Riboflavin Is Directly Involved in N-Dealkylation Catalyzed by Bacterial Cytochrome P450 Monooxygenases

Article abstract of DOI:10.1002/cbic.202000071

Like a vast number of enzymes in nature, bacterial cytochrome P450 monooxygenases require an activated form of flavin as a cofactor for catalytic activity. Riboflavin is the precursor of FAD and FMN that serves as indispensable cofactor for flavoenzymes. In contrast to previous notions, herein we describe the identification of an electron-transfer process that is directly mediated by riboflavin for N-dealkylation by bacterial P450 monooxygenases. The electron relay from NADPH to riboflavin and then via activated oxygen to heme was proposed based on a combination of X-ray crystallography, molecular modeling and molecular dynamics simulation, site-directed mutagenesis and biochemical analysis of representative bacterial P450 monooxygenases. This study provides new insights into the electron transfer mechanism in bacterial P450 enzyme catalysis and likely in yeasts, fungi, plants and mammals.

Full text of DOI:10.1002/cbic.202000071

Combining Chemistry and Biology
Accepted Article
Title: Riboflavin Is Directly Involved in the N-Dealkylation Catalyzed by
Bacterial Cytochrome P450 Monooxygenases
Authors: Chengchang Zhang, Meiling Lu, Lin Lin, Zhangjian Huang,
Rongguang Zhang, Xuri Wu, and Yijun Chen
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: ChemBioChem 10.1002/cbic.202000071
Link to VoR: https://doi.org/10.1002/cbic.202000071
01/2020

10.1002/cbic.202000071
ChemBioChem
Riboflavin is directly involved in N-dealkylation catalyzed by
bacterial cytochrome P450 monooxygenases
Dr. Chengchang Zhang,a ‡ Dr. Meiling Lu,a ‡ Dr. Lin Lin,b Prof. Dr. Zhangjian Huang,a
Prof. Dr. Rongguang Zhang,b Dr. Xuri Wu,a* Prof. Dr. Yijun Chena*
a Laboratory of Chemical Biology and State Key Laboratory of Natural Medicines, China
Pharmaceutical University, 639 Longmian Avenue, Nanjing, Jiangsu Province, 211198,
China
b National Center for Protein Science and Shanghai Institute of Biochemistry and Cell
Biology, Chinese Academy of Sciences, 333 Haike Road, Shanghai 201210, China
*Corresponding author
‡Contributed equally
1
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
Abstract
As a vast repertoire of enzymes in nature, bacterial cytochrome P450 monooxygenas-
es require an activated form of flavin as a cofactor for the catalytic activity. Riboflavin is
the precursor of FAD and FMN that serve as indispensable cofactors for flavoenzymes. In
contrast to previous notion, here we describe the identification of an electron transfer
process directly mediated by riboflavin for the N-dealkylation by bacterial P450
monooxygenases. The electron relay from NADPH to riboflavin and then via activated
oxygen to heme was proposed based on the combination of X-ray crystallography, mo-
lecular modeling and molecular dynamics simulation, site-directed mutagenesis and bio-
chemical analysis of representative bacterial P450 monooxygenases. This study provides
new insights into the electron transfer mechanism in bacterial P450 enzyme catalysis and
likely in yeasts, fungi, plants and mammals.
Key Words: Cytochrome P450 Monooxygenase; Riboflavin; Enzyme catalysis; Dealkyl-
ation; Electron Transfer
2
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
Introduction
Cytochrome P450 monooxygenases ubiquitously occur in nature. P450s catalyze a va-
riety of oxidative reactions, especially in the secondary metabolism in microorganisms
and plants and drug metabolism in humans. [1] Recently, bacterial P450s have become a
major focus on the discovery of novel functionalities from the efforts of directed evolu-
tion and protein engineering. [2] Currently, there are two major electron transfer mecha-
nisms for P450 enzymes. [3] One is a system in which an NAD(P)H-dependent and FAD-
containing reductase shuttles electrons to ferredoxin, which in turn transfers the electrons
to P450 enzymes. Another is more general, which utilizes an NAD(P)H-dependent flavin
mononucleotide (FMN) and flavin adenine dinucleotide (FAD)-containing reductase for
the electron supply. Although most catalytic systems of P450s fit into above electron
transfer pathways, several “unusual” electron transfer chains for different P450s have al-
so been found, indicating that there still is much to be uncovered regarding the diversity
of electron transfer for P450 enzymes. [3] The most well-known and biotechnologically
useful example for bacterial P450s is the two-domain fused and self-sufficient P450-BM3
from Bacillus megaterium, and this P450 enzyme contains a catalytic domain (BMP) and
a FMN and NADPH-dependent reductase domain (BMR). [4] With regard to its electron
transfer process, there have been quite a few studies. For example, a hinge-shortened cy-
tochrome P450 reductase variant produces a stable complex of heme and hydroxyl radical
and can support direct electron transfer from FMN to heme; [5] the FMN domain of BM3
undergoes large conformational changes to bring FMN- and heme-domains close enough
(8.8 Å) for facilitating the fast electron transfer in BM3. [6] These studies have generated
3
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
new knowledge on how the electrons are transferred during the catalysis by P450 en-
zymes. However, presently there is no general consensus on how flavin cofactors partici-
pate in the catalytic reactions.
