Design and synthesis of selective CYP1B1 inhibitor via dearomatization of α-naphthoflavone

Article abstract of DOI:10.1016/j.bmc.2018.11.045

Selective cytochrome P450 (CYP) 1B1 inhibition has potential as an anticancer strategy that is unrepresented in the current clinical arena. For development of a selective inhibitor, we focused on the complexity caused by sp³-hybridized carbons and synthesized a series of benzo[h]chromone derivatives linked to a non-aromatic B-ring using α-naphthoflavone (ANF) as the lead compound. Ring structure comparison suggested compound 37 as a suitable cyclohexyl-core with improved solubility. Structural evolution of 37 produced the azide-containing cis-49a, which had good properties in three important respects: (1) selectivity for CYP1B1 over CYP1A1 and CYP1A2 (120-times and 150-times, respectively), (2) greater inhibitory potency of >2 times that of ANF, and (3) improved solubility. The corresponding aromatic B-ring compound 59a showed low selectivity and poor solubility. To elucidate the binding mode, we performed X-ray crystal structure analysis, which revealed the interaction mode and explained the subtype selectivity of cis-49a.

Full text of DOI:10.1016/j.bmc.2018.11.045

Accepted Manuscript
Design and synthesis of selective CYP1B1 inhibitor via dearomatization of α-
naphthoflavone
Makoto Kubo, Keiko Yamamoto, Toshimasa Itoh
PII:
S0968-0896(18)31894-7
https://doi.org/10.1016/j.bmc.2018.11.045
BMC 14649
DOI:
Reference:
To appear in:
Bioorganic & Medicinal Chemistry
Received Date:
Revised Date:
Accepted Date:
9 November 2018
29 November 2018
29 November 2018
Please cite this article as: Kubo, M., Yamamoto, K., Itoh, T., Design and synthesis of selective CYP1B1 inhibitor
via dearomatization of α-naphthoflavone, Bioorganic & Medicinal Chemistry (2018), doi: https://doi.org/10.1016/
j.bmc.2018.11.045
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Design and synthesis of selective CYP1B1 inhibitor via dearomatization of -naphthoflavone.
Makoto Kubo, Keiko Yamamoto, Toshimasa Itoh*
Laboratory of Drug Design and Medicinal Chemistry, Showa Pharmaceutical University, 3-3165
Higashi-Tamagawagakuen, Machida, Tokyo 194-8543, Japan
*Corresponding author
Toshimasa Itoh: titoh@ac.shoyaku.ac.jp ; phone, +81 42 721 1581; Fax, +81 42 721 1580
Abstract
Selective cytochrome P450 (CYP) 1B1 inhibition has potential as an anticancer strategy that is
unrepresented in the current clinical arena. For development of a selective inhibitor, we focused on
3
the complexity caused by sp -hybridized carbons and synthesized a series of benzo[h]chromone
derivatives linked to a non-aromatic B-ring using -naphthoflavone (ANF) as the lead compound.
Ring structure comparison suggested compound 37 as a suitable cyclohexyl-core with improved
solubility. Structural evolution of 37 produced the azide-containing cis-49a, which had good properties
in three important respects: (1) selectivity for CYP1B1 over CYP1A1 and CYP1A2 (120-times and
150-times, respectively), (2) greater inhibitory potency of more than 2 times that of ANF, and (3)
improved solubility. The corresponding aromatic B-ring compound 59a showed low selectivity and
poor solubility. To elucidate the binding mode, we performed X-ray crystal structure analysis, which
revealed the interaction mode and explained the subtype selectivity of cis-49a.

1. Introduction
Cytochromes P450 (CYPs) constitute a superfamily of heme-containing enzymes capable of
catalyzing oxidative biotransformations of both endogenous and exogenous compounds such as
steroidal hormones, bile acids, various kinds of xenobiotics and drugs, and therefore are important for
biological drug discovery.1–4 The CYP1 family, CYP1B1, CYP1A1 and CYP1A2, notably concerns
the bioactivation of procarcinogenic compounds.5–8 Polycyclic aromatic hydrocarbons (e.g.,
benzo[a]pyrene and 7,12-dimethylbenz[a]anthracene) are representative CYP1 family substrates.9–12
These are procarcinogens, which are activated by the CYP1 family. Hence, the CYP1 family has been
recognized as a target for chemoprevention against these carcinogens.5,6
Figure 1. Metabolic pathways of 17-estradiol (E2) catalyzed by CYP1A1, CYP1A2 and
CYP1B1. CYP1 family and catechol-O-methyltransferase (COMT) are responsible for the E2
metabolism. 2-Hydroxylation of the E
2OHE ), and which is then methylated by COMT to produce 2-methoxy-estradiol (2OMeE2).
CYP1B1 affords 4-hydroxy-estradiol, and which is gradually oxidized to estradiol-3,4-quinone (E2-
,4-quinone).
2
regulated by the CYP1A1/1A2 affords 2-hydroxy-estradiol
(
2
3

The CYP1 family also plays an important role in the metabolism of endogenous estrogens,13–15 17-
1
6
estradiol (E2) and estrone (E1), which cause estrogen-dependent cancers such as breast cancer and
ovarian cancer (Figure 1). 2-Hydroxylation of E2 regulated by CYP1A1 and CYP1A2 has been
recognized as a major inactivation pathway of E2, which affords 2-hydroxy-E2 (2OHE ) as a
2
metabolite,14,17 and methylation by catechol-O-methyltransferase (COMT) affords an inactivated
18
).19 In contrast to 2OHE
,
metabolite. On the other hand, CYP1B1 produces 4-hydroxy-E2 (4OHE
2
2
4
OHE
retains strong estrogenic activity17,20 and is gradually oxidized to quinone compounds,21,22
2
which are tumor initiators, because they are easily converted to oxidized products with resistance to
18
COMT-mediated inactivation. The major metabolite of E2 is 2OHE in normal human tissues, but in
2
some tumor tissues, there are high levels of 4OHE
owing to over-expression of CYP1B1.23,24 For
2
these reasons, 4-hydroxylation by CYP1B1 has been recognized as a mechanism of carcinogenesis.
Moreover, CYP1B1 is involved in the metabolism of some anticancer agents25–27 (e.g., Docetaxel,
Tamoxifen and Flutamide) used for the treatment of hormonal cancers and is thought to cause drug
resistance. These reports indicate that selective CYP1B1 inhibition could be an anticancer strategy
unrepresented in the current clinical arena.

