Parallel synthesis of N-biaryl quinolone carboxylic acids as selective M1 positive allosteric modulators

Article abstract of DOI:10.1016/j.bmcl.2009.11.100

An iterative analog library synthesis approach was employed in the exploration of a quinolone carboxylic acid series of selective M₁ positive allosteric modulators, and strategies for improving potency and plasma free fraction were identified.

Full text of DOI:10.1016/j.bmcl.2009.11.100

Bioorganic & Medicinal Chemistry Letters 20 (2010) 531–536
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Parallel synthesis of N-biaryl quinolone carboxylic acids as selective M1
positive allosteric modulators
a
a
a
a
Feng V. Yang a, William D. Shipe a, , Jaime L. Bunda , M. Brad Nolt , David D. Wisnoski , Zhijian Zhao ,
James C. Barrow a, William J. Ray b, Lei Ma b, Marion Wittmann b, Matthew A. Seager b,
Kenneth A. Koeplinger c, George D. Hartman a, Craig W. Lindsley a
*
a Department of Medicinal Chemistry, Merck Research Laboratories, 770 Sumneytown Pike, PO Box 4, West Point, PA 19486, USA
b Department of Alzheimer’s Research, Merck Research Laboratories, 770 Sumneytown Pike, PO Box 4, West Point, PA 19486, USA
c Department of Drug Metabolism, Merck Research Laboratories, 770 Sumneytown Pike, PO Box 4, West Point, PA 19486, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
An iterative analog library synthesis approach was employed in the exploration of a quinolone carboxylic
acid series of selective M1 positive allosteric modulators, and strategies for improving potency and
plasma free fraction were identified.
Received 9 October 2009
Revised 18 November 2009
Accepted 19 November 2009
Available online 24 November 2009
Ó 2009 Elsevier Ltd. All rights reserved.
Keywords:
M1
Positive allosteric modulator
One of the pathological hallmarks of Alzheimer’s disease (AD) is
the progressive degeneration of the cholinergic neural system.1
The cholinergic neurons release acetylcholine and are critical for
higher brain function and memory, so their deterioration leads to
irreversible loss of cognitive faculties. Cholinesterase inhibitors such
as donepezil, galantamine, and rivastigmine were developed based
on this hypothesis and were the first successful symptomatic treat-
ments for AD.2 The efficacy of cholinesterase inhibitors is modest, of
brief duration, and limited by tolerability, but indicates that enhanc-
ing cholinergic function is an effective strategy to combat AD.
The muscarinic receptors are metabotropic, class A, G-protein
coupled receptors (GPCRs) activated by acetylcholine.3 The M1
receptor is one of the five muscarinic receptor subtypes (M1
through M5) that have been identified to date, and its high expres-
sion in the hippocampus, striatum, and cortex suggest a role in
learning and memory.4 Direct activation of M1 as an approach to
treat the symptoms of AD has been explored in the clinical setting
with a number of nonselective muscarinic agonists.5 While cogni-
tive improvement was observed, the studies were limited by cho-
linergic side effects hypothesized to result from nonselective
action on other muscarinic receptor subtypes.
explored the possibility of identifying positive allosteric modula-
tors (potentiators).6,7 Allosteric binding sites are under less evolu-
tionary pressure than the orthosteric site, so potentiators of M1
may display good subtype selectivity. A high-throughput screen
for M1 potentiators identified compound 1 (Fig. 1), recently re-
ported by Ma et al., as a selective positive allosteric modulator of
M1.8 This N-benzylated quinolone carboxylic acid potentiates the
response of M1 to acetylcholine with no effect on other muscarinic
receptor subtypes, and was further characterized as a potential
lead compound. This Letter describes efforts to optimize 1 through
modifications to the quinolone core and the N1-benzyl substituent
employing an iterative analog library synthesis approach.
Compound potencies were determined in the presence of an
EC20 of acetylcholine in human M1-expressing CHO cells using cal-
cium mobilization readout on a FLIPR384 fluorometric imaging
plate reader.7,9 Plasma protein binding was determined using the
High orthosteric site conservation across muscarinic receptor
subtypes has made agonist selectivity for the M1 receptor difficult
to achieve. As a strategy for attaining subtype selectivity, we
* Corresponding author. Tel.: +1 215 652 6527; fax: +1 215 652 6345.
E-mail address: william_shipe@merck.com (W.D. Shipe).
Figure 1. In vitro profile of quinolone carboxylic acid 1.
0960-894X/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2009.11.100

532
F. V. Yang et al. / Bioorg. Med. Chem. Lett. 20 (2010) 531–536
Scheme 1. General synthesis of quinolone carboxylic acids.
Figure 2. Representative scaffolds explored.