Typically, bacterial P450s indispensably require a flavin cofactor, either FMN or FAD,
for their electron transfer and catalytic activity regardless of being self-sufficient or the
requirement of a redox partner. [7] For instance, the heme domain of BM3 can be reduced
by NADPH and FMN to yield the same product as did holo-BM3. [8] Regarding flavin
cofactors, FAD and FMN are biosynthesized from their precursor riboflavin (Vitamin B2,
Figure 1), in which the isoalloxazine ring is a functional group to accept or transfer elec-
trons. Flavin cofactors play a pivotal role in the step-down reaction between obligatory
two-electron donors and obligatory one-electron acceptors, as they are capable of both
one- and two-electron redox chemistry. [9] In flavoproteins, oxygen can be activated by
the transfer of one electron from fully reduced flavin to O2. [10] As previously known, the
electrons originate from NAD(P)H that binds either transiently or permanently. [11]
Meanwhile, previous studies have indicated that riboflavin itself may possess the capabil-
ity of accepting and transferring electrons from NAD(P)H. Riboflavin could be used as
an electron acceptor in an artificial flavoenzyme consisting of NADH and manganese
porphyrin. [12] Furthermore, riboflavin can donate electrons from NADH to a quinone in
the NADH-driven sodium pump. [13] Moreover, riboflavin can transfer electrons in modi-
fied CYP 2B4 and directly bind CYP 2B4. [14] In the case of CYP 152, a bacterial P450
enzyme, riboflavin could be activated by light-excitation and the electron donor eth-
ylenediaminetetraacetate can serve as an electron source to complete the catalysis by
CYP 152. [15] Interestingly, in this artificial system, when riboflavin is free or covalently
4
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
bound and either NAD(P)H or EDTA is used as the electron source, the peroxygenase
activity is generally low, and the activity of CYP 152 remains significant even when high
concentration of catalase is added into the reaction system, suggesting that the oxidative
catalysis by CYP 152 is not related to possible peroxygenase activity. These collective
evidences have implicated that riboflavin can directly participate in the electron transfer
during P450 catalysis. However, how the electron transfer process is mediated by ribofla-
vin in the complex catalysis remains elusive.
In the present study, given the valuable applications of N-dealkylation by P450s, espe-
cially in the area of drug metabolism, [16] we chose this reaction based on the unusual
findings as a model to uncover the roles of riboflavin in bacterial P450s. After confirming
riboflavin-mediated catalysis by representative bacterial P450 monooxygenases, we were
able to identify a new electron transfer pathway in addition to previously proposed cata-
lytic mechanisms of P450s. This could enable our thorough understanding on various
P450s and their catalytic mechanisms.
5
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
Figure 1. Structural comparison of riboflavin, FMN and FAD.
Results
Riboflavin directly mediated the N-dealkylation by bacterial P450s
Given the occurrence of a large number of uncharacterized bacterial P450 proteins in
the databases, we first examined whether HmtS (GenBank: CBZ42153.1), [17] a P450 ho-
mologous protein proposed to be involved in the final step of the biosynthesis of himasta-
tin, can utilize riboflavin for the natural reaction. Because the natural substrate of HmtS
was difficult to obtain and inaccessible by chemical synthesis, we selected a series of
simplified structural analogues, such as D/L-tryptophan, R-indoline-2-carboxylic acid, L-
abrine, and others, to test the biaryl coupling. However, no biaryl products were found
regardless of using riboflavin or ferredoxin/ferredoxin reductase (FDX/FDR) as a redox
partner in the presence of NADPH. Unexpectedly, N-dealkylation products were ob-
6
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
served when riboflavin and NADPH were used as cofactors. Subsequently, after expres-
sion and purification of HmtS protein in a monomeric form with a molecular weight of
45 KDa (Figure S1), various substrates were screened for the dealkylation by HmtS in the
presence of riboflavin and NADPH by LC-MS/MS (Table S1). Although eight com-
pounds were identified to generate dealkylated products by HmtS, diphenhydramine was
found to be the best substrate (Figure S2). In addition, the dealkylation reaction appeared
both riboflavin- and NADPH-concentration dependent, and optimal pH and temperature
for the reaction were determined to be pH 7.4 and 34 °C respectively (Figure S3). Ac-
cording to the results of Figure S3 and referring to the previous report, [14] we chose the
modified conditions to detect the catalytic activity of the enzyme. After a preparative re-
action with HmtS, diphenhydramine, riboflavin and NADPH, 50 mg N-desmethylated
product was purified, and the chemical structure of the product was confirmed by NMR
(Figure S4 and S5). Thus, diphenhydramine was used as a model substrate for subsequent
examinations. Compared to FAD and FMN, riboflavin showed higher turnover number
(TON). Meanwhile, NADPH showed higher activity than NADH (Table 1). To rule out
the possibility of being a false positive, hemin, bovine serum albumin and bovine catalase
were used to replace HmtS. Under the same reaction conditions, no product was observed
in these non-catalytic components (Figure S6). These results strongly indicated that ribo-
flavin indeed participates in the oxidative catalysis by HmtS.
7
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
Table 1. Comparison of different microbial P450 enzymes on the N-dealkylation of
diphenhydramine.
TON [a]
Component
Component
P450 BM3
TON
NADPH
0
0
0
13
0
NADPH + riboflavin
HmtS + riboflavin
P450 BM3 + NADPH
BMP + NADPH
0
HmtS + riboflavin +
NADPH
18
BMP + riboflavin + NADPH
9
HmtS + FMN + NADPH
HmtS + FAD + NADPH
14
16
14
BMP + FMN + NADPH
BMP + FAD + NADPH
9
8
HmtS + riboflavin +
NADH
HmtT + riboflavin +
NADPH
43
52
CYP106A2 + riboflavin +
NADPH
29
37
HmtN + riboflavin +
NADPH
CYP107DY1 + riboflavin +
NADPH
[a] The reaction mixtures contained 4 μM purified enzyme, 400 μM substrate, 3 mM NADPH,
4 μM riboflavin in PBS buffer (pH 7.4) and were conducted for 4 h. Data are average values of
3 independent experiments. TON was calculated based on the peak areas of the prod-
uct by HPLC with a standard calibration curve over a period of 4 h.
