Investigation of the requirements for efficient and selective cytochrome P450 monooxygenase catalysis across different reactions

Article abstract of DOI:10.1016/j.jinorgbio.2019.110913

The cytochrome P450 metalloenzyme (CYP) CYP199A4 from Rhodopseudomonas palustris HaA2 catalyzes the highly efficient oxidation of para-substituted benzoic acids. Here we determined crystal structures of CYP199A4, and the binding and turnover parameters, with different meta-substituted benzoic acids in order to establish which criteria are important for efficient catalysis. When compared to the para isomers, the meta-substituted benzoic acids were less efficiently oxidized. For example, 3-formylbenzoic acid was oxidized with lower activity than the equivalent para isomer and 3-methoxybenzoic acid did not undergo O-demethylation by CYP199A4. The structural data highlighted that the meta-substituted benzoic acids bound in the enzyme active site in a modified position with incomplete loss of the distal water ligand of the heme moiety. However, for both sets of isomers the meta- or para-substituent pointed towards, and was in close proximity, to the heme iron. The absence of oxidation activity with 3-methoxybenzoic acid was assigned to the observation that the C[sbnd]H bonds of this molecule point away from the heme iron. In contrast, in the para isomer they are in an ideal location for abstraction. These findings were confirmed by using the bulkier 3-ethoxybenzoic acid as a substrate which removed the water ligand and reoriented the meta-substituent so that the methylene hydrogens pointed towards the heme, enabling more efficient oxidation. Overall we show relatively small changes in substrate structure and position in the active site can have a dramatic effect on the activity.

Full text of DOI:10.1016/j.jinorgbio.2019.110913

Journal Pre-proof
Investigation of the requirements for efficient and selective
cytochrome P450 monooxygenase catalysis across different
reactions
Matthew N. Podgorski, Tom Coleman, Rebecca R. Chao, James J.
De Voss, John B. Bruning, Stephen G. Bell
PII:
S0162-0134(19)30642-7
DOI:
https://doi.org/10.1016/j.jinorgbio.2019.110913
JIB 110913
Reference:
To appear in:
Journal of Inorganic Biochemistry
Received date:
Revised date:
Accepted date:
25 September 2019
26 October 2019
7 November 2019
Please cite this article as: M.N. Podgorski, T. Coleman, R.R. Chao, et al., Investigation
of the requirements for efficient and selective cytochrome P450 monooxygenase catalysis
across different reactions, Journal of Inorganic Biochemistry (2019), https://doi.org/
10.1016/j.jinorgbio.2019.110913
This is a PDF file of an article that has undergone enhancements after acceptance, such
as the addition of a cover page and metadata, and formatting for readability, but it is
not yet the definitive version of record. This version will undergo additional copyediting,
typesetting and review before it is published in its final form, but we are providing this
version to give early visibility of the article. 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.
© 2019 Published by Elsevier.

Journal Pre-proof
Investigation of the requirements for efficient and selective
cytochrome P450 monooxygenase catalysis across different reactions
Matthew N. Podgorski,[a] Tom Coleman,[a] Rebecca R. Chao,[a] James J. De Voss,[b] John B.
Bruning,[c] and Stephen G. Bell*[a]
[a] Department of Chemistry, University of Adelaide, Adelaide, SA 5005, Australia
E-mail: stephen.bell@adelaide.edu.au
[b] School of Chemistry and Molecular Bioscience, University of Queensland, St Lucia, Qld
4072, Australia
[c] School of Biological Sciences, University of Adelaide, SA 5005, Australia
* Corresponding Author; Stephen G. Bell; E-mail: stephen.bell@adelaide.edu.au; Fax, +61 8
8313 4380
1

Journal Pre-proof
Abstract
The cytochrome P450 metalloenzyme CYP199A4 from Rhodopseudomonas palustris HaA2
catalyzes the highly efficient oxidation of para-substituted benzoic acids. Here we determined
crystal structures of CYP199A4, and the binding and turnover parameters, with different meta-
substituted benzoic acids in order to establish which criteria are important for efficient catalysis.
When compared to the para isomers, the meta-substituted benzoic acids were less efficiently
oxidized. For example, 3-formylbenzoic acid was oxidized with lower activity than the
equivalent para isomer and 3-methoxybenzoic acid did not undergo O-demethylation by
CYP199A4. The structural data highlighted that the meta-substituted benzoic acids bound in the
enzyme active site in a modified position with incomplete loss of the distal water ligand of the
heme moiety. However, for both sets of isomers the meta- or para-substituent pointed towards,
and was in close proximity, to the heme iron. The absence of oxidation activity with 3-
methoxybenzoic acid was assigned to the observation that the C-H bonds of this molecule point
away from the heme iron. In contrast in the para isomer they are in an ideal location for
abstraction. These findings were confirmed by using the bulkier 3-ethoxybenzoic acid as a
substrate which removed the water ligand and reoriented the meta-substituent so that the
methylene hydrogens pointed towards the heme, enabling more efficient oxidation. Overall we
show relatively small changes in substrate structure and position in the active site can have a
dramatic effect on the activity.
Keywords Metalloenzyme; C-H bond abstraction; X-Ray crystal structures; Heme
monooxygenase; Enzyme catalysis; Molecular docking
2

Journal Pre-proof
1. Introduction
The heme-dependent cytochrome P450 (CYP) monooxygenase family of metalloenzymes
catalyzes the oxidation of biological molecules often with high selectivity.[1-3] Moreover, they
can catalyze C-H bond abstractions resulting in carbon hydroxylation at ambient temperatures
using a reactive ferryl Compound I (Cpd I) intermediate.[1, 4-8] The insertion of the oxygen
atom into the C-H bond is hypothesized to occur via a radical rebound mechanism after the Cpd I
intermediate abstracts a hydrogen atom from the alkyl substrate.[5, 9] Importantly, members of
this enzyme superfamily can catalyze a diverse range of other oxidative transformations
including epoxidations, heteroatom dealkylations, sulfoxidations and other more complex
reactions.[10-16]
Investigations into the substrate range of the soluble bacterial P450 CYP199A4 revealed
that the enzyme is predisposed to bind and rapidly oxidize para-substituted benzoic acid
substrates.[17-24] The fastest reaction catalyzed by CYP199A4 is O-demethylation of 4-
methoxybenzoic acid, performed at a rate of 1220 min-1 with a coupling efficiency of 91%.[17,
19] CYP199A4 is also able to bind and oxidize the bulkier substrate 3,4-dimethoxybenzoic acid
(veratric acid). In this instance, demethylation occurs only at the para position, affording 4-
hydroxy-3-methoxybenzoic acid; the meta-substituent is not attacked.[21] CYP199A4 does not
hydroxylate the aromatic ring of any of these substrates and also fails to oxidize benzoic
acid.[19]
Crystal structures of substrate-bound CYP199A4 have been solved to elucidate the
orientation of substrates in the binding pocket and examine the tendency of CYP199A4 to attack
only the para-substituent (Figure 1).[18, 19, 22] These crystal structures revealed that the
benzoate moiety, held almost perpendicular to the heme, forms both hydrophilic and
3