F. V. Yang et al. / Bioorg. Med. Chem. Lett. 20 (2010) 531–536
533
Table 1
Table 1 (continued)
M1 FLIPR data for select compounds
#
R1
M1 Pot IP (nM)a
R2
O
O
a
R
R
b
c
d
5
2 = H
3 = H
R2 = H
R3 = F
R2 = F
R3 = H
R2 = F
R3 = F
OH
N
R1
8
R3
#
R1
M1 Pot IP (nM)a
29
210
380
84
270
a
R
R
b
c
d
2 = H
3 = H
R2 = H
R3 = F
R2 = F
R3 = H
R2 = F
R3 = F
ND = not determined.
a
Values represent the numerical average of at least two experiments. Interassay
variability was 30% (IP, nM), unless otherwise noted.
14
15
16
1700
4000
ND
ND
ND
3000
ND
>100,000
13,000
24,000
22,000
equilibrium dialysis method in the presence of human and rat ser-
um. Compound 1 had an M1 inflection point (IP) of 820 nM (Fig. 1)
and a plasma free fraction that varied between human and rat
(0.9% and 4.3%, respectively). The pharmacokinetic profile of 1,
characterized by low clearance and excellent oral bioavailability
across preclinical species, was very encouraging and gave us confi-
dence to further pursue this structural series. Our efforts sought to
improve on the M1 potency of 1 while maintaining an acceptable
physicochemical profile.
We viewed the structureof lead compound 1 as a contextsuitable
for iterative parallel synthesis, so a general synthesis of quinolone
analogs was employed as outlined in Scheme 1. Anilines 2 were
heated with diethyl ethoxymethylenemalonate (3) in the absence
of solvent to provide enamines 4. These intermediates were dis-
solved in diphenyl ether and further heated to elicit cyclization to
the quinolone nucleus (5). Under basic conditions, 5 was regioselec-
tively alkylated with activated halides 6 to furnish N-substituted
quinolones 7, and subsequent basic hydrolysis of the ethyl ester
yielded the carboxylic acid targets 8. To increase the scope of our ef-
forts, we also used a route wherein primary amines 12 could serve as
a source of structural diversity. Here, o-fluorobenzoylchlorideswere
used to acylate ethyl 3-(dimethylamino)acrylate (10) to produce the
N,N-dimethyl enamines 11, which underwent smooth substitution
with amines 12. The resultant enamines 13 could be cyclized to
the quinolones 7 under basic conditions.
ND
F
F
17
18
ND
ND
1200
1000
1100
1300
860
610
Br
F
19
ND
1200
830
ND
20
21
22
23
24
18,000
4300
820
13,000
3400
1100
2700
980
ND
16,000
ND
SO2Me
CONH2
ND
Optimization of the substitution pattern on the quinolone core
required beginning the synthesis with diversely substituted ani-
lines 2 or benzoyl chlorides 9. Each scaffold was substituted with
a number of common benzyl groups at the N1-position to enable
a rough stack-ranking assessment across scaffolds. After canvass-
ing a variety of fused bicyclic scaffolds, in most cases derived using
the chemistry described above, we concluded that fluorine or
hydrogen atom substitution at the 5- and/or 8-position of a quin-
olone nucleus was preferred for M1 potency (Fig. 2).
2000
2100
1500
ND
OMe
OMe
2800
ND
4200
760
We then turned to the optimization of the N1-substituent, and
chose to operate on these four ‘preferred’ scaffolds. A broad selec-
tion of monomers was employed in the diversity-yielding step, and
the products were screened in the FLIPR assay. While it was diffi-
cult to make predictive generalizations on the basis of the data
(Table 1), some trends did appear. Benzyl substitution at the 1-po-
sition was strongly preferred over alkyl (14 vs 15), and branching
at the benzyl position was poorly tolerated (14 vs 16). Halogenated
and lipophilic substituents were better tolerated than polar groups
(17–19 vs 20–21). A para substituent on the benzyl group was pre-
ferred over meta- or ortho-substitution (22 vs 23, 24 vs 25–26, 27
vs 28). These data did little to differentiate the four quinolone scaf-
folds, and relatively few compounds matched the potency of the
HTS lead 1 (=22a). However, the four biphenyl compounds 29a–d
that combined lipophilicity with placement at the para position
represented a potency breakthrough.
CF3
CF3
25
26
7200
4300
2700
6900
ND
ND
1600
2300
CF3
27
28
1800
4100
2100
3600
2800
5900
1300
3200
OCF3
OCF3

