Mechanistic basis for the enantioselectivity of the anaerobic hydroxylation of alkylaromatic compounds by ethylbenzene dehydrogenase

Article abstract of DOI:10.1016/j.jinorgbio.2014.05.006

The enantioselectivity of reactions catalyzed by ethylbenzene dehydrogenase, a molybdenum enzyme that catalyzes the oxygen-independent hydroxylation of many alkylaromatic and alkylheterocyclic compounds to secondary alcohols, was studied by chiral chromatography and theoretical modeling. Chromatographic analyses of 22 substrates revealed that this enzyme exhibits remarkably high reaction enantioselectivity toward (S)-secondary alcohols (18 substrates converted with > 99% ee). Theoretical QM:MM modeling was used to elucidate the structure of the catalytically active form of the enzyme and to study the reaction mechanism and factors determining its high degree of enantioselectivity. This analysis showed that the enzyme imposes strong stereoselectivity on the reaction by discriminating the hydrogen atom abstracted from the substrate. Activation of the pro(S) hydrogen atom was calculated to be 500 times faster than of the pro(R) hydrogen atom. The actual hydroxylation step (i.e., hydroxyl group rebound reaction to a carbocation intermediate) does not appear to be enantioselective enough to explain the experimental data (the calculated rate ratios were in the range of only 2-50 for pro(S): pro(R)-oriented OH rebound).

Full text of DOI:10.1016/j.jinorgbio.2014.05.006

Mechanistic basis for the Enantioselectivity of the Anaerobic Hydroxylation
of Alkylaromatic compounds by Ethylbenzene Dehydrogenase
Maciej Szaleniec, Agnieszka Dudzik, Bartłomiej Kozik, Tomasz Borowski,
Johann Heider, Małgorzata Witko
PII:
DOI:
Reference:
S0162-0134(14)00143-3
doi: 10.1016/j.jinorgbio.2014.05.006
JIB 9526
To appear in:
Journal of Inorganic Biochemistry
Received date:
Revised date:
Accepted date:
10 September 2013
16 May 2014
18 May 2014
Please cite this article as: Maciej Szaleniec, Agnieszka Dudzik, Bartꢀlomiej Kozik, Tomasz
Borowski, Johann Heider, Malꢀgorzata Witko, Mechanistic basis for the Enantioselectivity
of the Anaerobic Hydroxylation of Alkylaromatic compounds by Ethylbenzene Dehydro-
genase, Journal of Inorganic Biochemistry (2014), doi: 10.1016/j.jinorgbio.2014.05.006
This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.

ACCEPTED MANUSCRIPT
Mechanistic basis for the Enantioselectivity of the Anaerobic Hydroxylation of Alkylaromatic compounds
by Ethylbenzene Dehydrogenase.
Maciej Szaleniec* [a], Agnieszka Dudzik[a], Bartłomiej Kozik[c], Tomasz Borowski[a], Johann Heider[b], Małgorzata
Witko[a]
[a] Jerzy Haber Institute of Catalysis and Surface Chemistry Polish Academy of Sciences Niezapominajek 8, 30-239 Kraków
*Maciej Szaleniec: Fax: (+48)124251923 E-mail: ncszalen@cyfronet.pl
[b] Laboratory for Microbial Biochemistry, Philipps University of Marburg, Karl-von-Frisch Strasse 8, D-35043 Marburg, Germany
[c] Department of Organic Chemistry, Jagiellonian University, Ingardena 3, 30-060 Kraków, Poland
Keywords: ethylbenzene dehydrogenase • molybdenum enzymes • chiral chromatography • enantioselective
reaction • DFT • QM:MM
Abstract
The enantioselectivity of reactions catalyzed by ethylbenzene dehydrogenase, a molybdenum enzyme that
catalyzes the oxygen-independent hydroxylation of many alkylaromatic and alkylheterocyclic compounds to
secondary alcohols, was studied by chiral chromatography and theoretical modeling. Chromatographic analyses of
22 substrates revealed that this enzyme exhibits remarkably high reaction enantioselectivity toward (S)-secondary
alcohols (18 substrates converted with >99% ee). Theoretical QM:MM modeling was used to elucidate the structure
of the catalytically active form of the enzyme and to study the reaction mechanism and factors determining its high
degree of enantioselectivity. This analysis showed that the enzyme imposes strong stereoselectivity on the reaction
by discriminating the hydrogen atom abstracted from the substrate. Activation of the pro(S) hydrogen atom was
calculated to be 500 times faster than of the pro(R) hydrogen atom. The actual hydroxylation step (i.e., hydroxyl
group rebound reaction to a carbocation intermediate) does not appear to be enantioselective enough to explain
the experimental data (the calculated rate ratios were in the range of only 2-50 for pro(S): pro(R)-oriented OH
rebound).
1. Introduction
Pure chiral alcohols are universal synthons for the industrial production of fine chemicals and can be used as a
starting material for the development of new drugs, odorants, food additives or even chiral nematic liquid crystals
[1-5]. The most popular methods for the synthesis of optically active alcohols include either the chemical or
biocatalytic asymmetric reduction of pro-chiral ketones or the kinetic resolution of racemic alcohols by a lipase.
1

ACCEPTED MANUSCRIPT
Although the aforementioned methods have already found their way into industrial applications, they are
characterized by a number of drawbacks, including reversibility in asymmetric ketone reduction, need to regenerate
the NADH/NADPH cofactor, 50% maximum yield of the catalytic process, and frequent low reaction
enantioselectivity, particularly in the case of whole cell biocatalysts or that of racemic resolution with lipases [6-8].
Therefore, the production of chiral alcohols by irreversible, catalytic hydroxylation of hydrocarbons is an interesting,
chemically complementary method for the established processes [9, 10]. Such a reaction can be catalyzed by
ethylbenzene dehydrogenase (EbDH), a promiscuous enzyme that catalyzes the chiral hydroxylation of many
alkylaromatic and alkylheterocyclic compounds to secondary alcohols (Error! Reference source not found.).
EbDH is a molybdoenzyme belonging to subfamily II of the DMSO reductase family [11, 12] that catalyzes the
oxygen-independent, stereospecific hydroxylation of ethylbenzene to (S)-1-phenylethanol ((S)-1). EbDH is the first
known enzyme to be capable of the direct anaerobic hydroxylation of non-activated hydrocarbons [11, 13], and it is
involved in the mineralization of ethylbenzene by Aromatoleum aromaticum, a denitrifying Betaproteobacterium
[14-16]. EbDH is an  heterotrimer with subunits of 96, 43, and 23 kDa that contains a molybdenum cofactor and
a heme b559 cofactor linked by a linear row of five iron-sulfur clusters [17]. The enzyme active site is deeply buried
in the  subunit and is composed of the molybdenum cofactor (MoCo) and Asp223 and His192 side chains, which
have been proposed to be involved in the catalytic process. Based on analogy to other molybdenum enzymes
within the DMSO reductase family that catalyze oxygen transfer reactions [18, 19], it was suggested that an oxo
ligand is coordinated at the molybdenum atom in the oxidized form of the catalytically active enzyme [17, 20]. To
date, there is no experimental evidence of this hypothesis from direct structural or spectroscopic studies, but we
have previously shown that mechanistic models based on this assumption yield plausible results [20, 21].
Therefore, we investigated the enzyme enantioselectivity using the same active-site geometry.
Our studies have shown that EbDH hydroxylates aliphatic methylene groups adjacent to aromatic or heterocyclic
rings in many different substrates, forming the respective secondary alcohols [22-26]. The implementation of chiral
chromatography revealed that EbDH exhibits 100% reaction enantioselectivity (within detection limits) for a range
of compounds, exclusively yielding (S)-alcoholic products. For some substrates, however, substantial amounts of
(R)-alcohols were also detected, although always as a minor fraction [24, 27].
Studies on the hydroxylation of ethylbenzene derivatives to secondary alcohols by biotransformation are relatively
limited. In most cases, monooxygenases from the cytochrome P-450 group are used for the production of desired
compounds. Examples of such biotransformations have been shown for bacteria (e.g., Bacillus megaterium [28])
2

ACCEPTED MANUSCRIPT
and a range of fungi (e.g., Aspergillus niger [29], Mortierella isabellina, Helminthosporium species, Cunninghamella
echinulata [30], Fusarium moniliforme [31, 32]). However, in almost all cases, the biotransformation of
alkylbenzene compounds lead to mixtures of alcoholic products oxidized at benzylic and non-benzylic carbon
atoms. In contrast to these monooxygenases, ethylbenzene dehydrogenase exhibits more regioselective activity,
catalyzing activation of the C-H bond only at secondary carbon atoms adjacent to aromatic or heterocyclic systems
[23, 24].
Taking these reasons into account, we have investigated the enantioselectivity of EbDH for a variety of substrates
using experimental and theoretical techniques. The elucidation of enantioselectivity in reactions catalyzed by EbDH
is an important step on the road to practical applications for this enzyme in the biotechnological industry [33]. The
stereochemistry of the products was investigated by chiral liquid and gas chromatography mass spectrometry (LC-
MS and GC-MS), while quantum chemical modeling was used to understand the experimentally observed
phenomena. To date, no structural data for the structure of the oxidized, and hence catalytically active, active site
of EbDH are available. However, quantum mechanics:molecular mechanics (QM:MM) calculations take into
account the entire  subunit of the enzyme and assume that a directly bound oxo ligand at the Mo atom has
allowed for the modeling of the catalytically active structure of the MoCo and the investigation of the influence of
the enzyme milieu on the reaction pathway, including factors responsible for the high (S)-specificity observed for
most of the studied compounds. Additionally, the reason behind the loss of enantioselectivity observed in the case
of 1,2-diethylbenzene is discussed, and a hypothetical mechanistic explanation is proposed.
3