Journal Pre-proof
hydrophobic interactions with residues in the binding pocket.[22] The substrate carboxylate
hydrogen bonds to the hydroxyl groups of S95 and S244 and forms a salt bridge with the
guanidinium of R92.[22] Additionally, it interacts with a molecule of water which is in turn
hydrogen-bonded to S244 and R243.[22]
If the substrate carboxylate is replaced with alternate functional groups, it abolishes the
ability of the enzyme to tightly bind the substrate.[23, 24] Van der Waals interactions exist
between the benzene ring and L98, A248, V181, and F185 and also between the para-substituent
and F182 and F298.[18] Consistent with the ≥95% shift to high-spin observed when 4-
methoxybenzoic acid binds, the crystal structure (PDB: 4DO1) revealed that the iron-bound
water is completely displaced by this substrate. The para-methoxy group is held in close
proximity to the heme iron (the carbon is 4.1 Å from the iron) and thus C-H bond abstraction
occurs solely at this position, resulting in O-demethylation.[22] When veratric acid binds (PDB:
4EGN), its benzoate moiety and para-methoxy group occupy equivalent positions to those of 4-
methoxybenzoic acid (Figure 1).[18] As a result, the para-methoxy group is close to the heme,
but the meta-methoxy group of veratric acid is oriented away from the heme and is too far (8.1
Å) from the iron to react.[18]
The activity of CYP199A4 towards multiple other substituted benzoic acid substrates has
been investigated (Figure S1).[17-21, 23, 24] 2,4-Dimethoxybenzoic acid was also solely
oxidized at the para position.[19] On the other hand, 2- and 3-methoxybenzoic acid were not
oxidized at all.[19] 3-Methoxybenzoic acid binds substantially less tightly than 4-
methoxybenzoic acid (Kd = 69 μM vs 0.22 μM) and only shifts the spin-state to 40% high-spin,
implying that the substrate is not bound in a suitable orientation relative to the heme to
efficiently displace the iron-bound water.[19] This is consistent with previous studies which
4

Journal Pre-proof
found that other meta-substituted benzoic acids bind weakly to CYP199A2, which shares 86%
sequence identity with CYP199A4.[22] For example, 3-chlorobenzoic acid was found to induce
a substantially smaller spin-state shift than the para isomer in CYP199A2 (10% vs 80%).[26] In
the absence of a CYP199A4 crystal structure with a substrate containing only a meta-
arrangement of substituents, it was proposed that 3-methoxybenzoic acid could bind to the active
site in a similar orientation to that of veratric acid (Figure 1). This would position the meta-
substituent to point away from the heme iron, rather than towards it.[19] If this was the case the
meta-substituent would be held too far from the iron to react. Of note is that 3,5-
dimethoxybenzoic acid, which must bind with one meta-methoxy group pointing away from the
heme and one methoxy group pointing towards it, is oxidized with low efficiency (at a product
formation rate of 7 μM (μM-P450)-1 min-1).[19]
CYP199A4 can catalyze reactions other than O-dealkylation such as the sulfoxidation of
4-methythiobenzoic acid, N-dealkylation of 4-methylaminobenzoic acid and the hydroxylation of
4-methylbenzoic (4-toluic) acid.[17, 18] It also oxidizes bulkier substrates such as 4-ethoxy-, and
4-isopropyl-benzoic acid.[17, 18] Crystal structures have been obtained of CYP199A4 in
complex with several of these substrates (PDB codes: 4-methylthio-, 5KT1; 4-methylamino-,
5U6W and 4-ethoxy-benzoic acid, 5U6T).
In contrast to 3-methoxybenzoic acid, both 3-methylamino- and 3-methylthio-benzoic
acids were oxidized by CYP199A4[17]; however, the activities were significantly lower than
those of the para isomers. We therefore set out to more fully investigate the ability of
CYP199A4 to bind and oxidize benzoic acid substrates with alternative substituents at the meta-
position. Docking and structural studies were performed to assess whether 3-substituted benzoic
acids bind with the meta-substituent pointing away from the heme. The substrates were chosen to
5

Journal Pre-proof
enable exploration of a range of different substituents which would alter the steric bulk and
reactivity at this position (Figure 2).
The results were compared with those of the para-substituted isomers and used to clarify
the binding mode of these substrates in order to rationalize the preference of CYP199A4 for
para- compared to meta-substituted benzoic acids. This provided a greater insight into how the
coordination environment of the heme iron and the substrate orientation relative to this cofactor
affects the catalytic activity in this important and ubiquitous superfamily of heme
metalloenzymes.
6

Journal Pre-proof
2. Experimental methods
2.1 General methods
Production and purification of CYP199A4, crystal growth and structure determination, substrate
binding analysis, in vitro NADH activity assays, chromatographic analysis of metabolites,
hydrogen peroxide determination and catalytic activity as well as the use of a whole-cell
oxidation system to generate metabolites for characterization were all performed using methods
which have been reported previously (full details also provided in the Supplementary
information).[17, 19, 21-23]
2.2 Docking of ligands into CYP199A4
To identify the likely binding mode of the substrates in the binding pocket of CYP199A4, the
ligands were docked into a crystal structure of CYP199A4 (PDB ID: 5U6W) using ICM-Pro
software, version 3.8-6a (Molsoft LLC, San Diego CA).[27] First, hydrogens were added to the
crystal structure and optimized. Charges were also automatically assigned to residues which
should be protonated/deprotonated at neutral pH and the orientation of histidine, proline,
asparagine, glutamine and cysteine residues was also optimized.[27, 28] Waters were deleted
from the structure with the exception of W618 which forms a hydrogen bond with the
carboxylate of bound benzoic acid substrates, as well as active site residues (Figure 1).[17, 18,
22] The co-crystal ligand (4-methylaminobenzoic acid) was then removed from the active site of
the enzyme and this region was used to define the binding site. The 4-methylaminobenzoic acid
ligand was modified to 3-methylaminobenzoic acid and this ligand was re-docked into the
binding pocket. The top scoring (lowest energy) pose had the 3-methylamino substituent
pointing down towards the heme. A positional restraint was imposed on the nitrogen atom in
7

