cycle[17]
,
whereas a strong binding of a potential substrate allows a high coupling efficiency.[22b]
13
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
Herein, only substrate orientations are shown that correspond to productive binding modes[32,
33b] and which are in agreement with catalytic data (Table 1 and 2). Docking of 1b into the
active site of the P450 BM3 WT revealed one binding mode that allows equally α-benzylic
hydroxylation or 1,2-epoxidation of arylic carbons (Figure 1A)[17, 18b] and formation of the
major product 2h (Table 2, entries 10 to 12).
Figure 1. Docking of 1b into the active site of P450 BM3 WT and the homology model of P450
BM3 variant M3.[32, 33b] A: 1b docked into the crystal structure model of P450 BM3 WT
(1BU7)[34] yielding either 2g or 2h (ΔGbind: -4.93 kcal mol-1). B: Docking of 1b into the active
site of P450 BM3 variant M3 (R47S, Y51W, A330F, I401M; ΔGbind: -5.74 kcal mol-1). Arrows
indicate the closest distance between heme-bound water ligand and closest arylic C-atoms of
1b.
The binding of 1b is comparably strong with -4.93 kcal mol-1[35], however the distance
between the water ligand and the closest C-atom exceeds 6 Å suggesting a slow activation[12e,
16] that was determined experimentally for P450 BM3 WT (Table 1, Entry 4). Docking of 1b
into the homology model of P450 BM3 variant M3 (Figure 1B) suggested not only a stronger
binding (ΔGbind: -5.74 kcal mol-1), but also a closer distance to the water ligand (4.6 Å). The
closer and stronger binding is in good agreement with the measured improvements (higher
14
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
coupling and higher PFR for P450 BM3 variant M3; Table 1, entry 6) during conversion of
1b. Both binding modes (P450 BM3 WT and variant M3) suggest equal hydroxylation of α-
benzylic and aromatic carbons (Figure 1). However, hydroxylation of 1b proceeded
preferentially at arylic positions (94 to 96%; Table 2), which opens the question how
chemoselectivity during hydroxylation of aromatic compounds is directed by P450 BM3. One
reason for the high selectivity could be that 1,2-epoxidation and re-aromatization is faster or
more efficient than hydride-rebounding via compound II that requires also more catalytic
steps[16-17] including formation of a carbon radical at the α-position (Scheme 2B). Bond
dissociation energies (BDE) of benzylic C-H bonds are similar for toluene (89 kcal mol-1) and
1b (88 kcal mol-1)[36] and generally lower than for secondary aliphatic carbons (95 kcal mol-1),
the preferred hydroxylation target for P450 BM3.[12e] However, it is on the molecular level
unknown why P450 BM3 WT prefers aromatic over benzylic hydroxylation for most
benzenes[2a, 12a, 12c, 15, 18b, 37] (Table 2; BDEs for benzylic H-bonds of o-, m-, p-xylene are 74,
81 and 71 kcal mol-1).[38] Previous investigations could not elucidate the reason on the
enzyme’s chemoselectivity during alkyl-benzene conversions. A common pattern throughout
literature on alkyl-benzene hydroxylation with P450 BM3 is, that whenever a primary
benzylic carbon has no o-substituent e.g. toluene, p-xylene, m-xylene or 1b (as opposed to o-
xylene and 1a; both yield significantly higher amount of benzylalcohols) and P450 BM3
retains active site residue F87, aromatic o-hydroxylation is the predominant reaction with
selectivity >95%.[2a, 12a, 12c, 15, 18b] Substituting F87 by alanine (removing π-π interactions)
leads to increased formation of benzylalcohols (45-83% for m-xylene and 94-99% for o-
xylene)[18b] and a significant decrease in aromatic hydroxylation of p-xylene.[2a] Moreover
decoy molecules can influence the formation of benzylalcohol slightly as shown for
conversion of toluene in presence of five different perfluorinated fatty acids (5-8% benzyl
alcohol formed, P450 BM3 WT).[12c]
15
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
Due to its non-symmetrical structure, the docking of 1a is more complex and complexity rises
due to the formation of six different mono-hydroxylated products (2a-f; Table 2). The
conversion of 1a by P450 BM3 WT leads predominantly to benzylalcohol formation (main
products are 2e and 2f, 61%, Table 2; entry 1). On the other hand, α-benzylic hydroxylation of
1a at position 4 (no o-substituent on position 3 or 5) occurs with ≤5% 2d formed (Table 2).
