A QSAR model for in silico screening of MAO-A inhibitors. Prediction, synthesis, and biological assay of novel coumarins

Article abstract of DOI:10.1021/jm0509849

This work explores the potential of the MARCH-INSIDE methodology to seek a QSAR for MAO-A inhibitors from a heterogeneous series of compounds. A Markov model was used to quickly calculate the molecular electron delocalization, polarizability, refractivity, and n-octanol/water partition coefficients for a series of 1406 active/nonactive compounds. LDA was subsequently used to fit a classification function. The model showed 92.8% and 91.8% global accuracy and predictability in training and validation studies. This QSAR model was validated through a virtual screening of a series of coumarin derivatives. The 15 selected compounds were prepared and evaluated as in vitro MAO-A inhibitors. The theoretical prediction was compared with the experimental results and the model correctly predicted 13 compounds with only two mistakes on compounds with activities very close to the cutoff point established for the model. Consequently, this method represents a useful tool for the "in silico" screening of MAO-A inhibitors.

Full text of DOI:10.1021/jm0509849

J. Med. Chem. 2006, 49, 1149-1156
1149
A QSAR Model for in Silico Screening of MAO-A Inhibitors. Prediction, Synthesis, and
Biological Assay of Novel Coumarins†
Lourdes Santana,*,‡ Eugenio Uriarte,‡ Humberto Gonza´lez-D´ıaz,‡,§ Giuseppe Zagotto,| Ramo´n Soto-Otero, and
Estefan´ıa Me´ndez-AÄ lvarez
Department of Organic Chemistry, Faculty of Pharmacy, and Department of Biochemistry, Faculty of Medicine, UniVersity of Santiago de
Compostela, 15872 Santiago de Compostela, Spain; Department of Pharmaceutical Sciences, UniVersity of PadoVa, 35131 PadoVa, Italy; and
Department of Drug Design, Chemical Bio-actiVes Center, Central UniVersity of Las Villas, 54830 Santa Clara, Cuba
ReceiVed October 3, 2005
This work explores the potential of the MARCH-INSIDE methodology to seek a QSAR for MAO-A inhibitors
from a heterogeneous series of compounds. A Markov model was used to quickly calculate the molecular
electron delocalization, polarizability, refractivity, and n-octanol/water partition coefficients for a series of
1406 active/nonactive compounds. LDA was subsequently used to fit a classification function. The model
showed 92.8% and 91.8% global accuracy and predictability in training and validation studies. This QSAR
model was validated through a virtual screening of a series of coumarin derivatives. The 15 selected
compounds were prepared and evaluated as in vitro MAO-A inhibitors. The theoretical prediction was
compared with the experimental results and the model correctly predicted 13 compounds with only two
mistakes on compounds with activities very close to the cutoff point established for the model. Consequently,
this method represents a useful tool for the “in silico” screening of MAO-A inhibitors.
Introduction
the 1950s of the antidepressant properties of MAOIs was a major
finding in the monoamine theory of depression. However, earlier
MAOIs introduced into clinical practice were abandoned due
to adverse effects, such as hepatotoxicity, orthostatic hypoten-
sion, and the so-called “cheese effect”, which was characterized
by hypertensive crisis. These drawbacks were thought to be
related to nonselective and irreversible enzyme inhibition.9,10
In recent years, efforts have been focused on the discovery
of reversible and selective MAOIs and this has led to a new
class compounds (third generation). A broad consensus exists
concerning the necessity for a search for novel MAOIs and the
study of the mechanism of action for MAOs.11,12 In any case,
despite considerable progress in understanding the interactions
of the two enzyme forms with their preferred substrates and
inhibitors, few general rules are available for the rational design
of potent and selective inhibitors.
It is clear that MAOs are tightly associated with the outer
mitochondrial membrane; consequently, procedures to yield pure
enzymes make use of detergents that are necessary to prevent
aggregation or precipitation. For this reason, crystallization
procedures are more difficult than for soluble proteins. This fact
explains the research bottleneck in terms of obtaining detailed
3D structures of MAOs. Fortunately, the availability of large
quantities of purified recombinant human MAOIs has provided
suitable crystals for the elucidation of the structure of MAO-
A13 and MAO-B14 isoforms. This in turn allowed accurate
modeling of MAO inhibitors and opened new possibilities for
the design of more selective drugs.