ACCEPTED MANUSCRIPT
2. Experimental Section
2.1 Sample preparation
Ethylbenzene dehydrogenase was purified from ethylbenzene-grown Aromatoleum aromaticum cells as previously
described [11, 23]. The enzyme assays were routinely conducted at an optimum pH of 7.5 at 30°C in 20 ml of 100
mM Tris/HCl containing 200 M ferrocenium (III) tetrafluoroborate and ethylbenzene dehydrogenase (100-300 l of
app. 1 mg/ml protein solution). The reactions were initiated by adding 100 l of a stock solution of the respective
substrate in tert-butanol (a list of substrates, their purities and producers is available in the Supporting Information).
After an overnight incubation, the reaction mixtures where extracted from the water phase by solid-phase extraction
using either C18 Polar Plus (J.T.Baker® from Avantor Performance Materials, US) or polystyrene/divinylbenzene
(PS/DVB) copolymers solid phase extraction (SPE) columns (Strata™-X from Phenomenex, US or the equivalent
Chromabond® HR-X from Macherey-Nagel, Germany), which were eluted with 0.5 ml of isopropanol. The only
exception to the procedure was the 2-ethyl-1H-pyrrole reaction mixtures (22). Due to the high polarity of the
product derived from this compound, we were unable to concentrate the sample and exchange the solvent by SPE.
Therefore, the enzyme was precipitated by the addition of 50% methanol, and it was removed by centrifugation.
The supernatant was directly analyzed by non-chiral reversed-phase (RP)-LC.
2.2. Organic synthesis of standards
Chiral standards for most products were commercially available (with the exception of the ones mentioned below;
see Supplementary Information for a list of compounds, their purities and producers). The racemic compounds 1-
(4-ethylphenyl)ethanol (VI), (1S,1’S)-1,1'-(benzene-1,4-diyl)diethanol (VIII) and 1-(3-methylphenyl)ethanol (XIII);
their enantiomerically pure (S)-stereoisomers, and the alcohol (S)-1-([1,1'-biphenyl]-4-yl)ethanol ((S)-XI) were
prepared experimentally by reducing the appropriate carbonyl compounds. Racemates of VI, VIII and XIII were
obtained via chemical hydrogenation of 4-ethylacetophenone, 1,4-diacetylbenzene and 3-methylacetophenone with
sodium borohydride in anhydrous ethanol at -10 – (-5) ◦C (1.2 molar excess or 2.2 molar excess of borohydride in
the case of the diketone). The enantiomerically pure alcohols (S)-VI and (S,S)-VIII were synthesized from
4-ethylacetophenone and 1,4-diacetylbenzene by enantioselective reduction with (−)-β-
chlorodiisopinocampheylboran [(−)-DIP-ChlorideTM][34, 35], whereas (S)-XIII and (S)-XI were obtained by
asymmetric enzymatic reductions of 3-methylacetophenone and 4-acetyl-1,1'-biphenyl, respectively, using
enzymes from carrot (Daucus carota) roots [36-38].
Enantiomeric excesses of the compounds (S)-VI (96% ee), (S)-XII (100% ee), and (S,S)-VIII (100% ee) were
determined by chiral HPLC analysis. The absolute configurations of VI, VIII and XII were assigned by comparison
of the measured optical rotations with appropriate literature values [39].
4

ACCEPTED MANUSCRIPT
2.3 LC/LC-MS
Chiral HPLC separations were performed with an Agilent 1100 System in a normal-phase system in a cellulose
tribenzoate polysaccharide chiral stationary phase (Daicel CHIRALCEL® OB-H column, 250x4.6 mm, 5 m). To
avoid changes in the sample concentrations during the analysis, the autosampler was thermostated to 8o C.
Depending on the polarity of analytes, different types of uniform isocratic programs were used with mobile phases
composed of mixtures of n-hexane/isopropanol in ratios ranging from 97:3 up to 70:30 (v/v) (see Table 2) at 25°C
and at a flow rate 0.5 ml/min. Products were detected by a diode array detector (Agilent 1100 DAD) with or without
coupling to an MS detector (VL MS spectrometer) using an atmospheric pressure chemical ionization (APCI) ion
source in positive or negative ion mode (depending on the compound – see Table S3 of the Supplementary
Information). For efficient APCI ionization in the normal-phase system, a post column addition of 25 mM
ammonium formate buffer containing 75% isopropanol solution was applied to each sample (1:1 v/v). More details
for the MS procedure development are available in the Supplementary Information. Both the scan and SIM modes
were used to detect the product ions of the mass spectra. The characteristic [M-OH]+ originating from fragmentation
of the alcohols by dehydroxylation dominated the mass spectra of the products and was most frequently monitored
in the SIM mode.
The absolute configuration of the reaction products was determined based on commercially available standards
(either (S) or (R) isomers) or self-synthesized standards. In some cases in which no commercial chiral standards
were available and no pure products were accessible by our synthetic reductive procedures (see organic synthesis),
hypothetic configurations of the products were established based on a previously obtained quantitative structure-
enantioselective retention relationship (QSERR) model [24] combined with the retention times of fractions from
racemic standards. This approach has proven to be effective in predicting the retention order of a wide range of
ethylbenzene derivatives. Notable exceptions include the reaction product of 4-ethylbiphenyl, 1-([1,1'-biphenyl]-4-
yl)ethanol (XI), alcohols generated from substrates exhibiting closed rings such as 2,3-dihydro-1H-inden-1-ol (XV)
or 1,2,3,4-tetrahydronaphthalen-1-ol (XVII) and alcohols with an unsaturated aliphatic side change (such as 1-
phenylprop-2-en-1-ol (III)), which exhibited a unexpected reverse retention order of the enantiomers (i.e., first the
(R)- then the (S)-alcohol) on Chiralcel-OB-H CSP. Our QSERR retention model was supplemented with the
observation of reverse retention patterns. In all of those cases as well as a few examples in which racemic alcohols
were also not available, only enantiomeric excesses (ee) were reported, and a hypothetical identification of
stereoisomers was proposed.
5

ACCEPTED MANUSCRIPT
Reaction mixtures from the conversion of 2-ethyl-1H-pyrrole (22) were analyzed using a CN column (Zorbax SB-
CN Narrow Bore RR 2.1 x 100 mm, 3.5 m) in reverse-phase mode (using a gradient water/methanol mobile
solvent system) with positive-ion APCI MS detection mode.
Identification of the products was either based on comparison with available standards (with the same retention
times, UV-vis spectra and mass spectra) or deduction from the mass spectra obtained by LC-MS and GC-MS
analysis (see Table S3 of the Supplementary Information for MS identification details). Enantiomeric excess was
determined by chiral LC when only racemic or no standards were available. Finally, in most of cases absolute
configurations were determined based on commercially available or synthesized chiral standards. For some
compounds, highly probable absolute configurations were assigned based on the QSERR model that was
developed to predict the retention order of enantiomers on a Chiralcel OB-H column [24]. The previously described
QSERR model was supplemented with our recent observation that alcohol enantiomers derived from substrates
with unsaturated bonds in their alkyl chains (e.g.,1-phenylprop-2-en-1-ol (III) or 1-phenylprop-2-yn-1-ol (IV)) or
substrates with closed alkyl ring structures (e.g.,2,3-dihydro-1H-inden-1-ol XV or 1,2,3,4-tetrahydronaphthalen-1-ol
XVII) exhibit a reversed order of elution from columns (i.e., first the (S) enantiomer followed by the (R) isomer). In
these cases, data showing probable absolute configurations of major products are provided in addition to data
reporting the enantiomeric excess.
2.4 GC-MS
Some reaction mixtures were additionally analyzed with GC-MS. The EI ionization mode that was used allowed for
a more efficient ionization for most of the analytes (i.e., both substrates and products) and provided characteristic
fragmentation patterns, revealing more structural information about the products. Separations were conducted on a
Cyclodex-B column (60 m, ID 0.25 mm, 0.25 m) using a Trace DSQ 1000 Thermo Finnigan GC-MS. Analysis was
performed in split mode (1:20) with helium as a carrier gas and 1 l injections. The temperature program was
adjusted to the polarity and volatility of the particular compounds but basically comprised 2 minutes of hold time (at
60 or 100°C) followed by a slow temperature gradient (3-7 o/min) up to 180°C with a final hold time of 5-10 minutes
at 180°C.
2.5. QM:MM modeling
For QM:MM modeling [40, 41], a two layered ONIOM method was used [42] with an electronic embedding (EE)
formalism. Electronic embedding incorporates the partial charges of the MM region into the quantum mechanical
Hamiltonian, which provides a better description of the electrostatic interaction between the QM and MM regions.
Calculations were conducted using the Gaussian 09 suit program with the B3LYP function and a 6-31G(d,p) basis
set for light atoms, the LANL2DZ effective core potential and a basis set for Mo and an AMBER force field for the
6