2
8
Figure 2. Binding mode of ANF in CYP1B1 (green) (A) (PDB: 3PM0 ), CYP1A1 (teal) (B)
29
30
(
PDB:4I8V ) and CYP1A2 (yellow) (C) (PDB: 2HI4 ). ANF (white), heme (magenta) and residues
are presented as stick models in atom type. Cavities of CYP1B1, CYP1A1 and CYP1A2 were
calculated with Surflex Dock program31 on Sybyl X2 (Tripos) and shown in cyan, orange and red,
respectively. The orientations of ANF are opposite between CYP1B1 and CYP1A1/1A2. ANF has a
- stacking interaction with a conserved phenylalanine in CYP1 family. The position of B-ring is
near the heme.
Crystal structures of hCYP1A2, 1A1 and 1B1 have been reported in holo form with a-naphthoflavone
(
ANF) as an inhibitor (Figure 2).28–30 Homology of the sequences of the active sites is high among the
CYP1 family members, and thus the environments of the substrate-binding sites are similar and are
composed of planar substrate-binding cavities. Several classes of CYP1 inhibitors have been reported,
including flavonoids, trans-stilbenes, coumarins, and alkaloids.5–7 In particular, the structural
evolution of ANF, a representative non-selective CYP1 inhibitor, produced the strongest CYP1B1
3
2
inhibitor. However, ANF derivatives and some other types of inhibitors have a disadvantage in that
2
their aqueous solubility is low owing to their rigid and planar structures created by the sp carbon
atoms.33

Solubility is an important physical property in drug development.34 Ishikawa and co-workers
successfully improved the solubility by a new concept,33,35 i.e., planarity-disruption by introduction of
3
6
ortho-substituents in a -naphthoflavone. For the design of selective CYP1B1 inhibitors, we focused
3
3
3
on sp -hybridized carbons for two-reasons: (1) high fraction sp (Fsp ) compounds have more potential
3
7
as drugs due to their increased complexity, which is expected to reduce promiscuity and increase
38
solubility, (2) the effect on selectivity and inhibitory potency of the 3-dimensional (3D) property has
not been investigated in the design of CYP1B1 inhibitors. Despite the planar CYP1 cavities, the 3D
design could possibly bind by an induced fit strategy because the environment around the heme is
formed by a flexible loop structure. For these reasons, we planned to design ANF-based CYP1B1
3
inhibitors having sp carbons to improve the selectivity among the CYP1 family and increase the
aqueous solubility.
Herein, we report the development of dearomatized ANF derivatives designed for selective CYP1B1
inhibition activity and improvement of solubility. To verify our strategy, we synthesized these
inhibitors as well as aromatic reference compounds, and measured their ethoxyresorufin-O-deethylase
(EROD) activities and solubilities in PBS. In addition, to evaluate the binding mode, we characterized
hCYP1B1 spectroscopically in the presence of an inhibitor and performed the co-crystallization of the
hCYP1B1 and inhibitor.

2
2
. Results and discussion
.1. Design of CYP1B1 Inhibitor
To improve the selectivity and solubility, we attempted to increase the complexity of ANF by
2
3
conversion of the sp carbons into sp carbons. ANF binds to the CYP1 family but the CD-ring is
oriented in the opposite direction between CYP1B1 and CYP1A1/1A2 (Figure 2). In the active site of
the CYP1B1, the CD-ring has a critical -stacking interaction with Phe231 conserved in the CYP1
family and the B-ring is oriented toward the heme. We selected the B-ring as a dearomatization site
3
based on this binding mode. Initially, we planned to find a suitable sp -ring and then optimize it by
adding substituents containing oxygen, sulfur and nitrogen atoms (Figure 3).
Figure 3. Our strategy for selective CYP1B1 inhibitor. Chemical Structures of ANF and design
inhibitors containing benzo[h]chromone skeletone (middle structure, dash line) are shown. Partial
numbering of carbons and ANF (A-D) ring identification (left structure) are reported. Fist stage of our
3
strategy is dearomatization of ANF to find suitable sp -ring (red sphere) against CYP1B1, and second
stage is optimization of substitute groups.

2.2. Chemical Synthesis
Scheme 1. Synthetic routes for Benzo[h]chromone Derivatives. Reagents and conditions: (i) TBSCl,
Et N, DMF, rt; (ii) NaOH, MeOH/THF, rt; (iii) (Boc) O, Et N, 1,4-dioxane/H O, rt; (iv) SOCl
pyridine, DCM, rt; (v) DCC, DMAP, DCM, rt; (vi) t-BuOK, THF, rt; (vii) conc. H SO , EtOH, 80 °C.
3
2
3
2
2
,
2
4
Benzo[h]chromone derivatives containing various B-ring structures 37-47 with cyclohexyl,
cyclopentyl, norbornyl, adamantyl or phenyl moieties were synthesized using 1’-hydroxy-2’-
acetonaphthone as a starting material (Scheme 1). Carboxylic acids 1-3, 8-11, and 14 were esterified
by the Shotten-Baumann reaction with 1’-hydroxy-2’-acetonaphthone. In the case of 6a-c, unintended
deprotection occurred, thus after prepared of silyl- or t-Boc-protected 6a-c, 7, and 13a,b from 4a-c, 5,

