Structural insights into monoamine oxidase inhibitory potency and selectivity of 7-substituted coumarins from ligand- and target-based approaches

Article abstract of DOI:10.1021/jm060183l

A new series of 3-, 4-, 7-polysubstituted coumarins have been designed and evaluated for their monoamine oxidase A and monoamine oxidase B (MAO-A and MAO-B) inhibitory potency. Substituents at position 7 consisted of a bridge of different physicochemical

Full text of DOI:10.1021/jm060183l

4912
J. Med. Chem. 2006, 49, 4912-4925
Structural Insights into Monoamine Oxidase Inhibitory Potency and Selectivity of 7-Substituted
Coumarins from Ligand- and Target-Based Approaches
Marco Catto,† Orazio Nicolotti,† Francesco Leonetti,† Andrea Carotti,† Angelo Danilo Favia,† Ramo´n Soto-Otero,‡
Estefan´ıa Me´ndez-AÄ lvarez,‡ and Angelo Carotti*,†
Dipartimento Farmaco-Chimico, UniVersita` di Bari, Via Orabona 4, 70125 Bari, Italy, and Grupo de Neuroqu´ımica, Departamento de
Bioqu´ımica y Biolog´ıa Molecular, Facultad de Medicina, UniVersidad de Santiago de Compostela, Santiago de Compostela, Spain
ReceiVed February 16, 2006
A new series of 3-, 4-, 7-polysubstituted coumarins have been designed and evaluated for their monoamine
oxidase A and monoamine oxidase B (MAO-A and MAO-B) inhibitory potency. Substituents at position 7
consisted of a bridge of different physicochemical nature linking a phenyl ring to the coumarin scaffold.
Structure-affinity and structure-selectivity relationships, derived through CoMFA-GOLPE and docking
studies, revealed the key physicochemical interactions responsible for the observed MAO-B and MAO-A
inhibitory potency and suggested the main structural determinants for high selectivity toward one of the
two enzymatic isoforms. The predictive power of our models was proved with the design of a new inhibitor
demonstrating an outstanding MAO-B affinity (pIC50 ) 8.29) and the highest MAO-B selectivity (∆pIC50
) 3.39) within the entire series of ligands examined herein.
Introduction
breakthrough in the field occurred only few years ago with the
resolution of the X-ray crystal structure of human MAO-B
(hMAO-B)10 and rat MAO-A (rMAO-A).11 The high-resolution
3D structure of the hMAO-B in complex with a large number
of selective inhibitors, covalently bound to the FAD cofactor,10
finally paved the way to the structure-based design of new
modulators of MAO-B activity. More recently, the X-ray
crystallographic structure of hMAO-A, covalently bound to the
MAO-A selective inhibitor clorgyline, has been reported.12 As
a result, a direct comparison of the binding modes of selective
MAOIs was made possible with the potential for a structure-
based design of new, more potent and selective hMAOIs.
Moreover, a recently published method for screening large series
of MAOIs on cloned hMAO-B13 (for cloned hMAO-A, a similar
approach is being developed by the same research groups)
should be particularly helpful for a more efficient and rapid
discovery of novel MAOIs, since accumulating evidence has
shown that inhibition data from rMAOs (the most used source
of MAOs) often strongly differ from data collected with similar
methods from human enzymes.14,15
Following an initial study on the mechanisms of neurotoxi-
cation of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)
by MAO-B,16 we recently carried out a series of 2D- and 3D-
QSAR studies of condensed (di)azines,17 geiparvarins,18 and
coumarins,19,20 generally acting as strong and selective MAO-
BIs. As for the last class of compounds, we prepared and tested
a very large array of inhibitors, which also proved to be potent
and selective MAO-BIs.19,20 Moreover, some of them showed
an interesting additional inhibitory activity toward acetylcho-
linesterase (AChE).21
Monoamine oxidase (MAO; EC 1.4.3.4, amine-oxygen
oxidoreductase) is a membrane-bound, FAD(FMN)-containing
enzyme involved in the oxidative deamination of exogenous
and endogenous amines, including neurotransmitters.1 MAO
exists as two distinct enzymatic isoforms, MAO-A and MAO-
B,2,3 which differ in amino acid sequence, substrate specificity,
sensitivity to inhibitors, and tissue distribution.4 MAO-A
preferentially deaminates the neurotransmitters serotonin, nore-
pinephrine, and epinephrine and is selectively inhibited by
clorgyline, whereas MAO-B acts preferentially on â-phenethy-
lamines and sterically hindered amines and is selectively
inhibited by selegiline (L-deprenyl)5 (Chart 1).
MAO became a neuropharmacological target in the late 1950s
with the discovery of the antidepressant properties of MAO
inhibitors (MAOIs), such as iproniazid and tranylcypromine
(Chart 1). Lack of selectivity, irreversible mechanism of action,
severe hepatotoxicity, and even fatal hypertensive crisis of the
first-generation MAOIs stimulated further research aimed at the
discovery of new, more selective, and less toxic drugs.6 Several
selective MAO-A inhibitors (MAO-AIs), acting as antidepres-
sants (brofaromine, clorgyline, moclobemide and toloxatone),
and selective MAO-B inhibitors (MAO-BIs), used alone or in
combination with L-Dopa in the therapy of Parkinson’s disease
(i.e., selegiline and rasagiline,5,7,8 Chart 1), were discovered.
The pronounced neuroprotective effect of selegiline and other
selective MAO-BIs in animal models suggested their possible
application in the symptomatic treatment of Alzheimer’s
disease.9
The most clinically relevant MAOIs belong to the class of
propargylamines, which share a common mechanism of action
implying the formation of a covalent bond with the flavin ring
of the FAD cofactor.5
Although significant advances in the design of selective
MAOIs have been achieved in the past 2 decades, a veritable
Beginning with these findings and keeping in mind all the
relevant structural information on MAOs, we designed other
structural variations on coumarins to explore in more detail the
structure-affinity and structure-selectivity relationships (SAFIRs
and SSRs, respectively) for in vitro MAO inhibition.
Our previous investigations19,20 highlighted significant changes
in MAO-B inhibitory potency and MAO-B/MAO-A (from here
on B/A) selectivity arising from the introduction of diverse
substituents at position 7. However, given that only few such
structural modifications have been made, we decided to expand
* To whom correspondence should be addressed. Phone: +39 080 544-
2782. Fax: +39 080 544-2230. E-mail carotti@farmchim.uniba.it.
† Universita` di Bari.
‡ Universidad de Santiago de Compostela.
10.1021/jm060183l CCC: $33.50 © 2006 American Chemical Society
Published on Web 07/07/2006

Inhibitory Potency and SelectiVity of Coumarins
Journal of Medicinal Chemistry, 2006, Vol. 49, No. 16 4913
Chart 1. Chemical Structures of Nonselective and Selective MAO-A (A) and MAO-B (B) Inhibitors
Table 1. MAO-A and -B Inhibitory Activities of Coumarin Derivatives
6-43
the investigation of the SAFIRs and SSRs through the synthesis
and in vitro biological evaluation of a novel series of 7-substi-
tuted coumarin derivatives. The substituents at position 7 consist
of a phenyl ring linked to a coumarin core by bridges of different
size, length, and lipophilic and electronic nature. Along with
the newly prepared coumarin derivatives (i.e., compounds 8,
15-17, 21, 22, 27, 29, (()-31, (-)-31, (+)-31, 33, 36, and 37,
Table 1), other previously reported 7-substituted coumarins, with
no, one, or two methyl groups in position 3 or 4 were considered
for a more complete analysis.
pIC50
g
compd
R3
R4
R7
MAO-A MAO-B ∆pIC50
6a
H
H
CH3
H
OCH2C6H5
CH3 OCH2C6H5
OCH2C6H5
5.17
5.71
5.26
6.16
6.41
4.38
5.42
5.21
4.66
6.35
5.02
5.15
6.39
5.80
5.86
5.00
4.73
7.12
7.33
7.90
7.15
5.34
5.45
4.50f
5.01
5.00
5.49
5.00
6.00
4.99
4.64
5.70
5.55
4.50f
4.00f
6.24
5.95
6.91
6.17
4.90
7.26
7.74
8.18
8.36
7.07
5.67
5.89
6.24
4.59
4.26
5.62
6.34
7.55
6.79
6.72
7.40
5.56
5.28
4.50f
4.77
5.00
4.50f
6.49
5.49
7.74
7.15
6.52
7.55
8.25
7.57
7.28
4.00f
6.06
6.95
7.24
8.55
8.48
8.94
8.52
8.29
2.09
2.03
2.92
2.20
0.66
1.29
0.47
1.03
-0.07
-2.09
0.60
1.19
1.16
0.99
0.86
2.40
0.83
-1.84
-2.83
-3.13
-2.15
-0.84
1.04
0.99
2.73
2.15
1.03
2.55
2.25
2.58
2.64
-1.70
0.51
2.45
3.24
7a
8
H
9a
CH3 CH3 OCH2C6H5
Chemistry
10a
11a
12b
13b
14b
15
H
H
H
H
H
H
H
H
H
H
H
H
H
H
CH2OC6H5
CH2NHC6H5
CH2SC6H5
CH2S(O)C6H5
CH2S(O2)C6H5
OS(O2)C6H5
CH2OC(O)C6H5
The synthesis of many novel coumarin derivatives was
performed starting from commercial 7-hydroxycoumarin (um-
belliferone, 1) and previously reported 3,4-dimethyl-7-hydroxy-
coumarin (2),19 7-amino-3,4-dimethylcoumarin (3),22 and 7-(bro-
momethyl)coumarin (4)19 (Chart 2).
3-Methylcoumarin derivative 8 was prepared, according to
the pathway reported in Scheme 1, from 7-hydroxy-3-methyl-
coumarin 5, prepared in turn from 2,4-dihydroxybenzaldehyde,
and benzyl bromide in refluxing ethanol.
Phenyl sulfonate 15 was prepared by reacting 7-hydroxy-
coumarin (1) with benzenesulfonyl chloride in pyridine. Ben-
zoate 16 was obtained from benzoic acid and 7-(bromomethyl)-
coumarin (4), following a standard synthetic procedure.
â-Phenethyl derivative 17 was synthesized from styryl derivative
1819 (Table 1) by catalytic hydrogenation over activated
platinum on carbon.
Diazotization of 7-aminocoumarin derivative 322 with sodium
nitrite in hydrochloric acid, followed by the nucleophilic
substitution with sodium benzylmercaptide, yielded thioether
21, which afforded sulfone 22 upon oxidation with potassium
permanganate (Scheme 2).
16
17
CH3 CH3 CH2CH2C6H5
CH3 CH3 tCHdCHC6H5
CH3 CH3 NHCH2C6H5
CH3 CH3 NHC(O)C6H5
CH3 CH3 SCH2C6H5
CH3 CH3 S(O2)CH2C6H5
CH3 CH3 OS(O2)C6H5
CH3 CH3 OS(O2)-4′-CH3-C6H4
CH3 CH3 OS(O2)-4′-NO2-C6H4
CH3 CH3 OS(O2)-4′-OCH3-C6H4
CH3 CH3 N(CH3)S(O2)-4′-CH3C6H4
CH3 CH3 OCH(CH3)C6H5
CH3 CH3 OC6H5
18a
19a
20a
21
22
23a
24a
25a
26a
27
28a
29
30b
CH3 CH3 OCH2C(O)C6H5
(()-31 CH3 CH3 OCH2CH(OH)C6H5
(-)-31 CH3 CH3 OCH2CH(OH)C6H5
(+)-31 CH3 CH3 OCH2CH(OH)C6H5
32a
33
CH3 CH3 OCH2CH2C6H5
CH3 CH3 OCH2CH2OC6H5
CH3 CH3 OCH2CH2CH2C6H5
CH3 CH3 OS(O2)CH2C6H5
CH3 CH3 OCH2S(O2)C6H5
CH3 CH3 OCH2-tCHdCHC6H5
34c
35d
36
37
38e
39a
40a
41a
42a
43
H
H
OCH2-tCHdCHC6H5
CH3 CH3 OCH2-3′-F-C6H4
CH3 CH3 OCH2-3′-Cl-C6H4
CH3 CH3 OCH2-3′,4′-F2-C6H3
CH3 CH3 OCH2-3′,5′-F2-C6H3
2.31
2.53
2.03
2.35
Sulfonamido derivative 27 was prepared by N-methylation
of the already reported 3,4-dimethyl-7-(4-tolylsulfonamido)-
coumarin19 with iodomethane in DMF.
CH3
H
OCH2-3′-Cl-C6H4
3.39
7-Phenoxycoumarin derivative 29 was prepared according to
the classical von Pechmann procedure by heating 3-phenox-
yphenol and ethyl 2-methylacetoacetate with acid catalysis
(Scheme 3).
a From ref 19. b From ref 20. c From ref 42. d From ref 43. e From ref
44. f Truncated value estimated from the percent of inhibition at two different
concentrations. g ∆pIC50 is the difference between pIC50 of MAO-B and
pIC50 of MAO-A.
Racemic benzyl alcohol 31 was obtained by reduction of the
phenone derivative 3020 with lithium aluminum hydride in THF.
The enantiomeric resolution of the racemic mixture of 31 was
performed by preparative HPLC on a Chiralcel OD column.
Compounds 33, 36, and 37 were synthesized from 3,4-
dimethyl-7-hydroxycoumarin (2) by nucleophilic substitution