Next, four additional P450-like proteins that lack the reductase domain, including
CYP107DY1 (UniProtKB: D5E3H2), [18] CYP106A2 (UniProtKB: Q06069), [19] HmtT
(GenBank: CBZ42154.1) and HmtN (GenBank: CBZ42148.1), [17] were further investi-
gated to examine their capability of catalyzing the dealkylation. All four enzymes were
active against diphenhydramine using riboflavin and NADPH as the cofactors and
showed higher TON than HmtS (Table 1), suggesting that riboflavin can serve as the co-
8
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
factor for a broad range of bacterial P450-like monooxygenases that usually lack the re-
ductase domain. To further confirm this finding and expand the scope of bacterial P450s
with riboflavin, the self-sufficient flavocytochrome P450 enzyme BM3 [20] was evaluated
by the comparison with its truncated mutant BMP lacking the reductase domain (BMR).
As expected, the full-length BM3 containing BMR domain was able to catalyze the N-
dealkylation of diphenhydramine while BMP was inactive. Intriguingly, BMP could cata-
lyze the reactions in the presence of NADPH when riboflavin was added into the reaction
mixtures, indicating that riboflavin could be used to functionally transfer electrons for the
catalysis by BMP. Although FAD and FMN were also accepted by BMP, riboflavin
seemed equally suitable as a electron donor (Table 1). These results confirmed again that
riboflavin indeed participates in the electron transfer and enzyme catalysis. Therefore, the
simplified catalytic system for bacterial P450s consisted of enzyme, substrate, NADPH
and riboflavin.
Riboflavin effectively served as a redox partner for bacterial P450s
It is well-known that the catalytic process by cytochrome P450 enzymes from various
sources requires a redox partner and NADPH reductase. [21] Additionally, the binding of
hydrogen peroxide to ferric heme is also described as the peroxide shunt pathway. How-
ever, most natural P450s lack the structural features required to activate H2O2, the effi-
ciency of the peroxide shunt pathway in driving P450 catalysis is usually not high. [22] To
gain more information on the catalytic process of HmtS mediated by riboflavin, we
aligned the amino acid sequence of HmtS with bacterial P450 enzymes from different
sources (Table S2). The sequence alignments indicated that HmtS contains a conserved
oxygen-binding motif GXXT and a Leu residue at 340 (Figure S7), which is different
9
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
from conserved Phe in the heme-binding motif of FXXGXXXCXG. This conserved Phe
residue in P450s is responsible for modulating the reducing potential of heme through the
interaction with conserved heme-biding residue Cys347. [23] Therefore, HmtS-L340F mu-
tant was constructed, expressed and purified. Based on the close similarity of the CD
spectra between HmtS and HmtS L304F mutant, this mutation did not change the tertiary
structure of the enzyme (Figure S8). The molecular weight of HmtS-L340F was deter-
mined to be 44 KDa by gel filtration chromatography (Figure S9), which is the same as
WT. When diphenhydramine and riboflavin were respectively used as the substrate and
as the cofactor, the kinetic behaviors, similar to other P450s, were complicated, in which
the product increased steadily in the first 4 h (Figure S10). Then, we compared riboflavin
with classic redox partner FDR/FDX for the catalysis by HmtS. The initial rate with ribo-
flavin for HmtS was lower than FDR/FDX, but the final conversion yield between these
redox systems was equivalent (Figure S10). We next measured H2O2 formation during the
catalysis by P450s with different redox partners. Significant amount of H2O2 was ob-
served in FDR/FDX system (Figure S11a), whereas no measurable H2O2 was detected in
the riboflavin system. Moreover, NADPH was significantly consumed in the riboflavin
system compared to blank (Figure S11b). These observations suggested that the faster
initial rate of FDR/FDX is likely due to the simultaneous generation of H2O2. Subse-
quently, to evaluate if the peroxide shunt pathway is involved in the catalysis by HmtS
with riboflavin, exogeneous addition of H2O2 was performed to make a comparison. The
activity of HmtS with riboflavin was generally higher than H2O2 (Figure S12). When cat-
alase was added into the H2O2-mediated reactions, only small amount of the product was
observed. On the other hand, the same amount of catalase did not significantly affect the
10
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
catalysis with riboflavin (Figure S12). Collectively, these results indicated that the reac-
tion mediated by riboflavin is not a result of producing H2O2, and suggested that ribofla-
vin might mediate a new electron transfer pathway during the catalysis of HmtS.
The enzymatic activity correlated to oxygen-binding and activation
As previously reported, [3] oxygen plays a key role in the formation of complex I in the
catalysis of many P450 enzymes. In order to study the role of oxygen in the catalytic pro-
cess of riboflavin, we chose to crystallize HmtS and HmtS-L340F. Both wild type (WT)
and L340F mutant of HmtS were crystallized and their crystal structures (PDB: 5Z9I and
5Z9J) were solved at 2.0 Å and 2.2 Å respectively (Figure S13 and Table S3). Both struc-
tures adopt classic prism-shape architecture, consisting of 12 α helices and 14 β strands.
The heme prosthetic groups are located in the typical binding pocket. Although overall
structures of two proteins are very close (Figure S14), the difference is found in the pock-
et for O2 binding, in which the mutation from Leu to Phe results in a larger pocket due to
the change of back-side region from heme. Meanwhile, the distance between heme and
residue Thr241, the conserved site for O2 binding and activation, [24] is obviously longer
in HmtS-L340F than that in WT (Figure 2A).
11
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
Figure 2. Comparison of HmtS and HmtS-L340F on the efficiency of utilizing O2. A)
The enlargement of O2 binding pocket by L340F mutant. B) UV-Vis. spectroscopic dif-
ferences on CO binding between HmtS and HmtS-L340F.