ACCEPTED MANUSCRIPT
MM region. The choice of density functional theory (DFT) follows from its suitability to adequately describe reaction
pathways at a reasonable computational cost and its prior success in the modeling of organic reactions with
transition metal catalysts [43-45]. The MM point charges for a substrate, cofactor and iron-sulfur cluster present in
the alpha subunit of EbDH were fitted to the electrostatic p2.7otential according to the Merz-Singh-Kollman method
[46, 47].
The initial QM:MM model was obtained from the EbDH crystal structure (PDB code: 2IVF) in which the
molybdenum cofactor was modified to attain the conformation of the oxidized Mo(VI) form. Molecules from the
crystallization buffer were removed. The structure of the oxidized alpha subunit was optimized using a mechanical
embedding approach, and the substrate was docked into the active site with Autodock 3.0 [48].
The QM:MM working model was composed of the entire  subunit, with the water molecules visible in the crystal
structure (15907 atoms, total charge -3). Two types of QM parts were used in independent models: a small portion
(QM1, Error! Reference source not found.a) composed of 52 atoms (charge 0) and a larger one (QM2, Error!
Reference source not found.b) composed of 168 atoms (charge +1). All residues penetrating a 5 Å radius zone
from the molybdenum cofactor were allowed to move during the optimization of the model (1742 atoms), while the
rest of the protein was kept frozen (see Fig. S1 of the Supplementary Information). The QM1 portion contained only
the molybdenum cofactor core, ethylbenzene and the Asp223 and His192 side chains (see Error! Reference source
not found.A). The QM2 portion added Ile86, Trp87, Val193, Ile217, Asn218, Ile221, Phe446 and Trp481 side chains to QM1,
thus providing a more adequate electronic environment for the reaction center and further distancing the QM/MM
border from the reacting species. QM2 calculations were conducted to determine whether the electronic
embedding did not introduce artificial overpolarization to the studied system. These calculations were conducted
due to the observed discrepancies in the reaction energy profiles obtained with mechanical and electronic
embedding approaches and QM1 (data not shown). The transition states were localized by potential energy scans
along approximate reaction coordinates, followed by full optimization of the TS geometries using the Berny
algorithm.
Finally, single-point energies were calculated for the QM1 and QM2 models with the triple-dzeta basis set lacv3p**.
Due to the size of the QM:MM system, it was not possible to calculate vibration corrections for electronic energy,
although they were computed for the QM1 portion supplemented with link atoms. The resulting zero-point energy
corrections were scaled by a factor of 0.98, which is suitable for B3LYP/6-31G(d,p).
2.6. Interaction energies
To explain the observed loss in enantioselectivity in the conversion of 1,2-diethylbenzene, four different
approximate TS1 structures were studied by the QM:MM technique using a mechanical embedding approach.
7

ACCEPTED MANUSCRIPT
Initial approximate TS1 geometries for the abstraction of each of the four benzylic H atoms of 1,2-diethylbenzene
were first derived from the pro(S) TS1 structure with ethylbenzene. The structure of each complex was then
optimized on the B3LYP/AMBER level of theory with constraints imposed on the C--H and H--(O-Mo) bond lengths
(1.3 and 1.35 Å, respectively). The obtained geometries of the approximate TS1 complexes were used to estimate
vdW and electrostatic interactions between the substrate and enzyme active site. For each complex, interactions
were evaluated between the substrate, the MGD cofactor and a set of 19 amino acid residues penetrating a 5 Å
radius around the substrate (see Supplementary Information). Due to the geometry of the transition state, that is,
the close contact between the hydrogen atom being transferred and the oxo ligand, the vdW interactions with MGD
were always high and dominated the overall interaction energy value. Therefore, to evaluate the impact of the
active site on accommodation of a particular substrate conformation, the interaction energy was also calculated
without the cofactor. The same approach was also used to quantify interactions for the pro(S) and pro(R)
conformations of the TS1 and TS2 transition states in the case of ethylbenzene hydroxylation.
Moreover, to investigate the degree of the overpolarization of the QM-region that affects the energy in the
electronic embedding QM:MM approach for reaction stationary points, we calculated the interaction energies
between the QM and MM portions of the system with and without model solvent correction. For this purpose,
charges obtained from single point QM1:MM calculations with the Merz-Singh-Kollman scheme on the B3LYP/6-
31G(d,p)/LANL2dz level of theory were used. The influence of electrostatic screening within the enzyme active site
was determined with an implicit distance-dependent dielectrics method with an  value of 4 [49]. All interaction
energy calculations were conducted in Discovery Studio 4.0 using CHARMm force field according to a previously
reported procedure [50].
2.7. Kinetic isotopic effect (KIE) and isotope studies
The kinetic isotopic effect was measured and calculated according to a previously described method [20]. The
experiments were conducted with d10-ethylbenzene and ethylbenzene at 35°C. Direct comparison of the reaction
velocities was used to determine the kH/kD ratio. B3LYP/6-31 g(d,p) calculations were performed only for the QM1
portion with optimized positions for link atoms, and the rest of the geometry was derived from QM:MM calculations.
Thermal corrections to energy were used to calculate the energy difference between S, pro(S) TS1 and pro(S) TS2
stationary points, and they were used to estimate the theoretical KIE. The Wigner correction was used to account
for tunneling effects, and imaginary frequencies were scaled with by a factor of 0.9620 [51].
Moreover, a conversion experiment with H218O was conducted that aimed to detect potential isotope fractionation
effects. The conversion of ethylbenzene was conducted in a buffer containing 37% of the water labeled with 18O.
8

ACCEPTED MANUSCRIPT
The reaction mixture was extracted and analyzed by GC-MS to establish the relative abundance of 124 and 122
m/z signals, which correspond to products with 18O and 16O insertions, respectively.
3. Results
3.1 QM/MM structural model of MoCo in the oxidized state
The EbDH MoCo is a molybdenum complex coordinated by four cis-thiolate (sulfur) groups of the bisMGD cofactor
and a carboxylate oxygen atom (O) of Asp223. The sixth ligand of Mo is modeled as a catalytically active oxo
ligand, which was introduced into the structure in place of a carboxylate oxygen atom from an acetate ligand in the
crystal structure of the reduced form of EbDH [17]. Although there is no structural or spectroscopic evidence for the
presence of this oxo ligand in the oxidized form of the EBDH MoCo, we assumed that a Mo=O ligand is highly
probable in the catalytically active enzyme as outlined in previous studies [20, 21, 50].
The overall geometry of the modeled complex (Error! Reference source not found.) is a distorted trigonal prism
with average Mo-S distances of 2.51 Å, a Mo-O-Asp223 distance of 1.96 Å and a Mo=O distance of 1.74 Å. The
oxo ligand forms a strong hydrogen bond with His192 (which was previously postulated to be actively involved in the
catalytic process [17]). Another important H-bond interaction is associated with the O atom of Asp223, which acts
as an H-bond acceptor for the NH3 group of Lys450. The only covalent bond linking the enzyme with the cofactor is
between Mo and O-Asp. However, the cofactor forms a multitude of H-bond interactions, particularly with the His
and Arg residues in the enzyme (see Supplementary Materials Fig. S2).
3.2 Reaction pathway
The predicted reaction pathway (Fig. 4) proceeds along similar steps as those reported in our previous modeling
studies [20, 21]. The oxidative half-reaction can be divided into two phases: i) C-H activation (TS1) and ii) OH-
rebound (TS2). The steric interactions of the active site with the bound substrate enforce the almost flat
conformation of ethylbenzene, with the ethyl group pointing toward the MGD-Q cofactor. This conformation
exposes the pro(S) H-atom to the Mo=O ligand. The activation of ethylbenzene takes place when the substrate
molecule approaches the oxo ligand with its methylene group. In TS1, the bond between the carbon and hydrogen
9

