Design and synthesis of highly potent and plasma-stable dimeric inhibitors of the PSD-95-NMDA receptor interaction

Article abstract of DOI:10.1002/anie.200904741

On the double: Dimerization of monomeric peptide ligands towards the PDZ domains of the protein PSD-95 (postsynaptic density 95) leads to potent inhibitors of protein-protein interactions with stability in blood plasma. Optimization of the length of the polyethylene glycol linker results in unprecedented affinity for inhibitors of the PDZ1-2 domain (see picture).

Full text of DOI:10.1002/anie.200904741

Angewandte
Chemie
DOI: 10.1002/anie.200904741
Protein–Protein Interactions
Design and Synthesis of Highly Potent and Plasma-Stable Dimeric
Inhibitors of the PSD-95–NMDA Receptor Interaction**
Anders Bach, Celestine N. Chi, Gar F. Pang, Lars Olsen, Anders S. Kristensen, Per Jemth, and
Kristian Strømgaard*
Protein–protein interactions (PPIs) mediate
numerous regulatory pathways, which are
vital for normal biological processes and
disease development.[1] Inhibitors of PPIs
are therefore of great value to unravel
complex cellular phenomena and as potential
therapeutics.[2] Postsynaptic density protein-
95/discs large/zonula occludens-1 (PDZ)
domains are part of scaffold and adaptor
proteins involved in the assembly of cellular
signaling complexes, typically by recognizing
the C terminal of the interacting proteins.[3]
They consist of about 90 amino acids and are
present in great number in multicellular
organisms; in humans 256 different PDZ
domains are found in 142 different proteins.[4]
Inhibition of PDZ-domain-mediated PPIs
represents a promising strategy for specific
therapeutic intervention of signaling events
rather than targeting entire signaling cas-
cades by receptor antagonists.[5]
Figure 1. a) PSD-95 facilitates Ca2+-mediated excessive NO production during excitotox-
Extensive activation of the N-methyl-d-
icity, which can be impaired by inhibition of the PSD-95–NMDA receptor interaction.
aspartate (NMDA) receptor by glutamate,[6]
b) Structural model of PDZ1-2, where two conserved histidine residues, important for
ligand binding, and their relative distance are highlighted.[11] SH3/GK=src homology 3/
as seen in excitotoxicity, plays a key role in
several brain diseases. However, develop-
ment of drugs that directly modulate the
NMDA receptor has been difficult.[5c,6,7]
guanylate kinase; src proteins cause sarcomas.
Instead, inhibition of the ternary complex of the NMDA
receptor, neuronal nitric oxide synthase (nNOS), and post-
synaptic density protein-95 (PSD-95) attenuates glutamate-
induced cell death by impairing the functional link between
Ca2+ entry and NO production without affecting vital synaptic
transmission (Figure 1a).[5e,8] Thus, PSD-95 constitutes a
promising target for neuroprotective drugs. Uncoupling of
PSD-95 and the NMDA receptor subunit, GluN2B, has been
achieved by a 20-mer peptide (Tat-N2B), which mimics the
C terminal of GluN2B and thereby inhibits the nNOS–
NMDA receptor linkage by binding to PDZ1 or PDZ2 of
PSD-95 (Figure 1). This peptide is currently in clinical trials as
a potential treatment of stroke.[8,9]
PDZ domains are known for their promiscuous ligand
recognition, and together with their plentiful representation
in the human genome it is likely that compounds directed
towards individual PDZ domains of PSD-95 are not very
selective.[10] Herein, we focus on dimeric ligands that bind
PDZ1 and PDZ2 of PSD-95 simultaneously. Dimerization
should not only increase affinity[11,12] but also enhance
selectivity,[5d] since the tandem PDZ1-2 motif (Figure 1b) is
structurally more unique than individual PDZ domains.
