Identification of PDZ ligands by docking-based virtual screening for the development of novel analgesic agents

Article abstract of DOI:10.1016/j.bmcl.2013.02.100

Disrupting the interaction between the PDZ protein, PSD-95, and its target ligands (such as the glutamate NMDA receptor or the serotonin 5-HT_2A receptor) was found to reduce hyperalgesia in various models of neuropathic pain. Here, we set out to identify lead molecules which would interact with PSD-95, and hence, would potentially display analgesic activity. We describe the virtual screening of the Asinex and Cambridge databases which together contain almost one million molecules. Using three successive docking filters and visual inspection, we identified three structural classes of molecules and synthesized a potential lead compound from each class. The binding of the molecules with the PDZ domains of PSD-95 was assessed by ¹H-¹⁵N HSQC NMR experiments. The analgesic activity of the best ligand, quinoline 2, was evaluated in vivo in a model of neuropathic pain and showed promising results.

Full text of DOI:10.1016/j.bmcl.2013.02.100

Bioorganic & Medicinal Chemistry Letters 23 (2013) 2624–2627
Contents lists available at SciVerse ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Identification of PDZ ligands by docking-based virtual screening
for the development of novel analgesic agents
Naoual Bouzidi a, Hemantkumar Deokar a, Alexandre Vogrig a, Benjamin Boucherle a, Isabelle Ripoche b,
Isabelle Abrunhosa-Thomas b, Liam Dorr c, Anne-Sophie Wattiez d, Lu-Yun Lian c, Philippe Marin e,
Christine Courteix d, Sylvie Ducki b,
⇑
a Clermont Université, UBP, CNRS UMR 6296, Institut de Chimie de Clermont-Ferrand, BP 10448, F-63000 Clermont-Ferrand, France
b Clermont Université, ENSCCF, CNRS UMR 6296, Institut de Chimie de Clermont-Ferrand, BP 10187, F-63174 Aubière, France
c NMR Centre for Structural Biology, University of Liverpool, UK-L69 3BX Liverpool, UK
d Clermont Université, Université d’Auvergne, Institut NeuroDol, BP 10448, F-63000 Clermont-Ferrand, INSERM, UMR 1107, F-63001 Clermont-Ferrand, France
e IGF, CNRS UMR 5203, INSERM U661, Universités Montpellier 1&2, F-34094 Montpellier, France
a r t i c l e i n f o
a b s t r a c t
Article history:
Disrupting the interaction between the PDZ protein, PSD-95, and its target ligands (such as the glutamate
NMDA receptor or the serotonin 5-HT2A receptor) was found to reduce hyperalgesia in various models of
neuropathic pain. Here, we set out to identify lead molecules which would interact with PSD-95, and
hence, would potentially display analgesic activity. We describe the virtual screening of the Asinex
and Cambridge databases which together contain almost one million molecules. Using three successive
docking filters and visual inspection, we identified three structural classes of molecules and synthesized
a potential lead compound from each class. The binding of the molecules with the PDZ domains of PSD-95
was assessed by 1H–15N HSQC NMR experiments. The analgesic activity of the best ligand, quinoline 2,
was evaluated in vivo in a model of neuropathic pain and showed promising results.
Received 7 February 2013
Revised 19 February 2013
Accepted 23 February 2013
Available online 14 March 2013
Keywords:
Docking-based virtual screening
PDZ proteins
Pain
Ó 2013 Published by Elsevier Ltd.
Analgesic activity
Since their discovery two decades ago, about 267 PDZ domains
(Post-synaptic density-95 PSD-95, Disc large Dlg, Zonula occludens
ZO-1) have been identified in the human genome. PDZ domains
regulate multiple biological processes such as transporter and ionic
channel activity and membrane-bound receptor-operated signal-
ling. These domains consist of 100 amino acid residues usually
Virtual screening provides a valuable alternative to high-
throughput screening (HTS) by bioassays,7 which is a useful but
expensive and time-consuming method to identify new lead com-
pounds. In the past decade, structure-based design methods for
virtual ligand screening have been successful in lead discovery
due to the increased structural information on drug target, the
availability of large commercial databases, advancements in com-
putational power, and improvements in the efficiency of docking
algorithms.8–10
composed of 6 b-strands (bA to bF), and two
B) (Fig. 1). PDZ-mediated protein–protein interactions (PPIs)
involve the binding of the extreme C-terminus of target proteins
a-helices (aA and
a
(PDZ ligand) to a hydrophobic groove between
a
B and bB, delimited
In the present study, we report the docking-based screening of
the Asinex and Cambridge databases to identify lead molecules
which could potentially interact with the PDZ1 or PDZ2 domains
of the PSD-95 protein.