ACCEPTED MANUSCRIPT
atoms is broken and a new bond with the oxo ligand is formed. The bond cleavage appears to be homolytic, as
spin density corresponding to a radical intermediate is observed on the activated substrate, while Mo is reduced to
the Mo(V) oxidation state. Additionally, the imidazole group of His192, which forms a strong H-bond interaction with
the oxo ligand, donates its proton to MoCo, resulting in the formation of a Mo(V)-H2O species. Careful analysis of
the spin density reveals that in TS1, 30% of the spin density of the second electron is already transmitted to the
pterin ligands. As a result, a positive charge is formed on the substrate. The activation phase is completed by the
cleavage of the C-H bond. At this stage, the second electron is already completely transferred from the activated
substrate to the Mo atom, yielding a carbocation intermediate product (charge +0.96, spin=0) and a reduced
cofactor with a Mo(IV) atom coordinated by a water molecule. The H2O is also stabilized by H-bond interactions
2
with the N and O2 atoms of His192 and Asp223, respectively. This result is different from our previous DFT
calculation performed with small clusters, which predicted the transfer of the second electron to MoCo to occur
later during the OH rebound process [20].
The second phase of the pathway is composed of the OH rebound process, which proceeds upon approach of the
carbocation intermediate toward the Mo-H2O group. In the transition state (TS2), the Mo-O bond is broken, and a
C-O bond is formed when one of the protons is donated back to the His192 residue. The charge of the activated
hydrocarbon in TS2 (0.72) has already decreased versus the intermediate product (I), indicating partial bond
formation with the OH group. Finally, the alcohol product is formed, which is bound to the enzyme by two the H-
bonds of the OH group of the product with Asp223 and His192. The QM:MM modeling procedure reproduced the
structure of the reduced cofactor known from the crystal structure. The overlay similarity score calculated by the
steric field fit for the atoms of the cofactor and the side chains of Asp223 and His192 was 0.97 (0.98 for just the
cofactor atoms) – see Supplementary Materials Fig. S3. Using the electronic embedding approach, an exclusively
monodentate coordination mode of the Asp223 residue to Mo was observed (with O1 forming a strong H-bond with
the NH3 of Lys450). However, when the enzyme-product geometry was optimized with only a mechanical
embedding approach (data not shown), a hexacoordinate Mo(IV) MoCo structure was obtained that was similar to
the bidentately coordinated NarGHI structure obtained by Bertero et al. [52]. However, the energy of this bidentate
structure exceeded that of the monodentate structure by app. 56 kJ/mol, as calculated by an ME approach (the
energies were related to the respective enzyme-substrate complexes optimized within ME and EE in the QM:MM
method).
The reaction cycle is closed by the release of the product, coordination of a water molecule to the Mo(IV) site and
two one-electron transfer steps coupled with deprotonations, which result in the reoxidation of the active site to the
10

ACCEPTED MANUSCRIPT
catalytically active Mo(VI)=O species and the transfer of electrons via Fe-S clusters to the heme b cofactor.
However, this part of the mechanism was not addressed in this study.
Analysis of the reaction energy profile (
Table 1, Fig. 5) shows that the process of C-H activation is associated with an energy barrier of 67.2 kJ/mol (48.5
kJ/mol with ZPE correction), which is in good agreement with the experimentally measured G#298 of app. 70
kJ/mol [53] . Moreover, it appears that the formation of a carbocation is highly favored by electrostatic interactions
within the enzyme, as a significant drop in energy is observed for the carbocation intermediate product (I). This
apparent stabilization was not observed in the QM:MM calculations conducted with a mechanical embedding
approach.
It was not possible, however, to precisely determine the energy of the stationary point of intermediate state I
because the reagent tends to drift away from the vicinity of the MoCo. A similar computational artifact was
previously reported by Kloer et al. for optimization of the enzyme-product complex, and it was attributed to the
presence of a large empty cavity devoid of a solvent close to the active site [17]. To prevent the reagent from
drifting away from the MoCo, an additional constraint between the benzylic carbon atom of the intermediate and the
oxo atom of the MoCo was introduced. However, this leads to some bias in the estimated energy of species I. In
addition, the QM:MM electronic embedding approach is prone to charge overpolarization, particularly in the case of
a small QM portion. As a result, an unrealistically low energy value (-92 kJ/mol) was obtained for I. Therefore, to
improve the accuracy of the computational method, we calculated single point energies for a model with a
significantly larger QM portion (QM2), where residues in contact with the reactant are included in the QM-region
and hence screen the region where the reaction takes place. This approach resulted in elevation of the TS1 energy
to 103.1 kJ/mol (84.4 with ZPE correction) and, most importantly, an increase in the energy of I to 11.2 kJ/mol (5.9
kJ/mol with ZPE correction).
The absolute energy of the OH rebound process was associated with a lower activation barrier (only 0.1 kJ/mol
above the initial ES energy) than that for the C-H cleavage reaction. If related to the energy of I, TS2 appears to
exhibit a higher energy barrier than TS1 (92 vs. 67.2 kJ/mol). However, for the bigger QM model, the TS2 energy
drops below the energy level of I (-0.3 vs. 11.2 kJ/mol). This result means that this barrier no longer exists or is
small (accounting for the potential stabilization of I in QM2:MM in the case of removal of a distance constraint and
geometry optimization in the QM2 model). Similar result was already observed in DFT cluster modeling of the
heterolytic high pH reaction pathway [20].
11

ACCEPTED MANUSCRIPT
The overall hydroxylation reaction is exergonic (-39.3 kJ/mol for QM1 and -29.9 kJ/mol for QM2) with respect to ES,
indicating that the ethylbenzene hydroxylation is effectively irreversible.
To evaluate the effect of electrostatic screening on the reaction energy profile, we calculated the interaction
energies between the reactant and all residues within a 5 Å radius using a CHARMm force field with the QM
portion atomic point charges calculated with the QM1:MM model. It should be noted that these calculations were
based on automatic parameterization of the cofactor; therefore, interactions between the reactant and cofactor
residues and with His192 are, at best, approximate. With this approach, we observed the stabilization effect present
in the QM:MM calculations. Electrostatic interactions computed with no dielectric screening gave an increasing
stabilization of the charged reactants with respect to S in the range of -150 kJ/mol for TS1, -335 kJ/mol for I and -
500 kJ/mol for TS2. These large stabilization contributions were reduced to -4 kJ/mol for TS1, -16 kJ/mol for I and -
40 kJ/mol for TS2 when a distance-dependent dielectric model was employed. These results, together with those
obtained with the QM2/MM model, indicate that the computationally viable QM1 portion of the QM/MM model is
indeed too small to correctly describe all important electrostatic effects within the catalytic center, and as a result, it
becomes overpolarized, that is, it is too small to properly reproduce electrostatic interactions between the QM and
MM regions. However, it is expected that this error affects the pro(S) and pro(R) transition states to the same
extent (see below) because similar partial charges are present on both reactants in their TSs.
Our previous studies indicated that both transition states may be energetically close, and the relative barrier
heights may shift with pH and according to the protonation states of amino acid residues within the active site (e.g.,
His192) [20]. The kinetic isotope effect proved to be an efficient tool for elucidating that problem. Therefore,
combined experimental and theoretical studies of KIE were used to supplement the QM:MM calculations (see
below).
3.3. Enantioselectivity
The EbDH reaction products from 22 different substrates were studied by chiral chromatographic analysis. The
compounds were identified, and the enantiomeric excesses of the major chiral alcohol products and most of their
absolute configurations were established (see Table 2). Recently, another 11 substrates of EbDH containing OH
and NH2 substituents in the aromatic ring have also been reported [27]. It was shown that the hydroxylation of para
substituted alkylphenols and alkylanilines is most likely associated with a different reaction mechanism that
involves a non-enantiospecific reaction of a water molecule with the intermediate product. However, for ortho- and
meta-substituted compounds, 99+% of (S)-alcohol products were detected.
12

ACCEPTED MANUSCRIPT
The hydroxylation of alkylaromatic and alkylheterocyclic compounds by EbDH presented in this paper exhibits
99+% (S)-enantioselectivity in 19 out of 22 reaction mixture cases, as characterized by chiral chromatography
(Table 2). Significant amounts of (R) alcohols were observed for only two substrates: 4-ethyl-1,1'-biphenyl (11, 52%
ee) and 1,2-diethylbenzene (14, 60% ee).
3.4 Modeling of enantioselectivity
The enantioselectivity of the EbDH reaction was investigated by QM:MM modeling of the two possible alternative
pathways of an ethylbenzene conversion, leading to either (S)- or (R)-1-phenylethanol. By this approach, we
intended to elucidate the mechanism involved in imposing the observed S stereoselectivity in the conversion of
most substrates.
3.4.1. QM:MM model setup
Ethylbenzene was used as a representative substrate, which was docked in the active site of the alpha subunit.
Because the active site of EbDH provides a rather large cavity for substrate binding [17], bound ethylbenzene can
attain various conformations that do not necessarily present the pro(S) hydrogen atom towards the Mo=O ligand.
Based on this observation, it appeared that the formation of the ‘enzyme-substrate complex’ could not be the main
factor imposing the strong enantioselectivity. To resolve this issue, we calculated the optimal geometries and
activation energies of the transition states in QM:MM models that incorporated differently sized QM portions into
the protein scaffold by electrostatic embedding (models QM1 and QM2, see Error! Reference source not found.
and Fig. S1). Energies associated with activation of either the pro(S) or pro(R) hydrogen atoms of ethylbenzene (1)
(pro(S) and pro(R) TS1) and those for the OH rebound steps in either the pro(S) or pro(R) orientation (pro(S) and
pro(R) TS2) were calculated. Compared with our previous calculation of the reaction mechanism based on DFT
with a small model of the active site [20], we observed a strong stabilization effect on the carbocation species in the
TS2 in the QM:MM models. Moreover, the calculated energy values from the QM:MM models are in greater
agreement with the experimentally obtained data than the previous values from the DFT models (see below).
3.4.2. Modeling results of pro(S) vs. pro(R) enantioselectivity
The activation energies for the pro(S)- and pro(R)-oriented models were computed as energy differences between
those of the respective transition states and the most stable enzyme-substrate conformation obtained for the
pro(S)-oriented complex (Table 3). A significant energy difference (E(pro(R)-pro(S))) was observed, particularly
for the activation of the C-H bond (15.7 kJ/mol for the QM1 model and 17.2 kJ/mol for the QM2 model). This means
13

