Structure-based design and profiling of novel 17β-HSD14 inhibitors

Article abstract of DOI:10.1016/j.ejmech.2018.05.029

The human enzyme 17β-hydroxysteroid dehydrogenase 14 (17β-HSD14) oxidizes the hydroxyl group at position 17 of estradiol and 5-androstenediol using NAD⁺ as cofactor. However, the physiological role of the enzyme remains unclear. We recently des

Full text of DOI:10.1016/j.ejmech.2018.05.029

Accepted Manuscript
Structure-based design and profiling of novel 17β-HSD14 inhibitors
Florian Braun, Nicole Bertoletti, Gabriele Möller, Jerzy Adamski, Martin Frotscher,
Nathalie Guragossian, Patrícia Alexandra Madeira Gírio, Marc Le Borgne, Laurent
Etouatti, Pierre Falson, Sebastian Müller, Günther Vollmer, Andreas Heine, Gerhard
Klebe, Sandrine Marchais-Oberwinkler
PII:
S0223-5234(18)30442-2
10.1016/j.ejmech.2018.05.029
EJMECH 10440
DOI:
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 16 March 2018
Revised Date: 15 May 2018
Accepted Date: 20 May 2018
Please cite this article as: F. Braun, N. Bertoletti, G. Möller, J. Adamski, M. Frotscher, N. Guragossian,
Patrí.Alexandra. Madeira Gírio, M. Le Borgne, L. Etouatti, P. Falson, S. Müller, Gü. Vollmer, A. Heine,
G. Klebe, S. Marchais-Oberwinkler, Structure-based design and profiling of novel 17β-HSD14 inhibitors,
European Journal of Medicinal Chemistry (2018), doi: 10.1016/j.ejmech.2018.05.029.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT
Trp192
Thr205
Leu191
Leu195
Met199
Structure-Based
13
Design
NAD+
Pro96
His93
Gln148
Gln150
Profiling
Ki= 9 nM (17�-HSD14)
good selectivity
no cytotoxicity
Ser141
Tyr158
Ala149
no ER� stimulation

ACCEPTED MANUSCRIPT
Structure-Based Design and Profiling of Novel 17β-HSD14 Inhibitors
Florian Braun☼∆, Nicole Bertoletti☼∆, Gabriele Möller‡, Jerzy Adamski‡§$, Martin Frotscher#, Nathalie
Guragossian†,††, Patrícia Alexandra Madeira Gírio††, Marc Le Borgne†, Laurent Etouatti†, Pierre
Falson††, Sebastian Müller≈, Günther Vollmer≈, Andreas Heine☼, Gerhard Klebe☼, Sandrine
Marchais-Oberwinkler☼*
☼ Institute for Pharmaceutical Chemistry, Philipps University Marburg, 35032 Marburg, Germany
‡
German Research Center for Environmental Health, Institute of Experimental Genetics, Genome
Analysis Center, Helmholtz Zentrum München, 85764 Neuherberg, Germany
§
Chair of Experimental Genetics, Technical University Munich, 85350 Freising-Weihenstephan,
Germany
$ German Center for Diabetes Research (DZD), 85764 Neuherberg, Germany
#
Pharmaceutical and Medicinal Chemistry, Saarland University, Campus C23, 66123 Saarbrücken,
Germany
† Bioactive Molecules and Medicinal Chemistry, EA 4446, Faculty of Pharmacy – ISPB, University
Lyon 1, 69373 Lyon, France
†† Drug Resistance & Membrane Proteins laboratory, UMR 5086 CNRS-UCBL1, IBCP, 69367 Lyon,
France
≈ Chair for Molecular Cell Physiology & Endocrinology, Faculty of Biology, Technical University
Dresden, 01062 Dresden, Germany
∆ F.B. and N.B. contributed equally to the work.
* Corresponding author: Phone: +49 6421 28 22880; Fax : +49 6421 28 28994; E-mail:
marchais@staff.uni-marburg.de
1

ACCEPTED MANUSCRIPT
ABSTRACT
The human enzyme 17β-hydroxysteroid dehydrogenase 14 (17β-HSD14) oxidizes the hydroxyl group
at position 17 of estradiol and 5-androstenediol using NAD+ as cofactor. However, the physiological
role of the enzyme remains unclear. We recently described the first class of nonsteroidal inhibitors for
this enzyme with compound 1 showing a high 17β-HSD14 inhibitory activity. Its crystal structure was
used as starting point for a structure-based optimization in this study. The goal was to develop a
promising chemical probe to further investigate the enzyme. The newly designed compounds revealed
mostly very high inhibition of the enzyme and for seven of them the crystal structures of the
corresponding inhibitor-enzyme complexes were resolved. The crystal structures disclosed that a
small change in the substitution pattern of the compounds resulted in an alternative binding mode for
one inhibitor. The profiling of a set of the most potent inhibitors identified 13 (Ki = 9 nM) with a good
selectivity profile toward three 17β-HSDs and the estrogen receptor alpha. This inhibitor displayed no
cytotoxicity, good solubility, and auspicious predicted bioavailability. Overall, 13 is a highly
interesting 17β-HSD14 inhibitor, which might be used as chemical probe for further investigation of
the target enzyme.
Keywords: 17β-hydroxysteroid dehydrogenase type 14; inhibitors; structure-based design;
crystallography
2

ACCEPTED MANUSCRIPT
1. INTRODUCTION
17β-hydroxysteroid dehydrogenase type 14 (17β-HSD14), also known as DHRS10 and retSDR3,
belongs to the Short-chain Dehydrogenase-Reductase (SDR) superfamily [1,2]. In vitro, the human
enzyme catalyzes the oxidation at the 17 position of two prominent steroid hormones, estradiol (E2)
and 5-androstene-3β,17β-diol (5-diol), in presence of the cofactor NAD+, thus converting them into
their less active analogs, estrone (E1) and dehydroepiandrosterone (DHEA), respectively [2].
However, considering the low turnover of the enzyme for E2 and 5-diol (0.024 and 0.017 min-1,
respectively) and the low binding affinity of both substrates (about 6 µM) [3], it is not really clear
whether the enzyme is only involved in steroid regulation or contributes to other pathways in vivo.
The enzyme was described by Sivik et al. [4] to show a broad distribution across various tissues as
evaluated by immunohistochemistry studies, while northern blot analyses demonstrated that the
enzyme is predominantly expressed in brain, liver, placenta and kidney [2,5]. 17β-HSD14 is a
cytosolic enzyme, and its quaternary structure consists of a homotetramer [2,3].
The crystal structures of human 17β-HSD14 in binary complex with NAD+ (PDB codes: 5JS6 and
5JSF) reveal that the enzyme has a rather lipophilic conical active site, which is narrow in the
proximity of the catalytic triad and open to the solvent at the other end [3]. In addition the binding
pocket is reduced in size by the penetration of Tyr253’ from an adjacent enzyme monomer of the
tetrameric assembly.
The first class of nonsteroidal inhibitors for this enzyme was recently described by us [6]. One of the
most promising inhibitors was compound 1 (Figure 1) [6]. It is a member of the phenyl(6-
phenyl)pyridin-2-yl methanone (PPPM) class and exhibits a strong inhibitory effect on the enzyme
s(Ki = 13 nM). The crystal structure of this inhibitor in complex with the enzyme (PDB code: 5L7Y )
shows that the inhibitor adopts a conformation with a complementary shape to the substrate binding
pocket, induced by the carbonyl group between the B and the C ring [6]. The compound is bound to
the active site via different interactions: The most relevant ones are two strong H-bond contacts
between the 3-OH group on the C ring and the catalytic residues Tyr154 and Ser141. Another H-bond
3