4914 Journal of Medicinal Chemistry, 2006, Vol. 49, No. 16
Chart 2. Starting Coumarin Derivatives
Catto et al.
Scheme 1a
Figure 1. Square plot of rMAO affinity (pIC50) and selectivity. Top-
left and bottom-right corners contain inhibitors with high selectivity
toward rMAO-B and rMAO-A isoenzymes, respectively. Two of the
most potent and selective inhibitors of each enzymatic isoform are
highlighted.
a Reaction conditions: (i) CH3CH2COONa, (CH3CH2CO)2O, piperidine,
∆; (ii) benzyl bromide, K2CO3, EtOH, ∆.
Scheme 2a
To avoid the loss of important information for comparative
molecular field analysis (CoMFA), estimated pIC50 (-log IC50)
values were used for those few inhibitors (see footnote f in Table
1) whose low solubility did not permit the determination of the
IC50. ∆pIC50, that is, the difference between pIC50 of MAO-B
and pIC50 of MAO-A, was used as a measure of B/A selectivity
in CoMFA. For a more immediate and efficient analysis of the
variation of both affinity and selectivity, inhibition data are
presented in Figure 1 as a plot of pIC50 of MAO-A (x-axis)
versus pIC50 of MAO-B (y-axis) using the same scale and range
for both axes (square plot). Thus, compounds with equal
affinities at both isoenzymes lie on the bisector (y ) x) of this
plot, whereas MAO-A or MAO-B selective inhibitors are
situated below or above the bisector, respectively. The distance
of their pIC50 values from the bisector is a direct measure of
their degree of selectivity. Four parallel lines were traced at
one ∆pIC50 unit distance above and below the bisector to enable
a straightforward location of inhibitors with an IC50(MAO-A)/
IC50(MAO-B) affinity ratio (selectivity) higher than 10 and 100
(i.e., ∆pIC50 > 1 and ∆pIC50 > 2, respectively). Inhibitors
endowed with both high MAO-A affinity and A/B selectivity
(i.e., 25) are located in the lowest right-hand corner of the plot,
whereas inhibitors with both high MAO-B affinity and B/A
selectivity (i.e., 8) are located in the upper left-hand corner. At
a first glance, the plot suggests that very potent and selective
MAO-B inhibitors are more represented in the analyzed set and
that no apparent relationship exists between MAO-A and
MAO-B affinities.
a Reaction conditions: (i) NaNO2, HCl; (ii) benzylmercaptan sodium
salt; (iii) KMnO4, AcOH, ∆.
Scheme 3a
a Reaction conditions: (i) H+, ∆.
Scheme 4a
In a previous report,19 a significant increase of MAO-B
inhibitory affinity of 7-benzyloxycoumarin lead 6 was observed
by introducing methyl groups at positions 4 and 3,4 (compounds
7 and 9, respectively) and introducing the fluorine (39) or
chlorine (40) atom at the meta position of the benzyloxy ring
of 9. We have now prepared the 3-methyl regioisomer of 7 (8),
which displayed significantly higher MAO-B affinity and B/A
selectivity: pIC50 ) 8.18 vs 7.74; ∆pIC50 ) 2.92 vs 2.03.
Moreover, compared to the 3,4-dimethyl congener 9, 3-methyl
derivative 8 exhibited close MAO-B affinity (pIC50 ) 8.18 vs
8.36) and significantly higher selectivity (∆pIC50 ) 2.92 vs
2.20). These important findings suggested a straightforward
access to highly potent and more selective MAO-B inhibitors.
a Reaction conditions: (i) K2CO3, DMF, ∆.
with appropriate bromoalkyl derivatives in DMF and potassium
carbonate as an acid scavenger (Scheme 4).
Results and Discussion
The chemical structures of MAOIs examined in this report
are shown in Table 1, along with their in vitro inhibitory
potencies and B/A selectivity. MAO inhibitory activities were
determined on rat brain mitochondria using a continuous
spectrophotometric assay based on the monitoring of the
oxidation rate of the nonselective, nonfluorescent MAO substrate
kynuramine to the fluorescent 4-hydroxyquinoline.51
The substitution of both hydrogens at positions 3 and 4 with
methyl groups led to an almost 10-fold increase in MAO-B and

Inhibitory Potency and SelectiVity of Coumarins
Journal of Medicinal Chemistry, 2006, Vol. 49, No. 16 4915
MAO-A inhibitory activities (compare 9 vs 6, 23 vs 15), with
the only exception being the cinnamyl derivatives (37 vs 38).
Isosteric variations of the oxymethylene linker in reference
compound 6 led to inhibitors 10-12, all endowed with lower
MAO-B affinity and B/A selectivity. Similar isosteric changes
on the 3,4-dimethyl lead compound 9 afforded similar results
for phenethyl (17) and benzylamino (19) derivatives, whereas
a decrease of MAO-B affinity (pIC50 ) 7.55 vs 8.36) associated
with a slight increase of MAO-A affinity (pIC50 ) 6.39 vs 6.16)
was observed for trans-styryl derivative 18. Interestingly, the
thiobenzyl derivative 21 showed a B/A selectivity higher than
isoster lead compound 9 (∆pIC50 ) 2.40 vs 2.20) as a result of
a stronger diminution of the MAO-A affinity (pIC50 ) 6.16-
5.00).
The influence of the linker’s length on MAO affinity was
assessed by preparing either lower (29) or higher (32 and 34)
homologues of lead compound 9, as well as the dioxoethylene
(33) and the cinnamyl (37) derivatives. Phenoxy derivative 29
showed a dramatic decrease of affinity at both MAO isoen-
zymes, while the ethoxy derivative 32 fully retained the activity
profile of reference ligand 9. Further elongation of the bridge
with a methylene (34) or an oxygen atom (33) decreased
MAO-B and, to a higher degree, MAO-A affinity, yielding less
potent but highly MAO-B selective inhibitors. Interestingly, the
long but more rigid cinnamyl derivative 38 displayed MAO
affinity values very similar to those of the corresponding
hydrogenated analogue 34.
The oxidation of thioether 12 to sulfone 14 induced a strong
decrease of MAO-B affinity and, consequently, B/A selectivity.
The same transformation performed on homologous thioether
21 afforded sulfone 22, which presented very poor affinities
for both MAO isoenzymes. The oxidation of thioether 12 to
sulfoxide 13 restored satisfactory MAO-B affinity and B/A
selectivity.
The biological evaluation of new sulfonic ester 15 confirmed
the inversion of selectivity already observed in the corresponding
homologue 23.19,23 This effect was even more pronounced in
the para-substituted phenyl sulfonates 24-26 reported in our
previous paper.19 The affinity and selectivity of other newly
synthesized sulfonyl-containing inhibitors, such as N-methyl-
tolylsulfonamide 27 and benzylsulfonate 35, demonstrate that
an intact sulfonate group, directly linked to the coumarin ring,
is a key structural requisite for high A/B selectivity. A further
support to this general rule was given by sulfone 22, which
exhibited an inverted B/A selectivity compared to the isosteric
sulfonate 23 (∆pIC50 ) 0.83 vs -1.84).
To further evaluate the modulation of affinity arising from
the introduction of polar oxygenated groups in the bridge,
benzoate (16) and phenone (30) derivatives were prepared.
Compound 30, the only ketonic derivative in the entire series,
displayed quite high MAO-B affinity and B/A selectivity.
Compound 16 was endowed with an affinity in the micromolar
range at both enzyme isoforms. It must be remembered that
benzoate 16 could be partly hydrolyzed during the inhibition
assay. This might explain the unexpectedly high drop of affinity
compared to more stable inhibitors with similar structural
features.
Figure 2. Incorporation of the molecular alignment used in CoMFA
analysis into the rMAO-B binding site.
the introduction of a hydroxyl group in the bridge at position 7
affording alcohol 31. To verify whether an enantioselective
interaction might occur at the enzyme binding sites with the
benzylic hydroxyl, the racemic mixture of 31 was resolved
through preparative chiral HPLC and the two enantiomers were
tested individually. The (+)-enantiomer revealed higher MAO-B
affinity and slightly lower MAO-A affinity than (-)-enantiomer.
This determined a significantly enhanced B/A selectivity
(∆pIC50 ) 2.55 vs 1.03) and suggested a possible enantiose-
lective interaction of the benzyl alcohol moiety at the MAO-B
binding site. Additional studies are warranted for a complete
comprehension of the enantioselective interactions of polar
substituents placed on the bridge at position 7.
3D-QSAR Studies. Although most of the observations
reported above surely helped to clarify some important aspects
of the SAFIRs and SSRs, an exact picture of the main
interactions taking place at the active sites of the two rat
isoenzymes was still missing. To overcome this limitation, a
combined modeling study was carried out through CoMFA,24
homology building, and docking simulations. The purpose of
this study was twofold: (i) to gain further insights on the
physicochemical nature and precise spatial location of the key
interactions of the 7-substituents responsible for the observed
enzyme affinities and, more importantly, selectivities and (ii)
to confirm, complement, and better interpret the results of
previous studies conducted by some of us on different sets of
coumarin derivatives.19,20
Besides classical CoMFA of MAO-A and MAO-B affinities,
expressed as pIC50, an additional 3D-QSAR analysis was
executed to model directly B/A selectivity, measured as the
difference between pIC50 of MAO-B and pIC50 of MAO-A
(∆pIC50). Molecular alignment, the most critical step in CoMFA,
was made by adopting the same criteria reported in our previous
work19 and taking advantage of the knowledge of the space
accessible to 7-substituents to the binding site of MAO-B (see
Figure 2).
Along with classical CoMFA interaction fields (i.e., steric
and electrostatic fields) computed within SYBYL,25 the lipo-
philic field, calculated by the CLIP program,26 was also used
as an additional descriptor to better explain the variance in the
biological data. The results of the PLS analyses, obtained by
Finally, the effect of branching in the bridge on MAO affinity
was studied with the synthesis of ethoxy derivative 2819 and
alcohol 31. The introduction of the methyl group on the bridge
of compound 9 leading to 28 determined a remarkable decrease
of MAO-B affinity (about 2 orders of magnitude), and as a
consequence, B/A selectivity strongly decreased (∆pIC50 ) 1.04
vs 2.20). A similar, but less intense, effect was observed with

