Indole- and indoline-based kainate analogues with antagonist activity at ionotropic glutamate receptors

Article abstract of DOI:10.1016/j.bmcl.2005.05.098

A conformationally constrained, indole-based kainate analogue was designed based on Gouaux's X-ray structure of kainic acid bound to an iGluR2(S1S2) construct, a structural model for AMPA/kainate ionotropic glutamate receptors. In contrast to the parent kainic acid, a potent agonist, this compound, along with three structurally related analogues derived from synthetic intermediates, exhibited antagonist behavior towards KAR expressed in oocytes, a result that is rationalized by molecular modeling studies.

Full text of DOI:10.1016/j.bmcl.2005.05.098

Bioorganic & Medicinal Chemistry Letters 15 (2005) 3942–3947
Indole- and indoline-based kainate analogues with
antagonist activity at ionotropic glutamate receptors
Xiaohong Shou,a Ricardo Miledib and A. Richard Chamberlina,*
aDepartment of Chemistry, University of California, Irvine, CA 92697, USA
bDepartment of Neurobiology and Behavior, University of California, Irvine, CA 92697, USA
Received 14 June 2004; revised 18 May 2005; accepted 24 May 2005
Available online 6 July 2005
Abstract—A conformationally constrained, indole-based kainate analogue was designed based on GouauxÕs X-ray structure of
kainic acid bound to an iGluR2(S1S2) construct, a structural model for AMPA/kainate ionotropic glutamate receptors. In contrast
to the parent kainic acid, a potent agonist, this compound, along with three structurally related analogues derived from synthetic
intermediates, exhibited antagonist behavior towards KAR expressed in oocytes, a result that is rationalized by molecular modeling
studies.
Ó 2005 Elsevier Ltd. All rights reserved.
L-Glutamate represents the primary excitatory neuro-
transmitter in the mammalian central nervous system,
playing an essential role in mediating basal excitatory
synaptic transmission and many forms of synaptic plas-
ticity, such as long-term potentiation and long-term
depression, that are thought to underlie learning and
memory.1,2 Glutamate controls these functions by bind-
ing to ligand-gated ion channels (ionotropic receptors,
iGluRs) and G-protein-coupled metabotropic receptors,
initiating a conformational change in the cation-con-
ducting pore and thus an influx of cations into the post-
synaptic neuron.3 The ionotropic glutamate receptors
are subdivided into three categories, AMPARs, KARs,
and NMDARs, based on the selective pharmacologic
agonists a-amino-3-hydroxy-5-methyl-4-isoxazole-pro-
pionic acid (AMPA), kainic acid (kainate, KA), and
N-methyl-D-aspartate (NMDA), that, respectively, acti-
Figure 1. Examples of ionotropic glutamate receptor (iGluR) agonists
and antagonist.
vates each group with some specificity (Fig. 1). The
study of the specific roles of the AMPARs and KARs
has been hindered by the similarity between the two
There have been many efforts to characterize the molec-
(40% sequence homology) and the presence of multiple
ular interactions between the iGluRs and various ago-
subunits, iGluR1–4 for AMPARs and iGluR5–7,
nists and antagonists with which they interact. A
KA1–2 for KARs,4 that comprise each group.
landmark in this area was the successful crystallization
and structure determination by Gouaux et al.5a of the
iGluR2 ligand-binding core (S1S2) complexed with kai-
nic acid. This group has also reported X-ray structures
of the iGluR2-S1S2 construct with several other
agonists, including glutamate, AMPA, (S)-2-amino-
Keywords: Glutamate receptor; iGluR2; Kainic acid; AMPA; Agonist;
Antagonist; DNQX; Indole; Indoline.
