Rational design of a nonpeptide general chemical scaffold for reversible inhibition of PDZ domain interactions

Article abstract of DOI:10.1016/j.bmcl.2006.10.006

Novel small molecules were designed to specifically target the ligand-binding pocket of a PDZ domain. Iterative molecular docking and modeling allowed the design of an indole scaffold 10a as a reversible inhibitor of ligand binding. The 10a scaffold inhib

Full text of DOI:10.1016/j.bmcl.2006.10.006

Bioorganic & Medicinal Chemistry Letters 17 (2007) 549–552
Rational design of a nonpeptide general chemical scaffold for
reversible inhibition of PDZ domain interactions
Naoaki Fujii,a, Jose J. Haresco,b Kathleen A. P. Novak,a Robert M. Gage,d
Nicoletta Pedemonte,e David Stokoe,c Irwin D. Kuntza,b and R. Kiplin Guya,d, ,
*
aDepartment of Pharmaceutical Chemistry, University of California, San Francisco, CA 94143, USA
bLaboratory for Molecular Dynamics and Design, University of California, San Francisco, CA 94143, USA
cComprehensive Cancer Center, University of California, San Francisco, CA 94143, USA
dDepartment of Cellular and Molecular Pharmacology, University of California, San Francisco, CA 94143, USA
eLaboratory of Molecular Genetics, Giannina Gaslini Institute, Genova 16148, Italy
Received 9 September 2006; revised 2 October 2006; accepted 4 October 2006
Available online 6 October 2006
Abstract—Novel small molecules were designed to specifically target the ligand-binding pocket of a PDZ domain. Iterative molec-
ular docking and modeling allowed the design of an indole scaffold 10a as a reversible inhibitor of ligand binding. The 10a scaffold
inhibited the interaction between MAGI-3 and PTEN and showed cellular activities that are consistent with the inhibition of
NHERF-1 function.
Ó 2006 Elsevier Ltd. All rights reserved.
PDZ domains mediate crucial protein–protein interac-
tions that enforce localization and organization of
proteins in a variety of submembranous complexes asso-
ciated with cell signal mediators, including ion channels,
transmembrane receptors, and regulatory enzymes.1 We
have previously reported an irreversible inhibitor of the
interaction of a PDZ domain (1, Fig. 1).2 Because drug
utility and feasibility are valuable tools for studying cell
signaling, their corresponding reversible inhibitors are
of great interest. By using small molecules to inhibit
PDZ domain interactions, we will be able to evaluate
the cellular effect of those interactions by carrying out
a cell-based assay. For this purpose, the inhibitors must
be cell permeable and able to target PDZ domains
whose pharmacological functions are well known such
as those of MAGI.
One interaction of MAGI-3 adaptors is between the
second PDZ domain (PDZ2) of MAGI and PTEN,4
a lipid/protein phosphatase. The ITKV motif of the
PTEN carboxyl terminus coordinates the interaction
of MAGI-3 and PTEN; this interaction suppresses
PKB activity by colocalizing PTEN and its substrates.
Therefore, chemical disruption of this interaction
would be a unique way to investigate its role in coor-
dinating PKB signaling. In this study, we designed a
The MAGIs are a subfamily of the MAGUKs, which
are widely expressed and have six PDZ domains.3
Figure 1. Design of novel scaffolds mimicking the side-chain presen-
tation of the four carboxy-terminal residues of the PDZ domain’s
ligand. (1) Irreversible inhibitor based on spontaneous ionization of a
3-hydroxymethyl indole. (2) Designed scaffold and targeted variations.
R1 indicates variable replacement for the (0) ligand residue side-chain;
R2 indicates variable replacement for the (À1) ligand residue side-
chain; and R3 indicates variable replacement for the (À3) ligand
residue side-chain.
Keywords: PDZ domain; Protein–protein; Interaction; Drug design;
NHERF.
*
Corresponding author. Tel.: +1 901 495 5714; fax: +1 901 495
5715; e-mail: Kip.Guy@stjude.org
Present address: Department of Chemical Biology and Therapeutics,
MS 1000, St. Jude Children’s Research Hospital, 332 N. Lauderdale,
Memphis, TN 38105-2794, USA.
0960-894X/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2006.10.006