We selected the high-resolution structure of the PDZ1 domain
of PSD-95 complexed with cypin (PDB 2KA9) as the starting point
of our computational approach. The docking was performed on the
active site using the docking module in MOE (Molecular Operating
Environment).11 We developed a protocol for the docking-based
screening and searched the Asinex and Cambridge commercial dat-
abases (respectively 514,957 and 432,000 molecules)12,13 using
three successive hierarchical docking filters (Fig. 2).
by the GLGF loop of the PDZ domain.1
Several PDZ proteins have been associated with pathological
states (cancer, cystic fibrosis, schizophrenia, Parkinson’s and
Alzheimer’s diseases, chronic inflammatory and neuropathic pain)
and the therapeutic usefulness of inhibiting PDZ-mediated PPIs has
been clearly demonstrated by using interfering peptides.2
The PSD-95 protein contains three PDZ domains and has been
identified as a major partner of various membrane receptors
including the glutamate NMDA and the serotonin 5-HT2A recep-
tors.3 Disrupting the interaction between PSD-95 and its target
proteins was found to reduce hyperalgesia in various models of
neuropathic pain.4–6 Thus, identification of PSD-95 ligands could
potentially lead to the development of a novel class of analgesics.
(I) The Alpha Triangle method12 was used to identify molecules
with good affinity towards the target protein. Here the poses
are generated by superimposition of ligand atom triplets and
⇑
Corresponding author.
E-mail address: sylvie.ducki@ensccf.fr (S. Ducki).
0960-894X/$ - see front matter Ó 2013 Published by Elsevier Ltd.
http://dx.doi.org/10.1016/j.bmcl.2013.02.100

N. Bouzidi et al. / Bioorg. Med. Chem. Lett. 23 (2013) 2624–2627
2625
We hence identified five classes of molecules: indoles, nucleo-
tides, quinolines, N-acyl hydrazones and carbohydrates derivatives
linked to a phenyl group with polar substituents. Since we have
previously described indole derivatives as potential PDZ inhibi-
tors15,16 and because nucleotides interact with too many biological
targets, we focused our attention on the other three classes.
We defined a lead compound from each class which satisfied
three criteria: (i) the best docking pose should show interactions
with the GLGF loop of the PDZ domain, (ii) molecules should
be obtained from commercially available and affordable starting
material with easy synthesis, and (iii) they must be easily function-
alizable on various positions. In the quinoline family, we chose
molecule 2 which has similar substituents as indole 1 (Fig. 3), pre-
viously reported to interact with the PDZ1 domain of PSD-95. The
best choice in the carbohydrate-linked-to-phenyl-ring class was
the glucose analogue 3. In the N-acyl hydrazones family, we se-
lected triol 4 which displayed the best docking energy. The three
potential ligands 2–4 displayed interactions with the GLGF loop
of PSD-95 PDZ1 (Fig. 3, Table 1).
Figure 1. 3D representation of PSD-95 PDZ1 (PDB 2KA9). Helices in red, strands in
yellow, loops in green.
Quinoline 2 was prepared from aniline 5, ethyl glyoxylate and
propanal using the method described by Shimizu et al.17 The best
results for the cyclization were obtained with Yterbium triflate as
a Lewis acid with 20% yield. Saponification of compound 6 gave
the desired quinoline 2 in 35% yield (Scheme 1, steps a and b).
