Imidazo[1,5-a][1,2,4]-triazolo[1,5-d][1,4]benzodiazepines as potent and highly selective GABAA α5 inverse agonists with potential for the treatment of cognitive dysfunction

Article abstract of DOI:10.1016/j.bmcl.2009.08.027

In a search for GABA_A α5 ligands that combine high subtype binding selectivity with a marked inverse agonism imidazo[1,5-a][1,2,4]-triazolo[1,5-d][1,4]benzodiazepines were identified as a promising class. A short tandem reaction allowed rapid a

Full text of DOI:10.1016/j.bmcl.2009.08.027

Bioorganic & Medicinal Chemistry Letters 19 (2009) 5958–5961
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Imidazo[1,5-a][1,2,4]-triazolo[1,5-d][1,4]benzodiazepines as potent
and highly selective GABAA
of cognitive dysfunction
a5 inverse agonists with potential for the treatment
b
c
c
Bernd Buettelmann a, , Theresa M. Ballard , Rodolfo Gasser , Holger Fischer ,
Maria-Clemencia Hernandez b, Frédéric Knoflach b, Henner Knust a, Heinz Stadler a,
Andrew W. Thomas a, Gerhard Trube a
*
a F. Hoffmann-La Roche Ltd, Pharmaceutical Research Basel, Discovery Chemistry, CH-4070 Basel, Switzerland
b F. Hoffmann-La Roche Ltd, Pharmaceutical Research Basel, CNS Disease Biology Area, CH-4070 Basel, Switzerland
c F. Hoffmann-La Roche Ltd, Pharmaceutical Research Basel, Non Clinical Safety, CH-4070 Basel, Switzerland
a r t i c l e i n f o
a b s t r a c t
Article history:
In a search for GABAA a5 ligands that combine high subtype binding selectivity with a marked inverse
Received 7 July 2009
Revised 4 August 2009
Accepted 5 August 2009
Available online 11 August 2009
agonism imidazo[1,5-a][1,2,4]-triazolo[1,5-d][1,4]benzodiazepines were identified as a promising class.
A short tandem reaction allowed rapid access to this chemical series, thereby facilitating rapid SAR gen-
eration which guided the optimization process. Two compounds (10e and 11f) were found to be active in
an in vivo paradigm for cognitive improvement.
Ó 2009 Elsevier Ltd. All rights reserved.
Keywords:
GABA A
Inverse agonism
Cognitive dysfunction
c
-Aminobutyric acid (GABA) is one of the main neurotransmit-
and dependence.4Another marketed drug, Flumazenil 2, is charac-
terized as a subtype non-selective antagonist. It is clinically used in
the treatment of BZ-agonist overdosing.5
The therapeutic potential of subtype non-selective inverse ago-
nists (exemplified by FG7142, 3) as cognition enhancers has been
shown to be limited by severe side effects (anxiety, agitation).6,7
ters in human and rodent brain. It exerts its effects by interaction
with GABAA, GABAB and GABAC receptors. Among these, GABAA
receptors have a proven record as drugable targets, for example,
benzodiazepines, barbiturates and neurosteroids interact at this
receptor.1 GABAA receptors are functionally characterized as li-
gand-gated chloride channels with a GABA recognition site. Struc-
turally, they are defined as pentameric assemblies of different
subunits.2 Most mammalian GABAA receptors consist of
3 or 5 in conjunction with b2 or b3 and
a1, a2,
O
N
N
N
a
a
c
2 -subunits.3 These
CO2Et
N
subtypes also contain a recognition site for prototypical benzodi-
azepines = BZ (exemplified by diazepam 1, Fig. 1), which allosteri-
cally modulate the functional effect of GABA. Based on their
modulatory effects, BZ-site ligands are categorized as either ago-
nists (positive modulators, facilitating Clꢀ-flux), antagonists (‘neu-
tral’ modulators, normalizing Clꢀ-flux at basal level) or inverse
agonists (negative modulators, reducing Clꢀ-flux). For example, 1
N
Cl
F
O
diazepam 1
flumazenil 2
O
N
N
N
N
O
H
does not discriminate between
a1, a2, a3 and a5 containing GA-
N
BAA subtypes; functionally, it increases Clꢀ-flux. Compounds with
this profile are defined as non-selective agonists. A number of
drugs of this type are marketed as sedatives, anxiolytics and/or
anticonvulsants. Reported side effects are memory impairment
N
N
O
N
H
N
"α5IA-II" 4
"FG7142" 3
* Corresponding author.