[*] Dr. A. Bach, G. F. Pang, Dr. L. Olsen, Dr. A. S. Kristensen,
Prof. K. Strømgaard
Department of Medicinal Chemistry, University of Copenhagen
Universitetsparken 2, 2100 Copenhagen (Denmark)
Fax: (+45)3533-6040
E-mail: krst@farma.ku.dk
Homepage: http://www.farma.ku.dk/chembiol
C. N. Chi, Dr. P. Jemth
Department of Medical Biochemistry and Microbiology
Uppsala Universitet
Husargatan 3, ing. C11, 75123 Uppsala (Sweden)
[**] This work was supported by The Drug Research Academy, Faculty of
Pharmaceutical Sciences, University of Copenhagen (A.B.), The
Danish Council for Independent Research, Medical Sciences (K.S.
and L.O.), and the Swedish Research Council (P.J.). PSD-
We have previously shown that
a
pentapeptide
(1, IESDV), which corresponds to the extreme GluN2B
C-terminal sequence, is sufficient to bind the PDZ1 and
PDZ2 domains of PSD-95 with similar affinity as the GluN2B
95=postsynaptic density 95; NMDA=N-methyl-d-aspartate.
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/anie.200904741.
Angew. Chem. Int. Ed. 2009, 48, 9685 –9689
ꢀ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
9685

Communications
undecapeptide, while further truncation impairs binding.[13]
PDZ1 and PDZ2 have similar selectivity profiles,[10,13] and we
therefore focused on designing symmetric dimeric ligands. As
a starting point we used pentapeptide 1 as the ligand, and
selected monodisperse polyethylene glycol (PEG) linkers for
dimerization, since PEG linkers are structurally flexible and
known to impose favorable pharmacokinetic properties.[14]
Previously, dimeric peptide ligands have been synthesized
by activating polydisperse PEG diacids as pentafluorophenyl
(Pfp) esters followed by cross-linking of the peptides by
addition of Pfp-PEG diester in the presence of 1-hydroxy-
benzotriazole, which was repeated five times over five days.[15]
In our hands, the synthesis of dimeric ligands by this method
with monodisperse PEG diacids resulted in an average yield
of 21%. However, we found that direct coupling of PEG
diacids to the peptides using 2-(1H-benzotriazole-1-yl)-
1,1,3,3-tetramethyluronium hexafluorophosphate (HBTU)
allowed much shorter reaction times (5 ꢀ 30 min) and gave
increased yields (30%; Scheme 1).
Figure 2. FP inhibition curves for compounds 1 and 2 towards PDZ1-2
(a) and PDZ2 (b) of PSD-95.
Supporting Information). Thus, PEG linkers with n = 12, 8,
6, or 4 are favored over those with n = 2, 1, or 0. However, the
affinities measured in the FP assay for compounds 2–5 were
too high for determining accurate binding constants (Ki
values).[17]
To preclude the possibility that the PEG linker contrib-
utes to the increased affinity through a direct interaction with
PDZ, we first verified that the dimeric ligand containing the
nonbinding pentapeptide IEAAA (9) was inactive. More-
over, dimerization of nonbinding peptide IESDD with 1 to
provide the asymmetric dimeric ligand 10 resulted in a two- to
fourfold decrease in affinity relative to 1. Identical results
were obtained for the mono-PEGylated derivative of 1,
compound 11. These data clearly suggest that the PEG linker
itself possesses no affinity towards the PDZ domains of PSD-
95 (Table S1 in the Supporting Information).