550
N. Fujii et al. / Bioorg. Med. Chem. Lett. 17 (2007) 549–552
general chemical scaffold for PDZ-domain-binding
based on a canonical PDZ-domain structure and then
tailored the scaffold to mimic I-T/S-X-V to investigate
on MAGI-3.
Analysis of the structure of PSD-95 PDZ3 bound with
KQTSV ligand5 revealed the geomry of the ligand.
Inspection of the b-strand ligand revealed that its
conformation was uncharacteristic of other b strands.
The typical phi/psi angles of an antiparallel b strand
are 113°/180°. For the PDZ domain, these angles are
117.7°/110.4°, À106.6°/149.3°, and À101.9°/178.4°, for
the 0, À1, and À2 residues, respectively.
This uncharacteristic conformation allows the side-
chain of Val(0) to dip toward the hydrophobic cavity
in the PDZ-domain-binding groove. To preserve the ori-
entations of the carboxylate and the Val(0) side-chain,
we measured the distances and angles between the a car-
bons of residues 0 and À1 by using Weblab (Accelrys,
San Diego, CA) and then manually matched them to
those of simple heterocyclic rigid cores. Among the eval-
uated heterocycles, only indole or benzothiophene ade-
quately preserved the requisite distance and angle
restraints and accommodated all positions for functions
deemed necessary for the matching. On the basis of the
indole’s presence in a number of drugs and feasibility in
chemical modification for chemical library approach, we
focused on it. Scaffold 2 (Fig. 1) included positions R1
through R3 that projected various functions to optimize
the mimicry of peptide ligand side-chains into the PDZ-
domain-binding pocket. Initial docking experiments
with the indole scaffold 2 indicated that it could indeed
place critical functional groups in proximity to the ori-
entation of the b-strand conformation of bound ligand
in the crystal structure (Fig. 2). The R2 position of the
indole scaffold 2 has an optimal orientation for present-
ing the group to mimic the À1 side-chain or other func-
tional groups. However, PSD-95 PDZ3, whose structure
was used for these docking experiments, does not make
tight contacts with the Ser (À1) side-chain in the
KQTSV ligand. Our analysis of this structure indicated
similarly that the hydroxyalkyl groups evaluated at the
R2 position did not make contacts with the PDZ do-
main, nor did they appear to substantially contribute
to the overall scores. Introducing a hydroxyalkyl group
at the R2 position made the molecule too hydrophilic;
therefore, it would not be cell permeable. Thus, a simple
2-phenylethyl group was chosen as the R2 group to add
the proper hydrophobicity to make the molecule cell
permeable. The n-butyl group in 10a was considered a
suitable alternative for the M/I(À3) alkyl side-chain of
the M/I-TXV peptide. Compound 10a can successfully
make all of the following necessary contacts to mimic
the b strand: (1) a carboxylic acid mimics the C-terminus
of the peptide ligand; (2) a methyl group at the R1 posi-
tion fills the hydrophobic pocket that the Val(0) side-
chain normally occupies; and (3) a nonpeptide extension
at the R3 position places a hydroxy group within hydro-
gen bonding distance to histidine residue (Fig. 2D). This
potential is supported by DOCKs energy scoring:
À26.31 for 10a (Fig. 2C) versus À20.00 for the KQTSV
b strand (Fig. 2A).
Figure 2. Optimization of scaffold design by virtual library screening.
(A) KQTSV peptide (DOCK score: À20). (B) The results of substi-
tution of position R1 on an initially designed scaffold (DOCK score:
À30). (C) Designed reversible inhibitor 10a (DOCK score: À27). (D)
The overlay of targeted reversible inhibitor 10a (green) with the native
ligand, KQTSV peptide (blue).
We envisioned a synthetic route to 10a allowing full
variation of the highlighted variable groups by a library
approach (Fig. 3). The carboxyl group of 2-methyl-5-
nitrobenzoic acid 3 was protected, and the nitro group
was then reduced to give aniline 4. Iodination of 4 with
iodine monochloride gave a mixture of 5a and 5b. The
undesired product 5b was easily separated by chroma-
tography and recycled to 4. Iodoaniline 5a was coupled
with ethyl pyruvate in the presence of palladium acetate6
to give indole 6, which was easily converted to aldehyde
7. The aldehyde 7 is the foundation for developing the
library that has diverse substituents on its indole-1
(R2) and indole-2 (R3) positions by sequential electro-
philic and nucleophilic alkylation. The 2-phenylethyl
group was attached to the indole-1 position of 7, and
the n-butyl group was attached to the indole-2 position
and deprotected to give I-TXV mimetic 10a. Sodium salt
of 10a has good solubility (>30 mM) in water that does
not contain DMSO.
We have shown that pre-incubation with irreversible
inhibitor 1 blocks peptide-ligand binding and that 1
forms a stoichiometric irreversible adduct with the
MAGI-3 PDZ2 domain.2 In contrast, 10a did not affect
the pre-incubation. Longer incubation of the PDZ with
1 increased the inhibitory efficacy in a time-dependent