4916 Journal of Medicinal Chemistry, 2006, Vol. 49, No. 16
Catto et al.
Table 2. Statistics of the CoMFA-GOLPE Models of MAO-B
Inhibition (n ) 38)
investigated, and therefore, the new PLS models could be
considered complementary rather than alternative to previous
models.
model
field type
q2 a
ONCb
r2 c
SDEPd
Isocontour maps from the three-field PLS models VII (MAO-
B) and XIV (MAO-A) were developed to facilitate the spatial
location of the main interactions influencing the inhibitory
potency. The following color code was used to point out steric,
electrostatic, and lipophilic interactions: red/green to indicate
regions for unfavorable/favorable steric effects, gray/magenta
to indicate contour zones where a high electron density is
unfavorable/favorable to affinity, and cyan/yellow to display
zones where unfavorable/favorable lipophilic interactions may
occur. As can be seen in Table 1, the introduction of a halogen
atom in the meta position of the 7-benzyloxy group significantly
increased the binding affinities. Indeed, this observation is in
full agreement with the appearance of favorable signals in the
meta region in steric, electrostatic, and lipophilic isocontour
maps, indicated in Figure 3 by green, magenta and yellow areas,
respectively. Similarly, the negative effects on MAO-B inhibi-
tory activity exerted by the sulfonate group of derivatives 23-
26 were pointed out clearly in the three isocontour maps of
Figures 4. The poor MAO-B affinity of the phenoxy derivative
29 gave rise to unfavorable signals again in the three different
isocontour maps. Other significant signals came from specific
isocontour maps. The increased MAO-B affinity of monomethyl
and dimethyl derivatives in positions 3 and 4 was well explained
by a green zone in the steric map of Figure 3a, whereas the
high MAO-A affinity of p-nitrophenyl sulfonate 25 was
accounted for by magenta polyhedra in the electrostatic contour
map (Figure 4b). Sterically forbidden areas were impacted by
phenyl rings of the long inhibitors 33, 34, 37, and 38, which
were endowed with poor MAO-A affinity. The steric maps of
MAO-A (Figure 4a) furnished readily interpretable negative red
signals for the sulfoxide derivative 22, but surprisingly, no green
area for the OSO2 groups of the highly active sulfonic esters
23-26 was detected. Interestingly the different MAO-B affini-
ties of chiral alcohols 31 can be explained, at least in part, by
the electrostatic and lipophilic isocontour maps. Accordingly
the distomer (-)-31 and eutomer (+)-31 should have the R and
S absolute configurations, respectively.
I
II
Ste (S-B)
Ele (E-B)
Lipo (L-B)
0.692
0.805
0.674
0.830
0.677
0.834
3e
2
2
2
2
2
2
0.821
0.885
0.785
0.899
0.787
0.895
0.906
0.756
0.601
0.776
0.561
0.773
0.554
0.549
III
IV
V
VI
VII
Ele + Ste (ES-B)
Lipo + Ste (LS-B)
Ele + Lipo (EL-B)
Ste + Ele + Lipo (SEL-B) 0.837
a Leave-one-out squared cross-validated correlation coefficient. b Optimal
number of PLS components. c Squared correlation coefficient. d Standard
deviation of error of predictions. e Two-component statistics of model I:
q2 ) 0.615; r2 ) 0.735; SDEP ) 0.844.
Table 3. Statistics of the CoMFA-GOLPE Models of MAO-A
Inhibition (n ) 38)
model
field type
q2 a
ONCb
r2 c
SDEPd
VIII
IX
X
Ste (S-A)
Ele (E-A)
Lipo (L-A)
0.727
0.673
0.708
0.751
0.738
0.659
2
2
2
2
2
2
2
0.819
0.811
0.802
0.859
0.823
0.819
0.881
0.461
0.505
0.447
0.441
0.452
0.515
0.406
XI
Ele + Ste (ES-A)
Lipo + Ste (LS-A)
Ele + Lipo (EL-A)
XII
XIII
XIV
Ste + Ele + Lipo (SEL-A) 0.789
a Leave-one-out squared cross-validated correlation coefficient. b Optimal
number of PLS components. c Squared correlation coefficient. d Standard
deviation of error of predictions.
means of the leave-one-out cross-validation procedure27 and
GOLPE28 (for data pretreatment and analysis), are listed in
Tables 2, 3, and 4 for MAO-B affinity, MAO-A affinity, and
B/A selectivity, respectively.
Some differences from our previous CoMFA study on
coumarins19 were observed in the modeling of MAO-B and
MAO-A inhibition leading to PLS models I-VII and VIII-
XIV listed in Tables 2 and 3, respectively. Both MAO-B and
MAO-A affinities seem to be modulated by steric, lipophilic,
and electrostatic effects rather than by the electrostatic and
lipophilic ones, respectively, as found in our previous analysis.19
This apparent contradiction can be ascribed to the different data
sets used in the two 3D-QSAR studies. Indeed, while in our
previous CoMFA study most of the structural variations have
been made on the 7-benzyloxy group, in the current set of
inhibitors, more systematic variations in the 7-bridge were
Given that the main purpose of our research was the detection
of the molecular determinants of B/A selectivity, particular
Figure 3. CoMFA-GOLPE (MAO-B affinity) (a) steric, (b) electrostatic, and (c) lipophilic isocontour maps. (a) Contour levels are -0.0006 (red)
and 0.00035 (green). Inhibitors 9 (pIC50 ) 8.36), (+)-31 (pIC50 ) 7.55), (-)-31 (pIC50 ) 6.52), 29 (pIC50 ) 5.49), 27 (pIC50 ) 4.50), and 24 (pIC50
) 4.50) are shown to help interpretation. (b) Contour levels are 0.0018 (gray) and -0.00075 (magenta). Inhibitors 39 (pIC50 ) 8.55), 30 (pIC50
)
7.74), 21 (pIC50 ) 7.40), 36 (pIC50 ) 6.06), 12 (pIC50 ) 5.89), and 15 (pIC50 ) 4.26) are shown to help interpretation. (c) Contour levels are
0.0011 (yellow) and -0.0008 (cyan). Inhibitors 9 (pIC50 ) 8.36), 34 (pIC50 ) 7.28), 28 (pIC50 ) 6.49), 42 (pIC50 ) 8.52), 29 (pIC50 ) 5.49), and
27 (pIC50 ) 4.50) are shown to help interpretation.

Inhibitory Potency and SelectiVity of Coumarins
Journal of Medicinal Chemistry, 2006, Vol. 49, No. 16 4917
Figure 4. CoMFA-GOLPE (MAO-A affinity) (a) steric, (b) electrostatic, and (c) lipophilic isocontour maps. (a) Contour levels are -0.00055 (red)
and 0.0009 (green). MAO-A inhibitors 24 (pIC50 ) 7.33), 26 (pIC50 ) 7.15), 16 (pIC50 ) 5.02), 22 (pIC50 ) 4.73), 29 (pIC50 ) 4.50), and 38 (pIC50
) 4.00) are shown to help interpretation. (b) Contour levels are 0.00099 (gray) and -0.00073 (magenta). Inhibitors 25 (pIC50 ) 7.90), 41 (pIC50
) 6.91), 42 (pIC50 ) 6.17), 36 (pIC50 ) 5.55), and 21 (pIC50 ) 5.00) are shown to help interpretation. (c) Contour levels are -0.0037 (yellow) and
0.0001 (cyan). Inhibitors 24 (pIC50 ) 7.33), 26 (pIC50 ) 7.15), 22 (pIC50 ) 4.73), and 29 (pIC50 ) 4.50) are shown to help interpretation.
Table 4. Statistics of the CoMFA-GOLPE Models of B/A Selectivity (n
) 38)
model
field type
Ste (S-B/A)
Ele (E-B/A)
q2 a
ONCb
r2 c
SDEPd
XV
XVI
0.751
0.897
0.829
0.883
0.832
0.908
3e
2
2
2
2
2
2
0.867 0.840
0.941 0.539
0.881 0.696
0.932 0.575
0.834 0.690
0.944 0.510
0.926 0.560
XVII Lipo (L-B/A)
XVIII Ele + Ste (ES-B/A)
XIX
XX
Lipo + Ste (LS-B/A)
Ele + Lipo (EL-B/A)
Ste + Ele + Lipo (SEL-B/A) 0.889
XXI
a Leave-one-out squared cross-validated correlation coefficient. b Optimal
number of PLS components. c Squared correlation coefficient. d Standard
deviation of error of predictions. e Two-component statistics of model XV:
q2 ) 0.665; r2 ) 0.797; SDEP ) 0.974.
attention was devoted to the analysis of selectivity data. A closer
look at the square plot shown in Figure 1 provided a clear view
of the selectivity associated with each compound and allowed
easy detection of the most potent and highly selective MAO-A
and MAO-B inhibitors. However, additional and more important
insights on SSRs were gained from a CoMFA study of
selectivity data, expressed as ∆pIC50. The statistics of the
CoMFA-GOLPE models reported in Table 4 indicated that B/A
selectivity was governed mainly by electrostatic effects, whereas
lipophilic and, to a major extent, steric effects played a less
important role. The one-field electrostatic model XVI presented
the best statistics. Its very good fitting power (r2 ) 0.941) can
be easily appreciated from a plot of calculated versus experi-
mental ∆pIC50 selectivity values reported in Figure 5.
Isocontour maps, calculated from model XVI and displayed
in Figure 6, helped to interpret and locate at the 3D level the
electrostatic interactions modulating B/A selectivity. The ab-
sence of signals close to position 3 and/or position 4 of the
coumarin ring seemed to indicate that methyl groups in those
regions did not play a relevant role in modulating B/A
selectivity. This result apparently contrasted with the signifi-
cantly higher B/A selectivity observed for 7-benzyloxy-3-
methylcoumarin (8) compared with 7-benzyloxy-4-methyl (7),
7-benzyloxy-3,4-dimethyl (9), and the 7-benzyloxy-3,4-unsub-
stituted (6) derivatives. The lack of a definite signal in position
3 may be due to the fact that there was only one 3-methyl
derivative in the data set and that only very strong signals,
coming from substantial changes in ∆pIC50 selectivity values,
were selected and represented in the isocontour maps. The close
Figure 5. Scatter plot of the experimental vs calculated B/A selectivity
values (∆pIC50) (from model XVI, Table 4). The bisector represents
the case of perfect correlation (r2 ) 1).
Figure 6. CoMFA-GOLPE (B/A selectivity) electrostatic isocontour
maps. Contour levels are 0.002 (gray) and -0.00083 (magenta).
MAO-B selective inhibitors 38 (∆pIC50 ) 3.24), 34 (∆pIC50 ) 2.64),
42 (∆pIC50 ) 2.35), 41 (∆pIC50 ) 2.03), 11 (∆pIC50 ) 1.29), and 29
(∆pIC50 ) 0.99) are shown along with the MAO-A selective inhibitor
23 (∆pIC50 ) -1.84) to help interpretation.
inspection of Figure 6 allowed a straightforward interpretation
of the influence of substituent electronic properties on selectivity.
The different electron density localized on the R and â positions
of the bridge connecting the coumarin ring to the phenyl ring
seemed to be responsible for the observed B/A selectivity.
Indeed, strongly electronegative atoms (i.e., oxygen and sulfur)
in the R position may favorably contact the magenta region and
therefore improve B/A selectivity (i.e., 8). Close contacts in
the same zone may also furnish a sound explanation for the
significant fall of B/A selectivity caused by the inversion of