Journal Pre-proof
order to dock the ligand in an alternative configuration with the 3-methylamino moiety pointing
away from the heme. The docking ‘score’ was then used to evaluate the docked poses, with a
value of < -32 indicating a good fit.[27, 28] The other ligands were docked into CYP199A4
using the same method.[29]
8

Journal Pre-proof
3. Results
3.1 Substrate binding and catalytic activity of CYP199A4 towards meta-substituted benzoic
acids
The ability of CYP199A4 to bind and oxidize 3-methyl- and 3-formyl-benzoic acid was
explored. Compared to 3-methoxybenzoic acid, which induces a 40% spin-state shift in the heme
iron to the high-spin form upon binding to CYP199A4 (Figure S2),[19] both 3-methyl- and 3-
formyl-benzoic acid induced smaller type I shifts: 30% and 10%, respectively (Table 1, Figure
S2, Table S2). These spin-state shifts are also lower than those reported for the corresponding
para isomers and similar to those previously reported for 3-methylthio- and 3-methylamino-
benzoic acid.[17] 4-Methylamino-, 4-methylthio- and 4-methyl-benzoic acid all induce 70%
shifts to high-spin,[17] consistent with the para-substituent being held in close proximity to the
heme and resulting in loss of the iron-bound water. By way of contrast, 4-formylbenzoic acid
induced a smaller spin-state shift (25%), though this is larger than that induced by 3-
formylbenzoic acid (Table 1, Figure S2, Table S2).
Based on the weak binding reported for 3-methoxy-, 3-methylamino- and 3-methylthio-
benzoic acid, the substrates investigated here were predicted to bind weakly to CYP199A4.[17]
3-Methylbenzoic acid binds more than 100-fold less tightly than 4-methylbenzoic acid, and 3-
methylamino- and 3-methylthio-benzoic acids (Kd = 89 μM, compared to 31 and 33 μM,
respectively, Table 1, Figure S3). In line with its lower spin-state shift, 4-formylbenzoic acid
bound with a lower affinity than the other para-substituted benzoic acids. 3-Formylbenzoic acid
binds even more weakly than the other meta-substituted benzoic acids studied (Kd = 416 μM,
Figure S3) and ~9-fold more less tightly than the corresponding para isomer (Table 1). This data
indicates that all the meta-substituted benzoic acids fit less well in the CYP199A4 active site
9

Journal Pre-proof
compared to the corresponding para isomers. It has been hypothesized that this weaker binding
is due to lower complementarity between the enzyme active site and the meta-substituted
benzoic acids.[19]
The smaller spin-state shifts induced by the meta-substituted benzoic acids implied that the
substrate is not bound close enough to the heme to efficiently displace the iron-bound water.
Both 3-formyl- and 3-methylamino-benzoic acid slightly red shifted the Soret maximum (by 0.5
nm) despite giving type I difference spectra, implying that the heteroatoms of these substituents
may interact with a heme-bound water (Supporting information, Figure S2, Table S2). The spin
state experiments also suggest that NADH oxidation by CYP199A4 in the presence of these
substrates will be slow because electron transfer is gated by displacement of the water and the
related changes in spin state and redox potential.[17, 19, 30-33] If these substrates were to bind
exclusively with the meta-substituent on the side of the benzene ring which is further away from
the heme, we would predict little or no product to be formed. Therefore, in vitro NADH activity
assays were performed to evaluate whether these substrates are metabolized by CYP199A4
(Figure S4). The rates of NADH oxidation and product formation by CYP199A4 are
substantially lower with the meta-substituted benzoic acids than the equivalent para isomers
(Table 1). Whereas 3-methoxybenzoic acid is not oxidized by CYP199A4, HPLC and GC-MS
analysis revealed the presence of oxidation metabolites arising from CYP199A4-catalyzed
oxidation of all the other meta-substituted substrates. We have previously reported some of these
including 3-aminobenzoic acid and 3-methylsulfinylbenzoic acid, which arise from N-
dealkylation and sulfoxidation of the nitrogen- and sulfur-containing substrates, respectively.[17]
In vitro turnovers with 3-methylbenzoic acid oxidized NADH at a rate of 78 μM
(μM-P450)-1 min-1 (henceforth abbreviated to min-1), compared to a rate of 444 min-1 with the
10

Journal Pre-proof
para isomer (Table 1). CYP199A4 has previously been reported to convert the para isomer
exclusively into 4-hydroxymethylbenzoic acid.[18] CYP199A4 similarly converted 3-
methylbenzoic acid into a single product (Figure 3). The mass of the TMS-derivatized product
(m/z = 296.10) is consistent with the calculated mass of doubly derivatized 3-
hydroxymethylbenzoic acid (m/z = 296.1264). The product could also conceivably be 4-hydroxy-
3-methylbenzoic acid, arising from hydroxylation of the benzene ring. However, the retention
time of authentic 4-hydroxy-3-methylbenzoic acid was different to that of the P450 product,
discounting the possibility that the substrate was hydroxylated at the para aromatic C-H bond of
the benzoic acid moiety (Figure S5 and S6). To verify that the product was 3-
hydroxymethylbenzoic acid, this compound was chemically synthesized via reduction of 3-
formylbenzoic acid with NaBH4. The CYP199A4 oxidation product and synthesized 3-
hydroxymethylbenzoic acid co-eluted via HPLC (Figure S6). The coupling efficiency for this
reaction was 48%, and the product formation rate 38 min-1. CYP199A4 therefore oxidizes 3-
methylbenzoic acid at a ~10-fold slower rate than it hydroxylates 4-methylbenzoic acid.
The CYP199A4 system oxidized NADH in the presence of 3-formylbenzoic acid at a rate
(10.6 min-1) roughly equal to the leak rate (the NADH oxidation rate in the absence of substrate;
~9 min-1). However, low levels of the carboxylic acid product (isophthalic acid) were detected
(Figure S7). The coupling efficiency and product formation rate were only 8% and 0.9 min-1,
respectively. CYP199A4 converts 4-formylbenzoic acid into terephthalic acid at a rate of 110
min-1 with a coupling efficiency of ~83% (Figure 4). This demonstrates that CYP199A4 is
capable of catalyzing efficient aldehyde oxidation. Control reactions with these aldehyde-
containing substrates and the others showed that no product was detected when the P450 was
omitted from the reaction mixture (Figure 4 and S7).
11