Interestingly, this was independent of the P450 BM3 catalyst and supports the hypothesis of
an o-CH3 directed α-hydroxylation. Docking of 1a into the P450 BM3 WT structure proposes
three major orientations with strong binding energies (Figure 2; A: -5.60 kcal mol-1; B: -5.53
kcal mol-1; C: -5.37 kcal mol-1) as well as a comparably close distance of 1a to the heme-
bound water ligand (Figure 2; A: 3.37 Å; B: 3.34 Å; C: 3.01 Å).
Figure 2. Docking of 1a into the active site of P450 BM3 WT (1BU7).[34] A: Binding
orientation (ΔGbind of -5.60 kcal mol-1) allowing hydroxylation preferentially at the α-benzylic
C-2 (product: 2f). B: Binding orientation (ΔGbind of -5.53 kcal mol-1) allowing α-hydroxylation
preferentially at α-benzylic C-1 (product: 2e). C: Binding orientation (ΔGbind of -5.37 kcal mol-
1) allowing aromatic hydroxylation of 1a (product: 2c). Arrows indicate the closest distance
between the heme-bound water ligand and C-atoms of 1a.
This would allow a faster initiation of the catalytic cycle[16] as compared to 1b, which was
determined experimentally (Table 1). The three binding modes showed in Figure 2
16
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
preferentially lead to formation of 2f (A), 2e (B), and 2c (C) which, after biochemical
characterization, comprise in total 75 to 84% (Table 2, entry 1 and 2). We conclude that 1a
binds with high probability in the suggested modes. In the P450 BM3 WT crystal structure
model docking, residue A330 does not directly interact with 1a as the aliphatic side chain is
pointing away from the substrate binding pocket (Figure 2, A-C). The unexpected high number
of improved variants (76%, see SI) obtained after site-saturation mutagenesis of position 330,
suggests that sterically more demanding amino acids (18 in total) provide interactions
essentially for improved conversion of 1a (Table 1). The homology model of P450 BM3 variant
M3 places the best fitting rotamer of F330 toward the substrate binding pocket/substrate
channel thus leading to additional and beneficial aromatic interaction(s) between amino acids
F87/F330 and the benzene substrate.[35] 1a docked into the homology model of P450 BM3
variant M3 yielded an even stronger binding of -6.33 kcal mol-1 (Figure 3, A), whereas the
distance between the closest C-atom and heme iron was increased (5.04 Å) as compared to the
P450 BM3 WT (Figure 2, A-C).
Figure 3. Docking of 1a into active site of P450 BM3 variant M3 (R47S, Y51W, A330F,
I401M). A: Binding orientation (ΔGbind of -6.33 kcal mol-1) allowing hydroxylation of 1a at the
α-benzylic C-1 (product: 2f). B: Binding orientation (ΔGbind of -5.01 kcal mol-1) allowing
aromatic hydroxylation of 1a (position C-3; product: 2b). C: Binding orientation (ΔGbind of -
17
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
5.51 kcal mol-1) allowing aromatic hydroxylation at C-5 or C-6 (product: 2a or 2c). Arrows
indicate the closest distance between heme-bound H2O ligand and the targeted C-atom of 1a.