4918 Journal of Medicinal Chemistry, 2006, Vol. 49, No. 16
Catto et al.
Table 5. Active Site Residues of Both Human MAO Isoenzymes with
Divergent Residues in Italic Font and with Differing Residues in Rat in
Parentheses
and methods adopted by GOLD for energy calculations,
conformational search and clustering, and energy ranking is
briefly presented in the Experimental Section, whereas a fully
detailed description may be found elsewhere.36 An in-depth look
at the conformer population of inhibitor 9 generated during the
docking simulations into the rMAO-B active site revealed that
a convergent binding mode was largely adopted. In the majority
of the docking poses the coumarin ring was located in front of
the FAD moiety with the coumarin carbonyl facing Y435 (data
not shown). A hierarchical cluster analysis, carried out on the
best bound conformers generated during the docking runs,
revealed that all the molecules could be collected in one
prevalent geometric group within an rms value of 1.96Å and a
fitness score37 of 52.28 kJ/mol for the best scored (most stable)
complex. Additional docking studies were subsequently carried
out to more deeply analyze key regions of the binding pocket,
in particular those near the FAD cofactor, to find other
energetically viable binding modes. For this purpose, the
coumarin ring of the best bound conformer was turned 180°
around and a number of physical constraints (see Experimental
Section) were adopted before starting the second series of
docking runs. The resulting docking poses, showing a differently
oriented coumarin ring but similarly bound 7-benzyloxy group,
maintained high fitness scores (in the range 46.6-49.9 kJ/mol).
The small energy differences among the fitness scores from
constrained and unconstrained docking (<5 kJ/mol) suggested
that the binding interactions were most likely driven by the
benzyloxy group, which acted as the molecular anchor of the
inhibitor, whereas the coumarin ring might assume two different
binding topologies (data not shown). Surprisingly, among the
best scored docking poses, GOLD finds no hydrogen bond (HB)
between the oxygen atom of the benzyloxy group and an
eventual HB donating group in the enzymatic cleft. This result
was unexpected because the formation of a HB might have
explained the selective inhibition of MAO-B by 7-benzyloxy-
coumarin derivatives and the decrease of affinity observed when
the benzyloxy oxygen atom was substituted by unsaturated and
saturated carbon and sulfur atoms (compare the affinities of 9
vs 17, 18, and 21). A careful visual analysis of the docking
pose of 9 revealed that its ether oxygen was indeed relatively
close to the phenolic group of Y326 and, therefore, potentially
able to form a HB. This observation prompted us to run further
docking simulations by increasing the sensitivity of GOLD to
detect such a HB (see Experimental Section). Under these
experimental conditions, HB was indeed found in several viable
docking poses featuring an energetically acceptable fitness score
(nearly 47-48 kJ/mol), irrespective of the orientation of the
coumarin ring. The high MAO-B affinity and B/A selectivity
of 7-benzyloxycoumarins might therefore be plausibly accounted
for by the formation of this HB. Nevertheless, the formation of
an alternative network of HBs linking benzyloxy oxygen to
suitably positioned HB donating residues of the enzyme (i.e.,
T201, I199, and Q206) through a bridge consisting of a structural
water molecule could not be ruled out. For this reason and with
the aim to verify whether even additional water molecules might
play a role in ligand binding to rMAO-B, a molecular mapping
of the enzyme binding site was undertaken. A reasonable
detection and 3D location of structural water molecules in
proteins can be efficiently performed by means of the software
GRID38 (see Experimental Section), which calculates the most
favorable interaction energies of a selected chemical probe
positioned at each node of a properly sized grid surrounding
the protein. The binding site of rMAO-B was subjected to GRID,
and an arbitrary cutoff value of -9 kcal/mol was chosen to
hMAO-B
P102
Y188
I198
I199
Q206
T314
M315
I316 (V316)
hMAO-A
hMAO-B
hMAO-A
A111
Y197
I207
F208
Q215
C323
M324
I325
Y326
T327
M341
F343
Y398
Y435
M436
I335
T336
M350
F352
Y407
Y444
M445
the oxymethylene bridge (6 vs 10; ∆pIC50 ) 2.09 vs 0.66).
Electronegative moieties in the gray region near the â position
accounted for the inverted, or strongly decreased, B/A selectivity
observed for highly MAO-A selective sulfonates 23-26 and
phenoxy derivative 29. Magenta regions on the left-hand side
of Figure 6 can be reached by phenyl rings of the long MAO-B
selective inhibitors 32-34 and 37-38 and by the halogen atoms
of inhibitors 39-42. The high B/A selectivity of cinnamyl
derivatives 37 and 38 may also be due to the increased electron
density of their alkenyl double bonds placed on the γ position
(magenta region) of the bridge connecting coumarin to the
phenyl ring.
Docking Studies. To help interpretation of SAFIRs and to
increase our understanding of the main binding interactions at
the MAO active sites, a docking study was performed on some
selective inhibitors. Since inhibition data referred to MAOs from
rat brain mitochondria, we first developed a 3D homology model
of rMAO-B, for which no X-ray crystallographic data are
available. Fortunately, this was not such a difficult task because
rat and human enzymes presented highly similar amino acid
sequences (88.6% sequence identity; 519 and 520 amino acids
for rat and human enzyme, respectively) and almost identical
amino acid residues in the binding sites where the only
difference resulted from the substitution of I316 (human) with
V316 (rat). A number of active site residues of the two human
MAO isoenzymes are reported in Table 5 to help the reader to
identify the exact correspondence with the rat isoforms and to
safely compare human and rat X-ray and docking models.
The 3D model of rMAO-B was constructed through com-
parative protein structure modeling within the Modeller 8.1
program29 after sequence alignment performed with Clustal_X30
(see Experimental Section).
Our docking investigation focused on inhibitor 9, which
showed the highest MAO-B affinity (pIC50 ) 8.36) along with
a remarkable B/A selectivity (∆pIC50 ) 2.20). The GOLD 2.2
program was used to carry out docking simulations,31 since in
several studies it yielded better performances compared to other
similar programs.32-35 It must be noted that a recent exhaustive
study of a large number of enzyme-inhibitor complexes,35
whose X-ray crystallographic structures were retrieved from the
PDB, showed that for the most applied docking programs,
including GOLD, a limited correspondence between experi-
mental and docking poses and, moreover, the various scoring
functions used in the tested programs were not able to calculate
and rank correctly the binding energies of many experimental
complexes. The lesson to be learned from such a study is that
results of docking simulations have to be taken into account
with much caution because different binding modes may result
even from structurally close ligands and, furthermore, the same
ligand may experience multiple binding poses within a relatively
close energy window.
Keeping in mind the above limitations, we conducted the
docking simulations with GOLD 2.2. An analysis of principles

Inhibitory Potency and SelectiVity of Coumarins
Journal of Medicinal Chemistry, 2006, Vol. 49, No. 16 4919
Figure 7. Docking pose of inhibitor 9 into rMAO-B binding site.
Amino acid residues and FAD cofactor are represented in stick form,
colored according to the atom code (C atoms in cyan and yellow for
amino acid residues and cofactor, respectively); structural water
molecules are represented as red balls. Differing residues with MAO-
A, which are P102/A111, I199/F208, T314/C323, and Y326/I335, are
highlighted as white sticks. Yellow dashed lines are drawn among
electronegative atoms likely involved in HB interactions.
Figure 8. Alternative docking pose of inhibitor 9 into rMAO-B
showing a HB between the oxygen atom of the ether bridge and Y326
(see text). Color codes are as in Figure 7.
Finally, a further docking run was carried out, again tuning
GOLD to check for a HB between the oxygen of the ether bridge
of 9 and the hydroxyl group of Y326, analogous to what was
previously done on the docking runs without water molecules.
Again, the most favorable docking poses showed the expected
HB at the expense of a limited loss of fitness score (nearly 5
kJ/mol). In this binding mode, the coumarin ring, as shown
Figure 8, lost the HB with WAT160 and was oriented differently
compared to that shown in Figure 7.
Interestingly, all the proposed binding modes of 9 are in full
agreement with the increased MAO-B affinity observed for
m-halogenophenyl derivatives whose halogens interacted with
a hydrophobic region delimited by I164, L167, F103, and W119.
In summary, our extensive docking simulations, conducted
with and without potential structural water molecules, suggested
the formation of a key HB between the benzyloxy oxygen and
the phenolic group of Y326, and this provided a significant clue
for the interpretation of the MAO-B selective inhibition and
MAO-B affinity modulation observed in 7-substituted cou-
marins.
An additional docking study was performed on the highly
potent and selective MAO-A inhibitor 25 by applying the same
methodology described for MAO-B selective inhibitor 9. Given
that water molecules had been deleted from the crystal structure
of rMAO-A,11 GRID mapping with water as a probe was not
performed because we judged the identification of reliable hot
spots for truly structural water molecules too risky and arbitrary.
Phenyl sulfonate 25 was docked into the binding site of
rMAO-A whose 3D structure was retrieved from the PDB
(coded as 1O5W). No significant insights about high MAO-A
affinity and B/A selectivity of phenyl sulfonate inhibitors were
gained from such a docking simulation. These disappointing
results prompted us to apply alternative modeling approaches.
Keeping in mind that GOLD explores a full conformational
flexibility for the ligand but not for the proteins, we first
undertook a molecular dynamics (MD) study on both MAOs
to detect eventual differences in the conformational mobility
of the amino acids in the two binding sites. Preliminary MD
data40 indeed suggested a much higher conformational flexibility
of Q215 in MAO-A than the corresponding Q206 in MAO-B.
detect the most favorable hot spots for a water molecule used
as chemical probe. Four water molecules, labeled as WAT23,
WAT82, WAT102, and WAT160 according to the numbering
reported in the hMAO-B crystallographic structure10 (coded as
1OJC in PDB), were intercepted in proximity of the FAD
cofactor. Two of them, WAT102 and WAT23, have already
been detected in the crystallographic analysis of hMAO-B,10
while a third, WAT160, was found below the FAD pyrimidine
ring, embedded in the π-system of the aromatic side chains of
Y398 and Y435.39 The fourth water molecule, WAT82, was
located about 3.70 Å apart from the R carbons of I198 and G205.
A docking simulation of inhibitor 9 into the rMAO-B binding
site containing the four water molecules generated two most
energetically favored geometric clusters, consisting of four and
six conformers, corresponding to two basic docking poses that
differed mainly in the topology of the coumarin ring. In the
docking pose represented in Figure 7, the coumarin ring adopted
the same topology, found previously in the unconstrained
docking, with the carbonyl oxygen of coumarin engaged in a
HB with WAT160. From an energetic point of view, only a
negligible variation of the fitness score (52.04 vs 52.28 kJ/mol)
resulted from the docking run with and without water molecules.
In the second alternative and isoenergetic binding mode (52.03
kJ/mol), the coumarin ring was turned 180°, as found previously
in the constrained docking simulation. Again, it must be stressed
that the expected HB between the benzyloxy oxygen and a
relatively close HB donating residue, notably Y326, was not
detected by GOLD among the top-ranking solutions. Therefore,
an additional investigation was conducted with GRID by first
sampling with the water probe the docked complexes developed
by including the four water molecules. The most favorable
interactions were located in proximity of the FAD cofactor, and
for much lower cutoff values only (-5.5 vs -9.0 kcal/mol),
the regions surrounding the ether bridge of 9 were sparsely
intercepted (data not shown).