and 12a,b, thesewere reacted by Steglich esterification to give the corresponding esters 18a-c, 19 and
4a,b, respectively. Subsequently, Baker-Venkataraman rearrangement39,40 of esters 15-25 with
2
potassium tert-butoxide provided 1,3-diketones 26-36, which were then cyclized by removal of the
silyl and t-Boc protecting groups using sulfuric acid to give the desired benzo[h]chromone derivatives
37-47.
Scheme 2. Synthesis of Cyclohexane Derivatives. Reagents and conditions: (i) MsCl, Et
3
N, DCM, rt;
(
ii) NaN
3
, DMF, 80 °C; (iii) NaBH
, THF/MeOH, 0 °C.
4
, NiCl
2
･6H
2
O, MeOH/THF, 0 °C; (iv) AcSH, K
2
CO
3
, DMF, 80 °C;
(v) K CO
2
3
In order to screen the substituents, we synthesized cyclohexane derivatives (Scheme 2). Four
cyclohexanol isomers 40a-d containing regioisomers and geometrical isomers were converted to
mesylates 48a-d, respectively. 4’-trans-Mesylate 48b was treated with sodium azide in N,N-
dimethylformamide (DMF) at 80 °C to give cis-azide 49a together with trans-azide 49b and 3-
cyclohexene 50, which was formed by elimination of the methanesulfonyl substituent, in low

selectivity. To improvement the selectivity, we varied the conditions such as reagent equivalents,
temperature, leaving groups and the counterpart trans-isomer 48b. Nevertheless, there were no
conditions that gave our desired selectivity. As a result, 49a,b were separated by MPLC, and the other
isomer 49c was obtained as a diastereomixture (cis:trans = 1:1). Synthesis of amines 51a-c was carried
out using sodium borohydride in the presence of nickel salts with corresponding azides 49a-c.
Treatment of mesylate 48b with thiocarboxylic acid in DMF produced diastereomixtures of thioesters
5
5
2a,b and 3-cyclohexene 50. Thus, compounds 52a,b were also separated by MPLC, and regioisomer
2c was obtained as a diastereomixture (cis:trans = 2:1). The synthesized thioesters 52a-c were
reduced with potassium carbonate to give thiols 53a-c.
Scheme 3. Synthesis of Alkylated Cyclohexane Derivatives. Reagents and conditions: (i) MsCl, Et N,
3
DCM, rt; (ii) NaN
3
, DMF, 80 °C; (iii) DMP, NaHCO
3
, DCM, 0 °C; (iv) Ohira-Bestmann reagent,
K
2
CO , MeOH, 0 °C-rt; (v) CuI, Cy
3
2
NH; (CH O) , dioxane, reflux.
2
n

As shown in Scheme 3, compounds 54-58 were synthesized from alcohol 41. The azide 55 was
obtained by the procedures described above for the synthesis of azide 49. The aldehyde 56 was
obtained by oxidation of alcohol 41 with Dess-Martin periodinate. “Seyferth-Gilbert homologation”
of aldehyde 56 was performed using the Ohira-Bestmann reagent and provided a geometrical mixture
of alkynes 57, which were separated into isomers 57a and 57b by MPLC. Conversion of alkynes 57a,b
to allenes 58a,b was achieved by applying the Crabbe Allene synthesis.41
Scheme 4. Syntheses of α-Naphthoflavone Derivatives. Reagents and conditions: (i) conc. HCl,
o
NaNO
2
, NaN
3
, THF/H
2
O, 0 C-rt; (ii) 47% HBr in H
2
O, AcOH, reflux; (iii) dimethyl thiocarbamoyl
chloride, NaH, DMF, 0-80 °C; (iv) 210 °C; (v) 10% NaOH in MeOH, rt.

3
For comparison with the sp -B-ring, we synthesized ANF derivatives 59-63 (Scheme 4). The azides
59a,b were obtained by treatment of amines 46a,b with hydrochloric acid, nitrous acid and sodium
azide. Freunderberg-Schonberg thiophenol synthesis was applied to convert the phenol to a thiophenol
as follows. The methoxy groups of 47a,b were deprotected using hydrogen bromide to afford phenols
60a,b, which were then treated with thiocarbamoyl chloride to give thiocarbamates 61a,b. Thermal
o
rearrangement of thiocarbamates 61a,b at 210 C provided carbamothioates 62a,b, which was
followed by hydrolysis with 10% sodium hydroxide in methanol to give thiophenols 63a,b.
2.3. Inhibitory Activities of Synthetic Compounds toward CYP1B1, CYP1A1 and CYP1A2
4
2
The inhibitory potency of the synthetic compounds was determined using the EROD assay, which
measures fluorescence intensity as the enzymatic activity. Screening of the inhibitory activity was
evaluated as the relative fluorescence intensity at 1 M compound (defined as % activity), when the
fluorescence intensity at 0.5% DMSO in the reaction solution was taken as 100%.
3
To verify suitable sp -B-ring structures for selective CYP1B1 inhibition, we first performed the
3
screening on synthetic benzo[h]chromone derivatives with simple sp -B-ring moieties, compounds 37
and 42-45 (Table 1). ANF as a reference aromatic compound completely inhibited CYP1B1, CYP1A1
and CYP1A2 activities at 1 M. Some compounds including ANF showed % activity of minus, it
might be that not only inhibition of CYP1B1 EROD-activity but also background enzyme in
4
3
3
microsomes as reported by Dutour et al. The bulky sp -B-ring groups in norbornene endo-44, exo-