Since the affinities of 2–5 were too high for accurate
determination of Ki values by FP, we performed isothermal
titration calorimetry (ITC), which provided reliable dissoci-
ation constant (Kd) values as well as information on thermo-
dynamic parameters (enthalpy DH, entropy DS, and Gibbs
energy DG) and the stoichiometry between ligand and
protein. The ITC experiments confirmed that compounds 2–
5 were significantly more potent than 6–8 and established
PEG4 as the optimal linker. Hence, compound 5 displayed
the lowest Kd value of only (32 Æ 1.7) nm and is thereby an
exceptionally potent inhibitor of PPIs in general and 100-fold
more potent than 1. Compound 4 (PEG6 linker) was almost
equipotent to 5, whereas 2 and 3 (PEG12 and PEG8,
respectively) were slightly less potent (Figure 3 and
Table S2 in the Supporting Information). In general, these
results support a recent finding that increasing the PEG-
linker length beyond its optimum value only results in a
Scheme 1. Synthesis of dimeric compounds 2–8. DIPEA=N,N-diiso-
propylethylamine, TFA=trifluoroacetic acid.
Initially, we synthesized dimeric ligand 2 by cross-linking
resin-bound peptide 1 with PEG12 diacid (n = 12; Scheme 1).
This linker has a linear distance of about 50 ꢁ, which we
considered to be a suitable starting point for enabling
simultaneous ligand binding to both PDZ domains (Fig-
ure 1b).[16] The ligand was evaluated for binding affinity in an
in vitro fluorescence polarization (FP) assay, which measures
displacement of a peptide probe (Cy5-GluN2B) by the
compound of interest.[13] A dramatic increase in affinity was
observed for dimeric ligand 2 (inhibition constant Ki ꢀ
0.1 mm) compared to 1 (Ki = (2.9 Æ 0.10) mm) when tested at
the tandem PDZ1-2 construct. In contrast, 2 showed affinities
similar to 1 at the individual PDZ1 and PDZ2 domains
(Figure 2 and Table S1 in the Supporting Information), which
suggests that 2 binds as a dimer towards PDZ1-2.
Encouraged by this substantial increase in potency, we
explored the importance of linker length, and peptide 1 was
dimerized with PEGn diacids where n = 8, 6, 4, 2, 1, and 0 to
give compounds 3–8, respectively (Scheme 1). Testing of 3–8
in the FP assay using PDZ1-2 revealed that compounds 3–5
displayed affinities similar to that of 2, whereas 6–8 showed a
stepwise increase in Ki (Figure S1 and Table S1 in the
Figure 3. a) Kd values in nm from ITC shown as meanÆstandard error
of the mean (SEM). b) Thermodynamic bar graphs for selected
compounds (DH, ÀTDS, DG; all in kcalmolÀ1) towards PDZ1-2.
9686 www.angewandte.org
ꢀ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2009, 48, 9685 –9689

Angewandte
Chemie
relatively weak effect on affinity, whereas shortening the
linker leads to a more substantial reduction in affinity.[16]
The ITC data also revealed a number of important
features of the dimeric ligands. First, the dimeric ligands bind
in a 1:1 relationship with PDZ1-2, whereas monomeric
peptide 1 binds in a 2:1 relationship as expected (Table S2
in the Supporting Information). Second, dimerization of 1
with appropriate PEG linkers, compounds 2–5, favors PDZ
binding by a substantial enthalpy decrease, which, however, is
associated with a considerable entropy penalty. A reduced
decrease in the enthalpy is seen for compounds 6–8, and
although the entropy penalties are concomitantly smaller,
binding affinities are reduced compared to those of 2–5, as
also observed in the FP assay. Third, it is noticed that the
entropy penalty in general increases with linker length
(Figure 3, and Figure S2 and Table S2 in the Supporting
Information).
To further characterize the binding mechanism, we
performed kinetic studies by using fluorescence intensity
measurements. By applying PDZ1-2 constructs with a trypto-
phan residue introduced in either PDZ1 (PDZ1*-2) or PDZ2
(PDZ1-2*), we determined the kinetic rate constants by
measuring the altered fluorescence properties as a result of
ligand binding.[18] It was observed that the dimeric ligand 5
increased the off-rate constants in the individual PDZ
domains two- to threefold, but 5 dramatically decreased the
off-rate constants by six- to ninefold compared to 1 at PDZ1-2
(Figure S3 and Table S3 in the Supporting Information). On-
rate constants were increased for 5 relative to 1, for both
PDZ2 and PDZ1-2 to a similar degree (data not shown). The
increased affinity of dimeric ligands
Scheme 2. Synthesis of dimeric compounds 15–17.
towards individual PDZ domains was seen, but a dramatic
increase in affinity towards PDZ1-2 was observed for dimeric
ligands 15–17 (Figure 4a and Table S1 in the Supporting
Information).