3-[3-hydroxy-5-(2-methyl-2H-tetrazol-5-yl) isoxazol-4-
yl]propionic acid (2-Me-Tet-AMPA), (S)-2-amino-3-(3-
*
Corresponding author. Tel.: +1 949 824 7089; fax: +1 949 824
8571; e-mail: archambe@uci.edu
0960-894X/$ - see front matter Ó 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2005.05.098

X. Shou et al. / Bioorg. Med. Chem. Lett. 15 (2005) 3942–3947
3943
carboxy-5-methylisoxazol-4-yl)propionic acid (ACPA),
(S)-2-amino-3-(3-hydroxy-5-tert-butylisoxazol-4-yl) propi-
onic acid (ATPA), (S)-2-amino-3-(4-bromo-3-hydroxy-
5-isoxazolyl)propionic acid (Br-HIBO), willardiines, as
well as with the antagonists 6,7-dinitroquinoxaline-
2,3-dione (DNQX) and 2-amino-3-[5-tert-butyl-3-
(phosphonomethoxy) 4-isoxazolyl]propionic acid (ATPO)
(Fig. 1).5 The crystal structures have provided in-
sight into which iGluR2 residues are important for ligand
affinity and specificity, and beyond that have afforded con-
vincing evidence for the proposal that ligand-induced
channel opening in the native receptor arises from closure
of a ÔhingeÕ between two domains (S1 and S2) of iGluR2.5e
This hypothesis is based, in part, on a good correlation be-
tween agonist activity in native receptors with the degree
of agonist-induced domain closure in the construct. A sali-
ent feature of this model is that full agonists induce full
closure, partial agonists induce partial closure, and antag-
onists block closure—and hence channel opening—by
interfering with crucial S1–S2 contacts that must form as
the hinge closes.
Figure 2. Proposed conformationally constrained kainate analogue 2.
tion, indole 2-carboxylates, including 2, have been
reported to bind to the glycine site on the NMDA
receptors.11
The parent structure 2 was conceived by inspection of a
Gouaux model (Fig. 3A), noting in particular the dispo-
sition of the five-membered ring substituents in the bound
kainic acid, which appeared to be reinforced by allylic
strain between the isopropenyl group and the C-3 carb-
oxymethyl side chain. Mindful of the fact that reduced
basicity resulting from an N-aryl substitution on the
a-amino group in the analogue could thwart the plan
on account of the potential reduction in receptor affinity
of an unprotonated amino group, there appeared to be
ample space for a phenyl ring that could replace an iso-
propenyl group and at the same time bias the pyrrolidine
ring toward any desired conformation (although the pyr-
rolidine envelope in the analogue would clearly be less
pronounced than that shown; see Fig. 3B for energy-min-
imized structures). In fact, filling the Ôempty spaceÕ in the
vicinity of E402 and P478 with the aryl ring appeared to
provide additional favorable van der Waals interactions
in our preliminary modeling studies.12
Aside from providing a paradigm for agonist and antag-
onist activities, this structural information serves as the
basis for the design of new ligands that might interact
with iGluRs with greater selectivity. The native recep-
tors consist of multiple subtypes, as discussed above,
that share a high level of sequence homology that results
in considerable overlap of ligand binding specificity;6,7
for example, kainic acid binds not only to receptors
comprised of the KAR subtypes but also to the AMPA
family that includes iGluR1–4. Employing a combina-
tion of structure-based ligand design, homology model-
ing, and library synthesis, we have embarked on a
project to identify new subtype-selective iGluR agonists
and antagonists, and in this communication we report
initial results on a newly identified class of antagonists
based on the structure of the classic agonist kainic acid.
The synthesis of dihydroindoline 2 commenced with the
formation of the 2,3-disubstituted indole 3 (Fig. 4),
which had previously been converted into similar dihy-
droindoline derivatives.13 Condensation of phenylhydr-
azine hydrochloride with ketoglutaric acid provided
the corresponding phenyl hydrazone, which underwent
Fischer cyclization in the presence of polyphosphoric
acid to generate indole 3.14 Attempts at hydrogenating
indole 3 were unsuccessful, but the corresponding Boc-
protected indole was smoothly reduced to give the race-
mic cis-dihydroindoline 4; it has been proposed that
electron-withdrawing N-substituents facilitate hydroge-
nations in such systems by reducing the electron density
of the indole ring.14 Epimerization of the cis-indoline
with KO(t-Bu) afforded a mixture of trans- and cis-dihy-
droindolines (8:1 ratio) in a total yield of 75%. The
trans-indoline 5 was separated and then converted into
the designed ligand 2 by deprotection and saponifica-
tion. Meanwhile, synthetic intermediates 3 and 4 were
deprotected and saponified to generate additional ana-
logues, indole 6 and the cis-dihydroindoline 7, respec-
tively. A fourth structurally related analogue, 8, was
also readily available from an abortive earlier route to
2; Friedel–Crafts reaction of the indole-2-carboxylate,
1. Design and synthesis
The activity of kainic acid analogues toward the KARs
depends heavily on the nature of the C-4 substituent.