Orientation B (-5.53 kcal mol-1) represents the closest binding orientation of 1a near the heme
iron (3.21 Å) in P450 BM3 variant M3 suggesting the preferential formation of 2b. The
proposed binding mode is supported by an increased formation of 3a (via 2a/2b
intermediates) in the GDH supported reactions as compared to P450 BM3 WT (Table 2, entry
2 and 8). Orientation C (Figure 3) binds 1a with -5.51 kcal mol-1 and a larger distance to the
heme iron. This would in theory result in higher formation of 2a or 2c, of which at least 2c
was increased (entry 1 vs. entry 7; Table 2) as compared to the P450 BM3 WT. A previous
docking of anisole into the active site of P450 BM3 WT was in good agreement with the
measured selectivity as well as catalytic activity and generated the hypothesis of a T-shape
orientation/interaction coordinated by active site residue F87.[12a] Binding modes A-C (Figure
3) suggest that 1a is stacked between residues F330 and F87 in a T-shape π-1a-π orientation
(aromatic sandwich).[35] The semi-rationally introduced phenylalanine generates an even
stronger substrate binding (as shown during docking; increased coupling efficiency for 1a/1b)
very likely through aromatic (π-π stacking) interactions[35], which improved conversion of 1a
and 1b by variant M3 (Table 1).
Assuming the binding modes for 1a (Figures 2 and 3) are in good correlation with the
catalytic data, we analyzed product patterns and alterations in chemo-selectivity between
P450 BM3 WT and variant M3. It is well accepted that aromatic hydroxylation of benzenes
proceeds via epoxidation of the aromatic ring followed by epoxide ring opening, NIH-shift
and re-aromatization to the respective phenol (Scheme 2A).[17, 18b, 37] In addition, methyl-
group shifts were previously observed during conversion of o-xylene with P450 BM3.[18b] For
conversion of 1a and 1b we did not observe CH3-shifts. Scheme 2 shows routes to the major
phenolic and benzylic alcohol products detected after the conversion of 1a with purified P450
18
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
BM3 WT and variant M3 (Table 2). When orienting the substrate with the aromatic carbon 4
(C-4) toward compound I, P450 BM3 variant M3 is able to catalyze 3,4- and 4,5-epoxidation
of which the latter would yield 2c as product (41%; Table 2, Entry 7; Scheme 2A). This
orientation is also supported by an increased formation of 2b and directly related to an
increased formation of 3a (35%, Table 2).
Scheme 2. Deduced substrate orientations and hydroxylation sites for conversion of 1a with
purified P450 BM3 (200 µM NADPH supplemented). A: 1a oriented with arylic carbon 4
toward compound I, leading preferentially to formation of 2b and 2c. B: 1a oriented with
benzylic carbons 1 and 2 toward compound I. Hydride abstraction leads to formation of an α-
carbon radical which after oxygen rebound yields 2e and 2f (major benzylalcohol products)
during conversion of 1a. Further information on benzene epoxidation, epoxide ring opening
and re-aromatization can be found elsewhere.[12c, 17, 18b, 37]
Interestingly, the formation of 2d (<2%) is not increased, indicating that aromatic
hydroxylation is generally favored in absence of an o-CH3-substituent (Whitehouse et al.
19
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
could show that mutation F87A increases benzylic hydroxylation).[18b] The P450 BM3 WT
produced less 2a and 2b (in total 13 to 17%) and only 3% of 3a (Table 2). From this we
assume that a different binding orientation of 1a is predominant in the P450 BM3 WT active
site as compared to P450 BM3 variant M3. This is in line with previous observations showing
that binding orientations of the substrate are possibly more relevant for chemoselectivity than
the intrinsic reactivity of the substrate.[12a, 18b] Scheme 2B shows the productive binding
orientations for α-benzylic hydroxylation of 1a. Through formation of the α-carbon radical at
position 1 or 2 the predominant benzylalcohol products 2e (40%) and 2f (21%) are formed by
P450 BM3 WT (Table 2, Entry 1). In case of P450 variant M3, the formation of 2e/2f is
decreased (29 and 9%) indicating again a different binding orientation of 1a in the active site
of BM3 WT as compared to P450 BM3 variant M2 and M3, which is supported by results
from docking experiments. Further reasons for the change in chemo-selectivity during 1a
conversions could be: a) movements of the flexible protein backbone and in particular residue
F87 during substrate binding and heme reduction[39] b) restricted space/π-π interaction
through introduction of A330F and c) binding of more than one substrate (steric crowding)[18b,
40], which possibly acts as a decoy.[12c] The ability of decoys molecules to influence selectivity
of P450 BM3 has been reported before.[12c, 41] Since the active site of P450 BM3 is large
enough to bind bulky substrates such as naphthalene or steroids[33a, 42], a second arene
molecule could fit into the active site thus acting as potential decoy. This would lead to
additional π-π interactions in the active site that potentially influence substrate orientation and
chemoselectivity. To solve these unanswered questions, a crystal structure of P450 BM3 WT
and M3 with 1a as ligand would be invaluable.