4920 Journal of Medicinal Chemistry, 2006, Vol. 49, No. 16
Catto et al.
Figure 9. Docking pose of inhibitor 25 into rMAO-A active site. Color
codes are as in Figure 7. The binding conformation on display showed
the coumarin ring facing the FAD isoalloxazine ring.
Figure 10. Alternative docking pose of inhibitor 25 into rMAO-A
active site. Color codes are as in Figure 7. The binding conformation
on display showed the coumarin ring close to the cavity entrance.
of MAOIs. However, its potential role in the binding of MAO-
A-selective phenyl sulfonate inhibitors deserves further inves-
tigation.
Therefore, docking simulations on rMAO-A were repeated
starting from a number of energetically viable conformations
of the Q215 side chain. Results from GOLD runs suggested
multiple docking poses characterized by quite diverse binding
modes but comparable fitness scores. Analysis of top-ranking
docking solutions indicated two opposite binding modes as
found for rMAO-B: one with the coumarin and the other with
the p-nitrophenyl ring facing the FAD isoalloxazine ring system.
The two binding poses most consistent with our SAFIR and
SSR models are reported in Figures 9 and 10. They featured
comparable fitness scores and a similar extended conformation
of the Q215 side chain. An in-depth analysis of the pivotal
interactions stabilizing the enzyme-inhibitor complex reported
in Figure 9 suggested that the higher MAO-A affinity and
selectivity of phenyl sulfonates might be ascribed to the likely
formation of a network of strong HBs linking the oxygen atoms
of the sulfonate moiety to hydrogen atom(s) of the NH2 group
of the Q215 side chain. This hypothesis, supported by our MD
simulations (data not shown), was based on the assumption that
an optimal HB distance exists between the proximal oxygen
atom of SO2 of inhibitor 25 and the NH2 group of Q215, which
was also engaged in a HB with the oxygen atom of the sulfonate
group attached directly to the coumarin ring. In addition, GOLD
detected further HBs established by the nitro group of the
p-nitrophenyl moiety of 25 with T336 and M324 and by the
lactonic oxygen of the coumarin ring with the phenolic group
of Y407. Notably, a closer view of the docked complex revealed
that 25 might interact also with C323, whose SH group seemed
to be placed at a proper distance to make a HB with an oxygen
atom of the nitro group. This may provide an additional
explanation for the increased MAO-A affinity and B/A selectiv-
ity shown by 25 compared to 23. As a matter of fact, this HB
was not detected by GOLD. Interestingly, in a slightly different
binding conformation, GOLD indeed detected a close contact
(van der Waals-like interaction) between C323 and the phenyl
ring of 25. Such an interaction has also been described by
Mattevi for the binding of clorgyline to h-MAO-A.12 Since C323
is one of the few amino acids that differentiate rMAO-A from
rMAO-B, its interaction with inhibitors, irrespective of its nature,
might in principle provide MAO-A selectivity to some classes
In an alternative binding mode of 25 (Figure 10), a different
network of HBs may be seen. It involves the formation of a
number of HBs among the sulfonate group, the two hydrogen
atoms of the NH2 group of Q215, and the phenolic group of
Y407. Moreover, GOLD showed that the nitro group was again
engaged in a HB but with Y197. It is worth noting that in both
docked complexes reported in Figures 9 and 10 two different
oxygen atoms of the sulfonate group seem to be involved in
the formation of HBs: one from the sulfonyl group and the
other from the coumaryloxy moiety (SO2O-coum), in agreement
with the higher MAO-A affinity observed for sulfonates 23-
26 compared with other sulfonyl-containing derivatives such
as 14 and 36. Both docking models also explained the significant
enhancement of affinity and MAO-A selectivity determined by
the introduction of the nitro group in the para position of the
phenyl ring of inhibitor 23. Two plausible and favorable
interactions could take place: (i) the formation of HBs with
HB donating groups in the enzyme; (ii) a π-π stacking
interaction with F208 or FAD. Our docking models also allowed
us to also interpret the low MAO-A affinity observed for the
long alkoxy (i.e., 33 and 34) and cinnamyl (i.e., 37 and 38)
derivatives, which with their phenyl rings might impact a zone
occupied by I325 and L337, and for inhibitor 29 whose phenoxy
ring could collide with the side chain of I335.
Comparison among CoMFA and Docking Models. The
overlay of the CoMFA isocontour maps on the corresponding
MAO-A and MAO-B active sites, limited to the steric maps
for easier interpretation, lent further support to the hypothesized
docking models. Figure 11 showed the steric maps, already
reported in Figure 4a, embedded into the MAO-B binding site.
Both the binding modes reported in Figures 7 and 8 might be
compatible with the imported CoMFA steric maps. However,
for the sake of simplicity, only that derived from Figure 8 was
displayed in Figure 11 (without water molecules). Amino acid
residues likely involved in steric clashes with low active
inhibitors (i.e., 29 and phenyl sulfonates 23-26) were indicated
in red, whereas in green were represented apolar side chains of

Inhibitory Potency and SelectiVity of Coumarins
Journal of Medicinal Chemistry, 2006, Vol. 49, No. 16 4921
Docking simulations of compound 43 into the active site of
rMAO-B showed almost identical binding modes and similar
fitness values compared to lead derivative 9. The fitness values
for the docking poses of compound 43 in rMAO-A indicated
less stable enzyme complexes, likely suggesting a higher
selectivity of the newly designed inhibitor 43 compared to
inhibitor 9. In line with our expectations, the evaluation of the
inhibitory affinity of compound 43 resulted in a very high
MAO-B affinity (pIC50 ) 8.29), a very low MAO-A affinity
(pIC50 ) 4.90), and as a consequence an outstanding B/A
selectivity (∆pIC50 ) 3.39). This selectivity value, even better
than predicted, was the highest within the entire series of
inhibitors examined herein.
Figure 11. Active binding site of rMAO-B showing the steric CoMFA
isocontour maps imported from Figure 3a. Inhibitors 40 (pIC50 ) 8.48),
9 (pIC50 ) 8.36), 29 (pIC50 ) 5.49), 24 (pIC50 ) 4.50), and 27 (pIC50
) 4.50) are shown to help interpretation.
Conclusion and Prospects
In this investigation we reported 3D-QSAR and docking
studies of a new series of substituted coumarin inhibitors
designed to explore in a systematic manner the substituent
effects at position 7. On the basis of previous19,20 and herein
reported results, the major structural determinants of rMAO
affinity and B/A selectivity were established for this biologically
important and easily accessible class of natural compounds.41
We demonstrated that rMAO affinity and selectivity can be
efficiently modulated by appropriate modifications of length,
size, and chemical nature of the substituents at position 7 and
by introducing methyl groups in positions 3 and 4. Our findings
pointed out a crucial, undisclosed role of the methyl group in
position 3 to obtain highly potent and selective MAO-B
inhibitors. This hypothesis was experimentally confirmed with
the design and biological evaluation of a new 3-methyl-
substituted coumarin (compound 43), which proved to be a
highly potent and the most selective MAO-B inhibitor within
the entire series examined in this work. The systematic structural
variation of the sulfur-containing linker at position 7 revealed
the pivotal role of the sulfonate group, directly linked to the
coumarin ring, for high MAO-A affinity and selectivity.
Figure 12. Active binding site of rMAO-A showing the steric CoMFA
isocontour maps imported from Figure 4a. MAO-A inhibitors 24 (pIC50
) 7.33), 16 (pIC50 ) 5.02), 22 (pIC50 ) 4.73), 34 (pIC50 ) 4.64), 29
(pIC50 ) 4.50), and 38 (pIC50 ) 4.00) are shown to help interpretation.
Chart 3. Chemical Structure of the Most B/A Selective
Inhibitor
CoMFA-GOLPE studies enabled the identification of the
physicochemical nature and spatial location of the main binding
interactions modulating the enzymatic affinity and selectivity.
Convergent results from CoMFA models and docking simula-
tions indicated that MAO affinities depended on electronic,
steric, and lipophilic characteristics of the 7-substituted cou-
marins and confirmed that the benzyloxy and sulfonate groups
presented the best structural profiles for optimal interactions
with MAO-B and MAO-A binding sites. Indeed, the docking
poses resulting from GOLD disclosed favorable interactions of
the benzyloxy group, which took place in a smaller cavity of
the MAO-B binding site, relatively far from the FAD cofactor,
whereas the coumarin ring binds in the major lipophilic cage
delimited by FAD, Y326, Y398, and Y435 of rMAO-B,
assuming two possible different conformations. Appropriate
docking simulations suggested that the ether oxygen of the
benzyloxy moiety may form a HB with the phenolic group of
Y326, and this may account for both the high MAO-B affinity
and B/A selectivity of 7-benzyloxycoumarins. The same interac-
tion cannot take place in the rMAO-A binding site, which
presents an isoleucine (I335) in place of a tyrosine (Y326).
F103, I164, W119, and L167, which constituted the lipophilic
region surrounding the meta and, to a lesser extent, para halogen
atoms of highly active inhibitors (i.e., 39-42). Figure 12 referred
to the steric map, already shown in Figure 4a, embedded in the
MAO-A binding site. Between the two binding topologies
suggested by docking studies, the one reported in Figure 9
seemed more compatible with the observed CoMFA steric maps.
Again, in red were highlighted the amino acid residues generat-
ing steric hindrance to low active MAO-A inhibitors such as
16, 22, 29, 34, and 38, whose aromatic rings might collide with
side chains of I325, I335, and L337. The low active MAO-A
inhibitor 22 might also affect the Q215 side chain with its
sulfonyl group.
Rational Design of a New Potent and MAO-B Selective
Inhibitor. To challenge the external predictive ability of our
models and to prepare an inhibitor with high MAO-B affinity
and improved B/A selectivity, we designed a novel coumarin
derivative on the basis of our CoMFA and docking models
developed herein and on a QSAR model derived some years
ago for meta-substituted benzyloxycoumarins.19 7-(m-Chlo-
robenzyloxy)-3-methylcoumarin (43, Chart 3), whose structure
is reported below, emerged as a simple and potentially strong
and selective MAO-B inhibitor.
As far as the MAO-A selectivity is concerned, the docking
model of the highly selective MAO-A inhibitor 25 in Figure 9
suggested that the higher affinity of the phenyl sulfonate
inhibitors for MAO-A might result from the formation of a
network of HBs linking the oxygen atoms of the sulfonate
moiety to the HB donating side chain of Q215 and the oxygen
atom within the coumarin ring to phenolic group of Y407.