44 and adamantane 45 showed poor inhibitory potency against CYP1B1 (% activity of 74-89%),
reflecting the planar cavity of the CYP1 family. Cyclohexane 37 was more potent than cyclopentane
42, and was a selective inhibitor of CYP1B1 at 25, 52 and 52% for CYP1B1, CYP1A1 and CYP1A2,
respectively, in spite of its large B-ring structure. These findings indicated that the planar cavity in
CYP1B1 adapted to accommodate the cyclohexane B-ring having 3D property, as we expected.
Although 1-cyclohexene 43 showed strong CYP1B1 inhibitory potency comparable to ANF, the
CYP1A1 activity was also completely inhibited at 1M of 43. On the other hand, 3-cyclohexene 50
did not completely inhibit CYP1B1 activity (% activity of 6.5 ± 6.7%). It is considered that the double
bond position in the cyclohexene ring near the benzo[h]chromone skeleton contributes to the
molecular planarity. Because the cyclohexyl-containing derivative has sufficient inhibitory potency,
we selected compound 37 as a lead compound reflecting the 3D characteristic and optimized the
activity by adding substituents on the cyclohexane B-ring.
Table 1. Inhibition of CYP1 family activity by benzo[h]chromone derivatives containing various B-
ring structures.
a
%Activity at 1 M compound
compd
X
X
X name
CYP1B1
CYP1A1
CYP1A2
3
7
1
cyclohexyl
25.4 ± 1.8
51.5 ± 2.4
51.8 ± 2.8

4
4
2
3
X
X
X
X
X
X
X
2
3
4
4
5
6
7
cyclopentyl
1-cyclohexenyl
5-norbornyl
5-norbornyl
adamantyl
55.3 ± 6.3
-12.9 ± 5.0
83.0 ± 7.0
74.0 ± 8.1
89.0 ± 2.8
-11.3 ± 2.7
28.0 ± 1.9
3.2 ± 0.6
34.5 ± 3.9
21.2 ± 5.4
34.0 ± 6.3
exo-44
61.0 ± 1.7
43.0 ± 5.0
53.9 ± 4.6
endo-44
0.0
± 5.7
4
5
77.8 ± 8.1
-10.1 ± 1.9
36.9 ± 4.7
ANF
phenyl
0.4
± 0.7
5
0
3-cyclohexeyl
6.5
± 6.7
38.1 ± 3.3
aEROD-activity by hCYP1 family in the presence of NADPH. The enzymatic activity at 0.5% DMSO
in the reaction solution was taken as 100%. The experiment performed in triplicate (±SE).
Screening results of the cyclohexane B-ring derivatives (Table 2) demonstrated that introduction of
alcohol (40a-d, 41), mesylate (48a-d) and amine (51a-c) groups regardless of regio- and geometrical
isomerism reduced the inhibitory potency against the CYP1 family, which implied that the
cyclohexane B-ring was surrounded by hydrophobic residues in the substrate-binding cavity.
Methylated 38 and isopropylated 39 had similar inhibitory potency to compound 37. It is suggested
that the C(4’) position was capable of a certain size of hydrophobic substituent. Notably, azide (49a-
c, 55), thioacetyl (52a-c) and thiol (53a-c) derivatives showed more potent inhibitory activity. In
particular, azide-containing derivatives 49a-c and 55 showed better CYP1B1 inhibitory activity and
selectivity. To conduct a fine structure activity relationship study, we evaluated compounds 58a,b
containing an allene structure instead of the azide group. Compounds 58a,b were also strong inhibitors
against CYP1B1, which suggested that rigid and linear structures improved the hCYP1B1 inhibition.
Table 2. Screening of inhibitory potency against CYP1 family by cyclohexane B-ring derivatives.

a
%Activity at 1 M compound
compd
R
isomer
CYP1B1
CYP1A1
CYP1A2
3
3
4
4
4
4
4
4
4
4
4
5
4
4
4
5
5
5
5
5
5
5
5
5
5
5
5
8
4'-CH
3
cis:trans = 3:1
23.9 ± 1.1
27.9 ± 1.4
90.4 ± 1.8
51.3 ± 1.1
79.4 ± 1.6
72.9 ± 1.1
56.9 ± 0.4
51.5 ± 0.8
72.5 ± 2.8
80.5 ± 1.0
59.9 ± 1.5
50.1 ± 0.8
-15.3 ± 1.6
50.2 ± 4.1
48.3 ± 1.2
87.3 ± 3.4
72.7 ± 0.9
71.3 ± 0.7
78.0 ± 2.0
85.9 ± 1.2
66.1 ± 1.0
69.9 ± 0.6
38.6 ± 0.9
66.5 ± 2.1
78.1 ± 1.7
39.6 ± 0.2
39.5 ± 2.9
36.1 ± 2.0
63.0 ± 7.1
b
9
4'-iPr
cis:trans = 4:1
N/A
0a
0b
0c
0d
1
4'-OH
4'-OH
3'-OH
3'-OH
cis
77.3 ± 4.9
76.2 ± 2.2
68.2 ± 3.5
73.9 ± 6.5
92.7 ± 13.4
57.7 ± 2.4
80.3 ± 3.7
79.1 ± 4.9
69.1 ± 1.7
96.4 ± 8.0
29.6 ± 7.1
46.5 ± 8.3
42.5 ± 3.5
22.3 ± 3.2
74.5 ± 4.8
95.4 ± 4.5
44.5 ± 1.0
42.7 ± 2.8
45.4 ± 3.1
trans
cis
trans
4'-CH
OH
cis:trans = 1:1
2
8a
8b
8c
8d
4
4'-OMs
4'-OMs
3'-OMs
3'-OMs
cis
trans
cis
trans
4'-CH
OMs
cis:trans = 1:1
2
9a
9b
9c
5
4'-N
4'-N
3'-N
3
3
3
cis
trans
-1.8
-4.0
2.2
± 4.1
± 4.3
± 0.5
cis:trans = 1:1
cis:trans = 2:3
cis
4'-CH
4'-NH
4'-NH
3'-NH
N
6.3
± 2.1
2
3
1a
1b
1c
2a
2b
2c
3a
3b
3c
7a
7b
2
2
2
81.6 ± 2.0
89.5 ± 0.7
64.7 ± 1.2
73.7 ± 2.0
84.2 ± 2.2
80.9 ± 2.6
trans
cis:trans = 1:1
cis
4'-SAc
4'-SAc
3'-SAc
4'-SH
-1.3
-4.7
4.4
± 4.0
± 5.5
± 0.9
± 0.7
± 0.2
± 0.3
± 0.4
3.8
1.1
± 1.1
± 0.7
trans
cis:trans = 2:1
cis
31.7 ± 1.4
0.5
± 1.4
b
0.5
3.9
1.3
0.0
± 0.3
± 1.3
± 0.4
N/A
4'-SH
trans
1.8
0.6
± 4.0
b
3'-SH
cis:trans = 3:1
cis
-2.5
0.9
N/A
21.0 ± 4.0
8.8 ± 2.2
4'-CCH
4'-CCH
11.3 ± 1.6
0.5 ± 0.8
trans
15.0 ± 1.1