ACCEPTED MANUSCRIPT
that ethylbenzene (1) was activated greater than 500-fold slower (>900-fold for the QM2 model) in the pro(R)-
orientation than in the pro(S)-orientation.
For the OH rebound process, the pro(S)-orientation was also more favorable, but the calculated energy differences
were smaller (1.4 kJ/mol for the QM1 model and 8.6 kJ/mol for the QM2 model). As a result, the OH rebound
reaction leading to (R)-1-phenylethanol proceeded only 2-fold slower for the QM1 model and 30-fold slower for the
QM2 model compared with the reactions leading to (S)-1-phenylethanol. Introduction of ZPE corrections (Table 3)
did not change the overall picture.
The interaction energies (see Table S5 in Supporting Information) between activated ethylbenzene in either the
pro(S) or pro(R) TS1 orientations and nearby residues were -63.7 and -48.6 kJ/mol, respectively. The pro(R)
configuration is destabilized by unfavorable vdW interactions with Ile221, Ile217 and His192 and the lack of a favorable
interaction with Trp87. This destabilization is slightly offset by a lack of unfavorable vdW interactions with Ile85 and
Tyr483 and a positive electrostatic interaction with Asn218
.
Analysis of the interaction energies of the pro(S)- or pro(R)-oriented intermediates in TS2 shows that both are of
very similar values (-74 and -73 kJ/mol, respectively) provided that the interaction with His192 is excluded. The
2
interaction energy involving the H atom of His192, which forms a H-bond with 1-phenylethanol, is predicted to be
2
artificially high due to close contact between the H and OH groups. The electrostatic stabilization energies are
also almost identical for both conformations (-39 and -41 kJ/mol for pro(S) and pro(R), respectively). The only
difference between the two binding conformations is an unfavorable steric interaction between the pro(R) reactant
and Trp481 (Fig. 7B), which is almost completely compensated a by favorable interaction with Trp87.
3.4.3 Kinetic isotope effect and isotope labeled studies
The KIE calculated for the pro(S) reaction pathway was 5.4 for TS1 and 0.8 for TS2 (with Wigner tunneling
corrections of 7.45 and 0.8, respectively). The experimental KIE measured at the pH optimum for EBDH (pH 7.5)
was 6.9 [20]. However, the QM:MM model contained a protonated His192, which should represent the situation at
pH values below the optimum and lead to a lower KIE value in the range of 3.6-6, which is consistent with our
previous experimental results [20].
The reaction performed in 37.4% H218O showed an increase in the 124 m/z isotopomer of the product to 36.2%,
while the 122 m/z isotopomer (M+ of 1-phenylethanol) accounted for 63.8%. This observation demonstrates that
EbDH catalyzes the transfer of an 18O atom from a water molecule to the organic product, as shown previously [54].
14

ACCEPTED MANUSCRIPT
Unfortunately, the precision of the available analytic method (standard GC-MS, see Fig. S4) did not allow us to
further match the experimental data to the small KIE expected between 16O and 18O (calculated KIE for TS1 =
1.006 and for TS2 = 1.00 with respect to S, and TS2 = 0.988 with respect to I).
3.4.4. Loss of enantioselectivity due to steric hindrances
The observed loss of enantioselectivity with 1,2-diethylbenzene (14) may be caused by a different effect as
rationalized via QM:MM and MM modeling. Theoretically, all four benzylic H atoms can be activated by the Mo=O
ligand: two pro(S) and two pro(R) hydrogen atoms. For each of these atoms, an approximate TS1 state was
optimized with the QM1:MM model with imposed distance constraints on the critical C-H and oxo-H bonds with
values taken from the TS1 state found for ethylbenzene (Fig. 86A). Two general substrate binding modes were
observed. The first, which involves the pro(S) and pro(R) H atoms of the ethyl substituent marked in blue (Fig. 8A
and 8C, respectively), is characterized by a flat orientation of the other ethyl group marked in green with respect to
the aromatic ring. The blue ethyl group either points toward the MGD-Q pterin for the pro(S)-bound state or toward
the MGD-P pterin for the pro(R)-bound state, which was similarly observed for the TS1 of ethylbenzene. The
second binding mode involves the ethyl substituent marked in green and is characterized by a perpendicular
position of the whole-bound substrate compared with the usual geometry of the reactant and a perpendicular
arrangement of the other ethyl group relative to the ring. The activated (green) ethyl group attains either a
perpendicular conformation for the activation of the pro(S) H atom or a parallel conformation for the activation of
the pro(R) H atom.
The analysis of the interaction energies (see Supporting Information Table S6) with residues surrounding the
substrate in various conformations revealed that the approximate TS1 presented in Fig. 8A and D exhibit similar
interaction energies with energy differences well within calculation error (3.8 kJ/mol). Moreover, for only those two
conformations, the interaction between the amino acid residues and the substrate (i.e., interaction energy excluding
MGD) attains negative energy values (-41 and -45 kJ/mol). Next in order of probability was the pro(S) approximate
TS1 complex depicted in Fig. 8B, with the interaction energy elevated by 170 kJ/mol due to unfavorable steric
interaction with His192. The final pro(R) conformation (Fig. 8C) was impossible due to extremely high energy
resulting from steric clashes with Ile85 and His192
.
These results need to be related to the calculated interaction energy of the TS1 of ethylbenzene in the pro(S)
conformation. The total interaction energy of ethylbenzene in TS1 is lower by app. 20 kJ/mol from both of the 1,2-
diethylbenzene conformations predicted to have reasonably low energy barriers (Fig. 8A and 8C). This result
explains that i) during the hydroxylation of the 14 energy barriers of transitions states for either the pro(S)- or
15

ACCEPTED MANUSCRIPT
pro(R)-bound molecule, should be similarly inflicted with substantial steric hindrances from the neighboring ethyl
group and ii) the increased steric hindrance elevates the energy of C-H activation, resulting in the observed slow
reaction rate. In light of these results, both the limited enantioselectivity and the poor conversion rate of 1,2-
diethylbenzene hydroxylation by EbDH [25] may be explained.
4. Discussion
The chiral chromatography results showed that EbDH exhibits 100% enantioselectivity for most of the investigated
compounds. This enzyme stereoselectively introduces an OH group at the benzylic carbon atom when either ethyl,
propyl, propenyl, propynyl, cyclopentenyl or cyclohexenyl substituents are present. The high enantioselectivity is
not usually compromised by ortho and meta substituents regardless of their nature [27].
For the modeling of the EbDH reaction pathway, it was assumed that the catalytically active species conducting C-
H activation is an oxo ligand of the molybdenum atom. This hypothesis was based on structural and EXAFS data
from DMSO reductase derived by Schindelin et al. [55] and George et al. [56] who showed that oxidized MoCo in
DMSO reductase contains one oxo ligand [19]. However, it should be emphasized that other enzymes in the
DMSO reductase family exhibit different organization of their active sites. The enzyme most related to EbDH with a
solved structure of its oxidized form is nitrate reductase from E. coli; however, this enzyme has been described in
two structurally conflicting forms. Whereas the NarGH structure reported by Jormakka et al. (PDB code 1R27) [18]
contains either an oxo or a hydroxyl ligand to MoVI, the NarGHI structure (PDB code 1Q16) solved by Bertero et al.
[52] contains a bidentate Asp ligand to MoVI instead of an oxo/hydroxyl ligand. In the absence of another known
structure of a closely related enzyme and having only the structure of the reduced form of EbDH available (PDB
code 2IVF), we have based our modeling of the oxidized form on the fact that EbDH likely retains the monodentate
binding mode of Asp223 because it is stabilized by H-bonding with the -amino group of Lys450. In nitrate reductase
1
2
the Lys450 is substituted by His546 which apparently forms a much weaker H-bond with Asp-O due to longer N -
O1 distance (3.35 Å for 1R27 and 3.66 Å for 1Q16 vs. 2.95 Å for EbDH in 2IVF and 2.73 Å in QM:MM S geometry
– Fig. 3). It seems that in case of EbDH the Lys450 is so far removed from the reaction center that a major
conformational contortion severing the H-bond interaction is unlikely. The sixth coordination position of the known
EbDH structure is occupied by an acetate from the crystallization buffer, which was interpreted to occupy the
product binding site [17]. Taking into consideration the fact that EbDH transfers the oxygen from water to an
organic compound and that the process is highly enantioselective, it is highly probable that the acetate (the
respectively bound alcohol product) is replaced by a water molecule, which in turn is deprotonated to form an oxo
(or hydroxyl) ligand. Our recent modeling studies showed that the presence of a Mo=O ligand is essential for
16