Di-hydroxylation of 1a/1b or mono-hydroxylation of 2a/2b and 2h yielded mainly 1,4-
dihydroxy compounds indicating a strong p-directing effect of phenolic-substituents (Scheme
20
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
3).[15, 30] Formation of trimethylresorcinol would be feasible, however requiring a methyl-
group shift[18b], which was not observed in this study.
Scheme 3. Proposed routes to formation of TMHQ (3a) from 2a (A) or 2b (B). Resorcinol was
not formed to a detectable amount therefore these pathways have to be excluded at the moment.
Di-hydroxylation of 1b yielded a single product that was identified as p-quinol 3b. For this,
we propose a route that starts via 3,4-epoxidation of 2h followed by epoxide ring opening and
21
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
finally affording the resonance stabilized p-quinol (Scheme 4), which cannot re-aromatize and
therefore remains stable in solution.
Trimethylresorcinol or 3a were not formed to a detectable amount and would require in both
cases a methyl-group shift which is (in theory) feasible due to the lower BDE (bond
dissociation energy) of C-C bonds compared to C-H bonds.
Scheme 4. Proposed route to formation of 4-hydroxy-2,4,6-trimethylcyclohexa-2,5-dien-1-one
(3b) during conversion of 1b/2h
.
Conclusions
Alternative enzymatic routes for vitamin E synthesis are important to match increasing
demands. Here, we show the first direct aromatic hydroxylation of pseudocumene (1a) and
mesitylene (1b) in water, with O2 and under mild reaction conditions to access five key
phenolic α-tocopherol synthons including a direct route to TMHQ (3a) and the first
biocatalytic synthesis of the p-quinol 3b. The P450 BM3 WT enzyme displayed highest
selectivity for aromatic hydroxylation of 1b (94%), whereas 1a was hydroxylated at all six
potential positions. A novel P450 BM3 variant (R47S, Y51W, I401M, A330F) had a boosted
activity (23-fold) and improved coupling efficiency (3-fold) resulting in a 70-fold higher TOF
for conversion of 1a. For 1b even a 230-fold improved TOF compared to P450 BM3WT
22
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
could be measured. Interestingly, conversions of 1a with the engineered P450 BM3 variants
M2 and M3 displayed altered chemoselectivity compared to P450 BM3 WT with increased
formation of phenols as well as additional secondary hydroxylations. The unexpected over-
oxidations yielded 3a directly from 1a (or 3b from 1b), which is a significant step toward
more sustainable synthesis of vitamin E as the reaction can be performed in one-pot without
the need for intermediate purification or additional catalysts. A product formation of 0.12 g L-
1 for 3a (1a as substrate) is already a good starting point for further reaction and process
engineering. Despite that a TTN of 5268 for 1a is a good value for conversion of a non-
natural substrate, additional enzyme engineering is necessary to improve operational
performance and chemo-selectivity. Based on docking results and catalytic characterization
we conclude that the mutation A330F provides new aromatic interactions (‘π-X-π’) that
improve binding and conversion of alkyl-benzenes. In agreement with previously published
data on xylene hydroxylation, we observed that chemo-selectivity might not only guided by
the binding orientation, but also the substrate structure plays a key role. A constant pattern is
visible in which the o-alkyl substituents (besides sterical effects) promote α-benzylic
hydroxylations whereas aromatic hydroxylation is pre-dominant (>90%) when no o-
substituent is present. It still remains unclear how P450 BM3 “decides” to catalyze aromatic
or benzylic hydroxylation. The potentially intrinsic influence of methyl groups e.g. via carbon
radical stabilization or altered BDE therefore seems to play an underestimated role during
catalysis that requires deeper investigations. Based on the well-accepted enzymatic
epoxidation mechanism of benzenes, epoxide ring opening (3a and 3b) is strongly directed by
phenolic substituents to obtain 1,4-hydroquinones with high selectivity. In summary, the
obtained performance of P450 BM3 variant M3 indicates that a biocatalytic route to vitamin E
has potential to become of commercial interest.
