Optimization of TRPV6 calcium channel inhibitors using a 3D ligand-based virtual screening method

Article abstract of DOI:10.1002/anie.201507320

Herein, we report the discovery of the first potent and selective inhibitor of TRPV6, a calcium channel overexpressed in breast and prostate cancer, and its use to test the effect of blocking TRPV6-mediated Ca²⁺-influx on cell growth. The inhib

Full text of DOI:10.1002/anie.201507320

.
Angewandte
Communications
International Edition: DOI: 10.1002/anie.201507320
German Edition: DOI: 10.1002/ange.201507320
Computer-Aided Drug Design
Optimization of TRPV6 Calcium Channel Inhibitors Using a 3D
Ligand-Based Virtual Screening Method
CØline Simonin, Mahendra Awale, Michael Brand, Ruud van Deursen, Julian Schwartz,
Michael Fine, Gergely Kovacs, Pascal Häfliger, Gergely Gyimesi, Abilashan Sithampari, Roch-
Philippe Charles, Matthias A. Hediger,* and Jean-Louis Reymond*
Abstract: Herein, we report the discovery of the first potent
and selective inhibitor of TRPV6, a calcium channel overex-
pressed in breast and prostate cancer, and its use to test the
effect of blocking TRPV6-mediated Ca2+-influx on cell growth.
The inhibitor was discovered through a computational method,
xLOS, a 3D-shape and pharmacophore similarity algorithm,
a type of ligand-based virtual screening (LBVS) method
described briefly here. Starting with a single weakly active seed
molecule, two successive rounds of LBVS followed by
optimization by chemical synthesis led to a selective molecule
with 0.3 mm inhibition of TRPV6. The ability of xLOS to
identify different scaffolds early in LBVS was essential to
success. The xLOS method may be generally useful to develop
tool compounds for poorly characterized targets.
indeed been shown to reduce cell growth, these compounds
only inhibited calcium transport at high micromolar concen-
tration, and were non-selective versus the close analogues
TRPV5 and store-operated calcium channels, precluding
a conclusive link between TRPV6 calcium transport function
and cell growth.[5]
TRPV6 is assumed to adopt the six transmembrane
domain structure typical of TRP family channels.[3g]
A
homology model derived from the recently reported rat
TRPV1 crystal structure suggests that a hydrophobic binding
pocket analogous to the binding site postulated for capsaicin
in TRPV1[6] and suitable for small molecule binding might
exist in TRPV6 (Supporting Information, Figure S1). The
absence of an experimental structure, however, precluded the
use of structure-based methods for ligand design.[1b,7] We
therefore set out to test if ligand-based virtual screening
(LBVS)[1c] might be used to discover a potent and selective
small molecule TRPV6 inhibitor starting from the available
weak TRPV6 inhibitors 1–5 as seed compounds (Figure 1D,
the covalent non-selective channel inhibitor 2-aminoethyl
phenylboronate was not considered).[5] Since it was not
known which part of these inhibitors contributed to their
TRPV6 activity, we searched for an LBVS method capable of
identifying similarities to part or the whole of a seed molecule,
if possible favoring rather small, ligand-efficient fragments for
screening,[8] and emphasizing scaffold hopping[9] since the
scaffolds of the weakly active seed molecules were probably
not optimal.