Journal Pre-proof
Finally, we assessed the binding and turnover of the ortho-substituted substrate 2-
methylbenzoic acid. CYP199A4 displayed weak affinity for 2-methylbenzoic acid (Kd = 611
μM) and only a small spin-state shift (~5%) was observed (Table 1, Figure S8). In line with the
small spin-state shift induced by this substrate, the NADH oxidation rate was only 13 min-1,
barely faster than the leak rate (Table 1). No product was detected by GC-MS or HPLC analysis
of the turnover mixture (Figure S9 and S10). This result reaffirms that CYP199A4 does not
oxidize ortho-substituted benzoic acids.[19]
Minimal levels of H2O2 were detected in all CYP199A4-catalyzed oxidations. However, it
is important to note here that when the rate of NADH oxidation is slow, as it is with the meta-
benzoic acid isomers, H2O2 generated via uncoupling may be lost before the reaction is
complete. Indeed, in control experiments we found that when ~300 μM H2O2 was incubated with
a mixture containing HaPux, HaPuR and CYP199A4 in Tris-HCl buffer (but no NADH,
substrate or catalase) or with just the HaPux [2Fe-2S] ferredoxin, almost all H2O2 was lost over a
60-minute period (Supporting information, Figure S11). Therefore, this method is unlikely to
provide an accurate measurement of the amount of H2O2 formed at the end of a slow reaction
(>15 min) and the values obtained are likely to be an underestimate of the actual amount
generated (Figure S11). This is an important observation when considering the quantitation of
the contribution of different uncoupling pathways in P450 reactions.
3.2 Docking of substrates into the active site of CYP199A4
The ability of CYP199A4 to oxidize all the meta-substituted benzoic acids with the exception of
3-methoxybenzoic acid led us to speculate that the orientation of these substrates in the binding
pocket of CYP199A4 may differ. We therefore decided to dock the meta-substituted benzoic
acids into the active site of CYP199A4 to predict their likely binding mode. ICM-Pro software
12

Journal Pre-proof
was used to dock all the substrates (except 3-formylbenzoic acid due to the lower binding
affinity of the aldehyde isomers) into the CYP199A4 structure (PDB: 5U6W).[27]
In all instances the top-scoring pose had the meta-moiety pointing towards the heme
(Figure 5, Supporting information and Table S3). For all the substrates, except 3-
methylthiobenzoic acid, an alternative (lower-scoring) pose having this substituent pointing
away from the heme was found. For 3-methylthiobenzoic acid the next lowest scoring pose also
had the meta-substituent pointing towards the heme.
In contrast, when veratric acid was docked into the active site, the top-scoring poses had
the meta-substituent pointing away from the heme, consistent with the crystal structure (Figure
S12). Docking of 2-methoxybenzoic acid was conducted as an example of an ortho-substituted
substrate and revealed that it preferred to bind with the ortho-methoxy group pointing away from
the heme (Figure S12).
For 3-methylaminobenzoic acid the methyl carbon was only 4.1 Å from the heme iron and
the sulfur of 3-methylthiobenzoic acid and methyl group of 3-methylbenzoic acid were at
distances of 4.2 Å and 4.0 Å, respectively (Figure 5 and Table S3). This could rationalize why
these substrates were oxidized at these positions. However, the top-scoring pose of 3-
methoxybenzoic acid, which is not oxidized by CYP199A4, also had the meta-substituent
pointing towards the heme, with the methyl group 4.0 Å from the iron. Another important
observation from these docking experiments was that the position of the benzene moiety of the
meta-substituted substrates was slightly different to that observed in the X-ray crystal structures
of the para-substituted benzoic acids (Figure S13). To more satisfactorily answer the question of
how these different meta-substituted benzoic acids bind, we decided to solve crystal structures of
these substrates bound to CYP199A4.
13

Journal Pre-proof
3.3 X-ray crystal structures of CYP199A4 with meta-substituted benzoic acids
CYP199A4 was co-crystallized with 3-methoxy-, 3-methylthio-, 3-methylamino-, 3-methyl- and
4-methyl-benzoic acid (Figure S14) and X-ray diffraction data were collected at the MX1 or
MX2 beamlines at the Australian Synchrotron at 100 K.[34, 35] Crystal structures of these
protein-ligand complexes were solved at resolutions of 1.60-1.89 Å (PDB codes 6PQ6, 6PQD,
6PRR, 6PQW and 6PQS). Data collection and refinement statistics are presented in Table S4.
The RMSD between Cα atoms of the superimposed structures is ≤0.178 Å, indicating that the
overall protein fold is virtually identical for all five complexes. These crystal structures also had
a similar fold to previously reported structures of CYP199A4 (Figure S15). Clear electron
density corresponding to the substrate permitted modelling of each ligand into each structure
(Figure 6). These substrates all bind with the meta-substituent pointing towards the heme in
agreement with the docking studies. Overall this rationalizes why, with the exception of 3-
methoxybenzoic acid, oxidation occurred at this position.
The crystal structures of the meta-substituted benzoic acids also demonstrated that the
water ligand was maintained with a degree of occupancy at the sixth coordination position (21-
90%; Table 1, Figure 6). The crystal structures also provide evidence that the water ligand and
the N, S, or O atoms of the meta-substituent of 3-methylamino-, 3-methylthio- and 3-methoxy-
benzoic acid, were interacting with this water molecule. The small red shift observed in the Soret
maximum with 3-methylaminobenzoic acid may result from such an interaction. 3-
Methylbenzoic acid, on the other hand, which has a more hydrophobic substituent, largely
displaced the water in the 3-methylbenzoate-CYP199A4 crystal structure (water occupancy =
21%).
14