ACCEPTED MANUSCRIPT
efficient proton abstraction and substrate activation, even if EbDH actually uses water molecules that are not
ligated to the Mo atom to hydroxylate some substrates in an alternative and less enantioselective reaction mode
[27].
The modeling of the reaction pathway with the QM:MM method in an electronic embedding scheme showed that
enzyme electrostatics support 2-electron oxidation of ethylbenzene. First, the cofactor geometry of the ES complex
is strained in such a way that the S1S2S3S4 dihedral angle (-20o) between the two pterins resembles an optimal
trigonal antiprismatic geometry that is characteristic for Mo(V) hexacoordinate complexes [57, 58]. Second, the
electrostatics of the enzyme strongly stabilizes the carbocation species, supporting the irreversible two electron
oxidation of the substrate, leading to the carbocation intermediate product I. In our previous studies using small
DFT clusters with a PCM model solvent, the intermediate product exhibited lower energy in the radical rather than
carbocation state. Moreover, in a preliminary QM:MM calculation with only mechanical embedding (data not
shown), a similar radical intermediate product was predicted. To exclude the influence of potential overpolarization
of the system by proximities that are too close to the MM boundaries on the chemical nature of I, the single-point
energy for I was also calculated with a larger system (QM2). In this system, the carbocation solution also resulted
in a more stable than radical solution. The approximate stabilization energy was estimated to be 70 kJ/mol
(calculated as the difference between Mo(V) triplet and Mo(IV) singlet solution).
The high enantioselectivity of the EbDH reaction is apparently enforced by a special design of the active site. C-H
activation preferentially proceeds in the pro(S)-orientation of the substrate, resulting in the abstraction of the pro(S)-
hydrogen atom and the subsequent OH rebound from that side, producing an (S)-alcohol. The alternative
orientation, with a pro(R)-hydrogen atom pointing toward the catalytically active Mo=O ligand with the alkyl side
chain in the opposite direction, is associated with steric interactions with the Ile217, Ile221, His192 and Asp223 side
chains (Fig. 6B). These interactions lead to distortion of the TS1 structure and a significantly higher activation
barrier for cleavage of the pro(R) C-H bond.
The energy difference between pro(S) and pro(R) C-H activation is correlated with positioning the ethyl side chain
into a preferential orientation towards the MGD-Q ligand of the MoCo cofactor in the active site of EbDH (Fig. 6A).
In such an orientation, the pro(S) H atom faces the Mo=O ligand, while the ethyl side chain is located in a
hydrophobic pocket formed by Ile85, Trp87, Trp481 and the pterin ring of MGD-Q. As a result, the activation of the C-
H bond is not prevented by steric hindrance. The O, H and C atoms involved in TS1 are positioned in an almost
17

ACCEPTED MANUSCRIPT
straight line in the pro(S) and pro(R) orientations (163o and 169o, respectively). In pro(S)-oriented TS1, the Mo—O,
O—H and H—C distances are 2.036 Å, 1.295 Å and 1.355 Å, respectively, which is similar to those in the
corresponding gas phase model obtained in our previous calculations (2.030 Å, 1.256 Å and 1.326 Å). In contrast,
in pro(R)-oriented TS1, the Mo—O and O—H distances are elongated to 2.076 Å and 1.345 Å. These elongated
bonds result from steric repulsion between the ethyl side chains turned in the direction of the MGD-P ligand of the
MoCo cofactor and its interactions with Asn218, Ile221 and Asp223. As observed in Fig. 6B, the ethyl side chain in the
pro(R) conformation is ‘touching’ the protein surface, even in the elongated, and thus less energetically favorable,
pro(R)-oriented TS1 configuration.
It should be underscored that KIE measurements with EBDH had a significant isotope effect associated with the
rate limiting step (in the range of 3-7, depending on pH) [20]. Although the calculated barrier value is higher for TS2
than TS1 (92 kJ/mol vs. 67.2 kJ/mol), we favor the conversion of the enzyme-bound substrate to TS1 as the more
likely rate limiting step from our KIE calculations. The experimentally obtained KIE values of deuterated C-H bonds
would only be compatible with TS1 as a rate limiting step. The observed KIE should be close to 1 or even reverse if
TS2 was a rate limiting step. The high energy value predicted for TS2 in our QM1/MM model appears to originate
from an overpolarization of the QM-region, which is a consequence of the small size of the QM1 portion. This
exaggerated electrostatic interaction leads to low energies for carbocation intermediate I and TS2; therefore, the
barrier heights connected with the two transition states cannot be readily compared. We have used two
approaches to unravel and alleviate this methodological weakness: i) we extended the QM portion of the model
(from QM1 to QM2) to account for the static screening of the reactant by nearby residues and ii) we used MM
calculations with a solvent model that simulates a screening effect. Both approaches led to similar conclusions, that
is, we showed that using the QM1 model with an electronic embedding scheme introduces artificial stabilization to
all stationary points that are characterized by significant charge separation (TS1, I, TS2). Calculations of single
point energies with the QM2:MM model result in an increase in the energies of TS1 by 36 kJ/mol and I by 103.2
kJ/mol, but the energy of TS2 remained unchanged. Although this result suggests that OH rebound is not
associated with a barrier, it is possible that optimization of I at the QM2:MM level allows for the stabilization of I with
respect to TS2. The lack of TS2 barrier was previously observed in a cluster model DFT simulations of reaction
pathway involving heterolytic C-H activation [20]. After formation of the carbocation I species the geometry
optimization led directly to the enzyme-product complex P.
Similar conclusions were drawn from MM calculations. The introduction of screening effects reduced the
electrostatic stabilization, particularly for I and TS2. The interaction energy for TS2 (-73.3 kJ/mol) was still 20
18

ACCEPTED MANUSCRIPT
kJ/mol more favorable than that in the case of I (-54 kJ/mol), which is in agreement with the results of the QM2/MM
calculations.
The energy differences involved in reaching TS2 are less pronounced between the pro(S) and pro(R) orientations
than those for TS1. This observation is consistent with the different geometric requirements imposed in the TS2
transition state by reaction with the Mo-OH group that allows the reactant to find a more favorable orientation in the
active site. The carbocation intermediate approaches the electron pair in the OH group in such an arrangement that
the Mo-O-C atoms form 122.5 and 126o angles, respectively, distancing the ethyl side chain from the residues of
the enzyme that elevated the energy state of the pro(R) TS1 by steric hindrance (see Fig. S5 of the Supplementary
Information).
In the pro(S) TS2 orientation, the C-O and O-Mo distances are 1.924 and 2.528 Å, respectively, while in the pro(R)
orientation they are 2.023 and 2.505 Å, respectively. Therefore, the pro(R) orientation of the intermediate
apparently does not significantly distort the TS2 geometry. Moreover, the ethyl side chain is located away from the
other amino acids and almost in the same place in both orientations (Fig. 7Error! Reference source not found.).
Increased energy values for TS2 in the pro(R) orientation were only observed for single-point calculations in the
QM2 model, indicating that this is caused by unfavorable electrostatic interactions with Trp87, Trp481 or Phe446
which were introduced into the QM2 model.
,
It appears that the intermediate product (I) is short lived for most EbDH substrates because the enantioselectivity of
the reaction is not compromised by the rather small differences in activation energies for the pro(S) and pro(R) OH
rebound processes. This is particularly true in light of the QM2 energies, suggesting that the OH rebound process
proceeds without an energy barrier. However, the reaction intermediate (radical or carbocation) in some cases may
rotate in the active site if the substrates allow strong radical/carbocation stabilization (and the I energy was
significantly lowered), leading to the increased formation of the (R)-alcohol. This theoretical possibility may explain
the loss in enantioselectivity in the conversion of some substrates, such as 4-ethyl-1,1'-biphenyl (11). However, at
this stage, this hypothesis is speculative and has not been addressed in our modeling studies.
Finally, the observed low enantioselectivity of 1,2-diethylbenzene (14) hydroxylation appears to have another
cause. This compound is a poor EbDH substrate (and a good competitive inhibitor of ethylbenzene hydroxylation
with a low Kic of 0.4 µM [25]) and is converted into a mixture of both enantiomers (60% ee). The substrate structure
principally allows the hydroxylation of either of its two benzylic groups. However, steric interactions with the
enzyme favoring activation of the pro(S) H atom in ethylbenzene compete with steric clashes between the two ethyl
substituents in this compound and active site residues. Structural modeling of the substrate approaching the Mo=O
19

ACCEPTED MANUSCRIPT
ligand (i.e., approximate TS1 structures) obtained in the QM:MM calculations for 14 suggests that the substrate can
be bound in a conformation exposing the pro(S) and pro(R) hydrogen atoms in the benzylic carbon atoms toward
the Mo=O ligand with similar interaction energies (see Fig. 8).
5. Conclusion
EbDH is potentially useful for biotechnological applications because it produces a range of secondary (S)-alcohols
with a high enantioselectivity without a need for using expensive cofactors (such as NAD(P)H). This enzyme can
catalyze the synthesis of at least 19 enantiopure alcohols (with >96% ee) from hydrocarbons. Modeling results
show that the high reaction enantioselectivity is imposed by the structure of the enzyme active site.
6. Table of Abbreviations
APCI
DFT
EbDH
ee
Atmospheric Pressure Chemical Ionization
Density Functional Theory
ethylbenzene dehydrogenase
enantiomeric excess
EE
electronic embedding
KIE
kinetic isotope effect
ME
mechanical embedding
MGD
MoCo
molybdopterin-guanine dinucleotide
molybdenum cofactor
QM:MM quantum mechanics: molecular mechanics
QSERR quantitative structure-enantioselective retention relationship
TS
transition state
ZPE
zero-point energy
Acknowledgements
The authors acknowledge the financial support of priority program 1319 of the German research foundation (DFG),
the excellence program LOEWE/Synmikro from the state of Hessen/Germany, computational grant
MNiSW/SGI3700/PAN/121/2006, the Marian Smoluchowski Krakow Research Consortium - a Leading National
Research Centre KNOW grant, Polish National Science Center under grant N N204 269038, and financial support
20

