Relating the structure, activity, and physical properties of ultrashort-acting benzodiazepine receptor agonists

Article abstract of DOI:10.1016/S0960-894X(02)00513-9

The ultrashort-acting benzodiazepine (USA BZD) agonists reported previously have been structurally modified to improve aqueous solubility. Lactam-to-amidine modifications, replacement of the C5-haloaryl ring, and annulation of heterocycles are presented. These analogues retain BZD receptor potency and full agonism profiles.

Full text of DOI:10.1016/S0960-894X(02)00513-9

Bioorganic & Medicinal Chemistry Letters 12 (2002) 3219–3222
Relating the Structure, Activity, and Physical Properties of
Ultrashort-Acting Benzodiazepine Receptor Agonists
Gregory J. Pacofsky,y Jeffrey A. Stafford,{ Richard F. Cox, Jill R. Cowan,
George F. Dorsey, Jr., Stephen S. Gonzales, Istvan Kaldor, George W. Koszalka,
George G. Lovell, Maggie S. McIntyre, Jeffrey H. Tidwell, Dan Todd,
Graham Whitesell, Robert P. Wiard and Paul L. Feldman*
GlaxoSmithKline, Five Moore Drive, Research Triangle Park, NC 27709, USA
Received 4 February 2002; accepted 7 June 2002
Abstract—The ultrashort-acting benzodiazepine (USA BZD) agonists reported previously have been structurally modified to
improve aqueous solubility. Lactam-to-amidine modifications, replacement of the C5-haloaryl ring, and annulation of heterocycles
are presented. These analogues retain BZD receptor potency and full agonism profiles.
# 2002 Elsevier Science Ltd. All rights reserved.
Our earlier work1 had identified shorter-acting anxioly-
tic-sedatives exemplified by benzodiazepinone 1. We
considered that a neuropharmacological profile of
agonism would be preserved through binding to the
GABAA receptor with a benzodiazepine (BZD) core
and diminution of activity would be controlled via rapid
degradation of the appended carboxylic ester by non-
specific esterases. The predictable, rapid onset of and
recovery from sedation that characterize these com-
pounds represent an alternative to midazolam (2).2,3
The entities described previously, however potent or
selective, lacked the physical properties required of an
adjunct for total intravenous anesthesia.
could neither disrupt the high-affinity binding to the
BZD receptor nor alter the full agonism profile. We
rationalized that a heteroatom-based moiety would
provide additional sites for coordination with solvent
and/or salt formation. One approach explored the use
of substituted amidines, as exemplified by the structure
of the hallmark benzodiazepine agonist, chlordiazep-
oxide (3), not only to introduce these properties but also
to provide a point of diversification in the context of
SAR development. We also considered replacement of
the C5-aryl group; while the 20-haloaryl subunit confers
potency, 20-pyridyl variants, such as bromazepam (4),4
display reasonable activities. Finally, fused heterocyclic
systems, as exemplified by 2 itself, were examined.
We directed our attention towards the discovery of
agents with pharmacology comparable to 1 and with
improved aqueous solubility. Initial efforts focused on
the incorporation of additional functionalities into the
benzodiazepinone nucleus of 1 and targeted aqueous
solubilities of >1 mg/mL at pH 3 for the purpose of
formulating an injectable solution. Clearly these changes
*Corresponding author. Tel.: +1-919-483-6255; fax: +1-919-483-
3762; e-mail: plf4602@gsk.com
yPresent address: Chrysalon Molecular Research, 11 Avenue Albert
Einstein, 69100 Villeurbanne, France.
{Present address: Syrrx, Inc., 10410 Science Center Drive, San Diego,
CA 92121, USA.
0960-894X/02/$ - see front matter # 2002 Elsevier Science Ltd. All rights reserved.
PII: S0960-894X(02)00513-9