Amine oxidases catalyze the oxidation of amines, diamines,
and polyamines. Depending on their ability to preferentially
recognize one of these substrates, amine oxidases can be divided
into monoamine oxidases, diamine oxidases, and polyamine
oxidases, respectively.1,2 Several different enzymes fall into the
amine oxidase class, and the classification of some of these
systems still remains ambiguous. More specifically, monoamine
oxidase (MAO) was the term introduced by Zeller in an attempt
to distinguish one specific group of enzyme members of the
class responsible for the oxidative deamination of mono-
amines.3,4
MAO exists in two isoforms (MAO-A and MAO-B), and each
isoenzyme has a flavin adenine dinucleotide cofactor covalently
linked to a cystein residue in the active center. Both MAO
subtypes are characterized by specific substrates and inhibitors.
MAO-A has a higher affinity for serotonin and norepinephrine,
whereas MAO-B preferentially deaminates phenylethylamine
and benzylamine. These properties determine the clinical
importance of MAO inhibitors. Selective MAO-A inhibitors are
used in the treatment of neurological disorders such as depres-
sion,5,6 whereas the MAO-B inhibitors are useful in the treatment
of Parkinson’s7 and Alzheimer’s deseases.8
All of these aspects have led to an intensive search for novel
MAO inhibitors (MAOIs), and this effort has increased con-
siderably in recent years. Overall, three generations of MAOIs
have been described. The first generation includes the irrevers-
ible and nonselective inhibitors. The second generation consists
of all the irreversible but selective MAOIs. The discovery in
Moreover, once the 3D crystallographic structure of MAOs
is resolved, the estimation of docking parameters for lead drug
candidates may involve time-consuming operations where large
libraries of chemicals must be investigated.
Consequently, novel approaches are also needed for the
efficient search for new inhibitors, and although a number of
reports exist on the quantitative structure-activity relationships
(QSAR) for MAOIs, these are, in general, restricted to the study
of congeneric families of compounds.15
† In memory of Prof. Cipriano Antonello.
* To whom correspondence should be addressed. Phone: +34-981-
563100. Fax: +34-981 594912. E-mail: qolsant@usc.es.
‡ Department of Organic Chemistry, Faculty of Pharmacy, University
of Santiago de Compostela.
§ Central University of Las Villas.
| Department of Biochemistry, Faculty of Medicine, University of
Santiago de Compostela.
University of Padova.
10.1021/jm0509849 CCC: $33.50 © 2006 American Chemical Society
Published on Web 12/21/2005

1150 Journal of Medicinal Chemistry, 2006, Vol. 49, No. 3
Santana et al.
In this sense, the development of QSARs using simple
molecular indices appears to be a promising alternative or
complementary technique to drug-protein docking, high-
throughput screening, and combinatorial chemistry techniques.
Almost all QSAR techniques are based on the use of molecular
descriptors, which are numerical series that codify useful
chemical information and enable correlations between statistical
and biological properties.16-18
Results and Discussion
Computational Methods. The MARCH-INSIDE approach
(Markovian chemicals in silico design) is based on the calcula-
tion of the different physicochemical molecular properties as
an average of atomic properties (ap). For instance, it is possible
to derive average estimations of molecular or group electro-
negativities [kø(G)], refractivities [kMR(G)], polarizabilities
[kR(G)], and logarithms of water/n-octanol partition coefficients
k
A large number of examples have been published in which
the use of molecular descriptors has become a rational alternative
to massive synthesis and screening of compounds in medicinal
chemistry.19,20 Indeed, experience has shown that the use of
models fitted with large data sets of chemicals works as well
as the use of models built from a series of homologous
compounds and is also a more general method that can be
applied in a broad spectrum of cases. The principal deficiency
in the use of some molecular indices concerns their lack of
physical meaning. In this respect, the introduction of novel
molecular indices must obey physicochemical laws in order to
ensure a theoretically rigorous interpretation of the results.18
(log P).