ACCEPTED MANUSCRIPT
from the ‘Biotransformations for pharmaceutical and cosmetics industry’ project No. POIG.01.03.01-00-158/09-00,
partially financed by the European Union within the European Regional Development Fund.
Supplementary information
Supplementary data contain the list of reagents used for chromatography and synthesis; (Tables S1 and S2), a
figure depicting the flexible portion of the QM:MM model; (Fig. S1), a figure showing the superposition of the
reduced MoCo from an X-ray structure and QM:MM modeling; (Fig. S2), a figure showing a network of H-bonds
stabilizing the cofactor structure; (Fig. S3), a figure showing GC-MS spectra of 1-phenylethanol from 18O labelled
experiment; (Fig. S4), a figure demonstrating the superposition of pro(R) TS1 and pro(R) TS2; (Fig. S5), an
additional details of the computational procedure and development of the normal-phase LC-MS method, details of
identification of the EbDH products (Table S3), energies used in the calculations of KIE (Table S4), and interaction
energies for ethylbenzene (Table S5 and S6) and 1,2-diethylbenzene (Table S7).
References
[1] P.J. Reider, CHIMIA 51 (1997) 306-308.
[2] Y. Gen-Sheng, X. Jiang-Yan, O. Zhi-Min, Y. Shan-Jing, Biotechnol. Lett. 31 (2009) 1879-1883.
[3] C. Suan, M.R. Sarmidi, J. Mol. Catal. B Enzym. 28 (2004) 111-119.
[4] R.L. Hanson, S. Goldberg, A. Goswami, T.P. Tully, R.N. Patel, Adv. Synth. Catal. 347 (2005) 1073-
1080.
[5] M. Kaneoya, M. Uchida, N. Yoshida, K. Miyazawa, 1-Phenyl-1-propanol derivatives Patent
Application: EP0354655 B1, CH DE FR GB IT LI SE, 14-02-1990
[6] J. Hasegawa, H. Nanba, Y. Yasohara, Asymmetric Catalysis on Industrial Scale, Wiley-VCH Verlag
GmbH & Co. KGaA, 2010, pp. 79-109.
[7] M. Breuer, K. Ditrich, T. Habicher, B. Hauer, M. Kesseler, R. Sturmer, T. Zelinski, Angew. Chem. Int.
Ed. Eng. 43 (2004) 788-824.
[8] R.N. Patel, Coordination Chemistry Reviews 252 (2008) 659-701.
[9] S. Pal, D.-H. Park, B.V. Plapp, Chem-Biol. Interact. 178 (2009) 16-23.
[10] K. Uwai, N. Konno, Y. Yasuta, M. Takeshita, Bioorg. Med. Chem. 16 (2008) 1084-1089.
[11] O. Kniemeyer, J. Heider, J. Biol. Chem. 276 (2001) 21381-21386.
[12] R. Hille, Chem. Rev. 96 (1996) 2757-2816.
21

ACCEPTED MANUSCRIPT
[13] H.A. Johnson, D.A. Pelletier, A.M. Spormann, J. Bacteriol. 183 (2001) 4536-4542.
[14] R. Rabus, M. Kube, J. Heider, A. Beck, K. Heitmann, F. Widdel, R. Reinhardt, Arch. Microbiol. 183
(2005) 27-36.
[15] G. Fuchs, M. Boll, J. Heider, Nat. Rev. Microbiol. 9 (2011) 803-816.
[16] J. Heider, Curr. Opin. Chem. Biol. 11 (2007) 188-194.
[17] D.P. Kloer, C. Hagel, J. Heider, G.E. Schulz, Structure 14 (2006) 1377-1388.
[18] M. Jormakka, D. Richardson, B. Byrne, S. Iwata, Structure 12 (2004) 95-104.
[19] G.N. George, J. Hilton, C. Temple, R.C. Prince, K.V. Rajagopalan, J. Am. Chem. Soc. 121 (1999)
1256-1266.
[20] M. Szaleniec, T. Borowski, K. Schuhle, M. Witko, J. Heider, J. Am. Chem. Soc. 132 (2010) 6014-
6024.
[21] M. Szaleniec, A. Salwinski, T. Borowski, J. Heider, M. Witko, Int. J. Quantum. Chem. 112 (2012)
1990-1999.
[22] M. Szaleniec, M. Witko, R. Tadeusiewicz, J. Goclon, J. Comput. Aided. Mol. Des. 20 (2006) 145-
157.
[23] M. Szaleniec, C. Hagel, M. Menke, P. Nowak, M. Witko, J. Heider, Biochemistry 46 (2007) 7637-
7646.
[24] M. Szaleniec, A. Dudzik, M. Pawul, B. Kozik, J. Chromatogr. A 1216 (2009) 6224-6235.
[25] D. Knack, C. Hagel, M. Szaleniec, A. Dudzik, A. Salwinski, J. Heider, Appl. Environ. Microb., vol.
78, 2012, pp. 6475–6482.
[26] M. Szaleniec, Pharmacol. Rep. 64 (2012) 511-531.
[27] A. Dudzik, B. Kozik, M. Tataruch, A. Wójcik, D. Knack, T. Borowski, J. Heider, M. Witko, M.
Szaleniec, Can. J. Chem. 91 (2013) 775-786.
[28] W. Adam, Z. Lukacs, C. Kahle, C.R. Saha-Möller, P. Schreier, J. Mol. Catal. B Enzym. 11 (2001)
377-385.
[29] S. Yadav, R.S.S. Yadav, S. Yadava, K.D.S. Yadav, Catal. Commun. 12 (2011) 781-784.
22

ACCEPTED MANUSCRIPT
[30] H.L. Holland, E.J. Bergen, P.C. Chenchaiah, S.H. Khan, B. Munoz, R.W. Ninniss, D. Richards,
Canad. J. Chem. 65 (1987) 502-507.
[31] A. Uzura, T. Katsuragi, Y. Tani, J. Biosci. Bioeng. 91 (2001) 217-221.
[32] A. Uzura, T. Katsuragi, Y. Tani, J. Biosci. Bioeng. 92 (2001) 381-384.
[33] M. Szaleniec, M. Tataruch, P. Nowak, J. Bryjak, Sposób otrzymywania chiralnych czystych alkoholi
alkiloaromatycznych i alkiloheterocyklicznych oraz układ reaktorowy do realizacji tego sposobu Patent
Application: P.406286Polska, 23.11.2013
[34] J. Chandrasekharan, P.V. Ramachandran, H.C. Brown, J. Org. Chem. 50 (1985) 5446-5448.
[35] P.V. Ramachandran, G.-M. Chen, Z.-H. Lu, H.C. Brown, Tetrahedron Lett. 37 (1996) 3795-3798.
[36] J.S. Yadav, S. Nanda, P.T. Reddy, A.B. Rao, J. Org. Chem. 67 (2002) 3900-3903.
[37] K. Nakamura, T. Matsuda, J. Org. Chem. 63 (1998) 8957-8964.
[38] M. Uchiyama, N. Katoh, R. Mimura, N. Yokota, Y. Shimogaichi, M. Shimazaki, A. Ohta,
Tetrahedron: Asymmetry 8 (1997) 3467-3474.
[39] H. Doucet, T. Ohkuma, K. Murata, T. Yokozawa, M. Kozawa, E. Katayama, A.F. England, T.
Ikariya, R. Noyori, Angew. Chem. Int. Edit. 37 (1998) 1703-1707.
[40] A. Warshel, M. Levitt, J. Mol. Biol. 103 (1976) 227-249.
[41] S. Metz, W. Thiel, Coordin Chem Rev 255 (2011) 1085-1103.
[42] S. Dapprich, I. Komaromi, K.S. Byun, K. Morokuma, M.J. Frisch, J. Mol. Struc-Theochem 461
(1999) 1-21.
[43] D. Rutkowska-Zbik, M. Witko, E.M. Serwicka, Catalysis Today 169 (2011) 10-15.
[44] D. Rutkowska-Zbik, M. Witko, E.M. Serwicka, Catalysis Today 91-2 (2004) 137-141.
[45] D. Rutkowska-Zbik, R. Tokarz-Sobieraj, M. Witko, J. Chem. Theory Comput. 3 (2007) 914-920.
[46] U.C. Singh, P.A. Kollman, J. Comput. Chem. 5 (1984) 129-145.
[47] B.H. Besler, K.M. Merz, P.A. Kollman, J. Comput. Chem. 11 (1990) 431-439.
[48] G.M. Morris, D.S. Goodsell, R.S. Halliday, R. Huey, W.E. Hart, R.K. Belew, A.J. Olson, J. Comput.
Chem. 19 (1998) 1639-1662.
23