3220
G. J. Pacofsky et al. / Bioorg. Med. Chem. Lett. 12 (2002) 3219–3222
Scheme 2. Directed metalation9 of 7 at low tempera-
ture and reaction of the derived lithio species with
2-pyridinecarboxaldehyde provided alcohol 8. Oxida-
tion with MnO2 gave the BOC-protected aminobenzo-
phenone and subsequent deprotection with HCl gave 9
in 54% yield over these three steps. As before,1 amide
bond formation, deprotection, and intramolecular
cyclization delivered 10 in 45% overall yield. Lactam 10
was converted into amidine 11 using the aforemen-
tioned two-step protocol.
Scheme 1. Reagents and conditions: (a) Lawesson’s reagent, PhCH3,
reflux;(b)RNH2,THFor1,4-dioxane,reflux;(c)LiOH,THF/MeOH/H2O.
Amidines 5a–h were prepared using the two-step pro-
cess shown in Scheme 1. In the event, lactam 1 was
reacted with Lawesson’s reagent5 to provide, after
removal of inorganic materials and trituration, the cor-
responding thiolactam in 65% yield. Other thionation
procedures (P2S5;6 P2S5, n-BuLi7) were explored, but
they required tedious purification and provided thio-
lactam in poor chemical yield. Attempts to activate the
lactam via conversion to its chloroimine (POCl3; PCl5)
returned starting lactam or led to intractable mixtures of
materials. With a reactive intermediate in hand, we turned
to displacements with primary amines. These transforma-
tions proceeded smoothly at reflux in THF or 1,4-dioxane
to give the desired amidines in moderate to good yield
(30–85%) after chromatography.8 Saponification of the
esters delivered the corresponding carboxylic acids.
Analogues with a fused imidazole ring were prepared as
shown in Scheme 2. Lactams 12 and 13 were converted
into the requisite amidines via the two-step procedure
described above using appropriately substituted
2-aminoalcohols. Swern oxidation of the free alcohol
present in 14 and 15 was followed by acid-catalyzed
intramolecular cyclization to give 16a–c and 17a–d in
good (50–75%) yield. The regioisomeric imidazoles,
(3S)-18a and its enantiomer (3R)-18b, were available
from midazolam (2) directly. In the event, a novel
Michael addition was employed to install the propio-
nate sidechain; the use of t-butyl acrylate optimally
provided the desired racemic adduct.10,11 The
sequence of protecting group removal and alkylation
of the intermediate acid delivered (ꢀ)-18. Chiral pre-
parative HPLC separation12 gave enantiopure13
methyl esters (3S)-18a and (3R)-18b. Saponification of
these esters provided the corresponding carboxylic
acids.
Syntheses of the C5-pyridyl target compounds 10 and
11 were achieved using the sequence outlined in
Scheme 2. Reagents and conditions: (a) (i) t-BuLi, THF, ꢁ78 to ꢁ20 ꢂC; (ii) 2-pyridinecarboxaldehyde, ꢁ78 ꢂC; (b) MnO2, CHCl3; (c) 4 M HCl/
dioxane; (d) FMOC-Glu(OMe)-Cl, CHCl3; (e) piperidine, DMF; (f) HOAc, DMF; (g) Lawesson’s reagent, PhCH3, reflux; (h) RNH2, THF or 1,4-
dioxane, reflux; (i) Swern oxidation; (j) p-TsOH, DMF; (k) KOt-Bu, CH2CHCO2t-Bu, THF/t-BuOH, ꢁ10 ꢂC; (l) TFA; (m) K2CO3, MeI, DMF; (n)
chiral HPLC separation.