kø(G) )
p (ø ) ø
j
(1)
(2)
(3)
(4)
∑
k
j
j ∈ G
kMR(G) )
p (MR ) MR
k j j
∑
j ∈ G
kR(G) )
p (R ) R
k j j
∑
j ∈ G
log kP(G) )
p (log P ) log P
k j j
∑
j ∈ G
In the particular case of MAOIs, electron delocalization and
its consequencesssuch as polarizability and hydrophobicitys
may be a determinant physical factor. This issue can be
explained by considering that two models have been proposed
for MAO catalysis and these involve an electron transfer or polar
nucleophilic mechanisms.11,14
It is possible to consider isolated atoms (k ) 0) in the estimation
0
0
0
0
of the molecular properties ø(G), MR(G), R(G), log P. In
this case the probabilities 0p(apj) are determined without
considering the formation of chemical bonds (simple additive
scheme). However, it is possible to consider the gradual effects
of the neighboring atoms at different distances in the molecular
backbone. To achieve this goal, the method uses a MM, which
determines the absolute probabilities pk(apj) with which the
atoms placed at different distances k affect the contribution of
the atom j to the molecular property in question. For example,
in the case of molecular polarizability
Quantum chemical calculations can be used to obtain a priori
descriptors for QSAR studies. Given that some of these quantum
properties are not observable, the best way to calculate them is
not uniquely defined. Consequently, it is likely that there are
many different schemes for such calculations, and none of these
is fundamentally more correct than another. Unfortunately, the
calculations are often also computationally too demanding for
large sets of molecules.21 In an effort to address this difficulty,
Bultinck et al.22 described the implementation of a computational
approach based on the electronegativity equalization principle
(EEP) to allow the very rapid calculation of atomic charges and
related molecular descriptors. According to the EEP described
by Sanderson, when molecules are formed, the electronegativi-
ties and other properties of the constituent atoms become equal,
coinciding with a fixed distribution for probabilities of finding
the electrons in the neighborhood of a specific atom in the
molecule at the steady state.23,24 Simpler and faster methods to
calculate molecular descriptors based on the idea of EEP are of
interest in terms of their application to very large databases of
compounds with the aim of finding druglike leads.
kR ) [p(R1)p0(R2)...p0(Rn)]‚
1p1,1 p1,2 p1,3
‚
1p1,n
1
1
k
R1
R2
‚
‚
Rn
1p2,n
1
1
1p2,1 p2,2 p2,3
‚
‚
‚
‚
n
‚
)
p (R ) R
k j j
‚
‚
‚
‚
‚
‚
‚
‚
‚
‚
∑
j)1
[
] [ ]
1pn,1
1pn,n
k
From left to right, the first term is R, which is the average
molecular polarizability of the molecule considering the effects
of all the atoms placed at distance k over every atomic
polarizability Rj. The vector on the left-hand side of the equation
contains the probabilities 0p(Rj) for every atom in the molecule,
without considering chemical bonds. The matrix in the center
of the equation is the so-called stochastic matrix. The values of
this matrix (1pij) are the probabilities with which every atom
affects the polarizability of the atom bonded to it. Both kinds
of probabilities 0p(Rj) and 1pij are easily calculated from atomic
polarizabilities (Rj) and the chemical bonding information:
Our research group has just introduced a novel series of
stochastic indices in the so-called MARCH-INSIDE approach.
The method is based on the use of Markov models (MM)25 to
codify useful chemical structure information in terms of
molecular electron delocalization, polarizability, refractivity, and
n-octanol/water partition coefficient.26-28
Rj
In this work we will explore the potential of MARCH-
INSIDE to seek a QSAR for MAO-A inhibitors from a
heterogeneous series of compounds. In the first step, the
aforementioned molecular descriptors were calculated for a large
series of active/nonactive compounds. Linear discriminant
analysis (LDA) was subsequently used to fit a classification
function. The QSAR developed was then validated with an
external predicting series by the resubstitution technique. Finally,
the model was used for the prediction of a novel generation of
MAO-A inhibitors, which were subsequently synthesized,
structurally characterized, and experimentally assayed.
0pij )
(6)
n
R
∑
k
k)1
δijRj
1pij )
(7)
n
δ R
∑
ik
k
k)1
The only difference is that in the probabilities 0p(Rj) we consider

QSAR Model for Screening of MAO-A Inhibitors
Journal of Medicinal Chemistry, 2006, Vol. 49, No. 3 1151
Figure 1. Structural sets of known MAO-A inhibitors used in the QSAR study.
isolated atoms by carrying out the sum in the denominator over
all n atoms in the molecule. On the other hand, for 1pij, chemical
bonding is taken into consideration by means of the factor δij.
This factor has a value of 1 if atoms i and j are chemically
bonded and is 0 otherwise.