For example, if the isopropenyl appendage in kainate
is hydrogenated, the resultant analogue dihydrokainate
(DHK) loses activity at the AMPARs and KARs, and
becomes a moderately active inhibitor at the sodium-
dependent transporter; conversely, the absence of a C-
4 appendage results in an analogue that is active as an
NMDA receptor agonist.8 This sensitivity toward C-4
substituent changes, together with the Gouaux model
of the iGluR2–kainate complex, provides a basis for
designing new kainate analogues that might pharmaco-
logically differentiate among the iGluR subtypes. In this
communication, we describe our preliminary studies on
a series of kainate-like analogues that are constrained to
resemble bound conformation (Fig. 2), with an isopro-
penyl methyl group surrogate R that eventually can be
structurally varied (R = H in these initial studies). Many
kainic acid analogues have been reported previously,
including those modified at C3–C5,9 and others with
various linkers as conformational constraints.10 In addi-

3944
X. Shou et al. / Bioorg. Med. Chem. Lett. 15 (2005) 3942–3947
Figure 3. (A) Stereoview of kainic acid in the binding pocket of iGluR2S1S2 construct, from a Gouaux X-ray structure.5 (B) Stereoview of indoline
analogue 2 docked and minimized in the binding pocket of iGluR2S1S2 construct.
Figure 4. Synthesis of indoline/indole ligands 2, 6, 7, and 8. Reagents and conditions: (a) H2O, rt; (b) HCl, PPA, MeOH (66%, two steps); (c) Boc2O,
DMAP, Et3N (77%); (d) Pd/C, H2, AcOH (85%); (e) KO(t-Bu), followed by 1N AcOH (66%); (f) LiOH, THF/H2O, followed by 1N AcOH (90% for
6, 80% for 8); (g) TFA; (h) LiOH, THF/H2O, followed by 1N HCl (40% for 2, 50% for 7, two steps); (i) MeOH, AcCl; (j) AlCl3, ethyl malonyl
chloride (64%, two steps).

X. Shou et al. / Bioorg. Med. Chem. Lett. 15 (2005) 3942–3947
3945
prepared from the commercially available indole-2-
carboxylic acid, followed by saponification provided
the indole analogue 8.13 All four analogues—the target
trans-dihydroindoline 2 and the structurally related
compounds 6, 7, and 815—were prepared in ample quan-
tities for bioassays.
the antagonist assay showed that ligands 6 and 8 at
100 lM inhibited approximately one-third of the current
induced by 50 lM kainic acid. While ligands 2 and 7
show little or no activity at these 100 lM concentra-
tions, blocking only 2% and 5% of the current, respec-
tively, they do block currents (7% and 12%,
respectively) at lower concentrations (30 lM) of kainic
acid (data not shown). For comparison, the very potent
antagonist DNQX blocked 30% of the current at a con-
centration of 0.1 lM.
2. Biological assays
The activities of compound 2 and the three structurally
related analogues 6–8 were screened in oocytes injected
with native RNA from rat cerebral cortex.16 Potential
agonist activity was assayed based on the membrane
current generated at concentrations of 100 lM for each
analogue against varying concentrations (10, 30, and
100 lM) of kainic acid as control (Fig. 5A). The same
compounds were screened for antagonist activity by
measurements of 50 lM kainate-induced membrane cur-
rents inhibited by the presence of each analogue
(100 lM) (Fig. 5B). For comparison, the very potent
benchmark antagonist DNQX was also assayed in this
system. Experiments were highly reproducible in six oo-
cyte preparations, with a standard error of only a few
percent, as indicated by the error bars in Figure 5B.