E-mail address: bernd.buettelmann@roche.com (B. Buettelmann).
Figure 1. Reference compounds: BZ-ligands with different profiles.
0960-894X/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2009.08.027

B. Buettelmann et al. / Bioorg. Med. Chem. Lett. 19 (2009) 5958–5961
5959
With knowledge generated by the use of genetically modified mice
N
N
meanwhile evidence has accumulated that a selective GABAA
a5
F
N
N
a
b,c
inverse agonists can be expected to be devoid of these side effects.7
Two conceptionally different approaches have been followed to de-
N+
63%
59%
Br
Br
Br
N
N
N
sign compounds with this profile: Selective GABAA
agonism may be achieved either by a high binding selectivity com-
bined with the desired modulatory effect at the GABAA 5 receptor
or by functional selectivity (i.e., inverse agonism at the GABAA
a5 inverse
6
7
8
a
N
N
N
a5
receptor with low efficacy at the other GABAA receptor subtypes).
Following the first approach, researchers at Merck in 2003 de-
scribed a series of compounds that combined the desired efficacy
N
N
N
d
3%-78%
N
Br
Br
NH
a
5 binding selectivity (up to 30-fold).8
N
with a significant GABAA
Subsequently, this group focused on functionally selective com-
pounds active in animal models of cognitive function (e.g., 5IA-
II, 4).9,10 A related compound
5IA was tested in humans and
NH
9
R1
10
O
R1
a
a
Scheme 1. Reagents and conditions: (a) 4-methylimidazole, K2CO3, DMSO, 90 °C;
(b) Eschenmoser’s salt, DMF, 100 °C; (c) methyl iodide, CH2Cl2, 4 °C; (d) R1–
CONHNH2, DMF, 120 °C.
shown to reverse alcohol induced memory impairment without
causing significant side effects.7
With the aim to identify GABAA a5 subtype selective BZ-ligands
10a) reduced GABAA
affinity at the GABAA
tor 10, leading overall to a marked selectivity increase. Steric bulk
directly linked to the heterocycle (as in 10b) led to an even further
a
5 receptor affinity threefold. However, also
endowed with significant inverse agonism, a high-throughput
screening campaign of the Roche compound collection was per-
formed. A subsequent lead generation process furnished imidazo-
triazolobenzodiazepine 5 as a promising candidate for further opti-
mization. Compared with reference compound 4 (Table 1), it is
a1, 2 and 3 receptors was reduced by a fac-
a
a
reduction of GABAA
uninteresting. Introduction of a small ether functionality (10c) re-
stored GABAA 5 receptor affinity, with an overall selectivity pro-
file comparable to the simple methyl substituted 10a. A marked
a5 receptor affinity, rendering the compound
characterized by a somewhat reduced GABAA
affinity combined with a significant binding selectivity versus GA-
BAA 1, 2 and 3 subtype receptors. As both compounds potently
modulate Clꢀ-flux of the GABAA
a5 receptor binding
a
a
a
a
GABAA
GABAA
a
a
5 receptor affinity increase combined with a significant
a
5 receptor (by ꢀ41% and ꢀ42%,
1, 2 and 3 receptor selectivity (almost 2 log units)
a
a
respectively), they are functionally characterized as inverse
agonists.11
was observed for the methylene linked cyclopropane 10d. Com-
pared to 5, however, each of these compounds were functionally
not very efficacious, they were only weak inverse agonists (efficacy
>ꢀ25%). With the aim to combine high binding selectivity with a
reasonable intrinsic efficacy, a large number of substituents were
screened. Methylene linked pyrrolidinone 10e had a selectivity
profile comparable to 10d with a threefold higher efficacy. Gener-
ally, methylene linked nonaromatic as well as aromatic heterocy-
cles were found to be highly binding selective. Notably, isoxazole
The known literature synthesis of compounds with an imidazo-
triazolobenzodiazepine skeleton (as in 5) relies on a long linear se-
quence which turned out to be unpractical for rapid SAR explora-
tion.12 Since we sought for a synthetic sequence that allowed a
late stage introduction of the 6-substituent, we developed an alter-
native approach that commenced with a nucleophilic aromatic
substitution of 5-bromo-2-fluoro-benzonitrile 6 by 4-methylimid-
azole (Scheme 1). The mixture of regioisomeric products was
recrystallizated to afford 7 as a pure regioisomer. Reaction with
Eschenmoser’s salt then introduced regioselectively the dimethyl-
aminomethyl-function. The desired leaving group on the heterocy-
cle could cleanly be obtained by reaction with methyl iodide (8).