ACCEPTED MANUSCRIPT
interaction between the 3-OH group on the A ring and Ala149 is also observed. The pyridine ring of
the inhibitor addresses the hydrophobic pocket formed by Trp192 and Leu195 (Figure 1).
Figure 1. Schematic representation of the binding mode of compound 1 in the binding pocket of
17β-HSD14. Amino acids with polar functional groups are colored in red. Prevalent hydrophobic
amino acids are colored in blue. The different pockets are highlighted with gray lines. H-bonds are
depicted as orange dotted lines.
The goal of this study was to characterize the binding mode of new inhibitors and their interactions
with 17β-HSD14 and to optimize the structure of compound 1 in order to develop a promising
chemical probe that could be used to further investigate the enzyme.
Structure-based optimization of 1 was first undertaken to extensively exploit potential new
interactions provided by the enzyme and thereby enhance the inhibitory activity of the compounds.
Then the newly designed compounds were synthesized, tested for 17β-HSD14 inhibition and for
seven of them the crystal structures of each corresponding inhibitor-enzyme complex were
determined. This structural information was crucial for a better understanding of the structure-activity
relationship (SAR), additionally disclosing a shifted binding mode in case of one inhibitor. The
profiling of the most potent inhibitors was subsequently initiated: Their selectivity toward three
relevant 17β-HSD subtypes and the estrogen receptor alpha (ERα) along with their cytotoxicity and
solubility were determined experimentally and their physicochemical properties were calculated.
4

ACCEPTED MANUSCRIPT
2. RESULTS AND DISCUSSION
2.1 Structure-Based Design of the new Inhibitors
The scaffold of compound 1 was used as a starting point for structure-based modifications with the
intention to identify the key interactions between the individual inhibitors and the protein and further
on to optimize the geometry of the inhibitor 1 for a higher affinity to 17β-HSD14.
Previous studies have shown that in the PPPM class, the interaction of the 3-OH group of the C ring
with Tyr154 from the catalytic triad is essential to achieve high inhibitory activity [6]. The increased
acidity of the 3-OH function, boosted by the substitution with an adjacent fluorine atom (in ortho
position to the OH group), correlates with a significant gain in inhibitory activity. Thus, the 3-OH/4-F
pattern at the C ring was maintained during the subsequent development process.
In the first optimization strategy (I) we focused on the replacement of the linking carbonyl group of 1
(Chart 1), acting as a bridge between the B and C ring. The crystal structure of the protein in complex
with 1 disclosed that the carbonyl group is important to induce the overall V-shape of the inhibitor
scaffold to perfectly match the architecture of the protein’s active site (Figure 2). Interestingly, the
carbonyl group does not form any direct interaction with neighboring amino acids of the protein.
Therefore, the role of this linker (“X”) was investigated by replacing the carbonyl group with polar
linker groups such as an ether oxygen, a secondary amine, or apolar groups such as a tertiary amine
and an ethenyl group. The geometry of the modified scaffold (angle between the B and C rings)
should be kept similar to that of 1 leaving the positions of the B and C rings virtually unchanged in
the protein. Furthermore, amide groups were introduced to enlarge the distance between the B and C
ring and to evaluate the consequences of shifting the pyridine B ring in the substrate binding site.
Chart 1. Optimization strategies (I) and (II) starting from compound 1.
F
F
5
6
F
6
7
B
C
R
N
OH
N
X
OH
N
OH
A
O
O
F
F
OH
OH
(I)
1
(II)
5

ACCEPTED MANUSCRIPT
Additionally, a relatively large hydrophobic pocket, next to the B ring in the direction of the exit
vectors at the 5 and 6 position of the pyridine core, is observed in the crystal structure of 17β-HSD14
with compound 1 (Figure 2). Therefore, another optimization strategy (II) was selected to address this
hydrophobic pocket by extending the pyridine core and fusing the A ring with the B ring resulting in a
quinoline-based scaffold. In addition, the rigidity of the quinoline scaffold compared to that of the
pyridine-based inhibitors [6] was anticipated to be beneficial for inhibitory activity [7]. Furthermore,
the establishment of another scaffold would perhaps give access to compounds with modified
pharmacokinetic properties. Hydrophilic and hydrophobic substituents can be introduced in 6 or 7
position of the quinoline moiety, which would enable to explore the chemical space in this region of
the binding pocket (Chart 1).
Figure 2. Close-up view of the binding pocket as formed in the crystal structure of 17β-HSD14
obtained as ternary complex with 1 and NAD+ (PDB code: 5L7Y) [6]. The protein is displayed in gray
by use of the solvent-excluded surface. The inhibitor is shown as stick model. Carbon atoms of the
ligand 1 are shown in yellow. The relatively large hydrophobic pocket adjacent to the B ring is
highlighted in green. All structural representations were prepared with PyMOL [8].
6