V
irtual screening can valuably assist drug discovery when-
ever a large body of information is already available, such as
the activity and selectivity profiles of hundreds of potent
inhibitors of the desired target or its high-resolution crystal
structure.[1] On the other hand, small molecule inhibitors are
often needed to understand the biological role of poorly
characterized targets, in particular to selectively shut down
their activity on a short time-scale and test if the effects
observed by genetic knock-out are indeed caused by loss of
function.[2] Herein, we report the application of virtual
screening to such a problem at the example of transient
receptor potential vanilloid 6 (TRPV6), a selective calcium
channel overexpressed in advanced prostate cancer tissues
and carcinomas of the colon, breast, thyroid, and ovary.[3]
A
The ligand overlap score (LOS), which quantifies the
spatial overlap between two molecules as the weighted sum of
atom pair proximity scores, was selected as a substructure-
independent metric of shape similarity between two mole-
cules.[10] LOS was computed separately for hydrophobic
atoms, hydrogen bond donor and acceptor atoms which
were previously found to be useful atom categories for
pharmacophore fingerprint design.[11] Only the lowest energy
conformer generated by CORINA[12] was used as 3D-model
of each molecule because sampling multiple conformers
would increase computational costs by at least 100-fold and
has been shown to not yield clear performance benefits in
related 3D LBVS methods.[13] The combined score was
maximized by alignment and iterative translation and rota-
tion of query relative to seed molecule along their principal
molecular axes (Figure 1A–C; Supporting Information, Fig-
ure S2). The resulting algorithm xLOS (atom category
extended Ligand Overlap Score) performed comparably
well to other 3D LBVS methods for the recovery of actives
from inactive decoy molecules in various sets of bioactive
decreased proliferation rate was observed for prostate and
breast cancer cells upon siRNA knockdown of TRPV6,
suggesting that cancer cell proliferation might be controllable
by inhibition of calcium transport through this channel.[4]
Although several small molecule inhibitors of TRPV6 have
[*] Dr. C. Simonin, Dr. M. Awale, M. Brand, Dr. R. van Deursen,
Dr. J. Schwartz, Prof. Dr. J.-L. Reymond
Department of Chemistry and Biochemistry
National Center of Competence in Research NCCR TransCure
University of Bern, Freiestrasse 3, 3012 Bern (Switzerland)
E-mail: jean-louis.reymond@dcb.unibe.ch
Dr. M. Fine, Dr. G. Kovacs, P. Häfliger, Dr. G. Gyimesi, A. Sithampari,
Prof. Dr. R.-P. Charles, Prof. Dr. M. A. Hediger
Institute of Biochemistry and Molecular Medicine
National Center of Competence in Research NCCR TransCure
University of Bern, Bühlstrasse 28, 3012 Bern (Switzerland)
E-mail: matthias.hediger@ibmm.unibe.ch
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/anie.201507320.
14748
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2015, 54, 14748 –14752

Angewandte
Chemie
Figure 1. LBVS with xLOS and its application to TRPV6. A) Atom pair proximity score. B) xLOS combines LOS for H-bond donor (d), H-bond
acceptor (a), and hydrophobic atoms (h). C) Alignment of seed (1, orange) and query (9, green) molecule after xLOS value optimization by
iterative translations and rotation along the principal molecular axes p1-p2-p3. D) Structures of known TRPV6 inhibitors. E) MW histogram of top
xLOS scoring compounds. F) Structure of hits 6–10 identified after the first round of LBVS. IC50 values were determined with pure compounds
(see the Supporting Information). G) Scatter plot of substructure similarity versus xLOS score for tested compounds (yellow) and active hits (red,
hit rate: 4%), and xLOS histogram of the 800000 cpds analyzed by LBVS.
molecules from the directory of useful decoys (Supporting
Information, Figure S3).[11,13b,14] Furthermore, xLOS showed
significant scaffold-hopping, indicated by the fact that bioac-
tive analogues were retrieved by xLOS independent of their
substructure similarity to the seed compound used for LBVS
(Supporting Information, Figure S4).[9] This effect may be
attributed to the fact that xLOS perceives shape and
pharmacophores by atom positions only without considering
connectivity patterns.
xLOS was used to score ~ 800000 purchasable small
molecules against seed inhibitors 1–5 (Figure 1D).[15] Because
xLOS penalizes non-overlapping parts of seed and query
molecule by scaling to molecular size (Figure 1B), high-
scoring molecules generally had a relatively small size
favorable for hit compounds (Figure 1E).[16] The five lists of
1000 top scoring hits were clustered to select diverse
molecules avoiding pan assay interference compounds,[17]
and 133 compounds were purchased for experimental eval-
uation. Testing inhibition of TRPV6-induced Ca2+ and Cd2+
influx in transiently transfected HEK293 cells (Supporting
Information, Figures S5,S6)[18] revealed five weakly active
hits (6–10, IC50 ~ 20 mm), with very different scaffolds com-
pared to the seed molecules (Figure 1F/G).