23
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
Experimental Section
Chemicals and reagents
All chemicals were obtained from Sigma Aldrich (Steinheim, Germany), abcr (Karlsruhe,
Germany) or AppliChem (Darmstadt, Germany). Oligonucleotides were obtained at HPLC
purity from Eurofins MWG Operon (Ebersberg, Germany). Plasmid extraction was done
using a commercial Kit (QUIAGEN QIAprep Spin Miniprep Kit, Hilden, Germany). Glucose
dehydrogenase (GDH) from Pseudomonas sp. and catalase from bovine liver were obtained
from Sigma Aldrich (Steinheim, Germany). Phusion DNA polymerase, DpnI and dNTPs were
purchased from Fermentas (St.Leon-Rot, Germany).
Preparation and expression of site-saturation mutagenesis library on position A330
The saturation mutagenesis library on position A330 was prepared using variant M2 (R47S,
Y51W, I401M) as template.[2a] Oligonucleotides for mutagenesis were designed according to
the manual of the QuikChange Site-Directed mutagenesis Kit manual with following
sequences: 5’- CAACTGCTCCTNNKTTTTCCCTATATGC-3’ (forward primer) and 5’-
GCATATAGGGAAAAMNNAGGAGCAGTTG-3’(reverse primer). Amplification of DNA
for site-directed mutagenesis was achieved using a two-step PCR protocol with a previously
reported PCR master mix composition and thermo cycler program.[2a] Successful
amplification by PCR was verified by separating products on 0.8% agarose gel. After DpnI
digest of template DNA for 4 h at 37 °C, plasmids were transformed into chemically
competent E. coli BL21 lacIq1 cells.[43] Sequencing of plasmids was done at MWG Eurofins
DNA (Ebersberg, Germany) followed by analysis using Clone Manager 9 Professional
Edition software (Scientific & Educational Software, Cary, NC, USA). Preparation of
microtiter plates (MTPs), P450 BM3 expression and lysate preparation for screening in MTP
format was done as described elsewhere in detail.[44]
24
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
Screening and selection of variants for improved conversion of 1a
The SSM library was screened with 1a as substrate to identify a P450 BM3 variant that
preferentially hydroxylates at arylic carbons (positions 3 and 6). For this two 96 well MTPs
(180 clones; 99.9% theoretical coverage)[29] were screened, by measuring the NADPH
oxidation rate and formation of phenolic products using the 4-AAP detection system.[28] Each
of the 96 wells contained 50 µL protein lysate, 10% (v/v) DMSO, 200 mM 1a (2 M stock in
DMSO) and 130 µL potassium phosphate buffer (KPi; 50 mM; pH 7.5). MTPs were incubated
for 5 min before supplementation of 50 µl NADPH (1 mM stock). Oxidation of NADPH was
measured at 340 nm in a Tecan Sunrise MTP reader (Tecan Group Ltd., Männedorf,
Switzerland). After full depletion of NADPH in the first wells, the reaction was quenched by
pipetting 25 μl quenching buffer (4 M urea in 0.1 M NaOH) into all wells. The regioselective
4-AAP assay for phenol detection was performed as described elsewhere.[28] After 30 min
incubation the absorbance was measured at 509 nm in a Tecan Sunrise MTP reader to
determine relative quantities of 2a and 2b by comparing these to the productivity of the
template used for site saturation mutagenesis (variant M2).
