A novel series of positive modulators of the AMPA receptor: Discovery and structure based hit-to-lead studies

Article abstract of DOI:10.1016/j.bmcl.2010.07.138

Starting from an HTS derived hit 1, application of biostructural data facilitated rapid optimization to lead 22, a novel AMPA receptor modulator. This is the first demonstration of how structure based drug design can be exploited in an optimization program for a glutamate receptor.

Full text of DOI:10.1016/j.bmcl.2010.07.138

Bioorganic & Medicinal Chemistry Letters 20 (2010) 5753–5756
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
A novel series of positive modulators of the AMPA receptor: Discovery
and structure based hit-to-lead studies
Craig Jamieson a, , Stephanie Basten b, Robert A. Campbell a, Iain A. Cumming a, Kevin J. Gillen a,
⇑
Jonathan Gillespie a, Bert Kazemier b, Michael Kiczun a, Yvonne Lamont a, Amanda J. Lyons a,
John K. F. Maclean a, Elizabeth M. Moir a, John A. Morrow a, Marianthi Papakosta a, Zoran Rankovic a,
Lynn Smith a
a Merck Research Laboratories, MSD, Motherwell, Lanarkshire ML1 5SH, UK
b Merck Research Laboratories, MSD, PO Box 20, Oss 5340 BH, Netherlands
a r t i c l e i n f o
a b s t r a c t
Article history:
Starting from an HTS derived hit 1, application of biostructural data facilitated rapid optimization to lead
22, a novel AMPA receptor modulator. This is the first demonstration of how structure based drug design
can be exploited in an optimization program for a glutamate receptor.
Received 29 June 2010
Revised 30 July 2010
Accepted 31 July 2010
Available online 13 August 2010
Ó 2010 Elsevier Ltd. All rights reserved.
Keywords:
AMPA receptor
Allosteric modulator
SBDD
Hit-to-lead
The
a
-amino-3-hydroxy-5-methyl-4-isoxazole-propionic acid
Our research efforts were aimed at the identification of a novel
series of AMPA receptor positive modulators which are of potential
utility in the treatment of the kinds of neurological disorders delin-
eated above. Following a high-throughput screening campaign
using a functional assay with GluA1 overexpressed in HEK cells7,
we identified compound 1 (Fig. 1) as a hit with confirmed activity
from a well characterized solid sample. Compounds were inactive
in the absence of glutamate, thus confirming their allosteric nature.
While compound 1 represented a tractable chemotype amena-
ble to optimization and was deemed to have satisfactory potency
(pEC50 = 6.7), further in vitro profiling revealed a number of defi-
ciencies that required attention. Specifically, measured solubility
was low (<1 mg/L) and there were attendant issues with micro-
(AMPA) receptors are ionotropic glutamate receptors which are
abundantly expressed in the central nervous system and are be-
lieved to facilitate the majority of fast excitatory amino acid neuro-
transmission.1 The role of AMPA receptors appears to be critical to
mediating synaptic plasticity and long-term potentiation (LTP), the
use dependent enhancement in synaptic efficacy which is thought
to encode various forms of learning and memory. AMPA receptor
modulators have been shown to enhance LTP and are, therefore,
under serious consideration as therapeutic agents for a range of
neurological disorders including schizophrenia, Alzheimer’s dis-
ease, Parkinson’s disease and ADHD.2,3
Subunits of the AMPA receptor are encoded by four distinct
genes (GluA1–4), with each subunit comprising four domains: N-
terminal, extracellular glutamate binding site (Ligand Binding do-
main, LBD), a transmembrane region and a C-terminal domain.4
In the last decade or so, significant advances have been made in
studying the three-dimensional structure of the AMPA receptor
with structures reported from a construct relating to the LBD
(S1S2)5 and latterly the full-length receptor.6
somal stability (Rat/human intrinsic clearance, Cli >270
mg protein).