3. Discussion
The results of the biological assays of compounds 2, 6, 7,
and 8 clearly show that they are KAR antagonists rather
than agonists, despite the fact that 2 was designed around
the structure of the agonist kainic acid, and that 2 has pre-
viously been reported to be inactive in KAR-binding as-
says.11b The antagonist activity of these analogues
prompted us to dock them in the binding site of the
iGluR2S1S2 construct, but beginning with the structure
of the complex of the known antagonist DNQX,8 rather
than with the kainic acid complex (the L-isomers of 2
and 7 were modeled, by analogy to natural kainate).
Docked and minimized as before,11 the analogues adopt
orientations similar to DNQX in the active site (Fig. 6;
for clarity, only analogue 6 is shown) in which the respec-
tive phenyl groups push residues E402, P478, T686, and
E705 farther apart compared with agonist complex
iGluR2S1S2–kainic acid. As mentioned above, Gouaux
has postulated that domain closure is crucial for activa-
tion of the receptor, but the binding of the 7-nitro group
of DNQX to T686 blocks the interaction of T686 with
E402, which is a critical contact in this process. Similarly,
the phenyl group of each analogue blocks this contact and
perhaps also stabilizes the open state of the clamshell-like
S1S2 core, both of which could contribute to the observed
antagonist activity.
In the agonist activity assay, there was no current ob-
served for any of the analogues at concentrations up
to 100 lM, while the control agonist kainic acid induced
currents at concentrations as low as 10 lM. In contrast,
Interestingly, all four analogues show antagonist behav-
ior at similar potencies, although indole 6 and 8 are
somewhat more potent than either the designed indoline
analogue 2 or its cis-isomer 7. In addition, the results
suggest that binding affinities for 6 and 8 are only mar-
ginally lower than that of kainic acid itself, but approxi-
mately 1000-fold less than that of the very potent, non-
selective antagonist DNQX (the potencies could be high-
er for enantiomerically pure analogues). In retrospect,
the antagonist behavior of these compounds is perhaps
not surprising, given the fact that these bicyclic aromatic
compounds are somewhat reminiscent of the DNQX
structure and appear to assume very similar orientations
in the binding site to that of DNQX. Specifically, a com-
parison of the probable binding interactions for each
ligand suggests that 6 and DNQX generally form better
contacts than does the L-indoline 2. The a-carboxyl
group of all ligands forms hydrogen bonds with the
guanidinium group of R485 and the main chain NH of
T480. Also, the a-amino group of the ligands can interact
with the carbonyl group of P478, the hydroxy group of
T480, and the side-chain carboxylate of E705. Some of
these important interactions appear to be absent in
Figure 5. Assays of indoline/indole ligands 2, 6, 7, and 8 in oocytes
injected with native RNA from rat cerebral cortex. (A) Agonist activity
assay: membrane current generated from the analogues compared to
kainic acid control. (B) Antagonist activity assay: percentage of current
induced by kainic acid (50 lM) in the presence of the analogues
(100 lM); no analogue = 100%; DNQX was tested at 0.1 lM.

3946
X. Shou et al. / Bioorg. Med. Chem. Lett. 15 (2005) 3942–3947
Figure 6. (A) Stereoview of antagonist DNQX in the binding pocket of iGluR2S1S2 construct X-ray structure. (B) Stereoview of indole analogue 6
docked and minimized in the binding pocket of an iGluR2S1S2 construct.
indoline 2 complex, as residues E705, R485, T480, and
P478 are several Angstroms farther from the ligand in
the minimized structures of these complexes. Likewise,
in the cis-indoline 7 complex, the essential residues are
even farther removed. On the other hand, most of these
interactions are maintained in the indole complexes 6
and 7 (not shown), as they are in the DNQX complex. It
also appears that a stronger p–p interaction between 6
and the aryl ring of Y450 may contribute significantly to
ligand binding, as compared to the indoline 2 complex,
in which Y450 is pushed farther away by its ligand a-car-
boxylate (not shown). While any computational study is
inherently speculative, we believe that these models
provide a reasonable working hypothesis for the observed
antagonist activity of this family ofanalogues, and further
studies are underway based on this model.