N. Fujii et al. / Bioorg. Med. Chem. Lett. 17 (2007) 549–552
551
lysates of several cancer cell lines. No biotinylated bands
were observed in Western analysis (streptavidin blot-
ting) of the biotin-10a treated lysate sample, whereas
biotin-12 visualized many protein bands under the iden-
tical conditions (data not shown). This result suggests
that 10a does not covalently modify any proteins, while
1 does.
If 10a binds to PDZ domains as an ITXV mimic, chem-
ical modifications analogous to mutations that block li-
gand binding should destroy its binding activity.
Variations of 10a with altered substituents correspond
to altered carboxyl terminus and Thr(À2) hydroxyl
group, respectively, in ITXV peptide. From this view-
point, 10b and 10c were designed as mimetics of those
inactive mutant peptides. Compound 10a had affinity
for the MAGI-3 PDZ2 domain, as demonstrated by its
ability to displace a peptide probe by fluorescent polari-
zation method,7 in a concentration-dependent manner
(Fig. 4B). The competition was observed with concentra-
tions of 10a in the range of >100 lM. AlphaScreen ener-
gy transfer assay8 also afforded a similar competition
curve (data not shown). Meanwhile, negative controls
10b and 10c, in which moieties required for the PDZ li-
gand interaction were disrupted, did not show any com-
petition (data not shown). The 10a scaffold (50 lM)
doubled the level of PKB kinase activity in the
HCT116 cell line that expresses wild-type PTEN; this
finding suggests that co-localization of PTEN is inhibited
in cells. This activity level is reasonable in comparison
with 1, which inhibited the MAGI-3/PTEN interaction
more potently (ꢀ50% reduction of binding response at
10 lM) and activated PKB (ꢀ3 times at 5 lM).2
Figure 3. Synthesis of 10a–c. (a) HOCH2C(CH3)2NH2 (3 equiv),
HBTU (1.2 equiv), DIPEA (2 equiv), DMF, 40 °C, 14 h; (b) SOCl2 (5
equiv), CH2Cl2, rt, 0.5 h; (c) Fe (7.8 equiv), NH4Cl (aq), EtOH, reflux,
3 h, 92% over 3 steps; (d) ICl (1.6 equiv), CaCO3 (3 equiv), MeOH,
H2O, rt, 1 h. 5a: 36%, 5b: 47%; (e) H2, Pd–C, MeOH, Et3N, rt, 1 h,
quant; (f) ethyl pyruvate (5 equiv), Pd(OAc)2 (0.2 equiv), DABCO (5
equiv), DMF, 105 °C, 50 min, 44%; (g) LiAlH4 (10 mol equiv), THF,
reflux, 2.5 h, 75%; (h) MnO2 (1.8 equiv), CH2Cl2, rt, 12 h, 76%; (i)
BrCH2CH2Ph (5 equiv), K2CO3 (10 equiv), DMF, 40 °C, 22 h, 93%; (j)
n-BuMgBr (5 equiv), 0 °C, 0.5 h, 87%; (k) i—MeI (large excess),
K2CO3 (3 equiv), acetone, rt, 2.5 d; ii—NaOH, H2O, MeOH, reflux,
7 h, 2 steps overall 89%; (l) MeI (large excess), K2CO3 (large excess),
acetone, rt, 1.5 h, 69%; (m) H2, 10% Pd–C, HCl (20 equiv), MeOH, rt,
1 h, 25%.
An important feature of the chemistry presented (Fig. 3)
is its feasibility to make diverse libraries highly variant in
the R2 and R3 positions to discover class- and domain-se-
lective inhibitors. Therefore, the scaffold described here-
in offers several opportunities for optimization toward
other specific PDZ domains. To investigate whether
10a can serve as a basis of such optimization for target-
ing other PDZ domains, we have focused on the first
PDZ domain (PDZ1) of NHERF-1.9 The NHERF-1
PDZ1 binds to the D-T/S-X-L motif of the b2-adrenergic
receptor to recycle it from the endocytosed fraction at the
cell surface10 and cystic fibrosis transmembrane conduc-
tance regulator (CFTR) to control channel gating.11 We
treated HEK293 cells with 10a; these cells stably ex-
pressed tagged b2-adrenergic receptor or a delta d-opioid
receptor chimera that had a carboxyl-terminus sequence
of b1-adrenergic receptor that is not recycled back by
NHERF-1.10 The expression level of these receptors on
the cellular surface was quantified. Only the expression
of the b2-adrenergic receptor was reduced in a dose-de-
pendent manner, suggesting inhibition of recycling
(Fig. 5A). The 10a inhibitor was then tested in short-cir-
cuit current experiments on human bronchial epithelial
cells in which the basolateral membrane had been perme-
abilized. The 10a inhibited CFTR-mediated chloride
transport in a weak but dose-dependent manner
(Fig. 5B). Together, these results suggest that 10a weakly
inhibits interaction at the NHERF-1 PDZ1 domain.
Structural analysis revealed the importance of hydrogen
Figure 4. (A) Comparison of the effects of reversible inhibitor 10a and
irreversible inhibitor 1. The protein and inhibitors were allowed to
equilibrate for the indicated times at 4 °C and were incubated with the
OregonGreen-PFDEDQHTQITTV peptide ligand for 0.5 h to mea-
sure the fluorescent polarization (FP). Data are reported as %
normalized by control experiments (no compounds). (B) Titration
curve for competitive inhibition of PDZ domain–ligand binding, as
measured by fluorescence polarization. The fluorescent peptide probe
and protein were allowed to equilibrate for 0.5 h with increasing
concentrations of 10a; mP = milipolarization units.
manner, whereas 10a did not (Fig. 4A). Therefore, the
inhibitory effect of 10a is reversible. To investigate if
10a non-specifically targets other proteins in cells, we
have prepared biotin-tagged2 10a to treat with crude