G. J. Pacofsky et al. / Bioorg. Med. Chem. Lett. 12 (2002) 3219–3222
3221
The lead benzodiazepinones1 were quite potent when
Table 2. Comparative data for BZDs
evaluated in our receptor-binding assay, however, the
amidine-containing analogues of 1 generally displayed
reduced potencies (Table 1). A modest preference was
noted for smaller appendages in the series of hydro-
phobic analogues. For example, compounds 5a and 5b
were somewhat less potent than 1 while incorporation
of a phenyl group (5c) or a branched alkyl (5d) into the
amidine decreased binding affinity dramatically. This
trend was evident with hydrophilic analogues as well.
Compounds 5e and 5f were markedly weaker binders
than 1, but, again, the bulkier heterocycle-containing 5g
and 5h were considerably less potent than 1. As satis-
factory performance in the receptor binding assay was a
prerequisite for continued investigation, only simply
substituted amidines offered an alternative to the lactam
1 and amidine 5a emerged as the lead in this series.
Receptor affinity for 5a had decreased by 20-fold rela-
tive to 1, but 5a now displayed an improved solubility
profile (13 mg/mL at pH 3) relative to both midazolam
(2, ꢃ8 mg/mL at pH 3) and 1 (<3 mg/mL at pH 3
(estimated value) ). Furthermore, this amidine elicited a
full agonism response in the rat loss-of-righting reflex
(LRR)14 assessment (Table 2) to confirm that structural
changes had not altered pharmacology. Also note-
worthy are the recovery times (initial=15 min,
total=22 min) observed with this compound in the
LRR. An identical response was observed when 5a was
dosed in water adjusted to pH3. These data positioned
the amidines, specifically 5a, as a potential replacement
for 1 and broadly supported the hypothesis of lactam-
to-amidine modification.
Compd
Solubilitya
Doseb
LRR Recoveryc
Initial
Total
1
2
5a
<3d
ꢃ8
13
25
30
25
50
25
25
25
25
25
25
25
9
23
15
9
17
6
23
7
13
9
24
121
22
16
23
21
33
10
15
11
24
10
1
16a
16b
16c
17a
17b
17c
18a
ND
ND
<1
ND
2
13
ND
20
aDetermined at pH 3;15 mg/mL; ND=not determined.
bBolus injection, mg/kg; vehicle=50% PEG/25% EtOH/25% saline.
cRecovery time in min. A compound was identified as inactive in this
model if LRR was not observed within 5 min following injection.
dEstimated value.
retained. While the recovery times (initial=9 min,
total=16 min) observed in the LRR were noteworthy and
provided a significant advantage over both 1 and 5a, the
unanticipated drop in its pH-dependent solubility (1 mg/
mL at pH 3) was disappointing. Interestingly, combining
the amidine and C5 replacement strategies in the form of
11 offered no advantage in terms of receptor affinity.
Both amidine- and C5-pyridyl-containing BZDs dis-
played reasonable pharmacological profiles, but neither
series provided entities that combined these properties
with aqueous solubility. We speculated that incorpora-
tion of an additional heterocycle into the BZD scaffold
might improve solubility without impacting other attri-
butes. Towards that end, 16a–c, 17a–d, and 18a and b
were synthesized and evaluated. Receptor binding affi-
nities paralleled those of the respective parent BZDs,
and the aforementioned preference1 for the (S)-stereo-
isomer prevailed with the midazolam-derived esters, 18a
and b, as well. Moderate selectivity (i.e., receptor
potency of ester versus acid) was observed with these
series, so compounds were progressed to the in vivo
LRR assessment to confirm their full agonism profiles.
Importantly, total recovery times observed with each of
these compounds was significantly shorter than the
recovery time observed in animals dosed with 2. The
decreased aqueous solubility of 16c, as compared to 2,
suggested that related methyl group positional isomers in
the 20-haloaryl series, especially 18a, would possess solu-
bility profiles similar to 16c. This observation together
with the synthetic protocol required to deliver 18a promp-
ted its removal from further consideration despite its per-
formance in the LRR assessment. In contrast, compound
17c presented a level of in vivo efficacy and aqueous solu-
bility commensurate with our objectives. However, like
compound 5a, its relatively weak receptor affinity would
likely necessitate administration of inconveniently higher
doses, and dose volumes, and reduce its clinical utility.
Concurrently, we explored the effects of replacing the
C5-haloaryl moiety with a C5-pyridyl appendage. Bio-
logical evaluation of 10 showed a significant drop in
binding affinity as well as selectivity, relative to the
parent C5-haloaryl BZD, but a full agonism profile was
Table 1. Receptor binding affinity of BZDs
Compd
R
R1
Kia
Selectivityb
984
1
2
7
2
3
5a
5b
5c
5d
5e
5f
5g
438
96
CH3
CH2CH3
CH2Ph
156
146
875
1012
303
208
792
1230
139
1075
7
7
8
79
60
98
130
11
CH2CH(CH3)2
CH2CH2OH
(CH2)2-4-imidazolyl
(CH2)2-4-pyridyl
CH2-4-pyridyl
5h
10
11
46
16a
16b
16c
17a
17b
17c
17d
18a
18b
H
H
CH3
H
H
CH3
CH3
H
CH3
H
H
CH3
H
64
156
74
176
156
147
40
CH3
226
28
Improving the physical properties of our lead benzodia-
zepinones was achieved through various manipulations
of the BZD scaffold. The three strategies of lactam-to-
amidine modification, C5-haloaryl replacement, and
3572
aRat BZD receptor, nM.
bSelectivity=Ki (acid)/Ki (ester).