Finally, it is interesting to note that one can sum only the
atoms included in a specific group (G) rather than all atoms. In
this way, we can approach specific classes of average properties
selection here of discriminant techniques instead of regression
techniques was determined by the lack of homogeneity in the
conditions under which these values were measured. As reported
in different sources, numerous IC50 values lie within a range
rather than a single value. In other cases, the activities are not
scored in terms of IC50 values but are quoted as inhibitory
percentages at a given concentration. Once the training series
had been designed, forward stepwise linear discriminant analysis
(LDA) was carried out in order to derive the QSAR for the
MAO-A inhibitory activity score (i-MAO-A):
such as the average electronegativity for sp3 carbon atoms (Csp3
)
or average polarizability for heteroatoms (Het). All calculations
were performed using the program MARCH-INSIDE,28 which
was developed in-house.
i-MAO-A ) -1.32ø1(Csat) - 2.56ø2(Het) + 0.46R0(Het) +
1.69R1(Het) + 1.16MR1 + 7.12MR2(Cunsat) -
General QSAR for MAO-A Inhibitors. The development
of a discriminant function29 that allows the classification of
organic compounds as active or inactive is the key step in the
present approach for the discovery of MAO-A inhibitors. It was
therefore necessary to select a training data set of MAO-A
inhibitors containing wide structural variability. Here we
consider a general data set of 1406 compounds, 674 of which
have MAO-A inhibitory activity. This group contains chemicals
with IC50 values e25 µM or in some cases compounds that
show an inhibition percentage of g50% at inhibitor concentra-
tions e25 µM. This series is composed at random of the most
representative families of MAO-A inhibitors taken from the
literature (Supporting Information); these include propargyl
(clorgyline analogues) (I), benzamide (moclobemide analogues)
(II), phenylethylamine (III), indole (IV), coumarin (V), thiox-
anthene (VI), oxadiazolidone (toloxatone analogues) (VII), and
diazoheterocyclic derivatives (VIII) (see Figure 1).
The remaining 732 compounds were a heterogeneous series
of inactive compounds, including members of the aforemen-
tioned families, with IC50 > 25 µM, along with many other
structural patterns extracted from international databases.30,31
It is common for compounds with even higher IC50 values to
be considered as active or moderately active, but we considered
25 µM to be a reasonable limit. The selection of higher break
point values to cluster chemicals by their MAO-A IC50 may
generate a series with a clearly disproportionate size and,
therefore, a vastly reduced number of active compounds. The
1.71MR0(Het) - 2.95‚log P0 - 5.6‚log P2(Cunsat) + 1.21
λ ) 0.36 F ) 273.93 p < 0.01
The statistical significance of this model was determined by
examining Wilk’s λ statistic, the Fisher ratio (F), and the p-level
(p). We also inspected the percentage of good classification in
training and validation experiments. LDA produced a classifica-
tion function that gave rise to an efficient separation of 92.8%
of the 1406 chemicals (training series) into two groups. The
model correctly classified 95.1% of 674 MAO-A inhibitors and
90.73% of the 732 inactive compounds. This is equivalent to
predicting the inhibitory probability of 641 MAO-A inhibitors
out of 674 and 664 inactive chemicals out of 732 (see Table
1).
Validation of the model was carried out by the resubstitution
approach. In this kind of validation technique the data set is
first split into one training subset and a predicting subset or
control series. Next, the model is derived again using only the
compounds in the training subset. In the third step, the
compounds in both the training and predicting subsets (not used
to seek the new model) are classified. The compounds in the
training and predicting subsets are then interchanged at random
several times (10 times in this case). Finally, the average
predictability for all of the interchanged series is reported. In
this study the model presented an average validation predict-

1152 Journal of Medicinal Chemistry, 2006, Vol. 49, No. 3
Santana et al.
Scheme 1a
Table 1. Training and Validation Results
MAO-A
percent
inhibitors
nonactive
total
Training Results
MAO-A inhibitors
nonactive
total
95.1
90.7
92.8
641
68
33
664
674
732
1406
Validation Results
MAO-A inhibitors
nonactive
total
93.6
90.1
91.8
631
72
43
660
674
732
1406
a Reagents and conditions: (a) AlCl3/CH2Cl2, rt; (b) Br(CH2)3 Br/K2CO3,
acetone, reflux; (c) potassium phthalimide/DMA, rt then CH3NHNH2/CHCl3,
rt for 2; Et2NH/EtOH, 65 °C for 3.
Figure 2.
Scheme 2a
ability of 91.8%. This result shows that the model not only has
a high predictability but also a high robustness to data variation
(see Table 1). The names, observed classification, predicted
classification, and subsequent probabilities for all 1406 com-
pounds in training and average validation are given as Sup-
porting Information.