Hits 6–10 were used as seeds for a second LBVS round
using both xLOS and a conventional substructure fingerprint
similarity search,[19] and an additional 90 analogues were
purchased and tested. While all analogues of hits 6–8 were
inactive, precluding optimization, 16 analogues of hits 9 and
10 (11a-p) were active, indicating that their common (4-
phenylcyclohexyl)piperazine scaffold was suitable for further
exploration. Both of these compounds originated from
a scaffold-hopping step from prolinol type seed 1 in the first
LBVS round. Separation of the cis/trans diastereomers in the
second round of LBVS hits 11a–g revealed a consistently
stronger activity in cis-diastereomers, with the best activity
reached with compound cis-11a identified by xLOS (Fig-
ure 2A/B; Supporting Information, Figure S7).
Inhibitor cis-11a was further optimized by synthesizing
analogues by reductive alkylation of various 4-aryl-cyclo-
hexanones with different mono-substituted piperazines and
analogues (Scheme 1). Variations in ring A led to stronger
inhibition by removing the methylene group between ring A
Angew. Chem. Int. Ed. 2015, 54, 14748 –14752
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 14749

.
Angewandte
Communications
Figure 2. Optimization of TRPV6 inhibitors. A) Scatter plot of substructure similarity (TSfp) versus xLOS score in 2nd round LBVS (highest value
relative to seeds 6–10). Hits defined as ꢀ70% inhibition of Cd2+ influx in hTRPV6 transfected HEK293 cells. Hit rate xLOS: 20%, Sfp: 8%.
B) Structure and IC50 values for Cd2+ influx of hits highlighted in (A). C) Synthetic analogues of cis-11a. Activity on hTRPV6 given as% inhibition
of Cd2+ influx at 5 mm (3 replicates) or IC50 value (7 concentrations”6 replicates), as meanÆSEM for nꢀ2. D) Structures and activities of
optimized inhibitors TRPV6 22a-c and diastereomeric mixtures (R)- and (S)-17.
and the piperazine core (12a), and moving the nitrogen to the
meta position (12b), while further substitutions of the
pyridine ring (12d–g) did not affect the activity (Figure 2C).
Note that the ortho-pyridine analogue 12c reported as sterol
isomerase inhibitor (cis: Ki = 50 nm, trans: Ki = 17 nm),[20] was
also weakly active as a TRPV6 inhibitor (cis: 72% at 5 mm).
Activity was reduced by modifying piperazine ring B (13–15).
Variations in cyclohexane ring C were also detrimental (16,
18–20), however ring expansion to a seven-membered ring
gave the more potent chiral analogue 17 as a mixture of four
stereoisomers. A synthesis starting from cycloheptanones
(R)-26 (99% ee) and (S)-26 (96% ee)[21] yielded two
inseparable pairs of diastereomers (R)- and (S)-17 with
Scheme 1. Synthesis of 11a and analogues.Conditions: a) PhMgBr,
comparable activities. Substitution of ring D was also
explored leading to improvements with the 2-methoxy (cis-
21b) analogues. Variations in rings A and D were combined
to obtain cis-22a-c as optimized, potent TRPV6 inhibitors
(Figure 2D; Supporting Information, Figure S8).
Inhibitor cis-22a was investigated in detail and showed
high selectivity against other calcium channels and related
TRP targets (Table 1; Supporting Information, Figure S1).