534
F. V. Yang et al. / Bioorg. Med. Chem. Lett. 20 (2010) 531–536
Unsurprisingly, 29a–d displayed a high degree of plasma pro-
phenyl ring (Scheme 2). Employing the N1-benzylation route de-
scribed in Scheme 1, p-bromobenzyl quinolone carboxylic esters
30 were prepared, and basic hydrolysis yielded the corresponding
carboxylic acids 31. These aryl bromides were combined with
substituted aryl pinacolboronate esters or boronic acids 32 under
Suzuki–Miyaura cross-coupling reaction conditions to produce
the target biaryl compounds 33.
tein binding, most likely owing to the lipophilicity of the biphenyl
substituent (Table 2). During the course of our efforts, we found
that high protein binding correlated with decreased CNS exposure
in vivo, so we next focused on addressing this limitation. To this
end, we designed biaryl compound libraries to explore variation
of the distal aryl group with para disubstitution on the proximal
Table 2
M1 FLIPR and plasma protein binding data for select compounds
R2
5
O
N
O
OH
8
R3
R4
R4
R1
#
R1
R2
H
R3
H
M1 Pot IPa (nM)
Human PBb
99.5
Rat PBb
98.8
c Log Pc
29a
H
H
H
H
F
210
2.78
29b
29c
29d
34
H
F
F
380
84
98.0
99.2
99.6
98.8
>99.9
97.8
98.7
99.5
96.3
>99.9
2.99
2.99
3.20
2.99
3.45
H
F
F
270
50
H
H
H
H
35
Cl
120
36
37
H
F
F
H
H
540
400
99.4
99.1
99.4
98.5
2.87
2.87
N
N
H
O
OH
38
39
40
41
H
H
H
F
F
H
H
H
H
99
110
76
97.6
98.1
99.8
99.0
98.3
99.6
99.8
98.5
1.66
2.69
3.28
3.18
N
H
H
N
F
O
NH
F
N
H
190
NH2
42
43
44
H
H
H
F
F
F
H
H
H
210
94
96.4
99.0
95.2
97.8
98.8
95.2
2.24
3.15
1.50
NMe2
NMe2
220
a
b
c
Values represent the numerical average of at least two experiments. Interassay variability was 30% (IP, nM), unless otherwise noted.
Percent bound.
c Log P values were calculated using the Cerius2 product marketed by Accelrys Software.