The affinity of compound 16 was too high to be accurately
quantified by FP, hence ITC was performed, which revealed
an exceptionally high affinity towards PDZ1-2 with a Kd value
of (9.8 Æ 1.6) nm (Figure 4b and Table S2 in the Supporting
Information). This corresponds to a 145-fold increase com-
pared to monomeric ligand 13, and compound 16 is to the best
of our knowledge the most potent PDZ domain inhibitor yet
described. Furthermore, a strong linear correlation with
slope > 1 between the logarithmic Kd values for monomeric
versus dimeric ligands was observed (Figure 4c), which means
that the effect (fold-change) of dimerization increases with
increasing affinity of the monomeric ligand (Figure 4c and
Table S4 in the Supporting Information).
FP, ITC, and kinetic studies demonstrated that the
remarkable affinities of the dimeric ligands result from
linking two ligands together, thereby enabling simultaneous
compared to monomeric ligands can
therefore be explained by the
decreased dissociation rate constants,
which supports the view that 5 simulta-
neously binds to both PDZ1 and PDZ2
in PDZ1-2. In agreement with the
dissociation kinetics and the ITC meas-
urements, compound 5 binds PDZ1-2 in
a 1:1 stoichiometry as determined by
fluorescence endpoints from the kinetic
data (Figure S4 in the Supporting Infor-
mation).
Having optimized the linker length
and verified the binding mechanism, we
focused on optimization of the ligand
moiety. We have previously carried out
structure–activity relationship (SAR)
studies of monomeric penta- and tetra-
peptides, which, for example, revealed
that substitution of serine with a threo-
nine moiety increases the affinity two-
fold.[13] Based on these studies, the
monomeric ligands IATAV (12),
IETAV (13), IATAMeV (14), and their
corresponding dimeric ligands (15–17)
were synthesized (Scheme 2) and tested
in the FP assay. As demonstrated for
compounds 2–5, no change in affinity
Figure 4. a) FP inhibition curves for compounds 12–17 towards PDZ1-2 (left) and PDZ2 (right)
of PSD-95. b) ITC data for compound 16 towards PDZ1-2. c) Relationship between log Kd values
for monomeric ligands and the corresponding dimeric ligands.
Angew. Chem. Int. Ed. 2009, 48, 9685 –9689
ꢀ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 9687

Communications
binding to both PDZ domains. Moreover, the PEG linker
Peptide-based ligands are generally subject to enzymatic
cleavage by proteases in vivo, which often is the major
limiting factor for advanced biological studies.[14] We there-
fore evaluated the stability of the ligands in blood plasma, and
found that the dimeric ligands showed superior stability
(Figure 6 and Table S5 in the Supporting Information). The
itself has no apparent affinity towards the PDZ domains. To
examine this in a structural context, we performed molecular
modeling studies. Two models of the PDZ1-2 structure, with
and without peptide ligand, have been generated by NMR
structure elucidation combined with modeling and molecular
dynamics simulations.[11,19] We observed that compound 2
with the long PEG12 linker is well accommodated into both
models, with a curled linker in the ligand-free model and an
extended linker in the other (Figure 5a). In contrast, the more
Figure 6. In vitro stability in human blood plasma at 378C for key
dimeric (2, 5, and 16) and monomeric ligands (1 and 13) and Tat-
N2B. Half-lives (T1/2 [h]) are shown in parentheses (ÆSEM).
monomeric peptides (1 and 12–14) were degraded relatively
fast with half-lives (T1/2) of less than 1 h, but dimerization with
PEG linkers led to a seven- to ninefold increase in T1/2 in
blood plasma (15, 16) or even complete resistance towards
degradation as measured over a time period of 6 days (2–5,
17). The PEGylated monomeric ligand (11) demonstrated an
approximately 14-fold increase in T1/2 relative to 1; thus, the
PEG linker is the primary factor in mediating the increased
stability. However, since compounds 2–5 are still significantly
more stable than 11, the dimerization per se contributes to
stability. Thus, both the presence of PEG linkers and the
dimerization of ligands promote plasma stability.