Substituting this with a wide range of acylhydrazides at 120 °C
led to intermediates 9 (identified by MS, not isolated), which under
prolonged heating condensed to the desired triazolo-annelated
imidazobenzodiazepine 10.13 Using differently substituted 2-fluor-
obenzonitriles, analogous products could be obtained (Table 3).
First we studied (Table 2) the influence of substituents at the
triazole (position 6). Formally adding a methyl to 5 (leading to
10f (selectivity vs GABAA
a1, a2 and a3 receptors >100) was one
of the most interesting compounds, as it also was categorized as
a partial inverse agonist (efficacy <ꢀ25%). Oxadiazole 10g was
the most binding selective compound. However, selectivity was
achieved only at the expense of functional activity (weak inverse
agonism). Overall, appropriately designed substituents positioned
at C-6 gave access to high binding selectivity. Substituents at this
position were tolerated at the
a5, but not at GABAA a1, a2 and
a3 receptors. The structure-efficacy-relationship was much less
well understood; purely lipophilic interactions seemed to reduce
inverse agonism.
Table 1
Binding and efficacy profile of 4 and 5
N
N
N
N
Br
3
N
5
6
Compd
Affinity KI (nM) GABA A
axb3
c
2 receptorsa
Efficacy GABA A
axb3c
2 receptorsa,b (%)
a5
a1
a2
a3
a5
a1
a2
a3
4c
5
1.5
6
1.5
136
1.5
186
1.5
132
ꢀ41
ꢀ42
ꢀ9
11
nt
10
nt
ꢀ16
a
Cloned rat or human receptor subunits were expressed in insect SF9 cells, human HEK293 cells or Xenopus laevis oocyts for 3H-flumazenil a1a2a3 and a5 binding or
electrophysiology.
b
Efficacy is determined as the percentage change of a submaximal (EC10) response to GABA.
c
Compound synthesized and tested at F. Hoffmann-La Roche Ltd. The results are in good agreement with those reported.10

5960
B. Buettelmann et al. / Bioorg. Med. Chem. Lett. 19 (2009) 5958–5961
Table 2
Binding and efficacy profile of variously substituted compounds 10
Compd
R1
Affinity KI (nM) GABAA
a
xb3c
2d
Efficacy GABAA
a
5b3c
2 receptorsd (%)
a5
a
1
a
2
a3
10a
10b
10c
10d
Me
t-Bu
CH2–OMe
CH2–Cyclopropyl
20
146
13
1421
3160
619
3160
1068
693
1624
1206
592
ꢀ17
nt
ꢀ3
5
538
407
407
ꢀ12
O
N
10e
10f
7
624
188
669
105
515
73
ꢀ34
ꢀ27
O
N
0.6
O
N
10g
0.9
469
317
277
ꢀ18
N
d
see Table 1.
Table 3
Binding and efficacy profile of analogues of 10f
N
N
N
N
R2
N
O
N
Compd
R2
Affinity GABAA
a5b3c
2d KI (nM)
SeðlneMctÞivity:
Efficacy GABAA
a
5b3c
2 receptorsd (%)
K1
K1ðnMÞ
a
xb3
5b3
c
2
2
a
c
versus
a
1
versus
a
2
versus a3
10f
11f
12f
13f
14f
15f
16f
Br
Cl
0.6
0.6
11
1.6
0.8
2.2
1.2
314
466
68
174
251
83
122
223
111
16
360
222
nt
ꢀ27
ꢀ18
ꢀ20
nt
ꢀ16
ꢀ13
ꢀ7
OCF3
H
Me
OMe
OHe
57
20
620
307
99
332
193
nt
d
e
See Table 1.
Synthesized by BBr3 mediated hydrolysis of 15f.