The in Silico Evaluation of Novel MAO-A Inhibitors. The
model developed in this study was used for the design of novel
MAO-A inhibitors. As described previously, a specific way to
new lead discovery involves several general common steps when
it is used in terms of a QSAR: (a) construction of a suitable
molecular database of compounds that either show MAO-A
inhibitory activity or do not, (b) calculation of the molecular
descriptors, (c) construction of the model, (d) estimation of the
biological activity using QSAR, (e) synthesis and characteriza-
tion of selected compounds, and (f) assay of the candidate
compounds in order to corroborate the predicted biological
activity.
As mentioned above, the coumarin analogues are a family
of natural and/or synthetic compounds with different pharma-
cological activities,32 one of which is MAO inhibitory
activity.33-35 In many cases, it is know that activity and
selectivity are determined by the nature of the substituent in
position 7 (V structural set in Figure 1).33,34 On the other hand,
the influence on MAO-A activity of amino groups in position
3 of coumarins (A analogues), as well as the furocoumarins (B
analogues) and tetracyclic coumarin derivatives (C analogues),
has not been explored in this field (Figure 2).
On the basis of the above information, and considering our
experience with this family of compounds, we designed and
calculated the molecular descriptors for new coumarin ana-
logues.
The designed compounds were evaluated by the QSAR
model. We selected 15 coumarin derivatives with different levels
of structural complexity and, of these, 11 were determined as
active and 4 as inactive.
Synthesis. Compound 1 (8-MOP) is commercially available
and compounds 2, 3, 5-9, 11, 12, 14, 16, 18, 20, and 24 were
efficiently synthesized according to Schemes 1-5. Compounds
2 and 3 were obtained from 8-MOP by following three-step
synthetic routes35-37 (Scheme 1). Reaction of 7-hydroxy-4-
methylcoumarin (4) with 2-chlorocyclohexanone or 2-chloro-
cyclopentanone under Williamson conditions gave compounds
5 and 8, respectively.38-40 These compounds were treated with
alkali to give compounds 6 and 9. Finally, compound 6 was
oxidized with DDQ to give benzofurocoumarin 738 (Scheme
2). Compounds 11 and 12 were obtained from 5-methoxy-
a Reagents and conditions: (a) 2-chlorocycloalkanone/K2CO3, acetone,
reflux; (b) 0.1 N NaOH, reflux; (c) DDQ, toulene, reflux.
resorcinol (10) through 7-hydroxy-5-methoxy-4-methylcoumarin
and 5-hydroxy-7-methoxy-4-methylcoumarin, respectively.39-41
In a similar way compound 14 was obtained from 2-methoxy-
resorcinol (13)42 (Scheme 3). Compound 1643 was prepared from
2,4-dihydroxy-3-methoxybenzaldehyde (15) and was N-acet-
ylated and treated with chloroacetone under Williamson condi-
tions to give compound 18 in 60% overall yield. Treatment of
18 in a strong alkaline medium led to the furan ring, and
subsequent hydrolysis of the amide group in an acidic medium
afforded furocoumarin 20 in 63% overall yield (Scheme 4).
3-Aminofurocoumarin 2144 was used to synthesize the corre-
sponding N-glycyl derivative 24 according to Scheme 5.
N-Acetylation of 21 with bromoacetyl chloride, subsequent
replacement of the bromo substituent by a phthalimide group,
and hydrolysis of the imide gave 24 in 45% overall yield.
MAO-A Inhibition Assay. The synthesized compounds were
tested for their inhibitory effect on MAO-A. The inhibition data
(expressed as IC50 values) are reported in Table 2. Compound
8 showed the highest activity, with an IC50 value of 0.04 µM.
The three analogues 3, 12, and 16 did not inhibit MAO-A even

QSAR Model for Screening of MAO-A Inhibitors
Journal of Medicinal Chemistry, 2006, Vol. 49, No. 3 1153
Scheme 5a
Scheme 3a
a Reagents and conditions: (a) CH3COCH2CO2Et/H2SO4, 25 °C; (b)
2-chlorocyclohexanone/K2CO3, acetone, reflux.
Scheme 4a
a Reagents and conditions: (a) BrCH2COCl/pyridine/toulene, rt; (b)
potassium phthalimide/DMA, rt; (c) CH3NHNH2/CHCl3, rt.