In conclusion, we have designed and synthesized remark-
ably potent dimeric inhibitors of the PSD-95–NMDA recep-
tor interaction by linking pentapeptide ligands with mono-
disperse PEG linkers. Optimization of linker length and the
peptide moiety guided by FP and ITC assays led to the
identification of compound 16 as the most potent compound
with a Kd value of (9.8 Æ 1.6) nm, which is an unprecedented
affinity for a PDZ-mediated PPI. Furthermore, this com-
pound represents a 1000-fold improvement in terms of
affinity compared to the clinical candidate, Tat-N2B.[13]
These dimeric ligands could serve as excellent molecular
tools for studying interdomain flexibility of the PDZ1-2
domain of PSD-95, and are candidates for further exploration
towards development of in vivo neuroprotective compounds.
Finally, our studies represent a general and versatile strategy
for targeting tandemly arranged PDZ domains while achiev-
ing high potency, selectivity, and blood-plasma stability.
Figure 5. Modeling studies of dimeric ligands. a) Compound 2 and
b) compound 5 in association with two different models of PDZ1-2.
Ligands are shown in magenta, the ligand-free model[11] in green, and
the ligand-bound model[19] in blue. In each case the PDZ1 domains of
the two models have been aligned.
potent ligand 5 with the shorter PEG4 linker only fits into the
ligand-free model with an extended linker, but is too short to
bind both PDZ1 and PDZ2 in the ligand-bound model
(Figure 5b). Thus, modeling confirms that the dimeric ligands
can bind both PDZ domains in PDZ1-2, which explains the
observed increase in affinity and decrease in enthalpy, as well
as the decreased off-rate constants compared to monomeric
ligands. Also, the accessible binding cavity of PDZ1-2 allows
binding of the peptide ligand without any apparent accom-
panying interactions with the PEG linker (Figure 5).
The modeling studies also provide insights into the basis
for the entropy penalties observed for the dimeric ligands,
which could arise from restraining the ligand, the protein, or
both. It has been suggested that binding of monomeric
peptide ligands induces interdomain mobility and protein
flexibility.[11,19] This interdomain mobility can only be facili-
tated for PDZ1-2 when bound to 2 but not 5, according to the
modeling studies, which in terms of protein flexibility suggests
that 2 binds more favorably than 5. However, ITC experi-
ments reveal that the entropy penalty paid by 5 is smaller than
that for 2, thus indicating that confinement of the ligand, and
not PDZ1-2, is more decisive in the observed entropy
penalties of the binding reaction. This is important not only
from a mechanistic point of view, but also in the design of
future dimeric ligands.
Received: August 25, 2009
Published online: November 24, 2009
Keywords: bridging ligands · dimerization · inhibitors · protein–
.
protein interactions · receptors
9688 www.angewandte.org
ꢀ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2009, 48, 9685 –9689

Angewandte
Chemie
[10] M. A. Stiffler, J. R. Chen, V. P. Grantcharova, Y. Lei, D. Fuchs,
[1] a) J. A. Wells, C. L. McClendon, Nature 2007, 450, 1001; b) T.
Berg, Curr. Opin. Drug Discovery Dev. 2008, 11, 666.
[2] T. U. Mayer, Trends Cell Biol. 2003, 13, 270.
[3] a) E. Kim, M. Sheng, Nat. Rev. Neurosci. 2004, 5, 771; b) R.