activity. A refined computational model suggests a mode
of binding that is reminiscent of that found among the
non-selective quinoxalinedione antagonists, such as
DNQX. The change to antagonist behavior (compared
to kainate) likely results, as proposed by Gouaux for
DNQX, from additional interactions of the indole aryl
ring that block the interdomain closure necessary for
channel opening. This indole structure, aside from being
an interesting antagonist in its own right, also serves as a
core scaffold for library synthesis of potential iGluR
receptor subtype antagonists, guided by our homology
models of iGluR1, 3–7 complexes (unpublished), and
the recently reported crystal structures of iGluR5 and
iGluR6 ligand binding cores.17
Acknowledgments
4. Conclusion
We are grateful for support from the National Institute
of Neurological Disorders and Stroke (NS-27600 to
A.R.C.).
A conformationally constrained kainate analogue was
designed based on GouauxÕs X-ray structure of an
iGluR2S1S2 construct complexed with kainic acid. It
was synthesized in racemic form and, along with three
other structurally related analogues, was screened for
activity in oocyte assays. Although the analogues were
originally designed as kainate-like agonists, their in-
dole-based structures, instead, resulted in antagonist
References and notes
1. Sheridan, P. H.; Weight, F. F.; Herman, B. H. In
Glutamate and Addiction; Herman, B., Ed.; Humana
Press: Totowa, 2003, pp 3–71.

X. Shou et al. / Bioorg. Med. Chem. Lett. 15 (2005) 3942–3947
3947
2. Shou, X.; Chamberlin, A. R. Exp. Opin. Ther. Patents
2004, 14, 471.
G. R. J. Med. Chem. 1992, 35, 1942; (b) Rowley, M.;
Leeson, P. D.; Grimwood, S.; Foster, A.; Saywell, K.
Bioorg. Med. Chem. Lett. 1992, 2, 1627; (c) Noe, C. R.;
Knollmueller, M.; Schoedl, C.; Berger, M. L. Sci. Pharm.
1996, 64, 577.
3. Moloney, M. G. Nat. Prod. Rep. 2002, 19, 597.
4. (a) Brauner-Osborne, H.; Egebjerg, J.; Nielsen, E. O.;
Madsen, U.; Krogsgaard-Larsen, P. J. Med. Chem. 2000,
43, 2609; (b) Lerma, J.; Paternain, A. V.; Rodriguez-
Moreno, A.; Lopez-Garcia, J. C. Physiol. Rev. 2001, 81, 971.
5. (a) Armstrong, N.; Sun, Y.; Chen, G. Q.; Gouaux, E.
Nature (London) 1998, 395, 913; (b) Jin, R. S.; Horning,
M.; Mayer, M. L.; Gouaux, E. Biochemistry 2002, 41,
15635; (c) Jin, R. S.; Gouaux, E. Biochemistry 2003, 42,
5201; (d) Hogner, A.; Kastrup, J. S.; Jin, R.; Liljefors, T.;
Mayer, M. L.; Egebjerg, J.; Larsen, I. K.; Gouaux, E.
J. Mol. Biol. 2002, 322, 93; (e) Armstrong, N.; Gouaux, E.
Neuron 2000, 28, 165; (f) Hogner, A.; Greenwood, J. R.;
Liljefors, T.; Lunn, M. L.; Egebjerg, J.; Larsen, I. K.;
Gouaux, E.; Kastrup, J. S. J. Med. Chem. 2003, 46, 214;
(g) Jin, R.; Banke, T. G.; Mayer, M. L.; Traynelis, S. F.;
Gouaux, E. Nat. Neurosci. 2003, 6, 803; (h) Lunn, M. L.;
Hogner, A.; Stensbol, T. B.; Gouaux, E.; Egebjerg, J.
J. Med. Chem. 2003, 46, 872.
12. Molecular dynamics and minimizations of the ligand/
iGluR2S1S2 construct complexes were carried out on a
Silicon Graphics Octane 2 computer using the Discover_3
module on Insight II version 2000. Dynamics were carried
out at 300 K. The residues of iGluR2S1S2 construct
˚
within 7.0 A of each ligand were allowed to relax.