ACCEPTED MANUSCRIPT
[49] T. Wang, R.C. Wade, Proteins 50 (2003) 158-169.
[50] D.H. Knack, J.L. Marshall, G.P. Harlow, A. Dudzik, M. Szaleniec, S.Y. Liu, J. Heider, Angew.
Chem. Int. Ed. Eng. 52 (2013) 2599-2601.
[51] E.P. Wigner, Z. Phys. Chem. B19 (1932) 203.
[52] M.G. Bertero, R.A. Rothery, M. Palak, C. Hou, D. Lim, F. Blasco, J.H. Weiner, N.C. Strynadka, Nat.
Struct. Biol. 10 (2003) 681-687.
[53] M. Szaleniec, Ethylbenzene Dehydrogenase - biocatalytic oxidation of alkylaromatic and
alkylheteroaromatic compounds, ICSC PAS, Kraków, 2006.
[54] H.A. Ball, H.A. Johnson, M. Reinhard, A.M. Spormann, J. Bacteriol. 178 (1996) 5755-5761.
[55] H. Schindelin, C. Kisker, J. Hilton, K.V. Rajagopalan, D.C. Rees, Science 272 (1996) 1615-1621.
[56] G.N. George, J. Hilton, K.V. Rajagopalan, J. Am. Chem. Soc. 118 (1996) 1113-1117.
[57] I.K. Dhawan, A. Pacheco, J.H. Enemark, J. Am. Chem. Soc. 116 (1994) 7911-7912.
[58] J.H. Enemark, J.J.A. Cooney, Chem. Rev. 104 (2004) 1175-1200.
24

ACCEPTED MANUSCRIPT
Table 1. The energy differences (E) and ZPE corrected energy differences, E+ZPE) calculated for the
QM1 and QM2 models. The energies were related to S, with exception of TS2 (QM1), for which the energy
was also related to the energy of the intermediate I (in parenthesis).
QM1
QM2
Stationary point
E [kJ/mol] E+ZPE) [kJ/mol] E [kJ/mol] E+ZPE) [kJ/mol]
S
TS1
I
TS2
P
0.0
67.2
-92.0
0.1 (93.1)
-39.3
0.0
48.5
-97.3
0.0
103.1
11.2
-0.3
0.0
84.4
5.9
-4.7
-20.9
-5.7 (91.6)
-31.3
-28.9
25

ACCEPTED MANUSCRIPT
Table 2 Substrates converted by EbDH. S/R [%], the amount of isomers identified based on chiral standards; ee,
the enantiometic excess; QSERR, the identification of a major fraction based on the QSERR model; IPA/n-hexane
–, the composition of the mobile phase in LC chiral chromatography
IPA/n-
No substrate
product
S [%]
R [%]
% ee
QSERR
hexane
Extended alkyl
I
1
2
3
4
ethylbenzene
100
100
0
0
100
100
100
100
1-phenylethanol
10/90
10/90
10/90
10/90
II
propylbenzene
1-phenylpropan-1-ol
III
3-phenylprop-1-ene
3-phenylprop-1-yne
S
1-phenylprop-2-en-1-ol
IV
100
0
1-phenylprop-2-yn-1-ol
Para-substituted
V
5
1-ethyl-4-methylbenzene
1,4-diethylbenzene
100
100
100
100
100
100
52
S
1-(4-methylphenyl)ethanol
5/95
VI
6
100
100
100
100
0
0
0
0
1-(4-ethylphenyl)ethanol
10/90
10/90
10/90
5/95
VII
7
rac-1-(4-ethylphenyl)ethanol
(1S)-1-(4-ethylphenyl)ethanol
1-ethyl-4-fluorobenzene
1-ethyl-4-methoxybenzene
(1S)-1,1'-(benzene-1,4-diyl)diethanol
VIII
8
(1S,1’S)-1,1'-(benzene-1,4-diyl)diethanol
IX
9
1-(4-fluorophenyl)ethanol
X
10
S
1-(4-methoxyphenyl)ethanol
10/90
3/97
XI
11 4-ethyl-1,1'-biphenyl
76
24
1-([1,1'-biphenyl]-4-yl)ethanol
Ortho and meta substituted
XII
12
13
1-ethyl-2-methylbenzene
1-ethyl-3-methylbenzene
100
100
S
1-(2-methylphenyl)ethanol
5/95
XIII
5/95
10/90
100
0
1-(3-methylphenyl)ethanol
grad
97-
90%
XIV
14 1,2-diethylbenzene
60
1-(2-ethylphenyl)ethanol
Bicyclic substrates
XV
15
16
2,3-dihydro-1H-indene
2,3-dihydro-1H-inden-1-ol
100
100
0
0
100
100
10/90
XVIa
5-methoxy-2,3-dihydro-1H-indene
A 5-methoxy-2,3-dihydro-1H-inden-1-ol
b 6-methoxy-2,3-dihydro-1H-inden-1-ol
S
S
10/90
5/95
5/95
5/95
XVII
17
18
19
1,2,3,4-tetrahydronaphtalene
2-ethylnaphthalene
1H-indene
1,2,3,4-tetrahydronaphthalen-1-ol
100
100
XVIII
1-(naphthalen-2-yl)ethanol
XIX
1H-inden-1-ol
100
26

ACCEPTED MANUSCRIPT
heterocyclic
XX
20
21
2-ethylfuran
100
100
0
0
100
100
1-(furan-2-yl)ethanol
10/90
10/90
XXI
2-ethylthiophene
1-(thiophen-2-yl)ethanol
XXII
RP
15/85
22 2-ethyl-1H-pyrrole
1-(1H-pyrrol-2-yl)ethanol
n.d.
RP – RP-LC-MS separation on a Zorbax CN column in a methanol/water system
n.d. – not determined
27

ACCEPTED MANUSCRIPT
Table 3. Calculated activation energies of the transition states involved in ethylbenzene hydroxylation for pro(S)-
and pro(R)-oriented substrate binding at the active site. The parameters shown are the (a) energies of transition
states, (b) differences between pro(R) and pro(S) transition state energies calculated for the respective model, (c)
energies of transition states with ZPE corrections, and (d) differences in transition state energies (pro(R) - pro(S))
with ZPE corrections included. The energies were related to the energy of the pro(S) ES calculated with the
respective QM1 or QM2 model. For TS2, the energies related to carbocation intermediate I calculated for the QM1
model are in parenthesis.
Transition state:
QM1 model
pro(S) TS1
67.2
pro(R) TS1
Pro(S) TS2
0.1 (92.1)
-5.7 (91.6)
Pro(R) TS2
1.5 (93.5)
-2.9 (94.4)
a
b
c
d
E QM1 [kJ/mol]
82.90
E(pro(R)-pro(S)) [kJ/mol]
E+ZPE) QM1 [kJ/mol]
E+ZPE)(pro(R)-pro(S)) [kJ/mol]
15.7
64.5
16.0
1.4
2.7
48.5
QM2 model
a
b
c
d
E QM2 [kJ/mol]]
103.1
84.4
120.3
17.2
102.0
17.6
-0.3
-4.7
8.3
2.5
E(pro(R)-pro(S)) [kJ/mol]
E+ZPE) QM2 [kJ/mol]
8.6
7.2
E+ZPE)(pro(R)-pro(S)) [kJ/mol]
28

ACCEPTED MANUSCRIPT
Figure Legends
Fig. 1 Oxidation of ethylbenzene by EbDH
Fig. 2. QM portions used in QM:MM modeling: A) QM1 was used in geometry optimization and B) QM2 was used
for single point calculations.
Fig. 3. Model of geometry-optimized MoCo in oxidized EbDH. Bond and H-bond lengths given in Å.
Fig. 4. The predicted reaction pathway for the oxidation of ethylbenzene by EbDH. Bond and H-bond lengths are
given in Å. A hydrogen atom cleaved from the substrate is bold-faced.
Fig. 5. The energy differences (E) calculated for: the QM1 (dashed line) and QM2 (dotted line) models. The
energies were related to S.
Fig. 6. Structures of TS1 (C-H activation step). A) pro(S)- and B) pro(R)-oriented reaction pathway. Substrates in
the pro(S) orientation fit into the enzyme active site, while the in pro(R) orientation, steric interactions between the
ethyl side chain of the substrate and the Ile217, Ile221 and His192 of EbDH cause a strong increase in TS1 energy.
Fig. 7 Structures of TS2 (OH rebound). A) pro(S)- and B) pro(R)-oriented reaction pathway. An activated
carbocation fits well into the enzyme active site in both orientations. Steric interactions with Trp481 slightly elongate
the C-O bond, from 1.92 to 2.02 Å in the pro(R)-orientation, which results in moderately higher TS2 energy.
Fig. 8 The geometry of four approximate TS complexes optimized for 1,2-diethylbenzene. The two ethyl
substituents of the substrate are colored in blue and green for clear distinction. Activation of A) the pro(S) H atom
from the blue ethyl group, B) the pro(S) H atom from the green ethyl group, C) the pro(R) H atom from the blue
ethyl group and D) the pro(R) H atom from the green ethyl group.
29