ACCEPTED MANUSCRIPT
2.2 Pan Assay Interference Compounds
In order to identify undesired compounds, which might bind nonspecifically to numerous biological
targets, all designed compounds (2-17, Tables 1-3) were validated in silico using the PAINS-remover
computer tool [9,10]. As none of them was assigned as unspecific binder, all were considered for
synthesis.
2.3 Physicochemical Parameters
Prior to synthesis, the physicochemical properties of all designed compounds (2-17, Tables 1-3) were
computed to only consider compounds, which match with a good predicted bioavailability profile.
The molecular weight was selected to fall into a range of 250-400 g/mol. The clogP, the total polar
surface area (tPSA), the number of rotational bonds and H-bond donors and acceptors were calculated
in silico and adhered to the Veber Rules [11] and the Lipinski’s rule of five [12]. The potential ability
of the designed compounds to cross the blood brain barrier (BBB) was also taken into account, as
17β-HSD14 has been reported to be present in the brain [2]. Especially for the pyridine derivatives 1-
5 as well as for the quinolines 9, 11, 13, and 17 the physicochemical properties fit well with the
profile defined by Pajouhesh and Lenz [13] for compounds showing a good BBB penetration (i.e.
MW < 450 g/mol; H-bond donor < 3; H-bond acceptor < 7; number of rotatable bonds < 8; pKa 7.5-
10.5; tPSA < 60-70 Å2). The calculated physicochemical parameters and experimentally determined
solubilities are summarized in Table S1 (Supporting Information, SI).
2.4 Chemistry
The synthesis of compounds 2-8 was achieved starting from the brominated pyridine derivatives 2c,
6c, and 8c (Scheme 1). The reaction of the 2,6-dibromopyridine 2c with the phenol derivative 2d
provided 2b using sodium hydride as base, whereas the reaction with the aniline derivative 3c
afforded the secondary amine 3b in the presence of lithium hexamethyldisilazane (LHMDS).
Activation of carboxylic acids 6d and 8c with thionyl chloride and subsequent reaction with the
appropriate aniline derivative (6c or 3c) resulted in the amides 6b and 8b, respectively. Suzuki
coupling reactions[14] of 2b, 3b, 6b, and 8b with 2-fluoro-3-methoxyphenylboronic acid gave the
7

ACCEPTED MANUSCRIPT
methoxy intermediates 2a, 3a, 6a, and 8a. Wittig olefination [15] of 1a [6] with
methyltriphenylphosphonium bromide afforded the ethenyl derivative 5a. Methylation of 3a and 6a
with methyl iodide using sodium hydride as base provided additionally the methoxy intermediates 4a
and 8a. The methoxy groups were cleaved with boron tribromide (for compounds 2-4 and 6-8) or with
boron trifluoride dimethylsulfide complex (for compound 5).
Scheme 1: Synthesis of compounds 2-8.a
F
OMe
F
OMe
a
or
b
or
c
d
+
Br
N
X
Br
N
X
Y
X = Br
6c X = NH2
Y = OH
X = O
2b
3b X = NH
2c
2d
3c Y = NH2
Y = CO H
6d
X = NHCO
6b
X = CO H
8c
2
2
8b X = CONH
F
F
g
or
h
N
X
OMe
N
X
OH
F
F
OMe
OH
X = O
X = O
2a
2
3
4
5
6
7
8
3a X = NH
X = NMe
X = NH
X = NMe
X = CCH2
f
f
4a
5a X = CCH2
X = NHCO
e
X = CO
1a
X = NHCO
X = NMeCO
X = CONH
6a
7a X = NMeCO
X = CONH
8a
a
Reagents and conditions: a) for 2b: 2c, 2d, NaH, DMF, 100 °C, overnight; b) for 3b: 2c, 3c,
LHMDS, 1,4-dioxane/THF 1:1, room temperature, 3 h; c) for 6b (6c + 6d), 8b (8c + 3c) i) SOCl2,
DMF, toluene, reflux, 4 h, ii) Et3N, DCM, 0 °C to room temperature, overnight; d) Cs2CO3,
Pd(PPh3)4, 2-fluoro-3-methoxyphenylboronic acid, DME/water (2:1), 80 °C, overnight;
e) MePh3P+Br−, n-BuLi, THF, 0 °C to room temperature, overnight; f) NaH, MeI, DMF, 0 °C to room
temperature, overnight; g) for 2-4 and 6-8, BBr3, DCM, -80 °C to room temperature, overnight; h) for
5, BF3·SMe2, DCM, room temperature, overnight.
Compounds 9-12 were synthesized in three steps starting from the bromoquinolines 9c, 11c, and 12c
and 2-bromonaphthalene (10c), respectively (Scheme 2). Lithiation with n-butyl lithium and
nucleophilic addition to 4-fluoro-3-methoxybenzaldehyde (9d) provided after oxidation with
8

ACCEPTED MANUSCRIPT
2-iodoxybenzoic acid the ketone intermediates 9a-12a. Compounds 11c and 12c were synthesized by
regioselective bromination of the appropriate quinoline N-oxides [16]. For compounds 9-11, the
methoxy groups were cleaved with boron tribromide, whereas for 12 boron trifluoride dimethylsulfide
complex was used.
Scheme 2: Synthesis of compounds 9-12.a
F
R
1
a
,
b
+
OMe
X
Br
O
R1 = H, X = N
9c
9d
R = H, X = CH
10c
11c
12c
1
R = Me, X = N
1
R = OMe, X = N
1
R
1
F
R
F
1
c or d
X
OMe
X
OH
O
O
R1 = H, X = N
R1 = H, X = N
9a
9
R = H, X = CH
R = H, X = CH
10a
11a
12a
10
11
12
1
1
R = Me, X = N
R = Me, X = N
1
1
R = OMe, X = N
R = OH, X = N
1
1
a
Reagents and conditions: a) n-BuLi, anhydrous THF, -80 °C to room temperature, 2 h; b)
2-iodoxybenzoic acid, anhydrous THF, 60 °C, 2 h; c) for 9-11, BBr3, DCM, -80 °C to room
temperature, overnight; d) for 12, BF3·SMe2, DCM, room temperature, overnight.
The synthesis of compounds 13-17 was enabled starting from precursor 13b (Scheme 3). For the
synthesis of 13b, 4-bromo-1-fluoro-2-methoxybenzene (13e) was used to prepare the corresponding
Grignard reagent. The latter was reacted with compound 13d[17] in a nucleophilic addition and
afforded, after oxidation of the formed alcohol with 2-iodoxybenzoic acid, precursor 13b. The
reaction of 13b with Zn(CN)2 using Pd(PPh3)4 as catalyst afforded the nitrile 13a which was
subsequently used to prepare the corresponding tetrazole derivative 14a in a microwave-assisted
reaction [18]. Amination of 13b was performed using similar conditions to the one described by Zeng
et al. [19] to afford compound 15a. Subsequent amide formation using Boc-glycine and HOBt/
EDC·HCl as coupling reagents provided compound 16a. Suzuki coupling of 13b with 4-methoxy-3-
9