552
N. Fujii et al. / Bioorg. Med. Chem. Lett. 17 (2007) 549–552
the Fujisawa Pharmaceutical Co., Ltd. (NF), the UCSF
Stuart Trust Fund (NF, KAN, RKG), the Sandler Re-
search Foundation, and the Sidney Kimmel Research
Foundation (NF, RKG).
Supplementary data
Synthetic procedure, analysis data, and protocols for
biochemical assay and cell-based assay. Supplementary
data associated with this article can be found, in the on-
line version, at doi:10.1016/j.bmcl.2006.10.006.
Figure 5. (A) Inhibition of receptor recycling in HEK-293 cells.
Receptors were stimulated by each agonist; the receptor tags were
labeled with antibodies (FLAG); and surface receptors were quantified
by flow cytometry. The 10a inhibitor was added 0.5 h prior to the
internalization step. The indicated concentration of 10a was carried
though all subsequent steps. (B) Short-circuit current in permeabilized
human bronchial epithelial cells expressing human CFTR. CFTR
expression was stimulated by 100 lM CPT-cAMP, and 10a was added
in the apical bathing solution. The current was continuously recorded
using a voltage clamp.
References and notes
1. Harris, B. Z.; Lim, W. A. J. Cell Sci. 2001, 114, 3219.
2. Fujii, N.; Haresco, J. J.; Novak, K. A.; Stokoe, D.; Kuntz,
I. D.; Guy, R. K. J. Am. Chem. Soc. 2003, 125, 12074.
3. Dobrosotskaya, I.; Guy, R. K.; James, G. L. J. Biol.
Chem. 1997, 272, 31589.
4. Wu, Y.; Dowbenko, D.; Spencer, S.; Laura, R.; Lee, J.;
Gu, Q.; Lasky, L. A. J. Biol. Chem. 2000, 275, 21477.
5. Doyle, D. A.; Lee, A.; Lewis, J.; Kim, E.; Sheng, M.;
MacKinnon, R. Cell 1996, 85, 1067.
6. Chen Cy, C.; Lieberman, D. R.; Larsen, R. D.; Verho-
even, T. R.; Reider, P. J. J. Org. Chem. 1997, 62, 2676.
7. Novak, K. A.; Fujii, N.; Guy, R. K. Bioorg. Med. Chem.
Lett. 2002, 12, 2471.
8. Wilson, J.; Rossi, C. P.; Carboni, S.; Fremaux, C.; Perrin,
D.; Soto, C.; Kosco-Vilbois, M.; Scheer, A. J. Biomol.
Screen. 2003, 8, 522.
9. Hall, R. A.; Ostedgaard, L. S.; Premont, R. T.; Blitzer, J.
T.; Rahman, N.; Welsh, M. J.; Lefkowitz, R. J. Proc. Natl.
Acad. Sci. U.S.A. 1998, 95, 8496.
bond interaction on the D(À3) of the PDZ1 ligand.12 On
the basis of that study, the design and evaluation of
another compound with an R3 substitution mimicking
the D(À3) side-chain are underway.
We have found another indole-2-carbinol scaffold com-
pound showing >10 times potency in antagonizing the
interaction of Dishevelled (DVL) PDZ domain than
that of MAGI-3 PDZ2 (unpublished data). This shows
high potential of 10a as seed scaffold for further optimi-
zation. This study will be reported in due course.
10. Gage, R. M.; Matveeva, E. A.; Whiteheart, S. W.; von
Zastrow, M. J. Biol. Chem. 2005, 280, 3305.
11. Raghuram, V.; Mak, D. D.; Foskett, J. K. Proc. Natl.
Acad. Sci. U.S.A. 2001, 98, 1300.
Acknowledgments
We thank Angela McArthur for scientific editing of the
manuscript. Financial support was provided in part by
12. Karthikeyan, S.; Leung, T.; Ladias, J. A. J. Biol. Chem.
2001, 276, 19683.