Although similarity searches in ChEMBL followed by
THF, reflux, 30 min (quant.). b) TsOH, toluene, reflux, 4 h. c) H2, Pd/C,
EtOAc, RT, 18 h. d) HCl, acetone, reflux, 18 h (25, 44% over 3 steps;
28, 42% over 3 steps). e) Me3SiCHN2, nBuLi, MeOH, SiO2, Et2O, THF,
À788C!RT, 1 h30 (48%). f) N2CHCO2Et, (R)- or (S)-3,3’-bis(trimethyl-
silyl)-1,1’-binaphtyl-2,2’-diol, Me3Al, toluene, À788C, 24 h. g) LiCl,
DMSO, 1608C, 3 h (32-39% over 2 steps). h) Secondary amine,
NaBH(OAc)3, DCE, RT, 20 h (4-49%). i) 1-((pyridin-4-yl)methyl)piper-
azine, NaBH(OAc)3, DCE, RT, 20 h (cis-11a, 17%; trans-11a, 5%; rac-
17, 22%; 18, 40%; 20, 26%). j) HCl, THF, RT, 21 h (65%). k) HCl,
H2O, reflux, 15 h (72%). l) ZnEt2, TFA, CH2I2, DCM, 2 h (47%).
14750 www.angewandte.org
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2015, 54, 14748 –14752

Angewandte
Chemie
Table 1: In vitro pharmacology profile of cis-22a on several ion chan-
inhibitor 1 by two successive rounds of LBVS, followed by
nels.[a]
optimization by synthesis. While achieving useful hit rates
(round 1: 4%, round 2: 20%), xLOS selected high scaffold
diversity and allowed transition from the weakly active seed
1 to a more favorable scaffold early in the optimization
campaign. In contrast to many virtual screening approaches
requiring large amounts of information to get started, xLOS
can start from a single seed and does not require detailed
structural information on its target or any pre-existing
structure–activity relationship data. It should also be noted
that scanning purchasable collections was not sufficient, and
that chemical synthesis was critical to obtain a potent
inhibitor in the later stages of optimization. The method
should be generally useful to support the investigation of
targets for which only weak or unselective inhibitors have
been documented.
Channel
% Inh. at 10 mm
TRPV1 (agonist effect) (h)
TRPV1 (antagonist effect) (h)
TRPV3 (antagonist effect) (h)
TRPV5 (antagonist effect) (r)
TRPM8 (agonist effect) (h)
TRPM8 (antagonist effect) (h)
Store-operated Ca2+ channels
À15.5Æ0.7[b]
29.6Æ8.6
À11.5Æ2.0
79.9Æ1.4 (IC50 =2.4 mm)
À7.4Æ1.4[b]
21.2Æ7.5
À10.1Æ6.0
[a] The agonist or antagonist effect of cis-22a on TRP channels was
measured in a functional assay by recording calcium mobilization. [b] %
of control agonist response. h, human; r, rat. See also Supporting
Information, Table S1.
experimental validation identified significant cross-inhibition
of dopamine, muscarinic, opiate m, and serotonin GPCRs
typical for aromatic piperazines and of the cardiac potassium
channel hERG (23–98% at 10 mm; Supporting Information,
Table S1), cis-22a was judged suitable for testing the effect of
TRPV6 inhibition on cell growth. The anti-proliferative
activity of cis-22a was investigated on TRPV6 expressing
T47D human breast cancer cells and TRPV6 non-expressing
SKOV3 ovarian carcinoma cells (Figure 3A). Treatment of
Acknowledgements
This work was supported financially by the Swiss National
Science Foundation, NCCR TransCure. MF and GG were
supported by the Marie Curie Actions International Fellow-
ship Program (IFP) TransCure.
Keywords: calcium channels · drug discovery · TRP channels ·
virtual screening
How to cite: Angew. Chem. Int. Ed. 2015, 54, 14748–14752
Angew. Chem. 2015, 127, 14961–14965
[1] a) G. Klebe, Drug Discovery Today 2006, 11, 580 – 594; b) T.