lL/min/
From the outset of the program, we had access to biostructural
data using the isolated S1S2 LBD GluA2 construct5 thus enabling an
understanding of how our compounds interacted with the recep-
tor.8 Figure 1 depicts hit compound 1 in complex with the S1S2
LBD GluA2 construct.9 Although our biological assay used the
GluA1 receptor (wild type GluA2 is not permeable to calcium), in
our hands X-ray crystallographic study of selected compounds in
both GluA1 and GluA2 constructs showed the structures to be
identical.10 Given that structures could be more expediently
obtained with GluA2, we selected this construct for routine exam-
ination. Modulators of the AMPA receptor bind remotely to the
⇑
Corresponding author. Address: Department of Pure
University of Strathclyde, 295 Cathedral St., Glasgow G1 1XL, United Kingdom. Tel.:
+44 141 548 4830; fax: +44 141 548 4822.
& Applied Chemistry,
E-mail address: craig.jamieson@strath.ac.uk (C. Jamieson).
0960-894X/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2010.07.138

5754
C. Jamieson et al. / Bioorg. Med. Chem. Lett. 20 (2010) 5753–5756
solubility (<1 mg/L) or metabolic stability (human and rat
Cli >270 L/min/mg protein). Insertion of a basic amine (6) or
l
introduction of a pendant alcohol (7) led to a more than 10-fold
reduction in potency.
In parallel, the contribution of the tetrahydroindazole system
(shown in green, Fig. 1B) to AMPA modulatory activity was exam-
ined (Table 2). Deletion of the fused cyclohexyl ring (8) resulted in
approximately a 30-fold reduction in potency presumably due to
loss of a non-specific hydrophobic interaction. Activity could be
partially restored by introduction of a 5-CF3 substituent (9), how-
ever, again this was accompanied by low solubility and micro-
somal stability (solubility <1 mg/L, Cli (human) = 174
lL/min/mg
protein, Cli (rat) = 200 L/min/mg protein). A fully aromatic system
l
(10) exhibited promising GluA1 activity, however, was again com-
promised by inadequate solubility (<1 mg/L) and less than optimal
microsomal stability (Cli (human) = 153 L/min/mg protein, Cli
(rat) = 91 lL/min/mg protein). As we expected from the biostruc-
ture of 1, having a correctly orientated trifluromethyl moiety is
essential for activity. Replacement with a methyl (11) or altering
the relative position (12) gave rise to inactive compounds.
The most effective means of meeting our optimization goals
emerged from modification of the amide region of 1. From consid-
eration of the biostructural data, a hydrophilic pocket adjacent to
the primary carboxamide motif (highlighted in red, Fig. 1B) could
be accessed by suitably functionalized entities. We, therefore, fo-
cused our efforts in making modifications to this region. Initially,
simple carboxamide derivatives were prepared (e.g., 13, 14, Table
3) which confirmed that substitution at this position would be tol-
erated. Encouraged by these results, we next prepared a number of
more functionalized derivatives with pendant solubilizing groups.
Evaluation of compound 15 in the GluA1 assay showed the com-
pound had modest potency compared to our progenitor hit 1. How-
ever, it was sufficiently potent to obtain biostructural data (Fig. 2)
which revealed that contact with an Asp residue at the side of the
hydrophilic pocket could be further optimized. We hypothesized
Figure 1. (A) Hit compound 1 in complex with S1S2 LBD of GluA2. Receptor surface
is depicted, colored by atom type. Several residues have been omitted from this and
subsequent figures for clarity. (B) Highlighted in green, red and blue are the three
regions targeted.
orthosteric site and essentially stabilize a protein–protein inter-
action between the receptor subunits of each ligand binding
domain.11 Analysis of our X-ray data revealed that the interactions
made with the receptor were primarily hydrophobic in nature,
with key contacts made at the distal ends of the molecule (CF3 moi-
ety and tetrahydrobenzothiophene system). From consideration of
the biostructural data, three principal regions of the molecule were
targeted (shown in red, green and blue) in order to address the sol-
ubility and metabolism issues (Fig. 1).
Modification of the tetrahydrobenzothiophene portion (shown
in blue, Fig. 1B) was expected to reduce lipophilicity and con-
comitantly improve potency and reduce propensity for hepatic
metabolism. However, it was also anticipated that modulatory
activity at the AMPA receptor may be negatively impacted by such
changes.