4922 Journal of Medicinal Chemistry, 2006, Vol. 49, No. 16
Catto et al.
2.18 (3H, d, J ) 1.4 Hz, CH3); IR (cm-1) 1718, 1622, 1254. Anal.
(C17H13ClO3) C, H.
Further HBs and/or a π-π stacking interaction with F208 could
be established by the p-nitrophenyl moiety of 25. An alternative
binding mode based on an opposite ligand binding topology
and a different HB network (Figure 10) cannot be ruled out. It
would be interesting to cocrystallize inhibitor 25 with rMAO-A
to lend support to our binding hypotheses.
7-Benzenesulfonyloxy-2H-chromen-2-one (15). An amount of
1.0 mmol of 7-hydroxycoumarin 1 (162 mg) was dissolved in 5
mL of anhydrous pyridine, and an amount of 3.0 mmol (530 mg,
383 µL) of benzenesulfonyl chloride was added. The mixture was
refluxed for 30 min, then poured into ice and acidified with diluted
HCl, and the precipitate collected: 98% yield; mp 134-135 °C,
In conclusion, the present study confirmed and reinforced
the validity of the interaction models proposed by our group
on MAO-B inhibitors19,20 and deepened our understanding of
the structural requirements for high MAO affinity and selectiv-
ity. The easy synthetic accessibility and potentially low toxicity
of coumarins make them “privileged scaffolds” for preparing
new inhibitors with improved in vitro affinity. However, it could
be expected that the extreme lipophilic nature of 7-benzyloxy-
3,4-substituted coumarin derivatives might reduce their in vivo
potency as a result of a limited water solubility and poor
bioavailability. Interestingly, the discovery of high MAO-B
affinity and selectivity of 3-methylcoumarins might in part
alleviate this problem, since the overall molecular lipophilicity
(log P) is lowered at least by 0.5 log unit compared with the
corresponding 3,4-dimethyl congeners. Finally, to obtain inhibi-
tors with definitely improved pharmacokinetic properties, new
and original classes of MAO-A and MAO-B selective inhibitors
have been designed and are currently being developed in our
laboratories.
1
from ethanol; H NMR 7.85 (2H, d, J ) 7.4 Hz, 2H-Ar), 7.70
(1H, t, J ) 7.4 Hz, H-Ar), 7.64 (1H, d, J ) 9.5 Hz, H-4), 7.55
(2H, t, J ) 7.8 Hz, 2H-Ar), 7.42 (1H, d, J ) 8.4 Hz, H-5), 7.03
(1H, dd, J ) 8.4, 2.2 Hz, H-6), 6.86 (1H, d, J ) 2.2 Hz, H-8), 6.39
(1H, d, J ) 9.5 Hz, H-3); IR (cm-1) 1734, 1378, 1193. Anal.
(C15H10O5S) C, H.
Benzoic Acid (2H-2-Oxochromen-7-yl)methyl Ester (16). A
solution of bromomethylcoumarin 4 and benzoic acid (10 mmol
each) and 30 mmol (3.88 g, 5.23 mL) of N,N-diisopropylethylamine
in 20 mL of anhydrous DMF was stirred for 2 h. The mixture was
poured into 20 mL of cold 0.1 N HCl, and the product collected
by filtration was purified by crystallization: 88% yield; mp 116-
1
117 °C, from ethanol; H NMR 8.09 (2H, dd, J ) 7.9, 1.2 Hz,
2H-Ar), 7.71 (1H, d, J ) 9.6 Hz, H-4), 7.60 (1H, t, J ) 7.4 Hz,
H-Ar), 7.51-7.47 (3H, m, H-5 and 2H-Ar), 7.43 (1H, s, H-8),
7.34 (1H, d, J ) 8.0 Hz, H-6), 6.43 (1H, d, J ) 9.6 Hz, H-3), 5.45
(2H, s, CH2); IR (cm-1) 1725, 1626, 1286, 1116. Anal. (C17H12O4)
C, H.
3,4-Dimethyl-7-(2-phenylethyl)-2H-chromen-2-one (17). An
amount of 1.0 mmol (292 mg) of trans-styryl derivative 18,19
dissolved in 5 mL of ethyl acetate, was hydrogenated under
atmospheric pressure for 6 h over activated platinum on carbon
(10%). After filtration and evaporation of the solvent, a solid was
obtained in quantitative yield: mp 119-120 °C, from chloroform/
n-hexane; 1H NMR 7.49 (1H, d, J ) 8.0 Hz, H-5), 7.30-7.06 (7H,
m, H-6, H-8, and 5H-Ar), 3.02-2.93 (4H, m, CH2-CH2), 2.39
(3H, s, CH3-4), 2.21 (3H, s, CH3-3); IR (cm-1) 1704, 1614, 1095.
Anal. (C19H18O3) C, H.
Experimental Section
Chemistry. Starting materials, reagents, and analytical grade
solvents were purchased from Sigma-Aldrich, Inc. Chromatographic
separations were performed on silica gel (15-40 mesh, Merck) by
flash methodology. Melting points (mp) were determined by the
capillary method on a Stuart Scientific SMP3 electrothermal
apparatus and are uncorrected. Elemental analyses were performed
on a Euroea 3000 analyzer for C, H, and N; the results agreed to
within (0.40% of the theoretical values. IR spectra were recorded
using potassium bromide disks on a Perkin-Elmer FT-IR spectro-
photometer; only the most significant IR absorption bands are
reported. 1H NMR spectra were recorded in CDCl3 (unless
otherwise specified) at 300 MHz on a Varian Mercury 300
instrument. Chemical shifts are expressed in δ (ppm) and coupling
constants J in hertz (Hz). The following abbreviations were used:
s (singlet), d (doublet), t (triplet), dd (double doublet), dt (double
triplet), m (multiplet). Signals due to OH protons were located by
deuterium exchange with D2O.
The syntheses of compounds 6, 7, 9-11, 18-20, 23-26, 28,
32, and 39-42 have been described in ref 19. Compounds 12-14
and 30 have been reported in ref 20. Compounds 34,42 35,43 and
3844 were synthesized according to published procedures with slight
modifications; their analytical and spectral data were in agreement
with those reported in the literature.
Synthesis of 3-Methylcoumarin Derivatives 8 and 43. Starting
7-hydroxy-3-methylcoumarin 5 was prepared via a classical Perkin
condensation from 2,4-dihydroxybenzaldehyde and propionic an-
hydride according to Gia et al.45 An amount of 1.0 mmol of 5 (176
mg) was refluxed with 2.0 mmol of either benzyl bromide or
3-chlorobenzyl bromide and anhydrous potassium carbonate (1.0
mmol, 138 mg) in 10 mL of absolute ethanol for 2 h. The warm
solution was filtered and cooled to room temperature to give a pure
crystalline solid.
7-Benzylthio-3,4-dimethyl-2H-chromen-2-one (21). An amount
of 3.0 mmol of 7-aminocoumarin 3 (567 mg) was suspended in 21
mL of 10% HCl at -5 °C. An amount of 3.0 mmol (207 mg) of
sodium nitrite was then added, followed after 10 min by benzyl
mercaptane sodium salt (9.0 mmol, 1.31 g). The mixture was
allowed to warm to room temperature and extracted with chloro-
form. The organic layers, after being washed with water and drying
over Na2SO4, were evaporated to dryness to give a red oil, which
was purified by flash column chromatography with chloroform/
petroleum ether (2:8 v/v) as eluent: 25% yield; mp 120-121 °C,
1
from ethyl ether; H NMR 7.43 (1H, d, J ) 8.4 Hz, H-5), 7.36-
7.21 (5H, m, 5H-Ar), 7.16 (1H, d, J ) 1.8 Hz, H-8), 7.13 (1H,
dd, J ) 8.4, 1.8 Hz, H-6), 4.18 (2H, s, CH2), 2.34 (3H, s, CH3-4),
2.18 (3H, s, CH3-3); IR (cm-1) 1703, 1598, 1081. Anal. (C18H16O2S)
C, H.
7-Benzylsulfonyl-3,4-dimethyl-2H-chromen-2-one (22). An
amount of 0.3 mmol of 21 (89 mg) was dissolved under magnetic
stirring in 2 mL of acetic acid and heated to 50 °C. An amount of
5 mL of a 3% water solution of KMnO4 was added dropwise within
5 h. After the mixture was cooled, the excess oxidant was
decomposed with sodium bisulfite and the solution was diluted with
10 mL of water and extracted with chloroform. The organic layers
were collected, washed with a solution of Na2CO3 and with water,
dried over Na2SO4, and evaporated to dryness, giving a crude
product that was purified by flash column chromatography using
chloroform as eluent: 48% yield; mp 238-240 °C (dec); 1H NMR
7.63 (1H, d, J ) 8.3 Hz, H-5), 7.52 (1H, d, J ) 1.8 Hz, H-8), 7.47
(1H, dd, J ) 8.3, 1.8 Hz, H-6), 7.34-7.22 (3H, m, 3H-Ar), 7.09-
7.07 (2H, m, 2H-Ar), 4.34 (2H, s, CH2), 2.41 (3H, s, CH3-4), 2.25
(3H, s, CH3-3); IR (cm-1) 1706, 1144, 1096. Anal. (C18H16O4S)
C, H.
7-Benzyloxy-3-methyl-2H-chromen-2-one (8). 60% yield; mp
1
129-130 °C; H NMR 7.45-7.30 (7H, m, H-4, H-5, and 5H-
Ar), 6.91-6.88 (2H, m, H-6 and H-8), 5.11 (2H, s, CH2), 2.17
(3H, s, CH3); IR (cm-1) 1706, 1611, 1250. Anal. (C17H14O3) C, H.
7-(3-Chlorobenzyloxy)-3-methyl-2H-chromen-2-one (43). 57%
yield; mp 128-29 °C; 1H NMR 7.44 (2H, d, J ) 5.8 Hz, 2H-Ar),
7.34-7.30 (4H, m, H-4, H-5, and 2H-Ar), 6.89 (1H, dd, J ) 8.5,
2.5 Hz, H-6), 6.85 (1H, d, J ) 2.5 Hz, H-8), 5.08 (2H, s, CH2),
N-(3,4-Dimethyl-2H-2-oxochromen-7-yl)-N,4-dimethylbenzen-
sulfonamide (27). N-(3,4-Dimethyl-2H-2-oxochromen-7-yl)-4-me-
thylbenzensulfonamide19 and sodium hydride (each 0.3 mmol) were
dissolved in 3 mL of anhydrous DMF. An amount of 1.0 mmol of

Inhibitory Potency and SelectiVity of Coumarins
Journal of Medicinal Chemistry, 2006, Vol. 49, No. 16 4923
methyl iodide (142 mg, 62 µL) was added, and the mixture was
heated at 40 °C for 4 h. The resulting suspension was poured into
ice, and the precipitate was collected and crystallized: 83% yield;
mp 148-149 °C, from ethanol; 1H NMR 7.54 (1H, d, J ) 8.7 Hz,
H-5), 7.41-7.39 (2H, m, 2H-Ar), 7.28 (1H, dd, J ) 8.7, 2.2 Hz,
H-6), 7.24-7.21 (2H, m, 2H-Ar), 6.82 (1H, d, J ) 2.2 Hz, H-8),
3.16 (3H, s, N-CH3), 2.40 (3H, s, CH3-Ar), 2.38 (3H, s, CH3-4),
2.27 (3H, s, CH3-3); IR (cm-1) 1708, 1613, 1348, 1168. Anal.
(C19H19NO4S) C, H, N.
(1H, d, J ) 8.5 Hz, H-5), 7.47-7.29 (5H, m, 5H-Ar), 6.93-6.74
(3H, m, H-8, H-6, and Ar-CHdCH), 6.40 (1H, dt, J ) 15,7, 6.3
Hz, Ar-CHdCH), 4.75 (2H, d, J ) 5.8 Hz, CH-CH2), 2.38 (3H,
s, CH3-4), 2.19 (3H, s, CH3-3); IR (cm-1) 1698, 1611, 1174, 1090.
Anal. (C20H18O3) C, H.
Computational Methods. Computational analyses were con-
ducted on a 16-node Linux cluster employing an openMosix
architecture composed by AMD Athlon XP 2400+ and Intel Xeon
2600 cpus. Molecules and models were displayed and manipulated
on a Silicon Graphics O2+ machine.
3,4-Dimethyl-7-phenoxy-2H-chromen-2-one (29). According
to the classical von Pechmann procedure,46 a solution of 3-phe-
noxyphenol (8 mmol, 1.49 g), ethyl 2-methylacetoacetate (20 mmol,
2.88 g, 2.83 mL), and a few drops of concentrated sulfuric acid
was kept at 120 °C for 7 h. The reaction mixture was poured onto
ice and extracted with ethyl acetate. The solvent was dried over
Na2SO4 and eliminated under vacuum, and the residue was purified
by column flash chromatography using chloroform/n-hexane (6:4
v/v) as eluent: 40% yield; mp 180-181 °C, from ethyl ether/n-
CoMFA-GOLPE. Inhibitor molecules were built from the
SYBYL fragment libraries.25 Geometrical optimization and charge
calculation were made by means of a quantum mechanical method
with the PM3 Hamiltonian. Conformational analysis and molecular
overlays were performed according to the procedures and methods
described previously.19,20 Default settings were used in CoMFA,
except for the “drop-electrostatic” option, which was set to “NO”.
The steric and electrostatic interaction energies were calculated at
grid points at a regularly spaced 3D lattice with an sp3 carbon probe
atom having a charge of +1 and a van der Waals radius of 1.52 Å.
The grid size had a resolution of 1.00 Å, and the region dimensions
were defined setting the molecular volume automatic mode in
standard CoMFA. Molecular steric, electrostatic, and lipophilic
interaction fields were imported into GOLPE,28 an advanced statistic
tool for both variable selection and model validation. The matrix
was pretreated by zeroing those data with absolute values smaller
than 0.1 kcal/mol and removing any variable with standard deviation
below 0.1.
For the derivation of 3D QSAR models, PLS analyses were
performed with the “leave-one-out” cross-validation procedure.27
The optimal number of components (ONC) in the PLS models
was considered as the one yielding the smallest standard deviation
of error predictions (SDEP) given that the increase of q2 was not
higher than 5% by adding a further component. For an easy
comparison of the fitting and predictive power of PLS models,
statistics relative to two-component models were also reported in
footnotes of Tables 2 and 4. The Smart Region Definition (SRD)
algorithm was used to group variables (number of seeds ) 1500,
critical distance ) 1.0 Å, collapsing cutoff ) 2.0 Å) and two
Fractional Factorial Design (FFD) variable selection runs then
followed.
Homology Modeling. The 3D model of the rMAO-B (entry
name of AOFB_RAT; primary accession number of P19643 at the
ExPASy proteomics server) was developed starting from the X-ray
structure of the hMAO-B (PDB code: 1OJC). The two allotropic
(homologues) enzymes presented highly close amino acid sequences
(519aa vs 520aa, 88.6% of sequence identity) and almost identical
amino acid residues in the binding site where the only difference
was the substitution of I316 in human by V316 in rat. The spatial
model of rMAO-B was constructed through homology modeling47
within Modeller (version 8.1)29 after performing sequence alignment
with standard options of Clustal_X (version 1.83).30 Among the
10 best solutions derived from Modeller, the one provided with
the lowest value of the Modeller objective scoring function was
selected for the subsequent docking simulations. The stereochemical
quality of this model, as well as the overall residue-by-residue
geometry, was controlled with Procheck (version 3.5.4). When the
resolution was set at 2.0 Å, the Ramachandran plot returned 94.4%
of residues in the core regions that represent the most favorable
combinations of æ-ψ angle values and the remaining 5.6% of
residues in diverse, but still energetically allowed, regions.48
Interestingly, the 3D structure of our model was in perfect
agreement with the model returned by the SWISS-MODEL, the
well-known automated protein homology modeling server, to which
the rMAO-B primary sequence was electronically submitted.49 The
interested reader is referred elsewhere for a deeper overview of all
the above-mentioned procedures.50 As expected, the fitting of the
R carbons of the binding site residues of theoretical (rMAO-B) and
experimental (hMAO-B) models gave a very low deviation (rms
) 0.0056 Å), indicating a substantial conservation of their spatial
position.
1
hexane; H NMR 7.54 (1H, d, J ) 8.8 Hz, H-5), 7.40 (2H, t, J )
7.4 Hz, H-3′ and H-5′), 7.21 (1H, t, J ) 7.4 Hz, H-4′), 7.07 (2H,
d, J ) 7.4 Hz, H-2′ and H-6′), 6.93 (1H, dd, J ) 8.8, 2.5 Hz, H-6),
6.85 (1H, d, J ) 2.5 Hz, H-8), 2.39 (3H, s, CH3-4), 2.20 (3H, s,
CH3-3); IR (cm-1) 1699, 1612, 1085. Anal. (C17H14O3) C, H.
7-(2-Hydroxy-2-phenylethyloxy)-3,4-dimethyl-2H-chromen-
2-one (31). To a solution of 0.75 mmol (234 mg) of phenone 3019
in 15 mL of anhydrous THF, an amount of 38 mg of lithium
aluminum hydride (1.0 mmol) was added in small portions, within
2 h, under magnetic stirring at 40 °C. The suspension was cooled
to room temperature, the excess of hydride cautiously decomposed
with ethyl acetate, the insoluble residue was filtered off, and the
solution was evaporated to dryness to afford a solid product in
1
quantitative yield: mp 129-130 °C, from ethanol; H NMR 7.48
(1H, d, J ) 8.8 Hz, H-5), 7.46-7.31 (5H, m, 5H-Ar), 6.86 (1H,
dd, J ) 8.8, 2.4 Hz, H-6), 6.79 (1H, d, J ) 2.4 Hz, H-8), 5.14 (1H,
d, J ) 8.1 Hz, CH-OH), 4.16-4.04 (2H, m, CH2-O), 2.68 (1H,
d, J ) 2.3 Hz, OH), 2.35 (3H, s, CH3-4), 2.17 (3H, s, CH3-3); IR
(cm-1) 3390, 1695, 1612. Anal. (C19H18O4) C, H.
Optical Resolution of the Racemic Mixture of 31. Pure
enantiomers of the racemic mixture of 31 were obtained by
semipreparative enantioseparation on a Chiralcel OD HPLC column
(250 mm × 4.6 mm i.d.), using n-hexane/ethanol (80:20 v/v) as
eluent (flow rate 1.0 mL/min; 20 °C; UV monitoring at 320 nm).
The first eluted compound (tR ) 19.5 min) corresponded to the
(-)-enantiomer, the second one (tR ) 23.5 min) to the (+)-
enantiomer. Measurement of specific rotation [R]D was made with
a Perkin-Elmer 341 polarimeter in a 1 dm cell at 20 °C at a
concentration of 1 g in 10 mL of CHCl3 and gave values of +30°
and -30° for the two optical antipodes.
Synthesis of 3,4-Dimethylcoumarin Derivatives 33, 36, and
37. Title compounds were prepared from 7-hydroxy-3,4-dimeth-
ylcoumarin (2), the appropriate bromoalkyl derivative, and potas-
sium carbonate (each 10 mmol) in 10 mL of anhydrous DMF by
heating at 120 °C for 1-4 h. After cooling, the mixture was poured
onto ice. Compound 36 was extracted with chloroform and purified
by flash column chromatography using chloroform/ethyl acetate
95:5 (v/v) as eluent. Compounds 33 and 37 were collected as
precipitate and recrystallized.
3,4-Dimethyl-7-(2-phenoxyethoxy)-2H-chromen-2-one (33).
1
98% yield; mp 176-177 °C from ethanol; H NMR 7.51 (1H, d,
J ) 8.8 Hz, H-5), 7.33-6.86 (7H, m, H-6, H-8 and 5H-Ar), 4.39-
4.34 (4H, m, CH2-CH2), 2.37 (3H, s, CH3-4), 2.19 (3H, s, CH3-
3); IR (cm-1) 1698, 1607, 1235, 1083. Anal. (C19H18O4) C, H.
7-(Benzenesulfonylmethoxy)-3,4-dimethyl-2H-chromen-2-
one (36). 48% yield; mp 169-170 °C; 1H NMR (DMSO-d6) 7.91
(2H, d, J ) 7.3 Hz, 2H-Ar), 7.76 (1H, t, J ) 7.3 Hz, H-Ar),
7.68-7.63 (3H, m, H-5 and 2H-Ar), 7.09 (1H, d, J ) 2.5 Hz,
H-8), 6.99 (1H, dd, J ) 8.8, 2.5 Hz, H-6), 5.70 (2H, s, CH2), 2.34
(3H, s, CH3-4), 2.05 (3H, s, CH3-3); IR (cm-1) 1699, 1612, 1144.
Anal. (C18H16O5S) C, H.
7-(3-Phenylprop-2-en-1-yloxy)-3,4-dimethyl-2H-chromen-2-
one (37). 50% yield; mp 142-143 °C from ethanol; 1H NMR 7.51