The glucosylamine derivative 3 was obtained in 50% yield, from
Docking-based screening
514,957 and 432,000 molecules
D
-glucose 8 by reaction with the commercially available 4-amino
Step I : Alpha Triangle Method
benzoic acid 7 (Scheme 1, step c), using the method described by
Satorelli et al.18
94,000 molecules (top 10%)
The N-acyl hydrazone 4 was prepared in two steps (Scheme 1,
steps d and e) from the commercially available ethyl 4-hydroxy-
benzoate 9, in 33% yield.19
Step II : Triangle Matcher Method
Compounds 2–4 were evaluated for their interactions with the
PDZ1 and PDZ2 domains of PSD-95.15N-labelled PDZ proteins were
produced and purified after subcloning and expression in Esherichi-
a coli BL21 (see Supplementary data). 1H–15N heteronuclear single-
quantum coherence (HSQC) NMR experiments were recorded in
the absence and presence of compounds 2–4.16
Ligand-induced chemical shift perturbations (CSP) in the
1H–15N HSQC of the PDZ domains verified the presence of binding
and identified the amino acids whose chemical environments were
affected by the bound compound, hence, locating its binding
site(s). The reference spectra of free PDZ domains in the buffer
and temperature condition used here were assigned during the
course of this work and compared to literature.20
9,400 molecules (top 10%)
Step III : Conformational analysis
940 molecules (top 10%)
Step IV : Visual inspection
5 structural classes
of molecules
Figure 2. Docking-based screening protocol.
Unfortunately neither compounds 3 nor 4 had any effect on the
NMR spectra of the PDZ domains and only quinoline 2 displayed
some binding ability (Fig. 4A). While weak CSP were observed on
the spectrum of PSD-95 PDZ1 (see Supplementary data), stronger
CSP were observed on the PSD-95 PDZ2 spectrum (Gly18, Ser20,
Ile21, Gly23, Lys 53, Ala47, Glu3, Ala96, Asn63 and Asn27,
Fig. 4B). These residues are located in the peptide binding groove,
hence suggesting that quinoline 2 is binding in the same region as
the natural peptide ligands.
The anti-hyperalgesic effect of quinoline 2 was evaluated on
chronic constriction nerve injury-induced (CCI) neuropathic pain
in rats. This model consists in a unilateral peripheral mononeurop-
athy induced by four loose ligatures of the sciatic nerve.21 Intrathe-
cal (i.t.) injections of quinoline 2 at 12.3, 24.5 and 122.6 ng/rat
induced a dose-dependent anti-hyperalgesic effect in CCI rats.
The highest dose abolished mechanical hyperalgesia 15 min after
injection (Fig. 5a). This anti-hyperalgesic effect was similar to that
induced by an interfering peptide corresponding to the 10
C-terminal amino acids of the 5-HT2A receptor (TAT-2A peptide,
100 ng/rat, Fig. 5b), which disrupts interactions between the
receptor and its PDZ partners.4
triplets of receptor site points. The receptor site points were
alpha sphere centres which represent locations of tight
packing. A random conformation was selected at each itera-
tion; a random triplet of ligand atoms and a random triplet
of alpha sphere centres were used to determine the pose.
At the end of this step, molecules were sorted on the basis
of their interaction energy and the top 10% (around 94,000
molecules) was kept.
(II) Triangle Matcher method14 generated poses by aligning
ligand triplets of atoms on triplets of alpha spheres in a more
systematic way than in the Alpha Triangle method. As for
step 1, the top 10% (9400 molecules) was kept.
(III) A conformational analysis for the 9400 molecules was car-
ried out and all output molecules were kept as input for this
step. In the forcefield-based refinement of protein–ligand
complex, the energy minimization of the system was carried
out using the conventional molecular mechanics setup. At
the end of this step the molecules were examined for poses
and binding energy. The top 10% (940 molecules) were visu-
ally inspected (Step IV) and classified on the basis of their
chemical structures.

2626
N. Bouzidi et al. / Bioorg. Med. Chem. Lett. 23 (2013) 2624–2627
O
HO
OH
N
HO
N
H
OH
O
O
O
2
1
OH
HO
H
O
H
N
N
HO
HO
N
OH
OH
O
HO
O
2
3
4
3
4
OH
Figure 3. Indole 1 and the best docked poses for the top ranked molecules: quinoline 2, glycosylamine 3, N-acyl hydrazone 4.