Expression, purification and biochemical characterization of P450 BM3
Expression of P450 BM3 in shake flasks and purification of the monooxygenase was achieved
as described elsewhere.[44-45] Cell-free lysates cell pellets were prepared according to a
published protocol.[15] Shortly, frozen cell pellets were lysed with lysozyme followed by
sonication, centrifugation and freezing of cell-free supernatant in liquid N2. Finally, remaining
water was removed overnight (16-20 h) under vacuum in a freeze-dryer. Kinetic
characterizations were done using purified P450 protein at concentrations of 0.3 µM (WT)
and 0.1 µM (P450 BM3 variant M2 and M3) which was determined by CO-binding assay
following the protocol by Omura and Sato.[46] NADPH oxidation activity and coupling
efficiency of the P450 variants and WT was measured as initial NADPH oxidation rates at
25
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
340 nm in 5 mL quartz glass cuvettes. For this, the reaction contained 10 mM of substrate
(1a/1b; pre-solubilized in DMSO), 1.5% DMSO (V/V) in potassium phosphate buffer (KPi;
50 mM; pH 7.5) in a final volume of 5 mL. After 5 min incubation, 200 µM NADPH were
supplemented and the oxidation of the cofactor was measured at 340 nm in a Varian Cary 50
UV spectrophotometer (Agilent Technologies, Darmstadt, Germany). After full depletion of
NADPH, products were extracted twice with 500 µL MTBE containing 1.5 mM 2,5-
dimethylphenol (2,5-DMP) as internal standard. Organic phases were dried over anhydrous
MgSO4 and combined for product analysis by GC-FID/-MS. Calibration curves for 2a-h were
prepared accordingly from analytical standards.
Conversion of 1a and 1b in presence of a cofactor regeneration system
Regioselectivity, product yields and total turnover number (TTN) were determined using a
GDH regeneration system for efficient regeneration of the NADPH cofactor.[47] If not stated
otherwise, conversions of 1a contained: 1 μM P450 BM3 variant, 3 U GDH, 60 mM glucose,
1400 U mL-1 catalase, 10 mM substrate, 2% DMSO (V/V), 400 μM NADPH, 10 mM ascorbic
acid in 50 mM phosphate buffer (pH 7.5). For the conversion of 1b, a 4 mL reaction mixture
was assembled containing 0.3 µM P450 BM3 WT or 0.1 µM of variant M2 or M3, 1.5%
DMSO (V/V), 10 mM substrate, 30 mM glucose, 200 µM NADPH, 1200 U mL-1 catalase and
12 U GDH.
GC-FID and GC-MS analysis
If not stated otherwise, products were extracted twice after 24 h conversion with 500 µL
MTBE containing 2,5-DMP (internal standard) as described in the previous section. Extracted
products were analysed by GC-FID and GC-MS. For GC-FID measurements, 1 µL of the
organic phase was injected into a GC-2010-Plus chromatograph (Shimadzu GmbH, Duisburg,
Germany) connected to a flame-ionization detector (FID) employing H2 as carrier gas.
26
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
Extracted compounds from conversion of 1a were separated on a Hydrodex-ß-TBDAc
column (Macherey Nagel, Düren, Germany). A temperature gradient was applied starting
from 120 °C for 8 min, 7 °C min-1 up to 210 °C and a final hold for 2 min at 210 °C.
Conversions of 1b were separated isothermally at 140 °C on a FS Supreme 5 capillary column
(Macherey Nagel, Düren, Germany; Figure S4). For GC-MS analysis a Shimadzu GC-2010
gas chromatograph (Duisburg, Germany) was used connected to a Shimadzu GC-MS
QP2010S mass detector (Duisburg, Germany). Products resulting from conversions of 1a
were separated using the following program: 100 °C for 1 min, 3 °C min-1 up to 150 °C, 15 °C
min-1 up to 250 °C, hold for 3 min (Optima 17 ms column). Further GC-MS analysis (1b
conversions) was performed also using an Agilent 7890A GC-system equipped with an
Agilent 5975C mass-selective detector. Separation was achieved using an Agilent HP-5MS
column (30 m x 320 μm, 0.25 μm film) and the following program: 100 °C/hold 0.5 min; 10
°C min-1 to 300 °C; hold 0 min.