Compared to WT, L340F mutant showed approximately 3-fold increase of conversion
(Table 2), which may be attributed to the enlargement of O2 binding site for more effi-
cient activation. Also, HmtS-L340F was more thermally stable (Figure S15). Additional-
ly, spectroscopic analyses indicated that CO binding for HmtS-L340F mutant was much
stronger than WT (Figure 2B) when dithionite was used as the reducing agent, suggesting
12
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
that the activity increase by the mutant should be a result of the enlargement of O2 bind-
ing pocket for more rapid O2 activation.[24] Meanwhile, when mutant HmtS-L340F was
treated with riboflavin and NADPH, the spectroscopic difference on CO binding showed
a weak absorbance at 450 nm (Figure S16), suggesting that riboflavin with NADPH are
capable of directly reducing the intrinsic ferric heme. Together, the results suggested that
the enlargement of oxygen-binding pocket is critical to the catalytic activity with ribofla-
vin.
Riboflavin mediated a new electron transfer pathway
To explore the electron transfer process by riboflavin, complexed structures of enzyme
and cofactors were built by molecular modeling [25] with the crystal structure of HmtS-
L340F as a template due to its higher activity. In the modeled structures, HmtS-L340F
can simultaneously accommodate and bind NADPH and riboflavin (Figure S17), suggest-
ing that two cofactors can transfer electrons in the active pocket. Then, these models were
used as initial conformations to conduct molecular dynamics (MD) simulation. According
to literatures, [26] 120-ns simulation for HmtS-L340F-riboflavin complex and 100-ns sim-
ulation for HmtS-L340F-riboflavin-NADPH complex were performed in Charm 3.6 field
(Table S4). The changes in root mean square deviation (RMSD) and root mean square
fluctuation (RMSF) indicated that both complexes can achieve equilibrium states in the
initial stage by thermal energy (Figure S18). In the simulation of HmtS-L340F-riboflavin
complex, riboflavin can occupy the active pocket to form hydrogen bonds with residue
Arg233 (Figure S19 and S20). However, there is a long distance (9-15 Å) between ribo-
flavin and ferric ion of heme during the simulation process (Figure S21). Interestingly,
when complexed with additional NADPH, the ribityl moiety in NADPH forms a new hy-
13
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
drogen bond with residue Asp74 (Figure S22). Subsequently, NADPH and riboflavin
gradually adjust their positions in the active pocket in the initial stage of the simulation
(Figure 3A).
Figure 3. MD simulation for the complexed models and the electron transfer process me-
diated by HmtS-L340F. A) Serial spatial positions between NADPH and riboflavin in
HmtS-L340F-riboflavin-NADPH complex. B) Distance fluctuations of C4 (NADPH) to
N1 (riboflavin) in HmtS-L340F-riboflavin-NADPH complex. C) Serial spatial positions
between riboflavin and heme in HmtS-L340F-riboflavin-NADPH complex. D) Distance
fluctuations of N5 (riboflavin) to Fe center in HmtS-L340F-riboflavin-NADPH complex.
Ribbon diagram of NADPH (violet) and riboflavin (yellow) is shown in the models.
14
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
Next, C4 of nicotinamide ring in NADPH moves closer to N1 of isoalloxazine ring of
riboflavin to give a short distance of 4.81 Å between NADPH and riboflavin. Later, ribo-
flavin gradually leaves from NADPH and moves to ferric ion of heme to yield an in-
creased distance between NADPH and riboflavin (Figure 3B). As shown in Figure 3C
and 3D, the distance between riboflavin and heme is obviously decreased to 7.50 Å away
from Thr241 residue after 100-ns simulation and remains unchanged. Consequently, N5
of isoalloxazine ring of riboflavin transfers electrons to the ferric iron to allow O2 activa-
tion, which is promoted by Thr241 residue, to accomplish the N-dealkylation reaction.
On the other hand, when the enzyme was complexed with the substrate in addition to ri-
boflavin and NADPH, no stable confirmations were obtained during the 100-ns simula-
tion and the substrate was not able to move close to heme (Figure S23), suggesting that
the enzyme is unable to simultaneously bind the substrate and cofactors. This indicated
that the dealkylation should be a sequential process. At the same time, NADPH exhibited
two forms in the simulation of the electron transfer process (Table S5), suggesting that
NADP+ may leave the active pocket after donating a pair of electrons to allow the en-
trance of the substrate for the completion of the catalytic cycle as proposed in Figure S24.
Taken together, the MD simulations suggest that NADPH directly reduces riboflavin,
which transfers electrons through O2 to heme and then proceeds substrate oxidation.
Biochemical analyses confirmed the electron transfer process
After excluding the possibility of dimerization (Figure S1 and S9) and the existence
of covalently bound flavins in the enzymes from their crystal structures, to verify the
proposed electron transfer process, we individually tested whether the substrate or cofac-
tors can directly bind the enzyme measured by Microscale Thermophoresis Assay (MST).
15
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
Although the binding affinity between HmtS and substrate or NADPH was relatively
weak, the affinity between HmtS and riboflavin was strong (Table 2), which is similar to
a previous report. [27] Meanwhile, the binding affinity of riboflavin was also significantly
stronger than FAD or FMN (Table 2), suggesting a possible role of steric hindrance.
Table 2. Biochemical analyses on the P450-riboflavin system
Kd values (μM) [a]
Mutant of HmtS
WT
TON
18
HmtS
HmtS-
L340F
82
NADPH
Diphenhydramine
Riboflavin
FMN
83
35
1.9
40
43
L340F
L340F,D74A
51
3
43
5.3
L340F,R233A
L340F,T241A
L340F,C347A
L340F,D74A,R233A
3
38
13
0
FAD
42
0.
[a] Binding affinity determined by Microscale Thermophoresis (MST). Kd values were calculated
by MO Affinity Analysis v2.2.4. Data are average values of 3 independent determinations. All
reactions were conducted under the same conditions. The reaction mixtures contained 4 μM puri-
fied enzyme, 400 μM substrate, 3 mM NADPH, 4 μM riboflavin in PBS buffer (pH 7.4) and were
conducted for 4 h.