F. V. Yang et al. / Bioorg. Med. Chem. Lett. 20 (2010) 531–536
535
Scheme 2. General synthesis of biaryl quinolone carboxylic acids.
A series of substituted biphenyl compounds was first made in
an attempt to improve plasma free fraction (Table 2). Halogen sub-
stitution on the phenyl ring proximal to the quinolone core im-
proved M1 potency (29a vs 34–35) but only marginally improved
free fraction in the case of 34. Inclusion of polar groups (36–39)
and fused bicyclic heterocycles (40–41) was generally ineffective,
Table 3
M1 FLIPR and plasma protein binding data for select compounds
R2
5
O
N
O
OH
8
R3
R4
R1
#
R1
R2
H
R3
F
R4
H
M1 Pot IPa (nM)
Human PBb
94.2
Rat PBb
97.8
c Log Pc
N
OH
45
250
1.88
N
N
46
47
F
H
F
H
H
61
94.5
95.3
94.9
95.4
1.84
1.63
H
130
NH2
N
N
48
49
50
H
H
H
F
F
F
H
H
H
48
51
63
98.7
98.2
99.3
99.3
99.1
99.5
2.19
0.94
2.36
NHMe
N
NMe
N
N
OMe
51
52
H
H
F
F
H
H
100
45
97.6
98.5
98.0
99.2
1.84
2.54
NMe2
N
N
53
54
H
H
F
H
F
830
510
87.3
95.4
96.5
96.3
1.21
1.72
N
N
N
H
OMe
N
55
56
57
H
F
H
H
H
F
F
F
56
41
37
94.1
95.6
98.0
89.8
98.0
98.6
1.43
1.64
1.64
NH
N
NH
N
H
NMe
a
b
c
Values represent the numerical average of at least two experiments. Interassay variability was 30% (IP, nM), unless otherwise noted.
Percent bound.
c Log P values were calculated using the Cerius2 product marketed by Accelrys Software.