ACCEPTED MANUSCRIPT
methylphenylboronic acid gave the methoxy intermediate 17a. All methoxy groups were cleaved with
boron tribromide to afford compounds 13-15 and 17. Compound 16a, after ether cleavage, was
additionally treated with trifluoroacetic acid to afford 16 as trifluoroacetate salt.
Scheme 3: Synthesis of compounds 13-17.a
F
OMe
F
OMe
a b
,
+
Br
N
Br
N
Br
O
O
13d
13e
13b
F
F
e
N
N
R
N
NC
N
R
NH
O
N
O
R = OMe
R = OH
R = OMe
R = OH
13a
13
14a
14
d
d
c
F
F
f
g
13b
H N
2
N
R
HN
N
R
2
O
O
O
NHR
1
R = OMe
R = OH
R = Boc, R = OMe
1 2
15a
15
16a
16
d
j
d,h
R1 = H, R2 = OH
F
Me
R
N
R
O
R = OMe
R = OH
17a
17
d
a Reagents and conditions: a) Mg, anhydrous THF, 60 °C, 2 h, 80 °C, 4 h; b) 2-iodoxybenzoic acid,
anhydrous THF, 60 °C, overnight; c) Zn(CN)2, Pd(PPh3)4, DMF, 120 °C, 3 h; d) BBr3, DCM, -80 °C
to room temperature, overnight; e) NaN3, NH4Cl, DMF, microwave irradiation, 30 W, reflux (open
vessel), 30 min; f) CuI, NH4OH, L-Proline, K2CO3, DMSO, 85 °C, 15 h; g) Boc-Gly-OH, HOBt,
EDC·HCl, DIPEA, DCM, 40 °C; h) TFA, DCM, 1 h, room temperature; j) Cs2CO3, Pd(PPh3)4, 4-
methoxy-3-methylphenylboronic acid, DME/water (2:1), 80 °C, overnight.
10

ACCEPTED MANUSCRIPT
2.5 Biological Evaluation of 17β-HSD14 Inhibitory Activity
The inhibitory activities of the synthesized compounds were evaluated by means of a
fluorescence-based assay, quantifying the fluorescence of NADH produced during the catalytic
reaction. The assay was carried out as already reported [3,6], using the purified, recombinantly
expressed enzyme, E2 as substrate, NAD+ as cofactor, and the inhibitors. Percentages of inhibition
(measured at an inhibitor concentration of 2 µM) as well as binding affinities (Ki) were determined.
For 9a, no Ki could be measured due to its limited solubility. Thus, its inhibitory activity was
expressed as percentage of inhibition at the highest soluble concentration of the inhibitor (100 µM).
Starting from inhibitor 1, the linker between the B and the C ring was investigated following our first
optimization strategy concept by replacing the carbonyl group with different functional groups. Ether
2 (Ki = 58 nM), amine 3 (Ki = 47 nM), and methylamine 4 (Ki = 42 nM) were shown to be equipotent
ligands with slightly decreased affinity compared to 1 (Ki = 13 nM). The introduction of an ethenyl
group resulted in a ten times more potent inhibitor 5 (Ki = 1.5 nM), whereas the presence of an amide
linker led to a strong decrease in inhibitory activity (6: Ki = 686 nM, 8: Ki = 336 nM) and the
methylated amide (7: Ki = 2030 nM) was even detrimental for 17β-HSD14 inhibition compared to 1
(Table 1).
11

ACCEPTED MANUSCRIPT
Table 1. 17β-HSD14 inhibition constant (Ki) of pyridine derivatives with different linkers between
the B and C ring.
F
C
B
N
X
OH
A
F
OH
1-8
17β-HSD14
Ki (nM)a
13 ± 5
compd
X
C=O
O
1
2
3
4
5
6
7
8
58 ± 12
NH
NMe
47 ± 7
42 ± 8
C=CH2
NHCO
NMeCO
CONH
1.5 ± 0.4
686 ± 54
2030 ± 180
336 ± 54
a
Recombinantly expressed, purified 17β-HSD14 enzyme, fluorimetric assay, [E2] = 32 µM,
[NAD+] = 1.2 mM, 25 °C. Mean and SD values of at least three independent experiments each with
three technical repeats.
In the second approach the pyridine B ring of compound 1 (Ki = 13 nM) was expanded and led to the
equipotent quinoline-based inhibitor 9 (Ki = 12 nM, Table 2). The presence of the 3-OMe group at the
C ring (9a, no inhibition at 2 µM; 21% inhibition at 100 µM) was detrimental for the inhibitory
activity when compared to the 3-OH analogue 9 (69% inhibition at 2 µM). The role of the nitrogen
atom of the quinoline moiety with respect to the inhibitory activity was further investigated by the
synthesis of the naphthalene derivative 10 (Ki = 6 nM), which was shown to have a similar inhibition
constant as the quinoline analog 9. The nitrogen might not establish any specific interaction with the
12

ACCEPTED MANUSCRIPT
protein, however, it enhances the solubility of the 9 over its naphthalene analogue 10 (~ four times,
Table S1, SI).
Table 2. 17β-HSD14 inhibition constant (Ki) of quinoline and naphthalene derivatives.
F
F
C
B
N
A
B
X
C
OH
A
R
O
F
O
OH
1
9a, 9, 10
17β-HSD14
Ki (nM)a
13 ± 5
Compd
X
R
-
-
1
N
N
OMe 21%b
12 ± 3, (69%)c
9a
9
OH
CH OH
6 ± 2
10
a
Recombinantly expressed, purified 17β-HSD14 enzyme, fluorimetric assay, [E2] = 32 µM,
[NAD+] = 1.2 mM, 25 °C. Mean and SD values of at least three independent experiments each with
b
c
three technical repeats. percent inhibition at 100 µM, standard deviation <10%. percent inhibition
at 2 µM, standard deviation <10%.
Furthermore, compound 9 (Ki = 12 nM) was decorated with different substituents in 6 or 7 position of
the quinoline moiety. A methyl group in 6 position (11: Ki = 6 nM) did not influence the inhibition,
whereas a hydrophilic OH-group (compound 12, Ki = 119 nM) reduced the potency ten times
compared to that of 9. In 7 position, the introduction of non-hydrophilic substituents (13: Ki = 9 nM;
17: Ki = 6 nM) as well as polar substituents (14: Ki = 10 nM; 15: Ki = 34 nM; 16: Ki = 23 nM) resulted
in inhibitors with similar potencies compared to 9 (Table 3).
13