Table 1
Docking of compounds 2–4 into PSD-95 PDZ1
Compounds
Interaction energy (kcal/mol)
Amino acid interactions
(H-Bonds)
Amino acid interactions
(other interactions)
2
3
4
ꢀ7.52
ꢀ8.52
ꢀ7.32
Leu17, Gly18, Phe19, Ile21
Gly18, Phe19, Ile21, Leu79
Ile21, Leu79
Leu17, Gly18, Phe19, Ile21, His72, Leu79
Gly18, Phe19, Ile21, His72, Val76, Leu79
Gly16, Leu17, Gly18, Phe19, Ser20, Ile21, Leu79
Energy of interaction (in kcal/mol). Localization of amino acids showing interactions with compounds 2–4 (from docking). Numbering uses the generic system 18Gly-19-
Leu-20Gly-21Phe for the PDZ domain.
a
b
O
OEt
O
HO
OH
N
NH2
N
O
O
O
O
O
6
2
5
7
OH
H2N
c
H
O
HO
N
HO
OH
OH
COOH
OEt
O
3
d
e
HO
HO
HO
H
N
H
N
N
NH2
O
O
9
O
10
4
HO
OH
Scheme 1. Synthesis of compounds 2–4. Reagents and conditions: (a) Yb(OTf)3, ethyl glyoxylate, propanal, CHCl3, O2, 90 °C, 16 h, 20%. (b) NaOH, MeOH/H2O, 70 °C, 12 h, 35%.
(c) -Glucose 8, EtOH, rt, 12 h, 50%. (d) Hydrazine, EtOH, reflux, 48 h, 50%. (e) 2,3-Dihydroxybenzaldehyde, EtOH, AcOH, reflux, 24 h, 66%.
D
Figure 4. (A) Overlay plot of 1H–15N HSQC spectra of free PSD-95 PDZ2 (black) and in the presence of compound 2 (red) (protein/ligand 1:10). (B) Chemical shift perturbation
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
2
(CSP) of PSD-95 PDZ2 resulting from compound 2. Dd (ppm) = ðDHÞ þ ð0:15DNÞ . Ie., the nitrogen difference should be scaled by 0.15.

N. Bouzidi et al. / Bioorg. Med. Chem. Lett. 23 (2013) 2624–2627
2627
PDZ domains. Quinoline 2 will be used as a lead compound for fur-
ther optimisation in order to develop anti-hyperalgesic agents.
Quinoline 2, 2.5 ng/rat
Quinoline 2, 12.3 ng/rat
Quinoline 2, 24.5 ng/rat
Quinoline 2, 122.6 ng/rat
(a)
13
12
11
10
9
***
Acknowledgments
***
***
***
i.th.
We would like to thank the French National Research Agency
(ANR) for funding of the project PDZ-CANPAIN and a postdoctoral
fellowship for B.B. and N.B.; the Regional Council of Auvergne
(Conseil Régional d’Auvergne) and the European Fund for Regional
Economic Development (FEDER) for a doctoral scholarship for A.V.
and A.S.W.; the PRES Clermont Université for a postdoctoral fel-
lowship for H.D.; the G.E. Healthcare/BBSRC for a CASE studentship
for L.D. The authors wish to thank Eric Chapuy (Inserm U1107 Neu-
ro-Dol, Clermont-Ferrand) for his technical assistance.
injection
***
*
***
**
***
8
7
6
5
Before Pre-ttt 15
CCI
30
45
60
90
120
(b)
Supplementary data
DMSO 20%, 10 µl/rat
TAT-2A 100 ng/rat
13
12
11
10
9
Quinoline 2, 122.6 ng/rat
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.bmcl.2013.02.
100.
***
***
***
i.th.
injection
***
***
*
References and notes
8
***
7
1. Doyle, D. A.; Lee, A.; Lewis, J.; Kim, E.; Sheng, M.; MacKinnon, R. Cell 1996, 1067,
85.
2. Ducki, S.; Bennett, E. Curr. Chem. Biol. 2009, 3, 146.
6
3. Bécamel, C.; Gavarini, S.; Chanrion, B.; Alonso, G.; Galéotti, N.; Dumuis, A.;
Bockaert, J.; Marin, P. J. Biol. Chem. 2004, 279, 20257.
4. Pichon, X.; Wattiez, A.-S.; Becamel, C.; Ehrlich, I.; Bockaert, J.; Eschalier, A.;
Marin, P.; Courteix, C. Mol. Ther. 2010, 18, 1462.