Based on sequence alignments (Figure S7), molecular modeling (Figure S17) and MD
simulation (Figure 3), among sixteen residues in HmtS for potential involvement in the
binding and catalysis as listed in Table S6, four important residues, including Cys347 for
heme binding, Thr241 for superoxide binding and activation, Arg233 putatively for ribo-
flavin binding and Asp74 putatively for NADPH binding, were identified and individual-
16
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
ly mutated to Ala to examine their influences to the activity. As anticipated, in addition to
conserved residues of Cys347 and Thr241, both Arg233 and Asp74 could remarkably af-
fect the activity of HmtS-L340F from the mutagenesis. In the complexed model consist-
ing of enzyme, NADPH and riboflavin, Arg233 is the site to bind riboflavin (Figure S19
and S20), and Asp74, similar to the situation of Asp632 in NADPH-cytochrome P450
oxidoreductase, [28] is responsible for the binding with NADPH (Figure S22). Not surpris-
ingly, the significant reduction of activity from the mutations of Arg233 and Asp74 was
in accordance with the results of MD simulation. Subsequently, double mutation of
R233A and D74A indeed resulted in the complete loss of activity (Table 2). The collec-
tive evidences demonstrated that Arg233 and Asp74 are essential residues for the binding
of the cofactors, which strongly supports the proposed mechanism of electron transfer
process mediated by riboflavin as illustrated in Figure 4.
17
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
Figure 4. Proposed mechanism for the electron transfer by riboflavin in bacterial P450
monooxygenases.
Discussion
Cytochrome P450 monooxygenases are heme-containing enzymes that use molecular
oxygen and NAD(P)H for oxidative insertion of one oxygen atom into an organic sub-
strate. Typically, P450 enzymes exhibit poor stability, low turnover number and require
electron transport protein(s) for their catalysis. Among different classes of P450s, class I
enzymes encompass most of known bacterial P450s in addition to a number of eukaryotic
18
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
mitochondrial P450s. This class of enzymes usually possess a characteristic of two P450
components, a FAD-dependent ferredoxin reductase and a ferredoxin containing an iron-
sulfur cluster and require NAD(P)H as a source of electrons. [3] Previously, HmtS was
proposed to be involved in the biosynthesis of himastatin, [17] presumably responsible for
the biaryl coupling. However, its actual role has not been confirmed in vitro. Given this
interesting reaction, we have attempted to verify the biaryl coupling in vitro, but the ef-
forts have been unsuccessful. On the other hand, when we screened various substrates for
N-dealkylation, HmtS exhibited notable activity. To test whether riboflavin could substi-
tute classic redox partner for the catalysis, [10-14] based on the structural features of widely
existed bacterial P450s that usually lack the reductase domain, we selected HmtS as a
model enzyme to probe its cofactor requirement. Surprisingly, HmtS indeed could direct-
ly utilize riboflavin and NADPH to accomplish the N-dealkylation, showing higher activ-
ity than FAD and FMN. With regard to the natural substrate and possible involvement in
the biaryl coupling by HmtS in the organism, we are uncertain at this point, which re-
quires further investigation. However, a thorough comparison revealed that this dealkyla-
tion reaction is mediated by a new electron transfer pathway, which is not the same as
any reported mechanisms for P450s. [29] Subsequently, based on sequence alignments with
other bacterial P450s, we found that residue Leu340 of HmtS is different from other en-
zymes. Thus, HmtS-L340F mutant was constructed to show higher activity. To elucidate
possible mechanism on the electron transfer, both HmtS and HmtS-L340F were crystal-
lized and their crystal structures were solved. Although the complexed structures of the
enzyme with riboflavin and NADPH could not be obtained after numerous attempts due
likely to lower binding affinity or transitional binding, we were able to identify the elec-
19
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
tron transfer process through MD simulations, and a new catalytic mechanism was then
proposed, which may have several implications for various P450s. Of note, same as most
other P450s, the activity of HmtS was relatively low, which might be a result of stability
issue and the complex electron transfer process required by multi-components. According
to previous report, [30] it is mechanistically possible that the present N-dealkylation might
be different from other types of oxidation catalyzed by P450s, which requires further
comparison and clarification.
As one of the common oxidative reactions, N-dealkylation can be catalyzed by various
oxygenases. However, regioselectivity is still an issue for many enzymes on their biocata-
lytic applications. By contrast, the present regioselective dealkylation by HmtS, ribofla-
vin and NADPH could be utilized to prepare specific amines to serve as versatile building
blocks for the synthesis of various pharmaceuticals and bioactive compounds. [31] In addi-
tion, this system could replace microsomal preparations and mammalian P450s for the
generation of relatively large quantity of drug metabolites in a short period of time for
preclinical evaluation of drug metabolism and toxicity, [32] which would be helpful to
speed up the process of drug discovery.
The studies on cytochrome P450 monooxygenases have increasingly been a spotlight,
[33], especially the exploration and discovery of the catalysis of novel reactions by protein
engineering. A recent report on the α-Amino C-H fluoroalkylation by engineered P450
enzymes [34] is just one of the excellent examples. Meanwhile, regardless of de novo de-
sign or rational engineering of enzymes, the catalytic mechanisms are a prerequisite.
Thus, we believe that our present finding can also provide a starting point for the design
20
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
and engineering of different P450 enzymes, and we anticipate that much more novel en-
zymes and reactions are yet to come.
In summary, we have demonstrated that riboflavin directly participates in the oxidative
catalysis by bacterial cytochrome P450 monooxygenases. Based on the collective evi-
dences, we have proposed a new electron transfer pathway for the N-demethylation reac-
tion by bacterial P450 enzyme, NADPH and riboflavin, which might be generally appli-
cable to numerous P450-like monooxygenases that lack the reductase domain. The pre-
sent work has established a foundation to better understand cofactor requirement, catalyt-
ic mechanism and substrate scope for numerous P450-like enzymes in nature.