3222
G. J. Pacofsky et al. / Bioorg. Med. Chem. Lett. 12 (2002) 3219–3222
heterocycle annulation provided novel alternatives to 1,
however the lead compounds from these efforts, 5a and
17c, lacked the receptor potency that was required. Our
work to combine all attributes (potency, agonism, solu-
bility) into a single chemical series will be the subject of
future disclosures.
8. Borer, R.; Gerecke, M.; Widmer, U. WO 96/23790, 1996.
CAN 125:221886.
9. (a) Gschwend, H. W.; Rodriguez, H. R. Org. React. 1979,
26, 1. (b) Fuhrer, W.; Gschwend, H. W.; J. Org. Chem. 1979,
44, 1133.
10. For a study of the C3-lithiation of diazepam, see: Reitter, B.
E.; Sachdeva, Y. P.; Wolfe, J. F. J. Org. Chem. 1981, 46, 3945.
11. For a C3-alkyation/decarboxylation protocol to prepare
racemic triazolo-BZDs, see: Hester, J. B., Jr.; Rudzik, A. D.;
VonVoigtlander, P. F. J. Med. Chem. 1980, 23, 643. It was
reported herein that substitution at this position significantly
reduced activity in the LRR assessment.
12. Chiral HPLC purification was performed using a Daicel
AD Column (5ꢄ50 cm, 20 mm packing). Elution with 20%
i-PrOH/hexanes, 50 mL/min flow rate and UV (270 nM)
detection gave retention times of 42 min for 18a and 72 min
for 18b.
Acknowledgements
We thank Bill Hall for performing the chiral HPLC
separation/purification of 18a and b. We thank Doug
Minnick and Mil Lambert for conducting and analyzing
data from the VCD experiments.
13. Absolute configurations were assigned using vibrational
circular dichroism (VCD) spectroscopy. Spectra were mea-
sured with a Bomem ChiralIRTM VCD spectrometer. Con-
formational searching was performed on model structures
using the Molecular Visualization Program (MVP).16 Ab initio
predictions of VCD spectra for model structures were per-
formed using Gaussian ’94.
References and Notes
1. (a) Stafford, J. A.; Pacofsky; G. J.; Cox, R. F.; Cowan, J.
R.; Dorsey, Jr., G. F.; Gonzales, S. S.; Jung; D. K.; Koszalka,
G. W.; McIntyre, M. S.; Tidwell, J. H.; Wiard, R. P.; Feld-
man, P. L. 2002, 12, 3215. (b) Feldman, P. L.; Jung, D. K.;
Kaldor, I.; Pacofsky, G. J.; Stafford, J. A.; Tidwell, J. H. WO
00/69836. CAN 134:4954.
14. Hester, J. B., Jr.; Rudzik, A. D.; Kamdar, B. V. J. Med.
Chem. 1971, 14, 1078.
2. Kanto, J. H. Pharmacotherapy 1985, 5, 138.
3. Shafer, A. Crit. Care Med. 1998, 26, 947.
4. Fryer, R. I.; Schmidt, R. A.; Sternbach, L. H. J. Pharm.
Sci. 1964, 53, 264.
15. Solubility was determined by visual inspection or UV
assay of a saturated sample. In both types of determinations,
solubility at pH 3 was verified using solubility at other pH and
spectrophotometric pKa values.
5. Thomsen, I.; Clausen, K.; Scheibye, S.; Lawesson, S. O.
Org. Synth. 1984, 62, 158.
6. Archer, G. A.; Sternbach, L. H. J. Org. Chem. 1964, 29, 231.
7. Goel, O. P.; Krolls, U. Synthesis 1987, 162.
16. Lambert, M. H. In Docking conformationally flexible
molecules into protein binding sites.; Charifson, P. S., Ed.;
Practical Application of Computer-Aided Drug Design; Mar-
cel Dekker: New York, 1997; pp 243–303.