4924 Journal of Medicinal Chemistry, 2006, Vol. 49, No. 16
Catto et al.
Docking Simulations. GOLD, a genetic algorithm-based soft-
ware,31 was used for a docking study selecting GOLDScore as a
fitness function. GOLDScore is made up of four components that
account for protein-ligand binding energy: protein-ligand hy-
drogen bond energy (external H-bond), protein-ligand van der
Waals energy (external vdw), ligand internal vdw energy (internal
vdw), and ligand torsional strain energy (internal torsion). Param-
eters used in the fitness function (hydrogen bond energies, atom
radii and polarizabilities, torsion potentials, hydrogen bond direc-
tionalities, and so forth) are taken from the GOLD parameter file.
The fitness score is taken as the negative of the sum of the energy
terms, so larger fitness scores indicated better bindings. The fitness
function has been optimized for the prediction of ligand binding
positions rather than the prediction of binding affinities, although
some correlation with the latter can be also found.37 The protein
input file may be the entire protein structure or a part of it
comprising only the residues that are in the region of the ligand
binding site. In the present study, GOLD was allowed to calculate
interaction energies within a sphere of a 12 Å radius centered on
phenolic oxygen atom of Y435 and Y444 in rMAO-B and rMAO-
A, respectively. In some docking runs, as indicated previously, the
following physical constraints were set: (i) between the carbamoyl
nitrogen atom of Q206 (MAO-B) and the oxygen atom at position
1 of the coumarin ring; (ii) between the oxygen atom of Y435
(MAO-B) and the carbon atom at position 4 of the coumarin ring;
(iii) between the â-carbon atom of L167 (MAO-B) and the atom
at position 3 of the phenyl ring; (iv) between the oxygen atom of
the ether bridge of 9 and phenolic group of Y326.
GRID Analysis. GRID collects a number of different chemical
probes for measuring molecular interaction fields.38 In the present
work, a water chemical probe was used to identify favorable
locations of functional water molecules within the homology model
of rMAO-B. A cube of 28 Å side, centered on the Y435 with a
grid size of 0.25 Å (NPLA ) 4), was used to calculate the
interaction energies. A cutoff of -9 kcal/mol (the most negative
energy value was -16.90 kcal/mol) allowed us to intercept four
water molecules corresponding to WAT23, WAT82, WAT102, and
WAT160 in the 1OJC hMAO-B crystal structure. The rMAO-B
including these four water molecules was subsequently used for
an additional docking run. The resulting best docked complex,
subjected to GRID, yielded a value of -17.41 kcal/mol for the
most negative interaction energy. A lower cutoff (-5.5 kcal/mol)
was necessary to locate water molecules in the proximity of the
ether bridge of 9 (see text).
students Leonardo Pisani and Teresa Fabiola Miscioscia for their
contributions to the experimental work.
Supporting Information Available: Elemental analyses of
target compounds. This material is available free of charge via the
Internet at http://pubs.acs.org.
References
(1) Tipton, K. F. Enzymology of monoamine oxidase. Cell Biochem.
Funct. 1986, 4, 79-87.
(2) Geha, R. M.; Rebrin, I.; Chen, K.; Shih, J. C. Substrate and inhibitor
specificities for human monoamine oxidase A and B are influenced
by a single amino acid. J. Biol. Chem. 2001, 276, 9877-9882.
(3) Johnston, J. P. Some observation upon a new inhibitor of monoamine
oxidase in brain tissue. Biochem. Pharmacol. 1968, 17, 1285-1297.
(4) Grimsby, J.; Lan, N. C.; Neve, R.; Chen, K.; Shih, J. C. Tissue
distribution of human monoamine oxidase-A and oxidase-B mes-
senger-Rna. J. Neurochem. 1990, 55, 1166-1169.
(5) Fowler, C. J.; Ross, S. B. Selective inhibitors of monoamine oxidase
A and B: biochemical, pharmacological and clinical properties. Med.
Res. ReV. 1984, 4, 323-358.
(6) Yamada, M.; Yasuhara, H. Clinical pharmacology of MAO inhibi-
tors: Safety and future. Neurotoxicology 2004, 25, 215-221.
(7) Kyburz, E. New developments in the field of MAO inhibitors. Drug
News Perspect. 1990, 3, 592-599.
(8) Drukarch, B.; van Muiswinkel, F. L. Drug treatment of Parkinson’s
disease. Time for phase II. Biochem. Pharmacol. 2000, 59, 1023-
1031.
(9) Riederer, P.; Danielczyk, W.; Grunblatt, E. Monoamine oxidase-B
inhibition in Alzheimer’s disease. Neurotoxicology 2004, 25, 271-
277.
(10) (a) Binda, C.; Newton-Vinson, P.; Hubalek, F.; Edmondson, D. E.;
Mattevi, A. Structure of human monoamine oxidase B, a drug target
for the treatment of neurological disorders. Nat. Struct. Biol. 2002,
9, 22-26. (b) Binda, C.; Li, M.; Hubalek, F.; Restelli, N.; Edmond-
son, D. E.; Mattevi, A. Insights into the mode of inhibition of human
mitochondrial monoamine oxidase B from high-resolution crystal
structures. Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 9750-9755.
(11) Ma, J. C.; Yoshimura, M.; Yamashita, E.; Nakagawa, A.; Ito, A.;
Tsukihara, T. Structure of rat monoamine oxidase A and its specific
recognitions for substrates and inhibitors. J. Mol. Biol. 2004, 338,
103-114.
(12) De Colibus, L.; Li, M.; Binda, C.; Lustig, A.; Edmondson, D. E.;
Mattevi, A. Three-dimensional structure of human monoamine
oxidase A (MAO A): Relation to the structures of rat MAO A and
human MAO B. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 12684-
12689.
(13) Novaroli, L.; Reist, M.; Favre, E.; Carotti, A.; Catto, M.; Carrupt, P.
A. Human recombinant monoamine oxidase B as reliable and efficient
enzyme source for inhibitor screening. Bioorg. Med. Chem. 2005,
13, 6212-6217.
Biological Assays. MAO inhibitory activity of compounds in
Table 1 was assessed using a continuous spectrophotometric assay,51
monitoring the rate of oxidation of the nonselective nonfluorescent
MAO substrate kynuramine to 4-hydroxyquinoline. Briefly, male
Sprague-Dawley rats (200-250 g) were sacrificed by decapitation.
The brains were immediately removed and washed in an ice-cold
isotonic Na2HPO4/KH2PO4 buffer (pH 7.40) containing sucrose.
A crude brain mitochondrial fraction was then prepared by
differential centrifugation52 and stored at -40 °C in an isotonic
Na2HPO4/KH2PO4 buffer (pH 7.4) containing KCl. MAO-A and
MAO-B activities in mitochondrial preparations (1 mg/mL) were
assayed using as selective and irreversible inhibitors (-)-L-deprenyl
(250 nM) and clorgyline (250 nM), respectively. After a preincu-
bation for 5 min with the assayed compound dissolved in DMSO
at a final concentration of 5% (v/v), kynuramine was added at a
concentration equal to the corresponding KM value (90 µM for
MAO-A and 60 µM for MAO-B). Then the rate of formation of
4-hydroxyquinoline was monitored at 314 nm for 5 min. Finally,
IC50 values were determined by nonlinear regression of MAO
inhibition vs -log of the concentration plots, using the program
Origin, version 6.0 (Microcal Software Inc., Northampton, MA).
(14) Geha, R. M.; Chen, K.; Shih, J. C. Phe(208) and Ile(199) in human
monoamine oxidase A and B do not determine substrate and inhibitor
specificities as in rat. J. Neurochem. 2000, 75, 1304-1309.
(15) Novaroli, L. Rational strategy and screening methods to identify
multifunctional hits for the treatment of PD and AD. Ph.D. Thesis,
No. 3643, Universite´ de Gene`ve, Faculte´ des Sciences, 2005.
(16) Altomare, C.; Carrupt, P. A.; Gaillard, P.; El Tayar, N.; Testa, B.;
Carotti, A. Quantitative structure metabolism relationship analyses
of Mao-mediated toxication of 1-methyl-4-phenyl-1,2,3,6-tetrahy-
dropyridine and analogs. Chem. Res. Toxicol. 1992, 5, 366-375.
(17) (a) Kneubuhler, S.; Carta, V.; Altomare, C.; Carotti, A.; Testa, B.
Synthesis and monoamine-oxidase inhibitory activity of 3-substituted
5H-indeno[1,2-c]pyridazines. HelV. Chim. Acta 1993, 76, 1956-1963.
(b) Kneubuhler, S.; Thull, U.; Altomare, C.; Carta, V.; Gaillard, P.;
Carrupt, P. A.; Carotti, A.; Testa, B. Inhibition of monoamine
oxidase-B by 5H-indeno[1,2-c]pyridazines. Biological activities,
quantitative structure-activity-relationships (QSARs) and 3D-
QSARs. J. Med. Chem. 1995, 38, 3874-3883. (c) Altomare, C.;
Cellamare, S.; Summo, L.; Catto, M.; Carotti, A.; Thull, U.; Carrupt,
P. A.; Testa, B.; Stoeckli-Evans, H. Inhibition of monoamine
oxidase-B by condensed pyridazines and pyrimidines: effects of
lipophilicity and structure-activity relationships. J. Med. Chem. 1998,
41, 3812-3820. (d) Thull, U.; Kneubuhler, S.; Gaillard, P.; Carrupt,
P. A.; Testa, B.; Altomare, C.; Carotti, A.; Jenner, P.; Mcnaught, K.
S. P. Inhibition of monoamine-oxidase by isoquinoline derivatives.
Qualitative and 3D-quantitative structure-activity-relationships. Bio-
chem. Pharmacol. 1995, 50, 869-877.
Acknowledgment. The Spanish authors gratefully acknowl-
edge the financial support of the Ministerio de Ciencia y
Tecnologia (Grant BFI2003-00493) and the contribution from
the ERDF (Madrid, Spain). The Italian authors thank MIUR,
Rome, for financial support (PRIN Project 2004) and Ph.D.
(18) Carotti, A.; Carrieri, A.; Chimichi, S.; Boccalini, M.; Cosimelli, B.;
Gnerre, C.; Carotti, A.; Carrupt, P. A.; Testa, B. Natural and synthetic
geiparvarins are strong and selective MAO-B inhibitors. Synthesis
and SAR studies. Bioorg. Med. Chem. Lett. 2002, 12, 3551-3555.