5
Before Pre-ttt 15
CCI
30
45
60
90
120
5. LeBlanc, B. W.; Iwata, M.; Mallon, A. P.; Rupasinghe, C. N.; Goebel, D. J.;
Marshall, J.; Spaller, M. R.; Saab, C. Y. Neuroscience 2010, 167, 490.
6. Garry, E. M.; Moss, A.; Delaney, A.; O’Neill, F.; Blakemore, J.; Bowen, J.; Husi, H.;
Mitchell, R.; Grant, S.; Fleetwood-Walker, S. Curr. Biol. 2003, 13, 323.
7. Tiziano, T. Comb. Chem. High Throughput Screening 2009, 12, 303.
8. Caporucio, F.; Rastelli, G.; Imbirano, C.; Del Rio, A. J. Med. Chem. 2011, 54, 4006.
9. Wang, W.; Wenig, J.; Zhang, X.; Liu, M.; Zhang, M. J. Am. Chem. Soc. 2009, 131,
787.
Figure 5. Anti-hyperalgesic effect of quinoline 2 in CCI rats (n = 8–18 rats per dose).
(a) CCI rats were i.t. injected with the indicated dose of quinoline 2. (b) CCI rats
were i.t. injected with either quinoline 2 (122.6 ng/rat), or TAT-2A (100 ng/rat) or
vehicle (20% DMSO 10 ll/rat). The vocalisation thresholds to paw pressure (Randall
and Selitto test22) were measured up to 120 min after injection. ⁄P <0.05, ⁄⁄P <0.01,
⁄⁄⁄P <0.001 compared with values measured before drug or vehicle injection (Pre-
treatment values).
10. Lengauer, T.; Rarey, M. Curr. Opin. Struct. Biol. 1996, 6, 402.
11. MOE 2010 Manual, 1997–2010 Chemical Computing Group Inc.
12. ASINEX Ltd, Moscow, Russia, http://www.asinex.com.
13. Computing Group Cambridge Structural Database System, Daresbury, UK,
http://cds.dl.ac.uk/Inc.
14. Idrus, S.; Sumo, U.; Tambunan, F.; Zubaidi, A. A. Bioinformation 2012, 8, 348.
15. Boucherle, B.; Vogrig, A.; Deokar, H.; Bouzidi, N.; Ripoche, I.; Thomas, I.; Marin,
P.; Ducki, S. Bioorg. Med. Chem. 2011, 19, 4346.
16. Vogrig, A.; Boucherle, B.; Deokar, H.; Thomas, I.; Ripoche, I.; Lian, L.-Y.; Ducki, S.
Bioorg. Med. Chem. Lett. 2011, 21, 3349.
17. Inada, T.; Nakajima, T.; Shimizu, I. Heterocycles 2005, 66, 611.
18. Wang, L.; Maniglia, C. A.; Mella, S. L.; Sartorelli, A. J. Med. Chem. 1983, 29, 1323.
19. Forman, D.; Yu, D. U.S. Patent 20,060,189,825, 2008.
We have performed a docking-based screening of the Asinex
and Cambridge databases to identify new lead compounds able
to interact with the PSD-95 protein. We identified in silico three
classes of molecules, quinoline, N-acyl hydrazone and substituted
carbohydrate. One molecule of each class was synthesized and
evaluated for interactions with two PDZ domains of PSD-95
(PDZ1 and PDZ2) using 1H–15N HSQC NMR experiments which al-
lowed us to identify quinoline 2 as the best ligand of the PDZ do-
mains of PSD-95. We also showed that this compound displayed
a significant anti-hyperalgesic activity in neuropathic rats (CCI),
similar to that induced by a natural peptide ligand of PSD-95
20. Piserchio, A.; Pellegrini, M.; Mehta, S.; Blackman, S. M.; Garcia, E. P.; Marshall,
J.; Mierke, D. F. J. Biol. Chem. 2002, 277, 6967.
21. Wang, L.; Piserchio, A.; Mierke, D. F. J. Biol. Chem. 2005, 280, 26992.
22. Randall, L. O.; Selitto, J. J. Arch. Int. Pharmacodyn. Ther. 1957, 111, 409.