ACCEPTED MANUSCRIPT
Table 3. 17β-HSD14 inhibition constant (Ki) of 6 and 7-substituted quinoline derivatives.
F
6
A
B
C
R
7
N
OH
O
9-17
17β-HSD14
Ki (nM)a
12 ± 3
compd
R
H
9
6-Me
6-OH
7-CN
6 ± 2
11
12
13
14
15
16
17
119 ± 7
9 ± 1
7-(tetrazol-5-yl)
7-NH2
10 ± 2
34 ± 7
7-NHGly
23 ± 2
7-(4-hydroxy-3-methylphenyl) 6 ± 2
a
Recombinantly expressed, purified 17β-HSD14 enzyme, fluorimetric assay, [E2] = 32 µM,
[NAD+] = 1.2 mM, 25 °C. Mean and SD values of at least three independent experiments each with
three technical repeats.
2.6 Crystal Structures Determination
Inhibitors of both, the pyridine and the quinoline series, showing the best inhibitory activity for
17β-HSD14, were selected for crystallization. Seven inhibitors could be co-crystallized in complex
with the protein resulting in crystals of sufficient quality for structure determination (2, PDB code:
5O6O; 4, PDB code: 5O43; 5, PDB code: 5O42; 9, PDB code: 5O6Z; 11, PDB code: 5O6X; 13, PDB
code: 5O7C and 16, PDB code: 5O72). All crystals exhibited tetragonal symmetry in space group
(I422) with the asymmetric unit formed by only one monomer, which assembled with the other three
symmetry equivalent molecules to form the biological relevant homotetramer. This was in accordance
with the findings of our previous study [3]. The structures obtained have a resolution ranging between
14

ACCEPTED MANUSCRIPT
1.35 Å to 1.91 Å. Data collection, processing and refinement statistics can be found in the Table S2A
and S2B (SI).
2.7 Binding Mode of the Pyridine-Based Inhibitors in Complex with 17β-HSD14 and SAR study
The superimposition of the crystal structures of the enzyme in ternary complexes with the pyridine-
based inhibitors 2, 4 and 5 revealed that these three inhibitors, despite a chemically varied linker
between the C and the B ring, bind in a similar mode as 1 (Figure 3A). All inhibitors, in spite of their
different linkers, adopt the same V-shaped conformation as observed for 1. For all three crystallized
inhibitors, the same conserved water molecule (W1) is found at the same position as described for the
other known pyridine-based inhibitors [6]. As expected, the ether oxygen (2, Figure 3B) is not
involved in any direct interaction with the protein. The methyl amine (4, Figure 3C) and the ethylene
moiety (5, Figure 3D) accommodate well in the hydrophobic pocket formed by Leu191, Trp192 and
Leu195 (amino acids shown in Figure 1). The enhanced flexibility of the ligand with the ether linker
is reflected in the split conformation of the A and B ring of 2 (Figure 3B). Remarkably, the slightly
different valence angle between the B and C ring (~ 1°) generated by the ethylene linker when
compared to the respective angle in 1 pushes the B ring of 5 into the direction of Leu195 leading to a
flip of the A ring plane of about 180° and resulting in an orientation of the OH and F substituents in
opposite direction to that in compound 1 (Figure 3A). The 3-OH group at the A ring of all the
inhibitors is within H-bond distance to the carbonyl backbone of Ala149 (d = 2.7-2.9 Å).
15

ACCEPTED MANUSCRIPT
Figure 3. Close-up view of the binding pocket of the four crystal structures of 17β-HSD14 obtained
as ternary complexes with NAD+ and the pyridine-based inhibitors: 1 (PDB code: 5L7Y, A) [6], 2
(PDB code: 5O6O, A and B), 4 (PDB code: 5O43, A and C), and 5 (PDB code: 5O42, A and D). The
carbon atoms of the ligands are shown for 1 in yellow, 2 in orange (two conformers), 4 in blue, 5 in
purple and NAD+ in light beige. (A) Superimposition of the four crystal structures. The protein is
displayed in gray by use of the solvent-excluded surface. Inhibitors and cofactor are shown as stick
models. The individual sites of the water molecule W1 are represented in the same color as the
corresponding inhibitor in the various structures. W1 corresponds to water molecule 508 in 5L7Y,
472 in 5O43, 488 in 5O42 and 501 in 5O6O (hidden behind the other water molecules) in the
respective crystal structures. In panels B, C and D, the amino acids within a distance of 4.6 Å are
16

ACCEPTED MANUSCRIPT
shown as thin lines with the carbon atoms in white. H-bond interactions are depicted as dotted lines.
Distances are given in Å.
Replacement of the carbonyl moiety (1) for the ethylene group (5) led to an almost ten-fold more
potent inhibitory compound 5 (Ki = 1.5 nM) compared to 1 (Ki = 13 nM). The crystal structure of 5
revealed that the ethylene linker is more deeply buried in the hydrophobic pocket defined by Leu191,
Leu195 and Trp192 compared to the carbonyl group. This linker is also beneficial for binding due to
its lower desolvation penalty compared to the carbonyl analog 1 whose polar oxygen is not involved
in any interaction at the protein binding site. The flipped A ring of 5 (compared to 1) led to an
improved geometry because this ring comes into proximity to the hydrophobic region formed by
Pro95, Pro96 and Leu195 (van der Waals interactions). In addition, the 3-OH group attached to the A
ring is moved to a more favorable H-bond distance (d = 2.7 Å) toward the carbonyl group of the
Ala149 backbone. All together these aspects might be responsible for the gain in 17β-HSD14
inhibitory activity of 5 compared to 1. Compounds 6, 7, and 8 lose affinity possibly owing to the
introduced polar amide linkers, which remain exposed to a predominantly hydrophobic environment.
In addition, the A rings of these inhibitors become partly solvent exposed, most probably due to the
fact that the compounds are rather long for the binding site, resulting in the loss of interactions of the
A ring with the protein. Furthermore, the important V-shape of these inhibitors 6, 7, and 8 could not
be achieved.
2.8 Binding Mode of the Quinoline-Based Inhibitors in Complex with 17β-HSD14 and SAR
study
From the crystal structures of 9, 13 and 16 (Figure 4), it became apparent that all these ligands
adopted the same binding mode. In addition, the C rings of the three quinoline-based inhibitors 9, 13,
and 16 are located in the same region compared to the C ring of pyridine-based inhibitors described so
far: The 3-OH group of the C ring experiences H-bond interactions with the catalytic Tyr154 (d = 2.4-
2.5 Å) and Ser141 (d = 2.4-2.5 Å) (Figure 4A, 4B and Figure 5). The quinoline core addresses the
hydrophobic region formed by Pro95, Pro96, Leu191, Trp192 and Leu195 (amino acids shown in
17