13. Collot, V.; Schmitt, M.; Marwah, P.; Bourguignon, J. J.
Heterocycles 1999, 51, 2823.
14. Robinson, J. R.; Good, N. R. Can. J. Chem. 1957, 35,
1578.
15. Spectra data of selected compounds, 2: 1H NMR
(400 MHz, D2O) d 3.06 (dd, J = 6.0, 17.1 Hz, 1H), 3.13
(dd, J = 5.8, 17.1 Hz, 1H), 4.18 (dd, J = 5.8, 5.8 Hz, 1H),
4.83 (d, J = 5.8 Hz, 1H), 7.53 (m, 4H); 13C NMR
(100 MHz, D2O) d 175.1, 172.2, 135.7, 135.1. 129.9,
129.6, 125.4, 118.9, 64.7, 42.7, 37.7; 6: 1H NMR
(400 MHz, MeOH-d) d 4.17 (s, 2H), 7.07 (m, 1H), 7.25
(m, 1H), 7.41 (dd, J = 0.8, 8.3 Hz, 1H), 7.41 (dd, J = 0.8,
8.2 Hz, 1H); 13C NMR (125 MHz, MeOH-d) d 175.9,
165.4, 137.8, 129.2, 126.2, 126.0, 121.2, 121.1, 116.9, 113.3,
31.3; 7: 1H NMR (500 MHz, D2O) d 2.88 (dd, J = 7.2,
17.2 Hz, 1H), 2.95 (dd, J = 5.4, 17.2 Hz, 1H), 4.25 (m,
1H), 5.06 (d, J = 9.0 Hz, 1H), 7.50 (m, 4H); 13C NMR
(125 MHz, D2O) d 175.6, 172.5, 136.4, 135.7, 130.5, 130.0,
126.1, 119.3, 64.6, 40.0, 35.6; 8: 1H NMR (500 MHz,
MeOH-d) d 7.29 (m, 1H), 7.37 (m, 1H), 7.53 (d,
J = 8.3 Hz, 1H), 8.09 (d, J = 8.2 Hz, 1H); 13C NMR
(125 MHz, MeOH-d) d 187.2, 167.7, 163.0, 137.4, 134.8,
128.4, 127.0, 124.7, 123.1, 115.1, 114.0.
6. Pentikainen, O. T.; Settimo, L.; Keinanen, K.; Johnson,
M. S. Biochem. Pharmacol. 2003, 66, 2413.
7. Mayer, M. L.; Olson, R.; Gouaux, E. J. Mol. Biol. 2001,
311, 815.
8. Sonnenberg, J. D.; Koch, H. P.; Willis, C. L.; Bradbury,
F.; Dauenhauer, D.; Bridges, R. J.; Chamberlin, A. R.
Bioorg. Med. Chem. Lett. 1996, 6, 1607.
9. (a) Moloney, M. G. Nat. Prod. Rep. 1999, 16, 485; (b)
Maeda, H.; Selvakumar, N.; Kraus, G. A. Tetrahedron
1999, 55, 943; (c) Baldwin, J. E.; Fryer, A. M.; Pritchard,
G. J. J. Org. Chem. 2001, 66, 2588; (d) Collado, I.;
Ezquerra, J.; Mateo, A. I.; Pedregal, C.; Rubio, A. J. Org.
Chem. 1999, 64, 4304; (e) Jefferies, I. Bioorg. Med. Chem.
Lett. 1992, 2, 1519.
16. Ragsdale, D. S.; Miledi, R. Proc. Natl. Acad. Sci. U.S.A.
1991, 88, 1854.
17. (a) Mayer, M. L. Neuron 2005, 45, 539; (b) Naur, P.;
Vestergaard, B.; Skov, L. K.; Egebjerg, J.; Gajhede, M.;
Kastrup, J. S. FEBS Lett. 2005, 579, 1154.
10. Collado, I.; Ezquerra, J.; Mateo, A. I.; Rubio, A. J. Org.
Chem. 1998, 63, 1995.
11. (a) Carling, R. W.; Leeson, P. D.; Moseley, A. M.; Baker,
R.; Foster, A. C.; Grimwood, S.; Kemp, J. A.; Marshall,