Tonikian, et al., PLoS Biol. 2008, 6, e239; c) W. Feng, M.
Zhang, Nat. Rev. Neurosci. 2009, 10, 87; d) P. Jemth, S. Gianni,
Biochemistry 2007, 46, 8701.
J. E. Allen, L. A. Zaslavskaia, G. MacBeath, Science 2007, 317,
364.
[11] J. F. Long, H. Tochio, P. Wang, J. S. Fan, C. Sala, M. Niethammer,
M. Sheng, M. Zhang, J. Mol. Biol. 2003, 327, 203.
[12] a) D. H. Williams, E. Stephens, D. P. OꢂBrien, M. Zhou, Angew.
Chem. 2004, 116, 6760; Angew. Chem. Int. Ed. 2004, 43, 6596;
b) R. H. Kramer, J. W. Karpen, Nature 1998, 395, 710; c) L. Tian,
T. Heyduk, Biochemistry 2009, 48, 264; d) H.-X. Zhou, Bio-
chemistry 2001, 40, 15069; e) H.-X. Zhou, J. Mol. Biol. 2003, 329,
1.
[13] A. Bach, C. N. Chi, T. B. Olsen, S. W. Pedersen, M. U. Røder,
G. F. Pang, R. P. Clausen, P. Jemth, K. Strømgaard, J. Med.
Chem. 2008, 51, 6450.
[14] J. M. Harris, N. E. Martin, M. Modi, Clin. Pharmacokinet. 2001,
40, 539.
[15] M. Paduch, M. Biernat, P. Stefanowicz, Z. S. Derewenda, Z.
Szewczuk, J. Otlewski, ChemBioChem 2007, 8, 443.
[16] V. M. Krishnamurthy, V. Semetey, P. J. Bracher, N. Shen, G. M.
Whitesides, J. Am. Chem. Soc. 2007, 129, 1312.
[17] X. Huang, J. Biomol. Screening 2003, 8, 34.
[18] S. Gianni, ꢁ. Engstrꢃm, M. Larsson, N. Calosci, F. Malatesta, L.
Eklund, C. N. Chi, C. Travaglini-Allocatelli, P. Jemth, J. Biol.
Chem. 2005, 280, 34805.
[4] I. Letunic, R. R. Copley, B. Pils, S. Pinkert, J. Schultz, P. Bork,
Nucleic Acids Res. 2006, 34, D257.
[5] a) K. K. Dev, Nat. Rev. Drug Discovery 2004, 3, 1047; b) L. L.
Blazer, R. R. Neubig, Neuropsychopharmacology 2009, 34, 126;
c) F. Gardoni, M. Di Luca, Eur. J. Pharmacol. 2006, 545, 2; d) W.
Wen, W. Wang, M. Zhang, Curr. Top. Med. Chem. 2006, 6, 711;
e) F. X. Soriano, et al., J. Neurosci. 2008, 28, 10696; f) N. X.
Wang, H. J. Lee, J. J. Zheng, Drug News Perspect. 2008, 21, 137.
[6] R. Dingledine, K. Borges, D. Bowie, S. F. Traynelis, Pharmacol.
Rev. 1999, 51, 7.
[7] a) L. Hoyte, P. A. Barber, A. M. Buchan, M. D. Hill, Curr. Mol.
Med. 2004, 4, 131; b) C. Ikonomidou, L. Turski, Lancet Neurol.
2002, 1, 383.
[8] M. Aarts, Y. Liu, L. Liu, S. Besshoh, M. Arundine, J. W. Gurd,
Y. T. Wang, M. W. Salter, M. Tymianski, Science 2002, 298, 846.
[9] Thomson Current Drugs (http://www.thomson.com).
[19] W. Wang, J. Weng, X. Zhang, M. Liu, M. Zhang, J. Am. Chem.
Soc. 2009, 131, 787.
Angew. Chem. Int. Ed. 2009, 48, 9685 –9689
ꢀ 2009 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 9689