ACCEPTED MANUSCRIPT
Figure 1). The primary amine substituent of 16 is in H-bond distance with the carbonyl group of the
side chain of Gln150 (d = 2.9 Å), while no interaction with the protein can be observed for the nitrile
substituent of 13. These three inhibitors showed high inhibition of the enzyme (Ki = 12 nM, 9 nM and
23 nM for 9, 13 16, respectively) in the same range as the pyridine derivatives. The additional
hydrophobic interactions of the quinoline core motif with the pocket formed by Leu191, Trp192,
Leu195 and Met199 may have compensated the loss of the A ring and the H-bond interactions with
Ala149, which were observed for compound 1 (e.g. 9: Ki = 12 nM vs. 1: Ki = 13 nM). The conserved
water molecule W1 is present in a similar position as described for the pyridine-based inhibitors.
Figure 4. Close-up view of the binding pocket of the superimposed crystal structures of 17β-HSD14
obtained in ternary complexes with NAD+ and different inhibitors. The protein is displayed in gray by
use of the solvent-excluded surface. Inhibitors and cofactor are shown as stick models. The carbon
atoms of the NAD+ are shown in light beige. The water molecules W1-W5 are represented in the same
color as the corresponding inhibitor of the individual structures. (A) The carbon atoms of the ligands
are shown for 1 (PDB code: 5L7Y) [6] in yellow and 9 (PDB code: 5O6Z) in blue. (B) The carbon
atoms of the ligands are shown for 9 in blue, 13 (PDB code: 5O7C) in green and 16 (PDB code:
5O72) in light orange. (C) The carbon atoms of the ligands are shown for 9 in blue and 11 (PDB code:
5O6X) in red.
18

ACCEPTED MANUSCRIPT
Figure 5. Close-up view of the binding pockets of the crystal structures of 17β-HSD14 in complex
with cofactor NAD+ and inhibitors. The carbon atoms are shown in blue for 9 (A, PDB code: 5O6Z),
green for 13 (B, PDB code: 5O7C), light orange for 16 (C, PDB code: 5O72) and red for 11 (D, PDB
code: 5O6X). The inhibitors and cofactor NAD+ are shown as stick models. The amino acids are
shown as thin lines with carbon atoms in white. The water molecule W6 is represented as red sphere.
W6 corresponds to water molecule 419 in 5O6X. H-bonds interactions are depicted as dotted lines.
Distances are given in Å.
19

ACCEPTED MANUSCRIPT
As already remarked in our previous work [6], replacing the 3-OH group at the C ring by a 3-OMe
moiety resulted in a strong drop of the inhibitory activity (9a, 21% inhibition at 100 µM), certainly
due to the missing H-bond interaction with the catalytic Tyr154.
No crystal structure could be obtained for inhibitors 14, 15, 12 and 17 in complex with 17β-HSD14
and NAD+. A model structure can provide insight on the interactions achieved by a ligand with the
protein and therefore can sustain the understanding of SAR studies. First to validate the method, the
quinoline-based inhibitor 9 was modeled into the crystal structure of 1. The simulation suggested a
binding mode for 9, which was identical to the co-crystal structure (Figure S3, SI), validating the
model structure. A model structure was created for 14, 15, and 17. From their simulations, these
ligands are not expected to perform any additional interaction compared to 9 (Figures S4A-D, SI).
The 6-OH group of 12 is located in a lipophilic environment, which is unfavorable. The predicted
interactions made by the ligands 12, 14, 15, and 17, as observed in the different models, correlate well
with their respective inhibition constant (Ki = 119 nM, 10 nM, 34 nM, , 6 nM, and 12 nM for 12, 14,
15, 17, and 9 respectively).
The crystal structure of compound 11 displays a deviating binding mode for this ligand when
compared to the other inhibitors described previously (Figure 4C). The entire compound is pushed
deeper into the active site and induces an altered conformation of the sidechain of Tyr253’ of the
neighboring monomer of the tetramer. The C ring of 11 is stacked behind Tyr253’. The 3-OH group
of the C ring remains within H-bond distance to the side chain OH group of the catalytic Ser141
(d = 2.4 Å). Furthermore, the 3-OH group of the C ring is also in H-bond distance with the carbonyl
group of the backbone of Pro184 (d = 3.4 Å). However, the important and direct H-bond contact
between this 3-OH group and the side chain of Tyr154, found in all other structures of 17β-HSD14 in
complex with inhibitors, is not present anymore. Instead, the carbonyl group of the ligand interacts
with Tyr154 mediated via the water molecule W6 (d= 2.7 and 2.8 Å). The 6-methylquinoline moiety
is completely buried in the hydrophobic pocket composed of Pro95, Pro96, Leu191, Trp192 and
Leu195 and, in addition, the quinoline nitrogen is within H-bond distance to the sidechain of Gln148
(d = 3.1 Å, Figure 5D). This deviating binding mode might result from steric repulsion of the added
20

ACCEPTED MANUSCRIPT
methyl group in position 6 with Met199 of the protein, which pushes the ligand deeper into the active
site of the enzyme. In the position described, the methylated quinoline moiety is able to form
optimum hydrophobic interactions with the pocket formed by Leu191, Tpr192, Leu195, Met199 and
Thr205 (Figure 5). As the deviating binding mode of the 6-methylquinoline derivative 11 did not
parallel a significant change in inhibition strength compared to 9 (Ki = 6 nM vs. Ki = 13 nM,
respectively), the cumulative hydrophobic effects and the new interaction with water W6 seem to
compensate the loss in the direct interaction to Tyr154.
Interestingly, the water molecule W6 is entrapped to mediate a contact between ligand and protein.
This water is also found at the same position in the 17β-HSD14-NAD+ complex structures with 11
and 13 and in the 17β-HSD14 holoenzyme complex with NAD+ (PDB code: 5JSF (S205 variant) and
5JS6 (T205 variant), Figure 7).
Figure 7. Superimposition of the crystal structures of the 17β-HSD14 in complex with the cofactor
NAD+ (PDB code: 5JSF (S205 variant) and 5JS6 (T205 variant)) and 17β-HSD14 in complex with
NAD+ and the inhibitors 11 (PDB code: 5O6X) and 13 (PDB code: 5O7C). The carbon atoms of 13
are colored in green and those of 11 in red. The inhibitors and cofactor NAD+ are shown as stick
models. The amino acids are shown as thin lines with carbon atoms in white. H-bonds are depicted as
dotted lines. The water molecule W6 is represented in red for 5O6X, in pink for 5JS6, in green for
5O7C and in blue for 5JSF, and corresponds to water molecule number 419, 424, 403 and 439,
21