Experimental section
Construction of plasmids and site-directed mutagenesis
PCR products containing fdr, fdx, cyp106A2, cyp107DY1, cyp102A1, BMP, hmtT,
hmtN and hmtS were each amplified and cloned into the Nde I/Xho I or Nde I/Hind III
sites of pET28a(+) or pET22b(+) (ClonExpress II One Step Cloning Kit Vazyme Biotech
Co. Ltd, China) to produce the recombinant constructs pFDR-ET28a, pFDX-ET28a,
pBM3-ET28a, pBMP-ET28a, pCYP106A2-ET28a, pCYP107DY1-ET28a, pHmtT-
ET28a, pHmtN-ET28a, pHmtS-ET22b and pHmtS-ET28a for the expression of FDR,
FDX, CYP106A2, CYP107DY1, P450 BM3, BMP, HmtT, HmtN and HmtS, respective-
ly. Mutagenesis was performed with pHmtS-ET22b as a template using a QuickChange
site-directed mutagenesis kit (Stratagene, La Jolla, USA) according to the manufacturer’s
protocol. Superoxide dismutase (SOD) from bovine erythrocytes was purchased from
Aladdin (Shanghai, China).
21
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
Enzyme activity assay
The reactions for the screening of various substrates were performed for 4 h using 4
μM enzyme, 3 mM NADPH, 150 μM substrate and 4 μM riboflavin in PBS buffer (137
mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 2 mM KH2PO4, pH 7.4). The reactions for
the screening of substrates were performed for 12 h using 4 μM enzymes, 3 mM
NADPH, 150 μM substrate and 4 μM riboflavin in PBS buffer. The reactions for
FDR/FDX system contained 4 μM purified enzyme, 150 μM substrate, 3 mM NADPH, 5
μM spinach ferredoxin NADP+ reductase, 1 μM spinach ferredoxin in PBS buffer
(pH7.4) and were conducted for 12 h for the screening of substrates. Most substrates were
dissolved in the buffer, and some with addition of 0.4% DMSO or methanol. The reaction
mixtures contained 4 μM purified enzyme, 3 mM NADPH, 4 μM riboflavin/FMN/FAD
and 400 μM substrate in PBS buffer (pH 7.4) to a final volume of 200 μL, and aliquoted
in triplicate into a 48-well plate for 4-h reactions. To avoid possible interference from
light, the plates containing reaction mixtures were covered and wrapped in foil. After
shaking the plates at 250 rpm, the reactions were quenched by addition of methanol (400
μL) at -20 °C. This mixture was then transferred to a microcentrifuge tube and centri-
fuged at 12,000 × g for 10 min. The supernatant was subjected to HPLC analysis. Con-
version and TON were calculated based on the peak areas of the product and substrate
with standard calibration curves by HPLC.
Microscale thermophoresis (MST) assay
Purified enzymes were labeled with primary-amine coupling of NT647 dye according
to the manufacturers’ protocol and purified from free dye with Superdex 25 resin in the
22
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
buffer (50 mM HEPES, pH 7.5, 100 mM NaCl) at 25 °C. Titration series were made with
1:1 dilution of ligands while maintaining a constant 50 nM NT-647 labeled enzymes. The
initial concentrations of diphenhydramine, NADPH, riboflavin, FAD and FMN were 2
mM, 1 mM, 250 μM, 250 μM and 250 μM, respectively. The sample was loaded into the
NanoTemper glass capillaries and microthermophoresis carried out using 20% LED pow-
er and 80% MST. Kd values were calculated using the mass action equation via the Nan-
oTemper software from duplicate reads of triplicate experiments.
UV-Visible spectroscopic analysis
CO binding spectra were recorded as described previously.[35] UV-Visible spectra of
HmtS and HmtS-L340F at 2 mg/ml concentration were recorded from 300 to 550 nm.
The ferric heme was reduced by adding freshly prepared saturated solution of dithionite
(Na2S2O4) to the protein solution, then was bubbled softly by CO gas, and the spectrum
of reduced enzyme was then recorded. To examine the capability of reducing heme by
riboflavin and NADPH, 1 mM NADPH and 4 μM riboflavin were added to the enzyme
solution, and then the spectra before and after softly bubbling CO gas were recorded.
Protein crystallization and structural determination
After the screening of crystallization conditions, HmtS crystals were obtained using a
reservoir solution of 22% (w/v) polyethylene glycol (PEG) 3350, 0.1 M ammonium ace-
tate, 0.1 M NDSB-256 and 0.1 M Bis-Tris (pH 5.5) and grown at 20 °C by the sitting-
drop vapor diffusion method, in which 1 μL of 6 mg/ml protein solution in 0.1 M NaCl, 1
mM DTT, 20 mM Tris-HCl buffer (pH 7.4) was mixed with an equal volume of the res-
ervoir solution. For the preparation of HmtS-L340F crystals, the reservoir solution con-
23
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
sisted of 20% (w/v) PEG3350, 0.15 M ammonium sulfate and 0.1 M Bis-Tris (pH 6.5).
Other solutions and experimental methods were same as those used for HmtS crystalliza-
tion.
Crystals were protected by using 25% glycerol as the cryoprotectant. The diffraction
data were collected at 19 U Beamline of Shanghai Synchrotron Radiation Facility
(SSRF), and were processed by HKL2000 package. [36] Molecular replacement was per-
formed by Phaser in CCP4 [37] to solve the structure of HmtS-L340F using the crystal
structure of HmtT (PDB: 4GGV) as the searching model. The structure of HmtS was
solved by molecular replacement using the mutant structure as a searching model. Re-
finements were carried out using Phenix. refine. [38] Coot was used for model buildings
and modifications. [39] For detailed data, see Supplementary Table S3. Crystal structures
of HmtS and HmtS-L340F were overlapped and the O2 binding pocket of HmtS and
HmtS-L340F was determined from the distances of heme-N237, heme-T242, heme-T241
using MOE2014.