Indeed, stripping back the fused cyclohexyl ring system to
furnish for example compound 2 (Table 1) resulted in around a
30-fold reduction in potency, albeit with improvement observed
in solubility (2 showed kinetic solubility of 10 mg/L). Other trun-
cated analogs (e.g., 3 and 4) showed a similar reduction in potency
without any improvement in solubility. Retention of the fused ring
system with incorporation of a heteroatom was next explored in
order to attenuate lipophilicity. Compound 5 displayed promising
activity at the receptor but lacked any appreciable degree of
Table 2
Modifications to the tetrahydroindazole region
H2N
H
O
R
N
O
S
a
Compds
R
pEC50
5.3
F
F
F
F
N
8
N
N
F
F
N
9
5.8
Table 1
F
F
Modifications to the tetrahydrobenzothiophene region
F
H2N
F
F
N
H
O
N
F3C
N
N
N
F
10
11
6.4
O
R1
S
N
R2
R2
N
R1
a
N
Compds
pEC50
<4.5
1
2
3
4
5
6
7
–(CH2)4–
6.7
5.2
6.1
6.2
6.7
5.3
5.5
H
Me
Et
H
Me
Me
F
F
F
N
N
12
<4.5
–CH2–CH2–O–CH2–
–CH2–CH2–NH–CH2–
CH2OH
H
a
a
Values are means of two experiments performed in duplicate.
Values are means of two experiments performed in duplicate.

C. Jamieson et al. / Bioorg. Med. Chem. Lett. 20 (2010) 5753–5756
5755
Table 3
observations from compound 15 (Fig. 3). The propylamine side
chain was shown to project into the hydrophilic pocket as before,
but now forming a salt-bridge with the side-chain of Asp 760.
The biostructural data also suggested that conformationally con-
strained analogs of 16 could be accommodated in the hydrophilic
region and this led to the design and synthesis of analogs 22–25
(Table 4) in an effort to improve brain penetration and oral
bioavailability.
Comparing 22 and 23 shows a preference for the (R) stereo-
chemistry on the pyrrolidine ring system. This was consistent with
our expectations based on the conformation of the propylamine
moiety required to facilitate the salt-bridge with Asp 760. The
azetidine analogs 24 and 25 show similar potencies at GluA1 com-
pared to 22.
Further profiling of 22 indicated that the compound had both
reasonable solubility (20 mg/L) and microsomal stability (human
Cli = 41 lL/min/mg protein, rat Cli = 50 lL/min/mg protein).
Assessment of permeability in the CaCo-2 assay suggested that
absorption might still be an issue (AÀB = 19 nm/s, BÀA = 22 nm/
s). In vivo pharmacokinetics in rat (2 mg/kg iv, 10 mg/kg po) indi-
Modifications to the tetrahydrobenzothiophene region
R
HN
N
H
O
F3C
N
N
O
S
a
Compds
R
pEC50
13
14
15
16
17
18
19
20
21
Et
cPropyl
6.5
6.2
5.2
6.4
5.7
5.9
5.8
7.3
7.0
–CH2CH2NMe2
–(CH2)3–NH2
–CH2CH2NHMe
–(CH2)4–NH2
–CH2CH2NH2
–CH2CH2OH
–CH2CH2NHSO2Me
a
Values are means of two experiments performed in duplicate.
that either modification of the linker length, or removal of one or
both of the N,N-dimethyl moieties could be beneficial to potency.
In compound 15, the binding mode of the tetrahydroindazole re-
gion was essentially identical to that observed with the progenitor
compound (1).
cated that 22 had low clearance and
a reasonable half-life
(Clp = 2.3 mL/min/kg, T1/2 = 3.8 h, Vss = 0.5 L/kg) and significantly
improved brain exposure (brain:plasma ratio = 1.03). However,
measured oral bioavailability was still low (3.8%) indicating that
These considerations led to the design of compound 16 which
had comparable potency to the original hit. Other amine deriva-
tives (e.g., 17–19), although more potent than the progenitor 15,
were not as active at the receptor compared to compound 16.