ACCEPTED MANUSCRIPT
respectively, in the different crystal structures. The latter two water molecules are hidden by the other
two waters in the superimposition of the four structures.
In the case of the protein complex with 11, the water molecule W1 is shifted upward due to the
different conformation of this inhibitor (Figure 4C). To be remarked, three additional conserved water
molecules (W2-W4) can be observed in the structures of 9, 13, 16, and 11, located in the pocket
usually occupied by the A ring of the pyridine-based inhibitors (Figure 4). In case of the complex with
16, one water molecule (W3) is displaced by the amino group of the glycine moiety of the inhibitor.
Compounds 9 and 11 differ only in the presence/absence of a methyl group at position 6 of the
quinoline ring. Their crystal structures in complex with the protein reveal that these two compounds
are stabilized by the protein via different interactions. This result was unexpected and highlights how
difficult it is to predict the binding mode of a compound using docking studies in case of highly
flexible proteins.
2.9 Comparison of the Structures of 17β-HSD14 and 17β-HSD10
17β-HSD10 and 17β-HSD14 share the highest sequence identity between the protein members of the
17β-HSD family (27.5 % as calculated with Cluster omega program from the UniProt database [20]).
Therefore, the selectivity aspect towards 17β-HSD10 should be taken into account for the designed
17β-HSD14 inhibitors.
The crystal structure of the 17β-HSD10 (PDB code: 1U7T [21]) has been reported in complex with
the inhibitor AG18051 [21], which is covalently linked to the cofactor NAD+ (NAD-
AG18051 adduct) and can be compared to the 17β-HSD14 in complex with the pyridine-based
inhibitor 1 (PDB code: 5L7Y [6], Figure 6).
Structural comparison of the Rossman-fold region between the two enzymes reveals that the cofactor
is found with a nearly identical geometry in the pocket, with a RMSD value of 0.6 Å after alignment
of the Cα of the protein (calculated with fconv [22]). The structure of the substrate binding site
deviates among both enzymes, although the position of the catalytic triad is preserved. The flexible
loop of 17β-HSD14 (αFG1 and αFG2 segments) is slightly longer than the one corresponding to
22

ACCEPTED MANUSCRIPT
17β-HSD10. In addition, 17β-HSD10 bears a solvent-exposed loop (LD loop, residues 94–114,
Figure 6), which is essential for the interaction of 17β-HSD10 with β-amyloid-protein (Aβ) [23] and
does not exist in 17β-HSD14. In both 17β-HSD10 and 17β-HSD14, the C-terminal tail is
unstructured. However, in the case of 17β-HSD14, the C-terminal tail is involved in the interactions
between two monomers. Moreover, His93 and Gln148 induce the V-shaping of the substrate binding
site of 17β-HSD14. For type 10, His93 is replaced by Ala95 and as consequence Gln162
(corresponding to Gln148 for 17β-HSD14) is flipped to the outer side of the binding pocket while the
side chain of Gln165 in 17β-HSD10 is oriented toward the center of the binding pocket. Taken
together, these facts result in a shallower binding pocket for 17β-HSD10 compared to the one of type
14. The inhibitors 1 and AG18051 occupy a similar region in type 14 and 10, respectively. However,
the inhibitor for 17β-HSD10 adopts a rather planar geometry, reinforced by the hydrophobic residues
Leu206 and Leu209 (Figure 6).
23

ACCEPTED MANUSCRIPT
Figure 6. Superimposition of the 17β-HSD14 (yellow, PDB code: 5L7Y) and 17β-HSD10 (orange,
PDB code: 1U7T) structures. The proteins are shown as ribbon models. The ligands and cofactors are
shown as stick models and their carbon atoms are colored as the corresponding protein structure. (A)
Overview of the complexes structures. (B) Close-up view on the flexible loops of 17β-HSD10 and
17β-HSD14 (αFG1 and αFG2). (C) Close-up view on the ligand binding pockets. (D) Chemical
structures of inhibitor 1 and the NAD-AG18051 adduct.
2.10 Selectivity toward 17β-HSD1, 17β-HSD2 and 17β-HSD10
The reference compound 1, used as the starting point of this study, was retrieved from a library of
known 17β-HSD1 and 17β-HSD2 inhibitors [24,25]. It was therefore important to validate whether
the preference for binding to 17β-HSD1 and 17β-HSD2 of the new inhibitors was maintained or
whether the optimized compounds showed good selectivity profiles in favor of 17β-HSD14. In
addition, as mentioned previously, the investigation of the selectivity profile with respect to binding to
17β-HSD10 is of high concern considering the high sequence identity between 17β-HSD 10 and
17β-HSD 14 enzymes.
We selected inhibitors that displayed good inhibitory activity for 17β-HSD14 (Ki < 100 nM) to
validate their selectivity profile toward 17β-HSD1, 17β-HSD2, and 17β-HSD10. The ability of the
selected 14 inhibitors to inhibit these enzymes was studied using a competitive assay, by quantifying
the formation of [3H]-E2 (for 17β-HSD1) and [3H]-E1 (for 17β-HSD2 and 10) after incubation of the
protein with labeled substrate, cofactor, and inhibitors [26]. The results of these assays are shown in
Table 4 and expressed as percentage of inhibition.
24

ACCEPTED MANUSCRIPT
Table 4. 17β-HSD1, 17β-HSD2, and 17β-HSD10 inhibitory activities of the most interesting
compounds.
17β-HSD1
17β-HSD2
17β-HSD10
compd
% inh. at 1 µMa
% inh. at 1 µMb
% inh. at 1 µMc
8
56
n.i.
1
2
3
4
5
9
69
57
99
88
6
100
82
17
n.i.
19
100
100
76
n.i.
n.i.
n.i.
n.i.
n.i.
n.i.
n.i.
n.i.
n.i.
n.i.
64
12
20
n.i.
n.i.
21
3
96
10
11
12
13
14
15
16
17
78
71
56
52
83
63
11
80
a
Placental 17β-HSD1 enzyme, cytosolic fraction, substrate [3H-E1 + E1] = 500 nM,
[NADH] = 0.5 mM, mean value of two determinations; standard deviation < 10%, n.i.: no inhibition.
b
Placental 17β-HSD2 enzyme, microsomal fraction, substrate [3H-E2 + E2] = 500 nM,
c
[NAD+] = 1.5 mM, mean value of two determinations; standard deviation < 10%. Recombinantly
expressed 17β-HSD10 enzyme, bacterial suspension, substrate [3H-E2] = 25 nM, [NAD+] = 0.75 mM,
mean value of three determinations; standard deviation < 10%, n.i.: no inhibition.
25