5
5
8a
8b
4'-Allene
4'-Allene
cis
3.1
± 1.0
± 0.6
5.5
3.0
± 1.0
± 0.9
4.3
± 8.2
± 6.8
trans
-0.4
-4.5
aEROD-activity by hCYP1 family in the presence of NADPH. The enzymatic activity at 0.5% DMSO
b
in the reaction solution was taken as 100%. The experiment performed in triplicate (±SE). Not
analyzed

Table 3. Screening of inhibitory potency against CYP1 family by ANF derivatives.
a
%Activity at 1 M compound
compd
R
CYP1B1
CYP1A1
CYP1A2
4
4
4
4
5
5
6
6
6
6
6
6
6
6
6a
6b
7a
7b
9a
9b
0a
0b
1a
1b
2a
2b
3a
3b
4'-NH
2
2
-1.2
0.2
±
0.8
0.6
0.3
0.6
0.9
0.8
0.2
0.9
1.2
1.0
0.3
0.6
0.4
1.4
2.1
7.6
4.0
8.7
2.3
1.4
6.3
8.2
± 1.7
-16.3 ± 2.1
3'-NH
±
±
±
±
±
±
±
±
±
±
±
±
±
± 1.0
± 2.4
± 0.2
± 2.2
± 1.1
± 1.1
± 0.5
-3.5
15.9
6.1
± 5.8
± 12.0
± 5.3
4'-OMe
3'-OMe
-0.5
1.9
4'-N
-0.5
0.7
-13.6 ± 5.0
3
3
3'-N
6.9
± 8.4
± 6.9
± 6.4
± 7.4
± 2.1
± 5.3
± 8.8
4'-OH
3.7
10.4
37.2
96.9
75.4
91.9
68.7
3'-OH
1.6
4'-OC(=S)NMe
3'-OC(=S)NMe
4'-SC(=O)NMe
3'-SC(=O)NMe
15.1
2.5
17.6 ± 0.4
3.1 ± 1.2
10.3 ± 1.7
3.8 ± 0.7
2
2
2
2
12.3
2.4
4'-SH
3'-SH
54.3
9.8
71.3 ± 3.2
18.7 ± 0.5
106.4 ± 9.6
56.2 ± 1.7
aEROD-activity by hCYP1 family in the presence of NADPH. The enzymatic activity at 0.5% DMSO
in the reaction solution was taken as 100%. The experiment performed in triplicate (±SE).
The simple transformation from aromatic to non-aromatic B-rings (Table 3) gives rise to a notable
difference in the CYP1 inhibitory properties. Most synthetic ANF derivatives, even those containing
alcohol (60a,b) or amine (46a,b) groups, strongly inhibited the enzymatic activities of CYP1B1,
CYP1A1 and CYP1A2 at 1 M, ≤10%, except for 60b of CYP1A2. Introduction of a thiol group

increased the inhibitory activity among the cyclohexane B-ring derivatives, but in the case of an
aromatic B-ring, it reduced the CYP1B1 inhibitory activity, e.g., 63a (54%) and 63b (9.8 %). These
differences demonstrated that the 3D-structural property is an important factor for the selectivity and
inhibitory activity among the CYP1 family.
3
From these results, we found strong CYP1B1 inhibitors having sp -B-rings that included azide, allene,
thiol and thioacetyl groups. To investigate their inhibitory potency against the CYP1 family in more
detail, we determined the IC50 values of the compounds (Table 4).
The compound 49a showed stronger inhibitory potency (IC50 of 4.4 nM for CYP1B1) than that of
ANF (IC50 of 10.6 nM). Moreover, among the evaluated compounds, 49a showed the highest selective
inhibition against CYP1B1 at 120-times over CYP1A1 and CYP1A2 (IC50 of 525 and 667 nM for
CYP1A1 and CYP1A2, respectively). trans-Isomer 49b had 100-times lower inhibitory activity
against CYP1B1 (IC50 of 134 nM) than cis-isomer 49a. Interestingly, all azide derivatives (49a-c, 55)
showed selectivity for hCYP1B1. Although they had a linear structure, allene derivatives showed a
different inhibitory activity profile than the azide derivatives. Stronger CYP1B1 inhibition was found
for the trans-isomer 58b (IC50 of 13.4 nM) than the cis-isomer 58a (IC50 of 68.1 nM). Moreover, it did
not have selectivity among the CYP1 family. These results revealed that the allene group and the azide
group are not equivalent in CYPs inhibition. Concerning other substituents, thioesters 52a,b exhibited
more inhibitory potency against CYP1B1 than thiols 53a-c, but less than 49a. Thus, we concluded that