Scior, A. Bender, G. Tresadern, J. L. Medina-Franco, K.
Martinez-Mayorga, T. Langer, K. Cuanalo-Contreras, D. K.
Agrafiotis, J. Chem. Inf. Model. 2012, 52, 867 – 881; c) K.
Heikamp, J. Bajorath, Chem. Biol. Drug Des. 2013, 81, 33 – 40;
d) M. Reutlinger, T. Rodrigues, P. Schneider, G. Schneider,
Angew. Chem. Int. Ed. 2014, 53, 582 – 585; Angew. Chem. 2014,
126, 593 – 596.
Figure 3. Antiproliferative activity of cis-22a and trans-22a on TRPV6
positive (T47D) and TRPV6 negative (SKOV3) breast cancer cells.
A) Expression of TRPV6 in different cell lines. B,C) Percent of control
cell growth at day 6 as a function of increasing inhibitor concentration.
[2] J. Blagg, P. Workman, Curr. Opin. Pharmacol. 2014, 17, 87 – 100.
[3] a) J. B. Peng, X. Z. Chen, U. V. Berger, P. M. Vassilev, H.
Tsukaguchi, E. M. Brown, M. A. Hediger, J. Biol. Chem. 1999,
274, 22739 – 22746; b) L. Yue, J. B. Peng, M. A. Hediger, D. E.
Clapham, Nature 2001, 410, 705 – 709; c) M. A. Hediger, E. M.
Brown, J. B. Peng, Patent, US6534642, 2003 and US7078178,
2006; d) D. E. Clapham, L. W. Runnels, C. Strubing, Nat. Rev.
Neurosci. 2001, 2, 387 – 396; e) K. A. Bolanz, G. G. Kovacs, C. P.
Landowski, M. A. Hediger, Mol. Cancer Res. 2009, 7, 2000 –
2010; f) C. Fecher-Trost, P. Weissgerber, U. Wissenbach,
Handb. Exp. Pharmacol. 2014, 222, 359 – 384; g) U. A. Hellmich,
R. Gaudet, Handb. Exp. Pharmacol. 2014, 223, 963 – 990.
[4] a) V. Lehen’kyi, M. Flourakis, R. Skryma, N. Prevarskaya,
Oncogene 2007, 26, 7380 – 7385; b) K. A. Bolanz, M. A. Hediger,
C. P. Landowski, Mol. Cancer Ther. 2008, 7, 271 – 279; c) M.
Raphal, V. y. Lehen’kyi, M. Vandenberghe, B. Beck, S. Khali-
monchyk, F. Vanden Abeele, L. Farsetti, E. Germain, A.
Bokhobza, A. Mihalache, P. Gosset, C. Romanin, P. ClØzardin,
R. Skryma, N. Prevarskaya, Proc. Natl. Acad. Sci. USA 2014,
111, E3870 – E3879.
T47D cells with cis-22a decreased cellular proliferation by
20% at 5 mm (IC50 = 25 Æ 10 mm), a concentration sufficient to
block TRPV6-mediated Ca2+ influx, while SKOV3 cells were
unaffected (IC50 > 50 mm, Figure 3B). By contrast, there was
no significant growth difference between the two cells lines
when treated with the less potent diastereomer trans-22a
(Figure 3C). The selective but significantly smaller effect of
cis-22a compared to siRNA knockdown (50% reduction in
cell growth)[4b] suggests that TRPV6 affects cell growth by
other mechanisms than the Ca2+ uptake from the extracellular
medium inhibited by cis-22a, in line with a recent report of its
role in store operated calcium entry.[4c]
[5] a) C. Landowski, K. Bolanz, A. Y. Suzuki, M. A. Hediger,
Pharm. Res. 2011, 28, 322 – 330; b) G. Kovacs, N. Montalbetti,
A. Simonin, T. Danko, B. Balazs, A. Zsembery, M. A. Hediger,
Cell Calcium 2012, 52, 468 – 480; c) A. Hofer, G. Kovacs, A.