Journal Pre-proof
The occupancy of the water ligand in the other meta-substituted benzoic acid crystal
structures closely correlated with the measured fraction of high/low-spin heme (Table 1). For
example, the 3-methylaminobenzoic acid-CYP199A4 complex is 90% low-spin and the
occupancy of the water ligand in the crystal structure is also 90%. The 3-methylthiobenzoic acid-
CYP199A4 complex is 70% low-spin and the refined water occupancy is 77%, while the 3-
methoxybenzoic acid-CYP199A4 complex is 60% low-spin and the water ligand occupancy is
50%. 4-Methylbenzoic acid, on the other hand, induces a larger spin-state shift to the high-spin
form (70%) and a water ligand is not observed in the crystal structure. The para-substituted
benzoic acids displace the water molecule and induce larger shifts to the high-spin ferric
state.[17] It should be noted that partial spin-state shifts and water occupancy in the crystal
structure have also been reported with different substrates and P450cam (CYP101A1).[36, 37]
To ascertain why we observe such an acute difference in the binding and activity of meta-
versus para-substituted benzoic acids, we next compared the structures of the different substrate-
bound analogues. When the structures of the meta-substituted benzoate-bound CYP199A4 were
superimposed with the equivalent para substrate-bound structures, only subtle differences in the
positions of active-site residues were observed (Figures 7 and 8). However, the benzene ring of
the meta isomers are all shifted up away from the heme relative to same moiety of the equivalent
para-substituted benzoic acids. This was in agreement with what was observed in the docking
experiments and the positions of 3-methoxy-, 3-methylthio-, 3-methylamino-, and 3-methyl-
benzoic acid in the crystal structure agreed with the top-scoring docked poses. When the position
of the substrates was compared in the X-ray and the modelled structures all the atoms, including
those of the substituent, were found to align (Figure S16). This suggests that energy
minimization modelling could be a valuable tool in understanding the binding of different
15

Journal Pre-proof
benzoic acids and related substrates in this family of P450 enzymes. This movement of the
substrate, when the position of substitution changes, seems to occur to avoid a steric clash
between the meta-substituent and the heme (Figure S17).
A common feature of all of the structures with meta-substituted isomers is that there is
more space directly above the heme center as the substituent is more offset from the area directly
above the iron (Figures 7 and 8). This is consistent with the partial occupancy of the 6th ligand in
these structures compared to the para-substituted benzoic acid counterparts, in which this water
was not present. Based on the productive catalytic activity and NADH oxidation rates observed
(Table 1) it appears that this water molecule can be displaced enabling the enzyme to bind
dioxygen and enabling turnover of the catalytic cycle albeit at a lower rate than for the equivalent
para isomer.
While the heme iron in 4-methylthiobenzoate-bound CYP199A4 is five-coordinate, when
3-methylthiobenzoate is bound a high-occupancy water ligand is present (Fe–O distance 2.3 Å;
Table S5). The sulfur of 3-methylthiobenzoic acid is 3.3 Å from this water. The methylthio
moiety of 3-methylthiobenzoic acid is rotated out of the plane of the benzene ring by 13˚ and that
of 4-methylthiobenzoic acid is rotated by 34˚. In both structures, the methylthio moiety is held
close to the heme, though the sulfur of 3-methylthiobenzoic acid is held nearer to the iron than
that of 4-methylthiobenzoic acid (4.3 vs 5.0 Å; Table S5). The methyl group of 3- and 4-
methylthiobenzoic acid is also held in close proximity to the heme iron (4.4 vs 4.5 Å). When
comparing the structures, we wanted to ascertain how close the substrate would be to the oxygen
of the reactive Cpd I intermediate (as well as the heme iron). It has been calculated that the Fe=O
bond length of Cpd I is 1.62 Å.[38, 39] This is in good agreement with the measured Cpd I Fe=O
bond length using EXAFS, which was found to be 1.67 Å.[40] Therefore, to estimate the likely
16

Journal Pre-proof
distances between the ferryl oxygen of Cpd I and the relevant atoms of the bound substrate, an
oxygen atom was modelled 1.62 Å above the heme iron directly opposite the Fe-S bond and we
assumed that the position of the substrate is not significantly changed after oxygen binding and
activation (Table S5). The measured distance between the Cpd I oxygen and the sulfur of 3-
methylthiobenzoic acid is only 3.2 Å. This is slightly shorter than the distance between the Cpd I
oxygen and the sulfur of 4-methylthiobenzoic acid (3.4 Å). For the meta-methylthio isomer the
FeIV=O–S angle is 126.2˚, whereas for the para isomer the angle is larger (162.5˚).
In the crystal structure of 3-methylaminobenzoate-bound CYP199A4, a water ligand to the
heme is present at 90% occupancy (Fe–O distance 2.2 Å; Table S5), and this water is hydrogen-
bonded to the substrate nitrogen (3.1 Å). The nitrogen and methyl group of 3-
methylaminobenzoic acid are 4.2 and 4.4 Å from the heme iron, respectively. The distance
between the Cpd I oxygen and methyl carbon of 3-methylaminobenzoic acid is 3.1 Å, which is
similar to that of the methyl carbon of 4-methylaminobenzoic acid (3.0 Å). The FeIV=O–CMe
angle was similar for both the meta and para isomers (129.9 vs 124.6˚). The nitrogen heteroatom
of the para isomer had a larger FeIV=O–N angle than 3-methylaminobenzoic acid (148.6 vs
126.8˚; Table S5).
In both the structures of 3-methoxybenzoic acid- and 4-methoxybenzoic acid-bound
CYP199A4, the methoxy groups are held above the heme. The O and CMe of 3-methoxybenzoic
acid are equidistant from the heme iron (4.4 Å), and the methoxy group is rotated 20˚ out of the
plane of the benzene ring. In contrast, the methoxy group of 4-methoxybenzoic acid lies almost
in the plane of the benzene ring, and the CMe is closer to the iron (4.1 Å) with the oxygen being
further away (5.2 Å). When the ferryl oxygen of Cpd I was considered, the CMe and O of 3-
methoxybenzoic acid are both 3.1 Å from the Cpd I oxygen, whereas the CMe and O of 4-
17