ACCEPTED MANUSCRIPT
2.10.1 Selectivity of the Pyridine Derivatives
The first approach to modify the scaffold of the pyridine derivatives resulted in the production of less
selective compounds (2-5), which showed even higher inhibition of 17β-HSD1 (57-99% 17β-HSD1
inhibition at 1 µM) and of 17β-HSD2 (82-100% 17β-HSD2 inhibition at 1 µM) compared to the
initial compound 1 (8% 17β-HSD1 inhibition and 56% 17β-HSD2 inhibition, both measurements at 1
µM).
The ethylene linker derivative 5 is highly active on 17β-HSD1 (88% at 1 µM) and 17β-HSD14
Ki = 1.5 nM) while the carbonyl linker derivative 1 is inactive on 17β-HSD1 (8% at 1 µM) but highly
active on 17β-HSD14 (Ki = 13 nM). It is striking that a minimal change in the structure of the
inhibitor leads to such a difference in 17β-HSD1 inhibitory activity and has no influence on
17β-HSD14 activity taking into account that both enzymes can bind the same substrate/product. A
rational explanation for the observed selectivity profile of the inhibitors toward 17β-HSD1 and
17β-HSD2 in structural terms appears rather difficult to establish. Due to the conformational
properties of the 17β-HSD1 binding pocket - with the flexible loop restricting the remote part opposite
to the catalytic triad, and the existence of several conformations of the flexible loop [6]- it is
extremely challenging to perform reliable docking studies. However, it can be imagined that the
pyridine derivatives do not bind in the same way to 17β-HSD1 and to 17β-HSD14, leading to
differences in the inhibitory activities. The lack of a crystal structure of 17β-HSD2 does not allow a
detailed conclusive comparison.
With respect to 17β-HSD10, only compounds 2 and 4 showed a weak inhibitory activity (17% and
19% inhibition at 1 µM, respectively) while all the other synthesized derivatives were inactive
suggesting an overall very convincing selectivity profile toward 17β-HSD10 (Table 4). This
selectivity might be explained by the rather flat geometry of the active site of 17β-HSD10, which
clashes with our V-shaped inhibitors.
Taking together the activity and selectivity profile of the pyridine derivatives, it appeared reasonable
to focus further investigations on the quinoline class with the methanone linker.
26

ACCEPTED MANUSCRIPT
2.10.2 Selectivity of the Quinoline Derivatives
The quinoline-based compounds 9-17, obtained from the second optimization strategy, resulted in
inhibitors which showed no significant 17β-HSD1 inhibitory activity at 1 µM. Only the naphthalene
derivative 10 fell out of this profile and exhibited higher 17β-HSD1 inhibitory activity (64%
inhibition at 1 µM).
A direct comparison of the 17β-HSD1 and 17β-HSD2 inhibitory activities with those of 17β-HSD14
is problematic as the assays were performed using different conditions. However in our previous
study, an 17β-HSD2 IC50 of 1 µM was assessed for compound 1 [6]. Taking into account the kinetic
constant of the assay (Km = 403 nM [27], substrate concentration = 500 nM), a calculated Ki (cKi) for
1 can be estimated to be about 450 nM using the Cheng-Prusoff equation [28]. The determined Ki for
1 in the 17β-HSD14 assay was 13 nM. Compound 1 is therefore about 35-fold more selective for
17β-HSD14 compared to 17β-HSD2. As compounds 13, 14, and 16 exhibited a 17β-HSD2 inhibition
in the same range as 1 (52-63% inhibition at 1 µM), it can be expected that they will be as selective as
1 toward 17β-HSD2, thus underlining the satisfactory selectivity discrimination of the quinoline-
based inhibitors against 17β-HSD1 and 17β-HSD2.
The affinity of quinoline-based inhibitors toward 17β-HSD10 was minor. Similar to the pyridine
derivatives, the selectivity toward 17β-HSD10 might be explained by the quite flat active site
geometry of the enzyme. It certainly does not accept kinked inhibitors, even if the angular shape of
the quinolines-based compound is less pronounced than the V-shape of the pyridines-based analogues.
Taken all together, inhibitors 13, 14, and 16 show a good selectivity profile toward 17β-HSD1,
17β-HSD2 and 17β-HSD10 exhibiting signatures comparable to that of the pyridine-based 1.
2.11 Activation of the estrogen receptor alpha (ERα)
The synthesized compounds are structurally derived from 1, which is a steroidomimetic (mimic of
estradiol). Estradiol is known to induce cell proliferation upon binding to the ERα in the estrogen
sensitive organs like the breast or the endometrium. It should therefore be verified if the newly
synthesized compounds might show undesired effects by activating the ERα, which would complicate
the interpretation of an in vivo experiment studying the physiological role of 17β-HSD14. The
27

ACCEPTED MANUSCRIPT
molecule should therefore not behave as an agonist of the ERα. Based on their 17β-HSD14 inhibitory
activity and selectivity profile toward 17β-HSD1, 17β-HSD2, and 17β-HSD10, compounds 1, 13, 14,
and 16 were selected to evaluate their potential agonistic effects in an estrogen receptor-dependent
reporter gene assay. This assay was dose-dependently performed in U2OS-ERα cells [29] as
previously described [30,31]. At the highest concentration (1000 times higher than the reference
concentration of 10-8 M E2), the pyridine derivative 1 induced a weak agonistic response (13%
compared to E2 which was set to 100%, Figure S1, SI), while the quinoline-based inhibitors 13, 14,
and 16 did not induce any significant estrogen-dependent response different from the control
(DMSO). Quinoline-based compounds 13, 14, and 16 were therefore considered as better chemical
probes compared to the pyridine-based 1.
2.12 Cytotoxicity evaluation
Seven compounds (1, 5, 11-14 and 16), selected on the basis of their 17β-HSD14 inhibitory activities,
were evaluated for their cytotoxic effects on the HEK293 cell line by a colorimetric assay (reduction
of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT)) [32]. Compounds 1, 5, 12,
14, and 16 displayed a very low cytotoxicity (IG50 > 100 µM). Due to insufficient solubility, 11 and
13 could not be tested at 100 µM, however, the cells showed 80% survival at a concentration of 20
µM, the maximally possible concentration tested (Figure S2, SI).
CONCLUSION
The goal of this study was to develop new 17β-HSD14 inhibitors as chemical probes to further
characterize the enzyme and to investigate the role of this protein. Starting from the pyridine-based
inhibitor 1, a structure-based approach was followed and 16 new inhibitors were synthesized, divided
in two classes: pyridine-based and quinoline-based derivatives. The pyridine-based compounds (1-8)
turned out to be highly potent (Ki of compound 5 is 1.5 nM) but non selective toward 17β-HSD1 and
17β-HSD2. The novel quinoline-based compounds (9-17) are very promising considering their
potency and selectivity profiles toward 17β-HSD1, 17β-HSD2, and 17β-HSD10, especially inhibitors
28