The hydrophilic region proved to be relatively tolerant of other
functional groups with alcohols typified by 20 and sulfonamide
derivatives such as 21 displaying impressive potencies but in com-
mon with many other members of the series lacked solubility or
microsomal stability.
In the series as a whole, compound 16 displayed the best bal-
ance of solubility (38 mg/L) and microsomal stability (human
Cli = 17 lL/min/mg protein, rat Cli = 77 lL/min/mg protein). Per-
meability as measured by CaCo-2 was deemed to be low, however
there was no indication of efflux (AÀB = 34 nm/s, BÀA = 55 nm/s).
In vivo pharmacokinetics in Wistar rats (2 mg/kg iv, 10 mg/kg
po) indicated moderate clearance and half-life in the iv leg
(Clp = 21 mL/min/kg, T1/2 = 3.2 h, Vss = 3.8 L/kg), however, as antic-
ipated from the CaCo-2 data, observed oral bioavailability was low
(7.7%). Brain to plasma ratio was low (0.05) which was reasoned to
be a function of both the number of hydrogen bond donors and
rotatable bonds present in the molecule.
Figure 3. X-ray structure of 16 in complex with the S1S2 LBD of GluA2.
Table 4
Constrained analogs of compound 16
R
X-ray crystallographic data on the complex of 16 with the S1S2
LBD revealed the binding mode was as anticipated based on our
HN
N
H
O
F3C
N
N
O
S
a
Compds
R
pEC50
NH
NH
22
6.6
23
24
25
5.8
6.5
6.2
H
N
H
N
a
Values are means of two experiments performed in duplicate.
Figure 2. Biostructure of compound 15 in complex with the S1S2 LBD of GluA2.

5756
C. Jamieson et al. / Bioorg. Med. Chem. Lett. 20 (2010) 5753–5756
O
NH2
R1
Compounds 8–12 were synthesized in a two-step fashion from
the available amino thiophene derivative 28 (Scheme 2). Acetyla-
tion with bromacetyl bromide followed by alkylation of the appro-
priate azole derivative furnished the requisite target molecules in
an expedient fashion.
Amide derivatives 13–25 were accessed as depicted in Scheme
3. Starting from cyclohexanone, Gewald cyclization furnished the
aminothiophene ester 30. Amide coupling provided intermediate
31 which was deprotected by acidolysis. Subsequent HATU medi-
ated amide formation provided amides 13–15 and 20/21. In the
case of 16–19 and 22–25, a Boc protected diamine derivative is
employed in the coupling step with the desired amine unmasked
in a final TFA deprotection step.
O
NH2
R1
R2
H
N
F
F
N
N
O
H2N
a
b, c
R1
S
O
R2
R2
26
S
F
1 - 7
27
Scheme 1. Reagents and conditions: (a) sulfur, cyanoacetamide, diethylamine,
EtOH, reflux 23–51%; (b) 3-trifluoromethyl-4,5,6,7-tetrahydroindazol-1-yl)-acetic
acid, PS-CDI,
l
W, 120 °C, 10–53%; (c) TFA/CH2Cl2, rt, 23% (for 6).
O
O
O
H2N
H2N
H2N
b
In summary, we have demonstrated how hit compound 1 was
rapidly optimized to improve physicochemical parameters and
metabolic stability resulting in compounds such as 16 and 22. A
central component of our expedient optimization trajectory has
been the application of structure based drug design and is the first
example of its use in a glutamate receptor system. Compounds
such as 22 are useful tools in further elucidating the role of AMPA
receptor modulators in debilitating neurological disorders such as
Alzheimer’s disease and Parkinson’s. In the following paper, we de-
scribe how 22 can be further optimized to an orally bioavailable
AMPA receptor modulator with promising in vivo activity.
a
H
N
H
Het
N
S
H2N
S
S
O
Br
O
8-12
28
29
Scheme 2. Reagents and conditions: (a) bromoacetyl bromide, Et3N, CH2Cl2, rt,
94%; (b) azole derivative, NaH, DMF, 0 °C to rt, 14–49%.