Inhibitory Potency and SelectiVity of Coumarins
Journal of Medicinal Chemistry, 2006, Vol. 49, No. 16 4925
(19) Gnerre, C.; Catto, M.; Leonetti, F.; Weber, P.; Carrupt, P. A.;
Altomare, C.; Carotti, A.; Testa, B. Inhibition of monoamine oxidases
by functionalized coumarin derivatives: Biological activities, QSARs,
and 3D-QSARs. J. Med. Chem. 2000, 43, 4747-4758.
(20) Carotti, A.; Altomare, C.; Catto, M.; Gnerre, C.; Summo, L.; De
Marco, A.; Rose, S.; Jenner, P.; Testa, B. Lipophilicity plays a major
role in modulating monoamine oxidase B (MAO-B) inhibition by
7-substituted coumarins. Chem. BiodiVersity 2006, 3, 134-144.
(21) Bruhlmann, C.; Ooms, F.; Carrupt, P. A.; Testa, B.; Catto, M.;
Leonetti, F.; Altomare, C.; Carotti, A. Coumarin derivatives as dual
inhibitors of acetylcholinesterase and monoamine oxidase. J. Med.
Chem. 2001, 44, 3195-3198.
(37) Verdonk, M. L.; Cole, J. C.; Hartshorn, M. J.; Murray, C. W.; Taylor,
R. D. Improved protein-ligand docking using Gold. Proteins:
Struct., Funct., Bioinf. 2003, 52, 609-623.
(38) Goodford, P. J. A computational procedure for determining energeti-
cally favorable binding sites on biologically important macromol-
ecules. J. Med. Chem. 1985, 28, 849-857.
(39) Reist, M.; et al. Unpublished data.
(40) Carotti, A.; Leonetti, F.; Stefanachi, A.; Catto, M.; Capaldi, C.;
Muncipinto, G.; Pisani, L.; Nicolotti, O. In silico design and focused
solid phase synthesis of new, selective enzyme inhibitors with
potential in cancer and neurological therapies. Sci. Pharm. 2005, 73,
S27.
(41) O’Kennedy, R., Thornes, R. D., Eds. Coumarins. Biology, Applica-
tions and Mode of Action; John Wiley and Sons: Chichester, U.K.,
1997.
(42) Rendenbach, B.; Weifenbach, H. Teschendorf, H.-J. Arylalkoxycu-
marine. Verfahren zu Ihre Herstellung und Diese Enthalende Thera-
peutische Mittel (Arylalkoxycoumarins. Their Production Procedures
and Therapeutics). Patent DE 3834861 A1, 1990 (BASF AG).
(43) Schlecker, R.; Schmidt, P.; Thieme, P. C.; Lenke, D.; Teschendorf,
H. J.; Traut, M.; Mu¨ller, C. D.; Hofmann, J. P.; Kreiskott, H. Neue
Sulfonsa¨ureester von Hydroxycumarinen, Ihre Herstellung und Sie
Enthaltende Arzneimittel (New Sulfonic Acid Esters of Hydroxy-
coumarins: Their Production and Medicinal Properties). Patent DE
3243158 A1, 1984 (BASF AG).
(22) Atkins, R. L.; Bliss, D. E. Substituted coumarins and azacoumarins.
Synthesis and fluorescent properties. J. Org. Chem. 1978, 43, 1975-
1980.
(23) Rendenbach-Mu¨ller, B.; Schlecker, R.; Traut, M.; Weifenbach, H.
Synthesis of coumarins as subtype-selective inhibitors of monoamine-
oxidase. Bioorg. Med. Chem. Lett. 1994, 4, 1195-1198.
(24) Cramer, R. D., III; Patterson, D. E.; Bunce, J. D. Comparative
molecular field analysis (CoMFA). 1. Effect of shape on binding of
steroids to carrier proteins. J. Am. Chem. Soc. 1988, 110, 5959-
5967.
(25) SYBYL, version 7.1; Tripos Inc. (1699 South Hanley Road, St. Louis,
MO 63144).
(26) Gaillard, P.; Carrupt, P. A.; Testa, B.; Boudon, A. Molecular
lipophilicity potential, a tool in 3D QSAR. Methods and applications.
J. Comput.-Aided Mol. Des. 1994, 8, 83-96.
(27) Dunn, W. J., III; Wold, S.; Edlund, U.; Hellberg, S.; Gasteiger, J.
Multivariate structure-activity relationships between data from a
battery of biological tests and an ensemble of structure descriptors:
the PLS method. Quant. Struct.-Act. Relat. 1984, 3, 131-137.
(28) Baroni, M.; Costantino, G.; Cruciani, G.; Riganelli, D.; Valigi, R.;
Clementi, S. Generating optimal linear PLS estimations (GOLPE):
an advanced chemometric tool for handling 3D-QSAR problems.
Quant. Struct.-Act. Relat. 1993, 12, 9-20.
(44) Ahluwalia, V. K.; Kumar, D.; Gupta, Y. K. Claisen rearrangement
of some 7-cinnamyloxybenzo-R-pyrones. Indian J. Chem. 1978, 16B,
579-583.
(45) Gia, O.; Uriarte, E.; Zagotto, G.; Baccichetti, F.; Antonello, C.;
Marcianimagno, S. Synthesis and photobiological activity of new
methylpsoralen derivatives. J. Photochem. Photobiol., B 1992, 14,
95-104.
(46) von Pechmann, H. Duisberg, C. U¨ ber die verbindungen der phenole
mit acetessiga¨ther (On the compounds of phenol with acetoacetate).
Chem. Ber. 1883, 16, 2119-2128.
(47) Marti-Renom, M. A.; Stuart, A. C.; Fiser, A.; Sanchez, R.; Melo,
F.; Sali, A. Comparative protein structure modelling of genes and
genomes. Annu. ReV. Biophys. Biomol. Struct. 2000, 29, 291-325.
(48) Laskowski, R. A.; Macarthur, M. W.; Moss, D. S.; Thornton, J. M.
Procheck. A program to check the stereochemical quality of protein
structures. J. Appl. Crystallogr. 1993, 26, 283-291.
(49) Schwede, T.; Kopp, J.; Guex, N.; Peitsch, M. C. SWISS-MODEL:
an automated protein homology-modelling server. Nucleic Acids Res.
2003, 31, 3381-3385.
(50) Lesk, A. M. Introduction to Protein Science Architecture, Function,
and Genomics; Oxford University Press: Oxford, U.K., 2004.
(51) Mendez-Alvarez, E.; Soto-Otero, R.; Sanchez-Sellero, I.; Lopez-
Rivadulla, L. M.; Lamas, M. Inhibition of brain monoamine oxidase
by adducts of 1,2,3,4-tetrahydroisoquinoline with components of
cigarette smoke. Life Sci. 1997, 60, 1719-1727.
(52) Soto-Otero, R.; Mendez-Alvarez, E.; Hermida-Ameijeiras, A.; Sanchez-
Sellero, I.; Cruz-Landeira, A.; Lamas, M. L. R. Inhibition of brain
monoamine oxidase activity by the generation of hydroxyl radicals.
Potential implications in relation to oxidative stress. Life Sci. 2001,
69, 879-889.
(29) Marti-Renom, M. A.; Stuart, A. C.; Fiser, A.; Sanchez, R.; Melo,
F.; Sali, A. Comparative protein structure modelling of genes and
genomes. Annu. ReV. Biophys. Biomol. Struct. 2000, 29, 291-325.
(30) Thompson, J. D.; Gibson, T. J.; Plewniak, F.; Jeanmougin, F.;
Higgins, D. G. The CLUSTAL_X windows interface: flexible
strategies for multiple sequence alignment aided by quality analysis
tools. Nucleic Acids Res. 1997, 24, 4876-4882.
(31) Jones, G.; Willett, P.; Glen, R. C.; Leach, A. R.; Taylor, R.
Development and validation of a genetic algorithm for flexible
docking. J. Mol. Biol. 1997, 267, 727-748.
(32) Schulz-Gasch, T.; Stahl, M. Binding site characteristics in structure-
based virtual screening: evaluation of current docking tools. J. Mol.
Model. 2003, 9, 47-57.
(33) Wang, R.; Lu, Y.; Wang, S. Comparative evaluation of 11 scoring
functions for molecular docking. J. Med. Chem. 2003, 46, 2287-
2303.
(34) Kellenberger, E.; Rodrigo, J.; Muller, P.; Rognan, D. Comparative
evaluation of eight docking tools for docking and virtual screening
accuracy. Proteins: Struct., Funct., Bioinf. 2004, 57, 225-242.
(35) Warren, G. L.; Andrews, C. W.; Capelli, A. M.; Clarke, B.; LaLonde,
J.; Lambert, M. H.; Lindvall, M.; Nevins, N.; Semus, S. F.; Senger,
S.; Tedesco, G.; Wall, I. D.; Woolven, J. M.; Peishoff, C. E.; Head,
M. S. A critical assessment of docking programs and scoring
functions. J. Med. Chem., in press.
(36) Westhead, D. R.; Clark, D. E.; Murray, C. W. A comparison of
heuristic search algorithms for molecular docking. J. Comput.-Aided
Mol. Des. 1997, 11, 209-228.
JM060183L