the most effective substituent for the cyclohexane B-ring derivative is the azide group. In the case of
aromatic B-rings, compounds 59a,b having an azide substituent at C(4’) or C(3’) showed strong
inhibitory potency against not only CYP1B1 but also CYP1A1. Therefore, these compounds had low
3
selectivity compared to the series of cyclohexane B-rings. In this way, our strategy using sp carbons
in the B-ring allowed us to discover the selective CYP1B1 inhibitor 49a.
Table 4. Inhibition of CYP1B1, CYP1A1, CYP1A2 activity and selectivity by hit compounds.
a
IC50 (nM)
IC50 ratio
compd
R
isomer
CYP1B1
CYP1A1
CYP1A2
1A1/1B1
1A2/1B1
Cyclohexyl B-ring
4
4
4
5
5
5
5
5
5
5
5
3
9a
9b
9c
5
4'-N
4'-N
3'-N
3
3
3
cis
4.4
525
667
120.4
7.7
153
8.8
trans
134.6
90.3
23.5
68.1
13.4
24.3
29.4
38.8
40.3
66.1
539.8
1036
1190
b
b
cis:trans = 1:1
N/A
N/A
4'-CH
N
cis:trans = 2:3
2
3
215.6
33.
432.8
N/Ab
36.4
9.2
18.5
2.7
8a
8b
2a
2b
3a
3b
3c
7
4'-Allene
4'-Allene
4'-SAc
4'-SAc
4'-SH
cis
0.48
1.02
2.7
trans
13.6
65.
b
cis
N/A
trans
15.9
76.2
146
991
0.54
1.97
3.6
33.7
17.1
b
cis
N/A
4'-SH
trans
689
b
3'-SH
cis:trans = 3:1
none
133.5
N/A
2.02
b
b
H
N/A
N/A
Phenyl B-ring
5
5
6
9a
9b
0a
4'-N
3
3
none
none
none
none
2.6
9.5
93.8
74.7
908.7
3.6
35.6
42.8
21
3'-N
1.8
27.1
43
15.5
0.99
5.1
4'-OH
H
43.4
10.6
ANF
53.8
68.4
6.5
aIC50 (CYP1A1)/IC50 (CYP1B1) or IC50 (CYP1A2)/IC50 (CYP1B1). Not analyzed.
b

2.4. Comparing Solubilities of ANF, 37, 49a and 59a
In order to examine the influence of dearomatization and addition of azide group on solubility, the
thermodynamic aqueous solubility of ANF and compounds 37, 49a, 59a was evaluated with the
44
saturation shake-flask method. ANF is known to have quite low solubility in aqueous solvent due to
its planar and hydrophobic structure. We also observed that ANF was nearly insoluble in PBS (pH
7.4), and therefore used a solution in which PBS and ethanol were mixed at 9:1 as the solvent for
solubility measurements. However, the solubility of ANF was still poor (2.63 M) in this solution.
Aromatic B-ring-linked azide 59a showed the worst solubility (<1 M) in this series. On the other
hand, dearomatization of the B-ring improved the aqueous solubility. Indeed, cyclohexane B-ring 37
was 3.7-times more soluble (9.65 M) than ANF. Cyclohexane B-ring-linked azide 49a retained this
solubility (2.67 M) compared with ANF. Lipophilicity was also important for water solubility.
Introduction of the azide (from ANF to 59a, from 37 to 49a) increased the CLogP (>0.17) with
decreasing solubility, which could be caused by its hydrophobicity. In contrast, dearomatization of the
B-ring (from ANF to 37, from 59a to 49a) also increased ClogP (>0.5) but improved solubility. Thus,
3
we demonstrated that the increasing complexity induced by the sp -B-ring contributes to the aqueous
solubility.

2.5. Spectroscopic Analysis
Figure 4. Spectral changes induced in hCYP1B1 upon binding of ANF (A), imidazole (B), 40b
3
+
3+
(
C), and 53a (D). Shown are absorbance spectra of Fe state (black line), Fe state w/ inhibitor (red
line), Fe2+ state (black dashed line) and Fe2+ w/ inhibitor (red dashed line). Concentrations of
hCYP1B1, ANF, imidazole, 40b and 53a are 1 M, 5 M, 10 mM, 120M, and 50 M. Insets (a) are
difference spectral changes observed during titration of hCYP1B1 with different ligands. Insets (b)
3
+
show plots of spectral change vs. ligand concentration. The types of spectral change were type 1 (Fe -
3+
3+
3+
5
th state, A), type 2 (N-Fe -6th state, B), reverse type 1 (O-Fe -6th state, C) and S-Fe -6th state (D).
Oxidative state of iron changed from ferric to ferrous by addition of Na
2
S
2 4
O .
In order to determine the manner of interaction, we measured the ultraviolet-visible (UV-vis)
absorbance changes of CYP1B1 introduced by the inhibitor. The absorbance spectrum of CYP is
characterized by the spin and oxidized states of heme45 and thus reflects the binding of a “ligand”.
CYP1B1 was expressed in Escherichia coli (E. Coli) and purified in the absence of ANF by applying