O
O
O
H
N
References and notes
O
O
b
a
S
1. Kew, J. N. C.; Kemp, J. A. Psychopharmacology 2005, 179, 4.
2. Marenco, S.; Weinberger, D. R. CNS Drugs 2006, 20, 173.
3. Zarate, J.; Manji, H. K. Exp. Neurol. 2008, 211, 7.
4. Mayer, M. L.; Armstrong, N. Annu. Rev. Physiol. 2004, 66, 161.
5. Armstrong, N.; Sun, Y.; Chen, G. Q.; Gouaux, E. Nature 1998, 395, 913.
6. Sobolevsky, A. I.; Rosconi, M. P.; Gouaux, E. Nature 2009, 462, 745.
7. HEK.GluR1(i) cells were maintained in DMEM supplemented with 10%
N
N
O
H2N
S
31
30
F
F
F
R
O
N
H
d, e
fetaclone II, 1% non-essential amino acids and 150 lg/mL hygromycin, at
37 °C/5% CO2. Twenty-four hour prior to the assay, the cells were harvested
H
N
with trypsin and seeded onto Costar 96-well clear bottomed black plates at a
S
density of 3.5 Â 104 per well. Cells were loaded with 5
lM fluo3-AM in DMEM
media in the absence of hygromycin and incubated at 37 °C/5% CO2 for one h.
N
O
N
After dye loading, the cells were washed once with 200 ll of low calcium
solution (10 mM hepes, pH 7.4, 160 mM NaCl, 4.5 mM KCl, 2 mM CaCl2, 1 mM
MgCl2, 10 mM glucose) containing 0.625 mM of probenecid to remove the dye.
13-25
F
F
F
Then 200
added 50
ll of low calcium solution was added to each well. The FLEXstation
Scheme 3. Reagents and conditions: (a) sulfur, tert-butylcyanoacetate, diethyl-
amine, EtOH, reflux, 95%; (b) (3-trifluoromethyl-4,5,6,7-tetrahydroindazol-1-yl)-
acetic acid, HATU, DIEA, CH2Cl2/DMF, rt, 75%; (c) TFA/CH2Cl2, rt, quant.; (d) amine/
Boc-diamine, HATU, DIEA, DMF, rt 10–70%; (e) TFA/CH2Cl2 (for 16–19 and 22–25),
15–40%.
ll of glutamate +/À test compound in high calcium solution (10 mM
Hepes, pH 7.4, 160 mM NaCl, 4.5 mM KCl, 20 mM CaCl2, 1 mM MgCl2 and
10 mM glucose) to each well and the ensuing response was monitored on
FLEXstation.
8. Crystals were grown as described in Ref. 5. Co-crystals were prepared by
soaking test compound (100 mM) with crystals for 24 h prior to isolation and
data collection. Coordinates and structure factors have been deposited in the
Protein Data Bank for complexes of compounds 1 (3O28), 15 (3O29) and 16
(3O2A).
9. The Pymol Molecular Graphics System (DeLano Scientific, Palo Alto, CA, 2002)
was used to generate Figures 1–3.
10. Maclean, J. K. F.; Kazemier, B. J. Mol. Biol., in press.
11. (a) Sun, Y.; Olson, R.; Horning, M.; Armstrong, N.; Mayer, M.; Gouaux, E. Nature
2002, 417, 245; (b) Ptak, C. P.; Ahmed, A. H.; Oswald, R. E. Biochemistry 2009,
48, 8594.
conformational constraint alone was not sufficient to enhance this
important parameter.
Synthetic approaches to the compounds discussed above are
outlined in Schemes 1–3. Compounds 1–7 were prepared through
a Gewald cyclization12 to prepare the requisite tertrahydrobenzo-
thiophene systems (27) followed by HATU mediated coupling to
furnish compounds 1–7.
12. Gewald, K.; Schinke, E.; Boettcher, H. Chem. Ber. 1966, 99, 94.