Journal Pre-proof
methoxybenzoic acid are 2.7 and 3.6 Å from the ferryl oxygen. The FeIV=O–CMe angle is similar
for both 3-methoxy- and 4-methoxy-benzoic acid (136.3 vs 140.7˚).
The methyl group of both 3-methyl- and 4-methyl-benzoic acid is held close to the heme
iron (4.4 vs 4.5 Å). A low-occupancy water ligand is present in the 3-methylbenzoate-
CYP199A4 structure with a Fe–O distance of 2.5 Å, but no water ligand was detected in the 4-
methylbenzoic acid structure. The benzene ring of 3-methylbenzoic acid is shifted away from the
heme relative to that of 4-methylbenzoic acid, though the carboxylate groups are held in similar
positions. The methyl of 3-methylbenzoic acid is held slightly further from the ferryl oxygen
than that of 4-methylbenzoic acid (3.2 vs 3.0 Å). The FeIV=O–CMe angle is larger for 4-
methylbenzoic acid than for 3-methylbenzoic acid (154.4 vs 130.4˚).
3.4 Substrate binding and catalytic activity of bulkier meta-substituted benzoic acids
The crystal structures of 3-methoxy-, 3-methylamino-, and 3-methylthio-benzoate-bound
CYP199A4 revealed that although the meta-substituent was positioned close to the heme, the
water ligand was not fully displaced. We hypothesized that substrates with bulkier meta-
substituents would more efficiently displace the water ligand and that CYP199A4 would display
higher oxidation activity towards them. We selected 3-ethoxybenzoic acid and 3-tert-
butylbenzoic acid and docked these into the active site of CYP199A4 (Figure S18). The top-
scoring poses have the meta-substituent positioned on the heme side of the substrate (Table S6).
However, docking 3-tert-butylbenzoic acid into CYP199A4 resulted in a steric clash between the
tert-butyl group and a heme nitrogen. This implies that this substrate may be too bulky to fit into
the active site.
18

Journal Pre-proof
We subsequently measured the spin-state shifts induced by these substrates. The minimal
spin-state shift (<5%) induced by 3-tert-butylbenzoic acid implies that the bulky tert-butyl group
hinders binding (Table 2, Figure S19). The para isomer, 4-tert-butylbenzoic acid, bound to
CYP199A4 with a 90% shift to the high spin form, but had lower binding affinity (39 ± 2 M,
Figure S20) than the other para-substituted benzoic acid substrates, such as 4-methoxybenzoic
acid (Table 1). In contrast to the other meta-substituted benzoic acids we tested, 3-ethoxybenzoic
acid induced an almost complete (85%) shift of the ferric iron to the HS state, indicating efficient
removal of the 6th water ligand (Figure S19). This spin-state shift is larger than that induced by
certain para-substituted benzoic acids (Table 1). However, 3-ethoxybenzoic acid was found to
bind substantially less tightly than the para isomer (Kd = 82 ± 2 vs 0.2 ± 0.1 μM, Figure S21).
There was no increase in the NADH oxidation rate above the leak rate (as defined in
Table 1) in the presence of 3-tert-butylbenzoic acid. No product was detected by HPLC analysis
of the reaction mixture (Figure S22). These observations are evidence that 3-tert-butylbenzoic
acid does not bind to CYP199A4. The corresponding para isomer is hydroxylated by CYP199A4
at a rate of 227 min-1 with 100% coupling efficiency and chromatographic analysis indicated it
was oxidized to a single metabolite (Table 2, Figure S23 and S24). The GC-MS of the TMS-
derivatized turnover confirmed the formation of a single alcohol product (m/z = 338.25 vs
expected 338.17, Figure S24). The product was prepared using a whole-cell oxidation
system[21] and, after conversion to the methyl ester, purified and characterized by NMR as 4-
(1,1-dimethyl-2-hydroxyethyl)benzoic acid (Supporting information, Figure S25). This arose
from methyl hydroxylation due to the absence of any benzylic hydrogens within the substrate.
On the other hand, the CYP199A4 system oxidized NADH upon addition of 3-
ethoxybenzoic acid at a moderately fast rate (208 min-1). Whereas CYP199A4 is unable to
19

Journal Pre-proof
oxidize 3-methoxybenzoic acid, the enzyme did catalyze the O-deethylation of 3-ethoxybenzoic
acid. The coupling efficiency and product formation rate were calculated to be 66% and 138 min-
1, and only 7% of the NADH consumed was measured as being channeled into H2O2 production
(Table 2). A single product, 3-hydroxybenzoic acid, was detected, indicating that no
hydroxylation occurred at the less reactive terminal CH3 group (Figure 9). CYP199A4 converts
the para isomer, 4-ethoxybenzoic acid, into 4-hydroxybenzoic acid at a rate of 527 min-1 and
with a coupling efficiency of 100%.[17] CYP199A4 therefore still exhibits substantially higher
activity towards 4-ethoxybenzoic acid than 3-ethoxybenzoic acid, which again demonstrates the
generality of the preference of this enzyme for para- over meta-substituted benzoic acids.
3.5 Crystal structure of 3-ethoxybenzoate-bound CYP199A4
We obtained a crystal structure of 3-ethoxybenzoate-bound CYP199A4 which was solved at 2.37
Å resolution (PDB code 6PRS, Figure 10, Table S7, Figure S26). It confirms that the meta-
substituent points down towards the heme and that the 6th iron water ligand is displaced, in
agreement with the 85% shift to high spin and the docked structure (Figure S27). There is a
change in the (CH2-O-C3-C4, Table S8) dihedral angle of the larger ethoxy substituent compared
to that of the methoxy substituent which results in the methylene group being held more over the
heme iron than the methyl group of the methoxy substituent (Figure 10, Table S8).
Both the methylene (CH2) and methyl carbons (CH3) of the meta-ethoxy substituent are
held in close proximity to the heme iron (3.6 and 4.2 Å, respectively, Table S8). When the
distances were measured between the substrate and the hypothetical position of the Cpd I
oxygen, the methylene carbon (CH2) is also held closer than CH3 (2.1 vs 2.7 Å). The
FeIV=O−CH2 and FeIV=O–CH3 angles are 155.3 and 147.7˚, respectively. C-H bond abstraction
20

Journal Pre-proof
occurs solely at CH2 by CYP199A4, with no hydroxylation being observed at the methyl group
(Table S8). This can be rationalized by CH2 being the closest carbon to the iron and the weaker
strength of the secondary methylene C-H bonds (especially those adjacent to oxygen)[1]
compared to the primary C-H bonds of the terminal methyl group. Overall this structure
highlights that certain meta-substituted benzoic acid substrates can efficiently bind to the active
site and thus remove the 6th water ligand of CYP199A4 and that this enables more efficient
catalytic oxidation.
21