28
previously reported methods. The steady state absorption spectra showed peaks at 416, 535 and 570
46
nm in the Soret, - and  regions, which are representative of low spin state spectra along with water
coordination 46 (Figure 4).
4
6
Addition of ANF induced a type 1 spectral change (Fig. 4A). The spectral characteristics suggested
that ANF excluded the sixth water ligand of the heme and thus interacted with CYP1B1 without a
32
water ligand in the co-crystal structure. In the same way, cyclohexyl azides 49a,b, 55, phenyl azides
59a, 59b, and compounds 37-39, 43, 50, 52a, 53b, 57a and 58a,b also showed type 1 spectral changes
(Table 5). Moreover, we observed not only type 1 spectral changes but three other types of
representative spectral changes: (1) imidazole and amine compounds 51a, 51c induced type 2 spectral
4
7
changes (Fig. 4B) that reflect the coordination of a nitrogen atom; (2) alcohols 40a-b and compounds
5
1b, 52b, 57b showed reverse (rev.) type 1 spectral changes48 (Fig. 4C) indicating coordination of
oxygen atoms; (3) thiol groups 53a,c induced a distinctive spectral change (Fig. 4D) reflecting
49
coordination of thiol groups. In regards to the oxidation state, the reduced CYP1B1 spectrum showed
the same features as other CYPs. 46,50 In this way, our synthetic inhibitors induced distinctive spectral
changes corresponding to the substituent structure at the B-ring. We confirmed by spectroscopic
analysis that the cyclohexane B-ring was located toward the heme, similar to ANF, despite the
nonaromatic B-ring.

Table 5. Spectral Interaction of Synthetic Inhibitors with hCYP1B1a.
wavelenght
Ksb
ΔA
compd
R
isomer
(peak/trough) (M)
(Amax-Amin
)
ΔA/Ks
type
Cyclohexenyl B-ring
4
5
3
0
H
H
1-cyclohexenyl 384/418
3-cyclohexenyl 384/418
1.1
3.1
0.098
0.033
0.0892
0.0104
1
1
Cyclohexanyl B-ring
3
3
3
4
4
4
4
5
5
5
5
5
5
5
5
5
5
5
5
5
7
H
384/416
5.9
0.019
0.035
0.046
>0.1945
0.032
0.075
0.053
0.059
0.0033
0.0062
0.0004
0.0005
N.D.d
1
8
4'-Me
4'-iPr
4'-OH
4'-OH
3:1
385/418
384/416
420/385
424/386
384/418
384/418
384/418
428/N.D.
417/389
442/418
384/418
424/384
384/418
384/417
384/418
384/416
424/387
384/417
384/415
5.7
1
9
4:1
124.1
>1218
62.7
2.6
1
0a
0b
9a
9b
5
cis
rev. 1
trans
cis
rev. 1
4'-N
4'-N
0.0284
0.0098
0.0152
1
3
3
trans
2:3
5.4
1
4'-CH
N
3.9
1
2
3
d
d
d
d
1a
1b
1c
2a
2b
3a
3b
3c
7a
7b
8a
8b
4'-NH
4'-NH
3'-NH
2
cis
N.D.
N.D.
N.D.
2
d
d
d
2
2
trans
1:1
N.D.
N.D.
N.D.
rev. 1
2
2.9
0.012
0.084
0.035
0.055
0.053
0.044
0.024
0.033
0.023
0.019
0.0040
0.0208
0.0052
0.0059
0.0060
0.0054
0.0033
0.0001
0.0009
0.0081
4'-SAc
4'-SAc
4'-SH
cis
4.0
1
trans
cis
6.7
rev. 1
c
9.4
1
4'-SH
trans
3:1
8.9
1
c
3'-SH
8.1
1
4'-CCH
4'-CCH
cis
7.3
1
trans
cis
263.9
25.4
2.4
rev. 1
4'-Allene
4'-Allene
1
1
trans
Phenyl B-ring
5
5
9a
9b
4'-N
3'-N
H
3
384/418
384/418
384/418
0.9
0.8
0.7
0.082
0.070
0.083
0.0879
0.0922
0.1281
1
1
1
3
ANF
Other Structure
Imidazole
430/408
1255.0
0.092
0.0001
2
a
b
c
Detection by difference spectrum in the presence of inhibitor. Spectral dissociation constant. S-
d
coordination. Not detected.

The binding affinity of the synthetic inhibitors was evaluated as the spectral dissociation constant
(Ks) and magnitude of spectral binding (ΔA/Ks) calculated from equilibrium titrations (Table 5).
Comparing the six-membered B-ring structures, the binding affinity was higher in the order of ANF,
1-cyclohexenyl 43, 3-cyclohexenyl 50 and cyclohexyl 37 (Ks: 0.7, 1.1, 3.1 and 5.9 uM, respectively),
with the same order of CYP1B1 inhibition. Spectroscopic analysis supported that the molecular
planarity has a critical influence on binding to CYP1B1.
Cyclohexyl azides 49a,b, 55 and phenyl azides 59a,b induced spectral shape changes (ΔA > 0.05)
and thus had high binding affinity for CYP1B1 (Ks: 0.8-5.4 M). The order of binding affinity was
the same as the order of inhibitory activity. Interestingly, the phenyl azide itself did not show binding
affinity or inhibition activity at all (data not shown). In comparison with the cis-49a and trans-49b
azides ( A/Ks of 0.0284 and 0.0098, respectively), the cis-58a and trans-58b allenes showed weaker
binding affinity (A/Ks of 0.0009 and 0.0081, respectively) derived from their small absorption
changes (A of ~0.02). Spectroscopic analysis also showed the difference between azide and allene.