In summary the potent and selective TRPV6 inhibitor cis-
22a with a (4-phenylcyclohexyl)piperazine scaffold was
obtained starting from the weakly active prolinol-based
Angew. Chem. Int. Ed. 2015, 54, 14748 –14752
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 14751

.
Angewandte
Communications
Zappatini, M. Leuenberger, M. A. Hediger, M. Lochner, Bioorg.
Med. Chem. 2013, 21, 3202 – 3213.
[15] J. J. Irwin, T. Sterling, M. M. Mysinger, E. S. Bolstad, R. G.
Coleman, J. Chem. Inf. Model. 2012, 52, 1757 – 1768.
[16] a) M. M. Hann, MedChemComm 2011, 2, 349 – 355; b) N.
Foloppe, Future Med. Chem. 2011, 3, 1111 – 1115.
[17] J. B. Baell, G. A. Holloway, J. Med. Chem. 2010, 53, 2719 – 2740.
[18] G. Kovacs, T. Danko, M. J. Bergeron, B. Balazs, Y. Suzuki, A.
Zsembery, M. A. Hediger, Cell Calcium 2011, 49, 43 – 55.
[19] a) P. Willett, Drug Discovery Today 2006, 11, 1046 – 1053; b) M.
Awale, J. L. Reymond, Nucleic Acids Res. 2014, 42, W234 –
W239.
[20] F. Berardi, C. Abate, S. Ferorelli, A. F. de Robertis, M.
Leopoldo, N. A. Colabufo, M. Niso, R. Perrone, J. Med. Chem.
2008, 51, 7523 – 7531.
[21] T. Hashimoto, Y. Naganawa, K. Maruoka, J. Am. Chem. Soc.
2011, 133, 8834 – 8837.
[6] E. Cao, M. Liao, Y. Cheng, D. Julius, Nature 2013, 504, 113 – 118.
[7] M. Lill, Methods Mol. Biol. 2013, 993, 1 – 12.
[8] A. L. Hopkins, G. M. Keseru, P. D. Leeson, D. C. Rees, C. H.
Reynolds, Nat. Rev. Drug Discovery 2014, 13, 105 – 121.
[9] G. Schneider, W. Neidhart, T. Giller, G. Schmid, Angew. Chem.
Int. Ed. 1999, 38, 2894 – 2896; Angew. Chem. 1999, 111, 3068 –
3070.
[10] a) S. K. Kearsley, G. Smith, Tet. Comput. Methodol. 1990, 3, 615 –
633; b) E. A. Kennewell, P. Willett, P. Ducrot, C. Luttmann, J.
Comput.-Aided Mol. Des. 2006, 20, 385 – 394; c) P. C. Hawkins,
A. G. Skillman, A. Nicholls, J. Med. Chem. 2007, 50, 74 – 82.
[11] M. Awale, X. Jin, J. L. Reymond, J. Cheminf. J. Cheminform.
2015, 7, 3.
[12] J. Sadowski, J. Gasteiger, Chem. Rev. 1993, 93, 2567 – 2581.
[13] a) S. Renner, C. H. Schwab, J. Gasteiger, G. Schneider, J. Chem.
Inf. Model. 2006, 46, 2324 – 2332; b) V. Venkatraman, V. I. Perez-
Nueno, L. Mavridis, D. W. Ritchie, J. Chem. Inf. Model. 2010, 50,
2079 – 2093.
Received: August 6, 2015
[14] N. Huang, B. K. Shoichet, J. J. Irwin, J. Med. Chem. 2006, 49,
6789 – 6801.
Revised: September 2, 2015
Published online: October 12, 2015
14752 www.angewandte.org
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2015, 54, 14748 –14752