Journal Pre-proof
4. Discussion
The cytochrome P450 family of enzymes catalyze many reactions which are important in
xenobiotic catabolism and secondary metabolite formation. The majority of cytochrome P450s
are thought to follow the consensus catalytic cycle observed with P450cam, as does CYP199A4.
The reactive Cpd I intermediate has been calculated to have similar reactivity across different
forms. We hypothesize that the observations reported here, with CYP199A4, would be
applicable to many other P450 enzymes. However, it is important to note that an increasing
number of P450 enzymes which catalyze varied reactions are being discovered, highlighting the
complex roles they play in metabolism and their potential for mechanistic and function
diversification.
CYP199A4 displays a strong preference for benzoic acid substrates with a para-
substituent. Despite having no activity towards 3-methoxybenzoic acid and 3-tert-butylbenzoic
acid, this enzyme nevertheless is capable of oxidizing certain meta-substituted benzoic acids.
The binding mode of meta-substituted benzoic acids revealed by the crystal structures has the
meta-substituent positioned close to the heme, not directed away as previously observed in the
disubstituted veratric acid structure. All the meta-substituted benzoic acid substrates studied here
bind with the benzene ring and carboxylate held in virtually identical positions (Figure S28).
While the distance of the meta-substituent to the heme iron/Cpd I model oxygen is often
comparable to that of the equivalent position in the para-substituted species, the altered position
of the benzene ring and change in substitution pattern results in there being more space directly
above the heme iron in the former set of substrates.
Within each pair of meta and para isomers, the extent of the spin-state shift correlated with
the NADH oxidation rate, suggesting that this step is somewhat gated by displacement of the
22

Journal Pre-proof
water. The observation of partial occupancy of the 6th heme iron water ligand in the crystal
structures of the meta-substituted benzoic acid substrates would result in a lower spin-state shift
to the high-spin form and slower turnover of the catalytic cycle which is consistent with the first
electron reduction being the rate determining step in the catalytic cycle. This is consistent with
previous observations with this enzyme and other related class I bacterial systems such as
P450cam. However, the extent of the spin-state shift does not always directly correlate with
NADH oxidation rate or product formation of CYP199A4 across all these substrates. Another
observation from this work is that hydrogen peroxide is consumed by the P450 system (by the
[2Fe-2S] ferredoxin) which can make its quantification in slow reactions difficult or unfeasible.
The binding orientations of 3-methylamino-, 3-methylthio- and 3-methyl-benzoic acid
enable one to explain why these compounds can be oxidized by CYP199A4. Despite their lower
activities, the coupling efficiency of these oxidation reactions were comparable to the equivalent
para isomers, in line with their similar distances and orientations relative to the heme iron. The
crystal structures and molecular docking experiments with the meta-substituted benzoic acids
provide excellent insight into why meta-substituted and larger substrates are not as good
substrates for the CYP199A4 enzyme. These would be beneficial in the future design of variant
forms of CYP199A4 to improve the oxidation activity of these and similar substrates. If more
space was generated at the top of the active site by mutating the hydrophobic residues in this
region (F182, F185 and V181, Figure 7 and 8) the meta isomer could bind in an orientation
which more efficiently displaces the 6th water ligand and improve the activity. Such changes may
also enhance the binding and activity of larger substrates such as substituted naphthoic acids.
However, such amino acid substitutions could result in the substrate binding with the meta-
23

Journal Pre-proof
substituent pointing into the space generated, which as with veratric acid, would place them
further away from the heme.
The results with the methylthio and methylamino analogues are perhaps unsurprising given
their higher reactivity. The sulfur of 3-methylthiobenzoic acid is in range for oxidation by the
ferryl Cpd I reactive intermediate. For sulfur-containing substrates, Cpd I reacts with the
heteroatom via direct oxygen transfer, or via single electron transfer followed by oxygen rebound
(we note that other intermediates of the P450 catalytic cycle have been proposed to catalyze
sulfoxidations).[8, 12] The observation that 3-methylaminobenzoic acid is oxidized by
CYP199A4 but 3-methoxybenzoic acid is not requires more detailed analysis. These substrates
have virtually identical orientations in the two crystal structures and the distances of the methyl
group and heteroatom of each of the substituents to the heme iron are comparable. The reactivity
of methylamines and the potential of N-dealkylation to occur via the alternative single electron
transfer (SET) pathway - which would involve initial reaction at the closer nitrogen atom- rather
than a hydrogen abstraction (HAT) mechanism, are likely to be important here.[12]
The CYP199A4 oxidations of 3-methoxy-, 3-ethoxy- and 3-methyl-benzoic acids (and
their para isomer equivalents) would all proceed via an initial C-H bond abstraction. Factors that
would affect whether Cpd I can abstract a hydrogen are the distance of the C-H bond from the
ferryl oxygen, the C-H bond strength and the angle of approach of the substrate (Figure 11).[41]
The C-H bond strength should not be problematic in any of these oxidations. 3-Methoxybenzoic
acid is not oxidized by CYP199A4 which indicates that it must be inappropriately positioned in
the active site for C-H bond activation despite turnover of the catalytic cycle (as judged by the
NADH oxidation rate, Table 1). Whereas 3-methoxybenzoic acid is not oxidized, CYP199A4 is
able to dealkylate the bulkier 3-ethoxybenzoic acid and hydroxylate the methyl C-H bonds of 3-
24

Journal Pre-proof
methylbenzoic acid. There is a significant change in the conformation of the meta-substituent in
the ethoxy structure which pushes the methylene group closer to the heme iron and changes its
orientation with respect to where the reactive oxygen atom would be located.
It has been proposed that a linear arrangement of the C-H-O atoms is ideal for hydrogen
abstraction[42] and that it is optimal for the substrate to approach the FeIV=O species at an angle
of ~130˚ (Fe-O-C angle).[41] In the 4-methoxybenzoic acid crystal structure, the methoxy group
and its hydrogens point towards the heme. However, the methoxy group of 3-methoxybenzoic
acid is oriented such that the hydrogens could not form a linear C-H-O arrangement with the
iron-oxo atom of Cpd I (Figure 11). The change in the conformation in the 3-ethoxybenzoic acid
side chain results in the C-H bonds on the methylene pointing towards the iron-oxo atom in
contrast to those on the respective methyl groups of 3-methoxy and 3-ethoxybenzoic acid (Figure
11).
It is important to note that while the benzoic acid moiety of the substrates seems to be
tightly confined within the active site, we cannot rule out that the orientation of the substituent
observed in the crystal structures may not be the same as in solution and its position relative to
the heme iron may be altered once oxygen binds or in the presence of the reactive Cpd I
intermediate. However, the crystal structure and biochemical data reported are in agreement with
theoretical work related to this family of enzymes (Figure 11).[41, 42] The presence of
heteroatoms, which can potentially interact with ligands bound to the heme iron, could affect
heme reduction or oxygen binding and activation thus altering different steps of the catalytic
cycle. In the 4-methoxy- and 3-ethoxy-benzoic acid structures, the methyl or methylene groups
are held significantly closer to the heme than the heteroatom. In contrast, in the 3-
methoxybenzoic acid structure, the methyl group and oxygen are held at almost equal distances
25