536
F. V. Yang et al. / Bioorg. Med. Chem. Lett. 20 (2010) 531–536
but aniline 42 represented an incremental improvement. While its
N,N-dimethyl analog 43 was highly protein bound, the homologous
basic tertiary amine 44 had a free fraction of nearly 5% in both hu-
man and rat.
quinolone core were found to be effective strategies for decreasing
plasma protein binding,11 but these compounds nevertheless
showed poor brain exposure, possibly as a result of the action of
CNS transporters. Further studies are underway to identify selec-
tive M1 allosteric modulators with improved CNS exposure.
In addition to providing improved M1 potency with greater con-
sistency, libraries incorporating heterocyclic biaryls were even
more fruitful in the search for compounds with decreased protein
binding (Table 3). A number of pyridines (45–52) were examined,
and among this group were two (46–47) with ꢀ5% free fraction.
Pyrimidines 53 and 54 also lowered protein binding significantly,
but with a considerable loss of potency (ꢀ10-fold with respect to
pyridine 46). Pyrazole 55 represents another example combining
excellent potency with appreciable free fraction. Inclusion of a
fluorine atom at the quinolone 5-position (56) or an N-methyl
group on the pyrazole (57) maintained potency, but had a detri-
mental effect on the free fraction. Comparison of calculated Log P
values for the biphenyls as a group versus the heterocyclic biaryls
largely demonstrates the expected inverse relationship between
c Log P and plasma free fraction. However, within the group of het-
erocyclic biaryls, 45 and 46 showed a higher free fraction than
might have been predicted on the basis of the c Log P, providing
some validation of our choice to pursue this series of compounds
with parallel synthesis.
There is evidence to suggest that drug concentration in cerebro-
spinal fluid (CSF) is a reasonable predictor for unbound drug con-
centration in the brain.10 A number of compounds in this series
were therefore orally dosed to rats at 10 mg/kg, and CSF was used
as a surrogate for in vivo assessment of CNS exposure. Compounds
44 and 46 are discussed as representative of the biphenyl and het-
erocyclic biaryl subseries, respectively. At a time point of 1 h, com-
pound 44 showed a concentration of only 11 nM in the CSF; 46 a
concentration of 24 nM. In a P-glycoprotein efflux assay, 44 and
46 both showed efflux ratios greater than 3, indicating that they
are substrates for P-gp. Thus, the lower than expected exposures
observed here may result from the action of these or other efflux
transporters at the blood–brain barrier.
Acknowledgments
We thank Wei Lemaire, Scott D. Mosser, Rodney A. Bednar,
Michael W. Marlatt, Kristi L. Hoffman, Bang-Lin Wan, Charles W.
Ross, III, Joan S. Murphy, Vincent van Nostrand, Kevin B. Albertson,
Ray T. McClain, Anna Dudkina, April Cox, and Christopher J.
Denicola for technical support of this work.
References and notes
1. (a) Terry, A. V.; Buccafusco, J. J. J. Pharmacol. Exp. Ther. 2003, 306, 821; (b) Geula,
C. Neurology 1998, 51, 18.
2. Hansen, R.; Gartlehner, G.; Webb, A. P.; Morgan, L. C.; Moore, C. G.; Jonas, D. E.
Clin. Interv. Aging 2008, 3, 211.
3. (a) Bonner, T. I.; Buckley, N. J.; Young, A. C.; Brann, M. R. Science 1987, 237, 527;
(b) Bonner, T. I. Trends Pharmacol. Sci. 1989, 11.
4. Levey, A. I. Proc. Natl. Acad. Sci. 1996, 93, 13541.
5. Bodick, N. C.; Offen, W. W.; Levey, A. I.; Cutler, N. R.; Gauthier, S. G.; Satlin, A.;
Shannon, H. E.; Tollefson, G. D.; Rasumussen, K.; Bymaster, F. P.; Hurley, D. J.;
Potter, W. Z.; Paul, S. M. Arch. Neurol. 1997, 54, 465.
6. (a) Christopoulos, A. Nat. Rev. Drug Disc. 2002, 1, 198; (b) Wess, J. Mol.
Pharmacol. 2005, 68, 1506.
7. For prior efforts to develop selective allosteric modulators of M1, see: (a) Marlo,
J. E.; Niswender, C. M.; Days, E. L.; Bridges, T. M.; Xiang, Y.; Rodriguez, A. L.;
Shirey, J. K.; Brady, A. E.; Nalywajko, T.; Luo, Q.; Austin, C. A.; Williams, M. B.;
Kim, K.; Williams, R.; Orton, D.; Brown, H. A.; Lindsley, C. W.; Weaver, C. D.;
Conn, P. J. Mol. Pharmacol. 2009, 75, 577; (b) Conn, P. J.; Jones, C. K.; Lindsley, C.
W. Trends Pharmacol. Sci. 2009, 30, 148. and references cited therein.
8. Ma, L.; Seager, M.; Wittmann, M.; Jacobsen, M.; Bickel, D.; Burno, M.; Jones, K.;
Kuzmick-Graufelds, V.; Xu, G.; Pearson, M.; McCampbell, A.; Gaspar, R.;
Shughrue, P.; Danziger, A.; Regan, C.; Flick, R.; Pascarella, D.; Garson, S.;
Doran, S.; Kreatsoulas, C.; Veng, L.; Lindsley, C. W.; Shipe, W.; Kuduk, S. D.; Sur,
C.; Kinney, G.; Seabrook, G. R.; Ray, W. J. Proc. Natl. Acad. Sci. 2009, 106, 15950.
9. Select compounds in this structural series were examined in functional assays
at other muscarinic receptors and showed no activity at M2, M3, or M4
receptors. They were also shown to be potent potentiators of the human M1
receptor, eliciting a robust leftward shift of the acetylcholine dose–response
curve at fixed concentrations of the potentiator.
Selective M1 positive allosteric modulators were discovered
based on N-benzylated quinolone carboxylic acid lead 1. Efforts
driven by parallel synthesis identified that fluorine atom substitu-
tion at the 5- and/or 8-positions of the quinolone ring system were
preferred in terms of M1 potency. Analog libraries also led to the
discovery that biaryl substituents at N1 further enhanced potency,
providing a 20-fold improvement over lead compound 1. Incorpo-
ration of a basic tertiary amine or certain heterocycles distal to the
10. Lin, J. Curr. Drug Metab. 2008, 9, 46.
11. Further studies explored compounds containing a heterocyclic aryl group
proximal to the quinolone core as another strategy to reduce protein binding:
Kuduk, S.D.; Di Marco, C.N.; Cofre, V.; Pitts, D.R.; Ray, W.J.; Ma, L.; Wittmann,
M.; Seager, M.A.; Koeplinger, K.A.; Thompson, C.D.; Hartman, G.D.; Bilodeau,
M.T. Bioorg. Med. Chem. Lett. 2009, 19 (manuscript following this one)
doi:10.1016/j.bmcl.2009.11.059.