1H- and 13C-NMR analysis
NMR spectra of 1a conversions were measured on a Bruker Avance I 400 MHz NMR
spectrometer. NMR spectra of 1b conversions were measured on a Bruker Avance III 300 MHz
NMR spectrometer. Chemical shifts are reported relative to TMS (δ= 0.00 ppm). Coupling
constants J are given in Hz.
Molecular modeling of P450 BM3 variants and docking of substrates
For visualization, substrate docking and modelling of protein structures the P450 BM3 WT
crystal structure (PDB ID: 1BU7) was used. The crystal structure model is available under
http://www.rcsb.org/pdb/home/home.do (accessed 2017-01-20). The structure models of
P450 BM3 M1 (R47S, Y51W), M2 (R47S, Y51W, I401M) and M3 (R47S, Y51W, A330F,
I401M) were based on the aforementioned crystal structure of P450 BM3 WT and generated
27
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
by using Yasara Structure version 11.6.16 software.[32] Mutation(s) were introduced using
swap option. The crystal structure of P450 BM3 as well as structure of the variants were
further rotamerized and energy minimized to correct the residues that have non-standard
torsion angles. Docking of substrates was carried out using VINA docking plug-in (25
independent runs)[48] within Yasara according to the default docking parameters from the
example directory. The simulation cell was set to be 5 Å from the heme iron of P450 BM3.
Visualization of the most meaningful docking results were prepared using Yasara software
package.
ASSOCIATED CONTENT
Supporting Information
The Supporting Information (SI) includes results of GC and NMR analysis.
Corresponding Author
*e-mail, u.schwaneberg@biotec.rwth-aachen.de
*e-mail, alexander.dennig@tugraz.at
*e-mail, aj.ruff@biotec.rwth-aachen.de
Author Contributions
The manuscript was written through contributions of all authors. All authors have given
approval to the final version of the manuscript. ‡These authors contributed equally.
ACKNOWLEDGMENTS
We would like to acknowledge the EU funded 7th Framework project OXYGREEN (FP7-
KBBE; Project Reference: 212281) for mesitylene related work. The work concerning
pseudocumene has received funding from the European Union (EU) project ROBOX (Grant
Agreement Nr. 635734) under EU’s Horizon 2020 Programme Research and Innovation
28
This article is protected by copyright. All rights reserved.

10.1002/chem.201703647
Chemistry - A European Journal
actions H2020-LEIT BIO-2014-1. The authors thank Dr. Iwona Kaluzna and Dr. Monika
Müller (both DSM, The Netherlands) for fruitful discussions and Ines Bachmann-Remy and
Prof. Dr. Walter Leitner (Institute of Technical and Macromolecular Chemistry, RWTH
Aachen) for NMR analysis. Furthermore we thank Marco Grull (RWTH Aachen) for
experimental assistance.
ABBREVIATIONS
TMP, trimethylphenol; NADP(H), nicotinamide adenine dinucleotide phosphate; TOF,
turnover frequency; WT, wild type; TTN, total turnover number; CFE, cell free extracts;
DMBA, dimethylbenzylalcohol; TMHQ, trimethylhydroquinone; MTP, microtiter plate.
REFERENCES
[1]
[2]
H. Gröger, Angew. Chem., Int. Ed. Engl. 2014, 53, 3067-3069.
a) A. Dennig, J. Marienhagen, A. J. Ruff, L. Guddat, U. Schwaneberg, ChemCatChem
2012, 4, 771-773; bR. J. Schmidt, Appl. Catal., A 2004, 280, 89-103.
a) E. Busto, R. C. Simon, W. Kroutil, Angew. Chem., Int. Ed. Engl. 2015, 54, 10899-
10902; b) T. Netscher, Vitam. Horm. 2007, 76, 155-202; c) T. Netscher, Angew.