Molecular modeling
Molecular modeling was performed using Molecular Operating Environment (MOE
2014.0901 i4w9). Atomic structures of receptor-ligand complexes for ligands
(diphenhydramine, NADPH and riboflavin) were built in MOE. In this study, anchor-
based conformers were prepared in MOE using HmtS-L340F module, and side-chain
conformations for the resulting molecules were explored using low mode sampling with
fixed scaffold. The active site was generated for HmtS-L340F using the MOE alpha site
finder. The ligand molecules were placed in the site using Triangle Matcher method, and
24
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
then ranked with London dG scoring function. One hundred best poses were retained and
further refined by energy minimization in the pocket, followed by rescoring with
GBVI/WSA dG scoring function. Docking these ligands into the simulated active pocket
of HmtS-L340F gave an output having heme approximately in the same orientation as
that observed in the crystal structure of HmtS.
Molecular dynamics simulation
To verify the modeled structures and probe the catalytic process, molecular dynamics
(MD) simulation for the complexes of HmtS-L340F-riboflavin and HmtS-L340F-
riboflavin-NADPH were started from initial conformations. The MD simulation was per-
formed with GROMACS package. The Charmm3.6 field was used for generating the to-
pology and coordinate files for each system. All small molecule force fields including
riboflavin, NADPH and heme in HmtS-L340F were generated by CGENFF force field
file, and then converted into the format supported by GROMACS package. Each struc-
ture was immersed in a regular dodecahedron box of TIP3P waters, resulting in the addi-
tion of 12,400 solvent molecules on average. Each system was neutralized through the
addition of counter ions and energy minimization. To avoid the dissolution of small mol-
ecules in the equilibrium process, protein, riboflavin and NADPH were classified as one
group, and solvent and ions as another group. Subsequently, the systems were gradually
heated from 0 K to 300 K for 5 ns of density equilibration in the NVT ensemble with a 2-
fs time step, then the systems were equilibrated in NPT ensemble at a target temperature
of 300 K and a target pressure of 1.0 bar with a 2-fs time step under periodic boundary
conditions. Totally, HmtS-L340F-riboflavin complex was simulated for 120 ns, and
HmtS-L340F-riboflavin-NADPH complex was simulated for 100 ns. The results were
25
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
analyzed, focusing on the changes of the distance between key atoms in the simulated
period and the conformational changes of small molecules and HmtS-L340F.
Author contributions
Y. C. and X. W. conceived and supervised the research. C. Z. and M. L. designed and per-
formed the experiments and analyzed the data, L.L. and R. Z. solved the crystal struc-
tures. Z. H. analyzed data. C. Z. and Y.C. wrote the manuscript. ‡ C. Z and M. L contrib-
uted to the work equally.
Corresponding Author
*yjchen@cpu.edu.cn, xuriwu@cpu.edu.cn.
ORCID
Yijun Chen: 0000-0002-4920-152X
Xuri Wu: 0000-0002-0933-5902
Additional information
Experimental procedures, characterization of small molecule compounds, UV-Visible
spectroscopy, additional data of HPLC, LC-MS/MS, structural data of the enzymes and
MD simulation are included in the supporting information.
Competing financial interests
The authors declare no competing financial interests.
Acknowledgements
This work was supported by the MOST (2018YFA0902000 and 2019ZX09721001-
0401), the “111 Project” from the Ministry of Education of China and the State Admin-
istration of Foreign Expert Affairs of China (No. 111-2-07), "Double First-
Class" University project (CPU2018GY36) and National Science Foundation of China
26
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
(No: 21778076). We are grateful to the staff of beamline BL19U at SSRF, Shanghai, Chi-
na, for the assistance on X-ray crystallography.
REFERENCES
[1] a) S. T. Jung, R. Lauchli, F. H. Arnold, Curr. Opin. Biotechnol. 2011, 22, 809-817; b)
J. D. Rudolf, C. Y. Chang, M. Ma, B.Shen, Nat. Prod. Rep. 2017, 34, 1141 -1172.
[2] a) Y. Wei, E. L. Ang, H. Zhao, Curr. Opin. Chem. Biol. 2018, 43, 1-7; b) 4
Guengerich, F. K. Yoshimoto, Chem Rev. 2018, 118, 6573-6655.
F. P.
[3] D. J. Cook, J. D. Finnigan, K. Cook, G. W. Black, S. J. Charnock, Adv. Protein. Chem.
Struct. Biol. 2016, 105, 105-126.
[4] P. S. Coelho, E. M. Brustad, A. Kannan, F. H. Arnold, Science 2013, 339, 307-10.
[5] M. Sugishima, H. Sato, Y. Higashimoto, J. Harada, K. Wada, K. Fukuyama, M. No-
guchi, Proc. Natl. Acad. Sci. U S A 2014, 111, 2524-2529.
[6] K. D. Dubey, S. Shaik, J. Am. Chem. Soc. 2018, 140, 683-690.
[7] A.W. Munro, H.M. Girvan, K.J. McLean, Biochim. Biophys. Acta 2007, 1770, 345-
359.
[8] S. Gonvindaraj, H. Li, T. L. Poulos, Biochem. Biophys. Res. Commun. 1994, 203,
1745-1749.
[9] M. M. Huijbers, S. Montersino, A. H. Westphal, D. Tischler, W. J. van Berkel, Arch.
Biochem. Biophys. 2014, 544, 2-17.
[10] a) R. Teufel, Arch, Biochem, Biophys. 2017, 632,20-27. b) R. Saleem-Batcha, F.