Table 4
Table 5
Pharmacokinetic parameters of 10e, 10f and 11f (in rats)
Property profiles of 10e, 10f and 11f
10e
10f
11f
10e
10f
1.4
39
4.6
34
0.2
11f
Solubility (
l
g/mL)f
>550
1.5
1.5
7
14
Half life t1/2 (h)
Cl (mL/min/kg)
Vss (L/kg)
F (%)
Brain/plasma
0.8
30
1.3
32
1.4
9.5
1.8
44
log D7.4
2.6
3.3
10/0
2.3
3.1
6/0
Permeation coefficient (10ꢀ6 cm sꢀ1) (PAMPA)
Cl (mL/min/mg protein)g (rat/human)
16/3
0.09
1.0
f
LYSA, for measurement details, see Ref. 15.
From incubations with liver microsomes.
g
(12f) was clearly detrimental:14 GABAA
duced almost 20-fold. Whilst a simple hydrogen (13f) gave a single
digit affinity at the GABAA 5 receptor, selectivity versus GABAA
1, 2 and notably 3 receptors was compromised. Interestingly,
at least in terms of GABAA 5 receptor affinity and selectivity,
a5 receptor affinity was re-
We then studied the influence of substituents at position 3 of
a
the imidazo-triazolobenzodiazepine keeping constant the methy-
lene linked methylisoxazole that conferred high binding selectivity
(Table 3). Formally replacing Br by Cl led to 11f which in terms of
a
a
a
a
the electronic nature of the 3-substituent does not play a role:
electron-donating methyl (14f) and methoxy (15f) combine high
GABAA
The desired inverse agonism, however, was reduced. Breaking
symmetry by introduction of the trifluoromethoxy substituent
a5 receptor affinity and selectivity was slightly superior.
r
-
GABAA
a5 receptor affinity with selectivity. In this respect, even

B. Buettelmann et al. / Bioorg. Med. Chem. Lett. 19 (2009) 5958–5961
5961
verse agonism may have a potential in the treatment of memory
impairment, 10e and 11f had been selected for an in-depth evalu-
ation on scopolamine-induced impairment of working memory in
the delayed match to position (DMTP) task (Fig. 2).16 Reference
compound 4 administered at 1.0 mg/kg ip 30 min prior to testing,
significantly reversed the working memory impairment induced
by scopolamine (Fig. 2A). Compound 10e significantly reversed
the scopolamine-induced impairment in a dose-related manner,
with full reversal achieved at 1.0–3.0 mg/kg ip 30 min post-admin-
istration (Fig. 2B). 10e was not active following po administration.
Compound 11f significantly, but partially, attenuated the scopol-
amine-induced working memory impairment at 1.0 mg/kg po at
60 min post-administration (Fig. 2C).
C
A
100
B
##
###
#
90
80
70
60
50
##
#
##
#
***
***
***
***
*
**
**
**
0
8
16
24
0
8
16
24
0
8
16
24
Delays (s)
Delays (s)
Delays (s)
Notably, in this experiment no behavioral impairment was ob-
served. From these results it can be concluded that even weak in-
Figure 2. Effect of 4, 10e and 11f on scopolamine (0.03 mg/kg s.c.) induced
impairment of working memory, measured by choice accuracy across delay interval
in the delayed match to position (DMTP) task. Key: s = vehicle; d scopolamine. (A)
Compound 4: h = 1.0 mg/kg ip; Ç = 1.0 mg/kg ip versus scopolamine. (B) Compound
10e versus scopolamine: Ç = 0.3, j = 1.0, Ç = 3.0 mg/kg ip. (C) Compound 11f vs.
scopolamine: Ç = 0.1, j = 0.3, Ç = 1.0 mg/kg po. Male Lister hooded rats (n=11–12)
were used in each study. Data are expressed as mean SEM. *p <0.05, **p <0.01, ***p
<0.001 versus vehicle; #p <0.05, ##p <0.01, ###p <0.001 versus scopolamine.
verse agonists at the GABAA
a5 receptor may have potential as
memory enhancing agents. In line with the referred literature we
found that high binding selectivity versus GABAA
receptors can prevent unwanted side effects.
a1, a2 and a3
References and notes
1. Bormann, J. Trends Pharmacol. Sci. 2000, 21, 16.
2. Simon, J.; Wakimoto, H.; Fujita, N.; Lalande, M.; Barnard, E. A. J. Biol. Chem.
2004, 279, 41422.
3. McKernan, R. M.; Whiting, P. J. Trends Neurosci. 1996, 19, 139.
4. Chouinard, G. J. Clin. Pharmacol. 2004, 65, 7.
a hydroxy (16f) substituent looks promising. In terms of efficacy,
however, none of the substituents tested were as effective as the
Br substituted structure 10f.