Chem., Int. Ed. Engl. 2014; d) R. C. Simon, E. Busto, N. Richter, V. Resch, K. N.
Houk, W. Kroutil, Nat. Commun. 2016, 7, 13323.
[3]
[4]
[5]
[6]
T. Oku, Y. Arita, T. Ikariya, Adv. Synth. Catal. 2005, 347, 1553-1557.
H. Fiege, Ullmann's Encyclopedia of Industrial Chemistry 2000.
a) R. Kannappan, S. C. Gupta, J. H. Kim, B. B. Aggarwal, Genes Nutr. 2012, 7, 43-52;
b) A. C. Lindhorst, J. Schütz, T. Netscher, W. Bonrath, F. E. Kühn, Catal. Sci.
Technol. 2017; c) M. Uyanik, H. Hayashi, K. Ishihara, Science 2014, 345, 291-294.
A. A. Agaev, K. M. Madatzade, Russ. J. Appl. Chem. 2004, 78, 683-684.
a) R. D. Chambers, P. Goggin, W. K. R. Musgrave, J. Chem. Soc. 1959, 1804-1807; b)
O. A. Kholdeeva, I. V. Kozhevnikov, V. N. Sidel'nikov, V. A. Utkin, Bull. Acad. Sci.
USSR, Div. Chem. Sci. 1989, 38, 1903-1907.
[7]
[8]
[9]
M. Tavanti, F. Parmeggiani, J. R. G. Castellanos, A. Mattevi, N. J. Turner,
ChemCatChem 2017.
[10] W. Adam, W. A. Herrmann, J. Lin, C. R. Saha-Möller, R. W. Fischer, J. D. G.
Correia, Angew. Chem., Int. Ed. Engl. 1994, 106, 2545-2546.
[11] G. A. Olah, T. D. Ernst, J. Org. Chem. 1989, 54, 1204-1206.
[12] a) A. Dennig, N. Lülsdorf, H. Liu, U. Schwaneberg, Angew. Chem., Int. Ed. Engl.
2013, 52, 8459-8462; b) F. Hollmann, I. W. C. E. Arends, K. Buehler, A. Schallmey,
B. Bühler, Green Chem. 2011, 13, 226-265; c) O. Shoji, T. Kunimatsu, N. Kawakami,
Y. Watanabe, Angew. Chem., Int. Ed. Engl. 2013, 52, 6606-6610; d) C. J. Whitehouse,
S. G. Bell, L. L. Wong, Chem. - Eur. J. 2008, 14, 10905-10908; e) C. J. Whitehouse,
S. G. Bell, L. L. Wong, Chem. Soc. Rev. 2012, 41, 1218-1260.
[13] A. Karich, M. Kluge, R. Ullrich, M. Hofrichter, AMB Express 2013, 3, 5.
[14] H. E. Schoemaker, D. Mink, M. G. Wubbolts, Science 2003, 299, 1694-1697.
[15] A. Dennig, E. Busto, W. Kroutil, K. Faber, ACS Catalysis 2015, 5, 7503-7506.
[16] B. Meunier, S. P. de Visser, S. Shaik, Chem. Rev. 2004, 104, 3947-3980.
[17] S. P. de Visser, S. Shaik, J. Am. Chem. Soc. 2003, 125, 7413-7424.
[18] a) C. J. Whitehouse, S. G. Bell, W. Yang, J. A. Yorke, C. F. Blanford, A. J. Strong, E.
J. Morse, M. Bartlam, Z. Rao, L. L. Wong, ChemBioChem 2009, 10, 1654-1656; b) C.
J. Whitehouse, N. H. Rees, S. G. Bell, L. L. Wong, Chem. - Eur. J. 2011, 17, 6862-
29
This article is protected by copyright. All rights reserved.