Stull, J. N. Sanders, B. S. Moore, B. A. Palfey, K. N. Houk, R. Teufel, Proc. Natl. Acad.
Sci. U S A 2018, 115, 4909-4914.
[11] M. J. Furst, F. Fiorentini, M. W. Fraaije, Curr. Opin. Struct. Biol. 2019, 59, 29-37.
[12] Y. Roux, R. Ricoux, F. Avenier, J. P. Mahy, Nat. Commun. 2015, 6, 8509.
27
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
[13] O. Juarez, M. J. Nilges, P. Gillespie, J. Cotton, B. Barquera, J. Biol. Chem. 2008,
283, 33162-33167.
[14] V. V. Shumyantseva, V. Uvarov, O. E. Byakova, A. I. Archakov, Arch. Biochem.
Biophys. 1998, 354, 133-138.
[15] M. Girhard, E. Kunigk, S. Tihovsky, V. V. Shumyantseva, V. B. Urlacher, Biotech-
nol. Appl. Biochem. 2013, 60, 111-118.
[16] N. L. Dang, T. B. Hughes, G.P. Miller, S. J. Swamidass, Chem. Res. Toxicol. 2018,
31, 68-80.
[17] a) J. Ma, Z. Wang, H. Huang, M. Luo, D. Zuo, B. Wang, A. Sun, Y. Q. Cheng, C.
Zhang, J. Ju, Angew. Chem. Int. Ed. 2011, 50, 7797-7802. b) Z. Guo, P. Li, G. Chen, C.
Li, Z. Cao, Y. Zhang, J. Ren, H. Xiang, S. Lin, J. Ju, Y. Chen, J. Am. Chem. Soc. 2018,
140, 18009-18015.
[18] M. Milhim, N. Putkaradze, A. Abdulmughni, F. Kern, P. Hartz, R. Bernhardt, J. Bio-
technol. 2016, 240, 68-75.
[19] D. Schmitz, J. Zapp, R. Bernhardt, Microb. Cell Fact. 2014, 13, 81.
[20] a) D. C. Haines, I. F. Sevrioukova, J. A. Peterson, Biochemistry 2000, 39, 9419-9429;
b) H. M. Girvan, T. N. Waltham, R. Neeli, H. F. Collins, K. J. McLean, N. S. Scrutton, D.
Leys, A. W. Munro, Biochem. Soc. Trans. 2006, 34, 1173-1177.
[21] F. Hannemann, A. Bichet, K. M. Ewen, R. Bernhardt, Biochim. Biophys. Acta. 2007,
1770, 330-344.
[22] (a) K. Piontek, E. Strittmatter, R. Ullrich, G. Gröbe, M. J. Pecyna, M. Kluge, K.
Scheibner, M. Hofrichter, D. A. Plattner, J. Biol. Chem. 2013, 288, 34767-34776; (b) K.
Kuhnel, W. Blankenfeldt, J. Terner, I. Schlichting, J. Biol. Chem. 2006, 281, 23990-
23998.
[23] J. P. Clark, C. S. Miles, C. G. Mowat, M. D. Walkinshaw, G. A. Reid, S. N. Daff, S.
K. Chapman, J. Biol. Inorg. Chem. 2006, 100, 1075-1090.
28
This article is protected by copyright. All rights reserved.

10.1002/cbic.202000071
ChemBioChem
[24] H. Yeom, S. G. Sligar, H. Li, T. L. Poulos, A. J. Fulco, Biochemistry 1995, 34,
14733-14740.
[25] A. Wadood, Bioinformation 2013, 9, 309-314.
[26] S. C. Dodani, G. Kiss, J. K. B. Cahn, Y. Su, V. S. Pande, F. H. Arnold, Nat. Chem.
2016, 8, 419-425.
[27] V. V. Shumyantseva, T. V. Bulko, N. A. Petushkova, N. F. Samenkova, G. P. Kuz-
netsova, A. I. Archakov, J Inorg Biochem. 2004, 98, 365-370.
[28] C. Xia, F. Rwere, S. Im, A. L. Shen, L. Waskell, J. P. Kim, Biochemistry 2018, 57,
945-962.
[29] F. P. Guengerich, ACS Catal. 2018, 8, 10964-10976.
[30] A. H. Meyer, A. Dybala-Defratyka, P. J. Alaimo, I. Geronimo, A. D. Sanchez, C. J.
Cramer, M. Elsner, Dalton Trans 2014, 43, 12175-12186.
[31] H. Kohls, F. Steffen-Munsberg, M. Höhne, Curr. Opin. Chem. Biol. 2014, 19, 180-
192.
[32] K. Purnapatre, S. K. Khattar, K. S.Saini, Cancer Lett. 2007, 259, 1-15.
[33] a) S. Nagano, T. L. Poulos, J. Biol. Chem. 2005, 280, 31659; b) T. Iyanagi, Biochim.
Biophys. Acta Bioenergetics 2019, 1860, 233-258.
[34] J. Zhang, X. Huang, R. K. Zhang, F. H.Arnold, J. Am. Chem. Soc. 2019, 141, 9798-
9802.
[35] T. Omura, R. Sato, J. Biol. Chem. 1964, 239, 2370-2378.
[36] Z. Otwinowski, W. Minor, Methods Enzymol. 1997, 276, 307-326.
[37] M. D. Winn, C. C. Ballard, K. D. Cowtan, E. J. Dodson, P. Emsley, P. R. Evans, R.
M. Keegan, E. B. Krissinel, A. G. Leslie, A. McCoy, S. J. McNicholas, G. N. Murshudov,
N. S. Pannu, E. A. Potterton, H. R. Powell, R. J. Read, A. Vagin, K. S. Wilson, Acta.
Crystallogr. D. Biol. Crystallogr. 2011, 67, 235-242.
29
This article is protected by copyright. All rights reserved.