Complimentary to the receptor parameters discussed above,
other molecular properties were also determined systematically
and were typically found to be in the desired range (Table 4).
Whilst pyrolidinone-substituted 10e combined low lipophilicity
with a remarkable solubility, its methylisoxazole substituted ana-
logue 10f was found to be more lipophilic (by 1 log unit) with an
acceptable solubility. Formally replacing Br in 10f by Cl leading
to 11f expectedly reduced lipophilicity while increasing solubility.
Permeation of all compounds was in the desired range, with a
trend of facilitated transport for more lipophilic compounds. Final-
ly, microsomal clearance (Cl) was low, with indications for
modulation.
Based on their favorable drug-like properties, 10e, 10f and 11f
were selected for a pharmacokinetic study (Table 5). 10e displayed
a short half life in rats with medium clearance and volume of dis-
tribution (Vss). Bioavailability (F) was tolerable, however, brain
penetration was suboptimal. The other Br substituted compound
(10f) had a somewhat longer half life and a threefold higher vol-
ume of distribution. This was rationalized in terms of the increased
lipophilicity (log D = 1.5 and 2.6, respectively). Interestingly, the Cl
substituted analogue 11f had an overall improved profile: notably
brain penetration was in the desired range.
5. Bentue-Ferrer, D.; Bureau, M.; Patat, A.; Allain, H. CNS Drug Rev. 1996, 2, 390.
6. Dorow, R.; Horowski, R.; Paschelke, G.; Amin, M.; Braestrup, C. Lancet 1983, 2,
98.
7. (a) Dawson, G. R.; Maubach, K. A.; Collinson, N.; Cobain, M.; Everitt, B. J.;
MacLeod, A. M.; Choudhury, H. I.; McDonald, L. M.; Pillai, G.; Rycroft, W.; Smith,
A. J.; Sternfeld, F.; Tattersall, F. D.; Wafford, K. A.; Reynolds, D. S.; Seabrook, G.
R.; Atack, J. R. J. Pharmacol. Exp. Ther. 2006, 316, 1335; (b) Nutt, D. J.; Besson, M.;
Wilson, S. J.; Dawson, G. R.; Lingford-Hughes, A. R. Neuropharmacology 2007, 53,
810.
8. Chambers, M. S.; Atack, J. R.; Broughton, H. B.; Collinson, N.; Cook, S.; Dawson,
G. R.; Hobbs, S. C.; Marshall, G.; Maubach, K. A.; Pillai, G. V.; Reeve, A. J.;
MacLeod, A. M. J. Med. Chem. 2003, 46, 2227.
9. Collinson, N.; Atack, J. R.; Laughton, P.; Dawson, G. R.; Stephens, D. N.
Psychopharmacology 2006, 188, 619.
10. Sternfeld, F.; Carling, R. W.; Jelley, R. A.; Ladduwahetty, T.; Merchant, K. J.;
Moore, K. W.; Reeve, A. J.; Street, L. J.; O’Connor, D.; Sohal, B.; Atack, J. R.; Cook,
S.; Seabrook, G.; Wafford, K.; Tattersall, F. D.; Collinson, N.; Dawson, G. R.;
Castro, J. L.; MacLeod, A. M. J. Med. Chem. 2004, 47, 2176.
11. We define compounds with an efficacy <ꢀ40% as ‘full inverse agonists’,
compounds with an efficacy >ꢀ40% and <ꢀ25% as ‘partial inverse agonists’and
those with >ꢀ25% and <ꢀ10% as ‘weak inverse agonists’.
12. Gerecke, M.; Kyburz, E.; Borer, R.; Gassner, W. Heterocycles 1994, 39, 693.
13. Buettelmann, B.; Knust, H.; Thomas, A. U.S. Patent 2007082890; Chem. Abstr.
2007, 146, 422016.
14. Shishkov, I. F.; Khristenko, L. V.; Vilkov, L. V.; Oberhammer, H. J. Phys. Chem. A
2004, 108, 4966.
15. Gonzalez, R. C. B.; Huwyler, J.; Boess, F.; Walter, I.; Bittner, B. Biopharm. Drug
Dispos. 2004, 25, 37.
16. Higgins, G. A.; Ballard, T. M.; Huwyler, J.; Kemp, J. A.; Gill, R. Neuropharmacology
2003, 44, 324.
To answer the question, whether GABAA
a5 receptor highly
binding selective compounds with only partial or even weak in-