2.6. X-ray Crystal Structure Analysis
Figure 5. Overall structure of compound 49a/CYP1B1 (A), and substrate binding site (B and C).
A shows superposition of 49a/CYP1B1 (cyan) and ANF/CYP1B1 (green). The 2Fo-Fc map (blue
mesh) positioned substrate binding site. The A-D ring of 49a (49a-core) conformed the electron
density map, but the localization of azide can not be confirmed (B). C shows conformational
comparison of 49a-core and ANF (green stick)/CYP1B1. (a) shows co-crystal of 49a/CYP1B1. (b)
shows 49a structure and 49a-core (dash line).
We co-crystallized compound 49a and hCYP1B1 to investigate the binding mode by X-ray crystal
structure analysis. Compound 49a was added after purification of CYP1B1. The complex of
compound 49a and CYP1B1 was crystallized by the hanging drop vapor diffusion method. In this way,

we successfully obtained a deep red crystal for the X-ray diffraction experiment and X-ray crystal
structure analysis (Figure 5). The data collection and refinement statistics are summarized in Table S1.
The overall structure of CYP1B1/49a (6IQ5. pdb) exhibits a canonical P450-fold and overlaps with
the structure of hCYP1B1/ANF (Fig. 5A). The electron density map of compound 49a exists in the
substrate-binding site around the heme, the same location that ANF occupies (Fig. 5B and 5C). In spite
of the transformation of the B-ring structure from phenyl to cyclohexyl, we confirmed the localization
of the CD-ring, which was in a position corresponding to ANF. The conformation of the cyclohexyl
B-ring, clearly different from that of ANF, was twisted against the planar surface of the CD-ring. While
we could confirm the location of the skeleton of compound 49a by X-ray crystal structure analysis, it
was difficult to determine of positioning of terminal nitrogen atom due to low resolution.

2.7. Docking Analysis
Figure 6. The localization site of azide substitute in CYP1B1. Docking model of 49a (A), 59a (B)
and 59b (C) are shown as stick model in white, blue and salmon, respectively. The electron density
map (blue mesh) observed by X-ray crystal structure analysis is displayed. Residues of CYP1B1 are
shown in green. D shows superposition of docking model (A-C) and the localization site of azide
indicated by black arrow. E is a view of D from the opposite side, which shows 49a and the localization
site of azide as translucent sphere model.
To elucidate the location of the azide substituent in compound 49a, we performed docking analysis
using the Surflex Dock program on Sybyl X2 (Tripos) software. We obtained two types of docking
models: (1) the B-ring is in a twisted conformation that is well fit to the electron density map of the
X-ray crystal structure of 49a (Fig. 6A), (2) a conformation similar to that of ANF in hCYP1B1 for

59a and 59b (Fig. 6B and 6C). All compounds were surrounded by hydrophobic amino acid residues
in similar with observed crystal structure. The CD-ring in both types of conformation formed a -
interaction with Phe231 and the B-ring had intimate interaction with Leu509. Three docking models
suggested that localization of the terminal nitrogen atom of the azide converged in a tiny hydrophobic
cavity constructed by Thr334, Val395 and Thr510 (Fig. 6C and 6D). Thus, we demonstrated that
hydrophobic interactions play an important role for the binding affinity and inhibitory potency of
compound 49a. However, CYP1A1 also conserved the residues of Thr321, Val382 and Thr497 as in
CYP1B1.
Figure 7. Docking model of compound 49a in CYP1A1. Docking model of 49a (white), the heme
(magenta) and residues (teal) are shown as stick models in atom type. Cavity of CYP1A1 is irastrated
in orange. Black arrow shows the locarization site of azide in CYP1B1. Azide substitute of 49a dose
not interact with the site in this model.

We constructed the docking model of compound 49a and CYP1A1 to discuss the selectivity (Figure
). In this analysis, because the benzo[h]chromone core was observed in a similar conformation as
7
was ANF in CYP1B1, we restricted the conformation of the CD-ring to be identical to that of the
ANF/CYP1A1 crystal structure. This docking simulation supports the reason why azide 49a lacks
CYP1A1 affinity. There is an absence of interaction between the azide substituent and the cavity as
judged by the distance between them. Thus, we concluded that the conformation of the
benzo[h]chromone core in CYP1A1 and CYP1A2 was unsuitable for compound 49a. In addition,
instead of Val395 as in CYP1B1, CYP1A2 has a Leu382 which is a bulkier residue, therefore it is
more difficult to accept the azide group.
The residue at position 395 in CYP1B1 has been reported to be critical factor for catalytic activity
and substrate recognition.51,52 We considered that the difference in binding affinity between cis-azide
49a and cis-allene 59a was caused by acceptability of this tiny site. As a reason for this, the terminal
carbon atom of the allene has two perpendicular hydrogen atoms. Although these reasons are not valid
for all our synthetic compounds, we deduced that the complexity caused by the combination of the
cyclohexane structure, the azide substituent and the benzo[h]chromone core effectively enforces the
desired selectivity for hCYP1B1 among the highly homologous CYP1 family.

3. Conclusion
We designed and synthesized a series of benzo[h]chromone derivatives with linked nonaromatic B-
rings to improve the selectivity for CYP1B1 and the solubility of ANF as a lead compound. EROD
assays demonstrated that cyclohexane-linked benzo[h]chromone 37 was suitable for hCYP1B1
3
inhibition by comparison of B-ring structures composed of sp carbons. Screening of substituents on
compound 37 revealed that the azide-containing cis-49a exhibited 120-times higher selectivity for
CYP1B1 than CYP1A1 and CYP1A2 and possessed stronger inhibitory potency than ANF. On the
other hand, the azide-containing naphthoflavone 59a showed strong inhibitory potency but low
selectivity. For this reason, we confirmed that the azide substituent increased the inhibitory potency
and the cyclohexane structure was effective for CYP1B1 selectivity. Moreover, we found that
compound 37 was more soluble in aqueous solution than ANF, and therefore demonstrated that the
conversion of the benzene B-ring to the cyclohexane B-ring improved the solubility. The solubility of
compound 49a decreased upon introduction of the azide substituent, but was still equal to or greater
than that of ANF. In this way, we synthesized a selective and strong CYP1B1 inhibitor without
decreasing the water solubility.
We then investigated the manner of binding between the synthetic inhibitor and CYP1B1.
Spectroscopic analysis showed that the synthetic inhibitors induced distinctive spectral changes in
CYP1B1 corresponding to the substituent structure at the B-ring, and thus suggested that the