Journal Pre-proof
from the iron. There is a possibility that this close approach of the oxygen heteroatom could
interfere with substrate oxidation.
5. Conclusions
Overall the crystal structures of CYP199A4 provide detailed information on the factors which
promote efficient catalytic monooxygenase activity for a range of substrates. In summary, this
study reveals that CYP199A4 does oxidize certain meta-substituted benzoic acids and provides
an understanding of why the activity towards these substrates is substantially lower than those of
the para isomers. The crystal structures of CYP199A4 with these substrates bound in the active
site provide important insights into the structural parameters, in this system, which define
whether C-H bond abstraction (or other P450-catalyzed activities) can occur efficiently. This
provides experimental validation of theoretical studies on other forms of these metalloenzymes
and, accordingly these observations could be extended to other P450s. This work highlights that
relatively minor changes in the substrate and its position in the enzyme active site relative to the
heme can have a dramatic effect on the activity of oxidation. Some P450 activities, for example
the C-H abstraction that results in O-dealkylation, were more sensitive to these parameters than
others which involve more reactive substituents (N-dealkylation and S-oxidation). These
considerations are essential when designing or evolving enzyme systems for efficient
biocatalytic reaction or understanding and interpreting the reactivity of computational docking or
structural studies of these enzymes. The range of monooxygenase reactions catalyzed by
CYP199A4 (Figure S1) infer that this system could be used to investigate other types of
cytochrome P450-catalyzed activities, which are important in drug metabolism or natural product
synthesis.
26

Journal Pre-proof
Abbreviations
AcCN, acetonitrile; BA, benzoic acid; Cpd I, iron(IV)−oxo porphyrin radical cation
intermediate; CYP, cytochrome P450 enzyme; CYP199A4, cytochrome P450 enzyme from
Rhodopseudomonas palustris strain HaA2; FAD, flavin adenine dinucleotide; FEM, feature
enhanced maps; GC-MS, gas chromatography-mass spectrometry; HaPuR, FAD containing
flavoprotein from Rhodopseudomonas palustris strain HaA2 (Uniprot: Q2ITV9); HaPux; [2Fe-
2S] ferredoxin from Rhodopseudomonas palustris strain HaA2 (Uniprot: Q2IU01); HAT,
hydrogen abstraction; HPLC, high performance liquid chromatography; HS, high spin; NADH,
reduced nicotinamide adenine dinucleotide; PFR, product formation rate; SET, single electron
transfer; TFA, trifluoracetic acid; TMS, trimethylsilyl; Tris, tris(hydroxymethyl)aminomethane.
Supplementary Information. Experimental methods, additional substrate binding parameters
and catalytic turnover activity data for CYP199A4, details of the distances and angles of
pertinent substrate atoms from the heme iron in the crystal structures of CYP199A4, spin state
shift assays, dissociation constant analysis, examples of NADH oxidation assays, H2O2 assays,
HPLC and GC-MS analysis of the in vitro turnovers, NMR spectra and details of isolated
products, computational docking studies, crystallography details and structures of the Cα chains
of different CYP199A4 crystal structures.
Acknowledgements
This work was in part supported by ARC grant DP140103229 (to JJDV and SGB). SGB
acknowledges the ARC for a Future Fellowship (FT140100355). The authors also acknowledge
the award of an Australian Government Research Training Program Scholarships (PhD to TC
27

Journal Pre-proof
and MPhil to MNP and RRC). We would like to thank the scientists at MX1 and MX2 beamline
for help with data collection. We acknowledge ANSTO for financial support and in providing
the facility used in this work.
28

Journal Pre-proof
References
[1] P.R. Ortiz de Montellano, Chem. Rev., 110 (2010) 932-948.
[2] T.L. Poulos, Chem. Rev., 114 (2014) 3919-3962.
[3] P.R. Ortiz de Montellano, in: P.R. Ortiz de Montellano (Ed.), Cytochrome P450: Structure,
Mechanism, and Biochemistry, Springer International Publishing, Cham, 2015, pp. 111-176.
[4] F. Ogliaro, S.P. de Visser, S. Cohen, P.K. Sharma, S. Shaik, J. Am. Chem. Soc., 124 (2002)
2806-2817.
[5] J. Rittle, M.T. Green, Science, 330 (2010) 933-937.
[6] P.K. Sharma, S.P. De Visser, S. Shaik, J. Am. Chem. Soc., 125 (2003) 8698-8699.
[7] B. Meunier, S.P. de Visser, S. Shaik, Chem. Rev., 104 (2004) 3947-3980.
[8] S. Shaik, S. Cohen, Y. Wang, H. Chen, D. Kumar, W. Thiel, Chem. Rev., 110 (2010) 949-
1017.
[9] X. Huang, J.T. Groves, J. Biol. Inorg. Chem., 22 (2017) 185-207.
[10] M.J. Cryle, J.E. Stok, J.J. De Voss, Aust. J. Chem., 56 (2003) 749-762.
[11] P.R. Ortiz de Montellano, J.J. De Voss, Nat. Prod. Rep., 19 (2002) 477-493.
[12] F.P. Guengerich, Chem. Res. Toxicol., 14 (2001) 611-650.
[13] E.M. Isin, F.P. Guengerich, Biochim. Biophys. Acta, 1770 (2007) 314-329.
[14] J.B. Behrendorff, W. Huang, E.M. Gillam, Biochem. J., 467 (2015) 1-15.
[15] M.T. Reetz, J. Am. Chem. Soc., 135 (2013) 12480-12496.
[16] V.B. Urlacher, M. Girhard, Trends Biotechnol., 30 (2012) 26-36.
[17] T. Coleman, S. Wong, M. Podgorski, J. Bruning, J. De Voss, S. Bell, ACS Catal., 8 (2018)
5915-5927.
29