Table 2. In Silico and in Vitro Evaluation of MAO-A Inhibition of
Selected Compounds
Ga
compd
IC50 (µM)
observed
predicted
pb active (%)
2
7
8
2.51
7.24
0.04
+
+
+
+
+
+
+
+
+
-
-
-
-
-
-
+
+
+
+
+
+
+
+
+
+
+
-
-
-
-
96.13
99.86
99.04
99.75
98.61
98.51
95.16
95.98
88.34
99.00
99.75
12.5
9
11.90
21.41
11.13
6.43
21.74
12.35
43.02
65.25
85.11
>200
>200
>200
11
14
18
20
24
5
6
1
3
12
16
0.45
1.39
42.30
a Reagents and conditions: (a) NH2CH2CO2H/Ac2O/Ac2ONa, reflux; 3
N HCl; (b) Ac2O/Ac2OH, 0 °C; (c) ClCH2COCH3/K2CO3/acetone, reflux;
(d) 0.1 M NaOH/EtOH, reflux; (e) (i) 3 M HCl/MeOH, reflux; (ii) NaHCO3.
a The observed and predicted groups for the selected compounds; -, a
given compound lies within the group; +, the IC50 < 25 µM for the observed
group and p > 50% for predicted group. b p is the subsequent probability
predicted by the model.
at the highest concentration tested (200 µM). Three other
compounds (1, 5, and 6) showed moderate activity, with IC50
values in the range 43-85 µM. All of the other assayed
compounds had IC50 values in the range 0.04-22 µM.
To further investigate the MAO-A inhibition mechanism, the
most potent inhibitor (8) was selected. The inhibition of MAO-A
by this compound (a sigmoid function) provided evidence for
the existence of reversible inhibition (Figure 3A). Kinetic
analysis was carried out using Lineweaver-Burk plots. The inset
(Figure 3B) shows the secondary plot of slope versus the
concentration of 8. The data indicate a noncompetitive inhibition
mechanism.
were predicted as active (compounds 5 and 6) were found to
have IC50 values of 43.0 and 65.2 µM, respectively, and although
these compounds are considered inactive by our cutoff point,
the values are in the range often considered as active. 8-MOP
(compound 1) and compounds 3, 12, and 16 were predicted to
be inactive by the model, and the experimental data agree with
this classification, even at the highest concentration tested (200
µM).
In conclusion, in the work described here, we developed a
MARCH-INSIDE methodology that allowed us to predict by
in silico screening the in vitro inhibitory MAO-A activity in a
rapid and efficient way. In addition, the significance of this
approach is that a QSAR model for the MAO inhibitors was
able to correctly classify, for the first time, a series of
compounds with different structural patterns. This ability
demonstrates that this is a general model and we can therefore
conclude that the MARCH-INSIDE approach represents a
promising tool for the discovery of active compounds as drugs.
Conclusions
A total of 11 compounds were predicted to be active in the
in silico evaluation carried out using the QSAR model developed
here; compounds were considered active when they had an IC50
value below 25 µM. Of these compounds, a total of nine were
confirmed to be active by in vitro tests, and one of these
compounds (8) had a very high activity. Two compounds that

1154 Journal of Medicinal Chemistry, 2006, Vol. 49, No. 3
Santana et al.
reaction mixture was heated under reflux for 20 h. The precipitate
was filtered off, the organic phase was concentrated to dryness,
and the residue was purified by FC using 1:1 toluene/ethyl acetate
as eluent to give 18: yield 860 mg (70.5%); mp 207 °C; 1H NMR
(CDCl3) 8.62 (s, 1H, H-4), 8.00 (bs, 1H, HN), 7.17 (d, 1H, H-5, J
) 8.5), 6.75 (d, 1H, H-6, J ) 8.5), 4.69 (s, 2H, CH2), 4.04 (s, 3H,
CH3O), 2.32 and 2.24 [2s, 3 + 3H, 2(CH3O)]; IR 3325, 3110, 2944,
1713, 1674, 1605, 1538, 1140. Anal. (C15H15NO6) C, H, N.
3-Acetylamino-8-methoxy-4′-methylfuro[3,2-g]coumarin (19).
To a solution of 18 (1 g, 3.39 mmol) in ethyl alcohol (200 mL)
was added 0.1 M NaOH (200 mL). The mixture was heated under
reflux for 3 h, acidified with HCl, concentrated to half-volume,
and left overnight. The resulting precipitate was filtered off and
purified by FC using 9:1 toluene/ethyl acetate as eluent to give 19:
1
yield 620 mg (66%); mp 248-249 °C; H NMR (CDCl3) 8.77 (s,
1H, H-4), 8.05 (bs, 1H, HN), 7.47 (d, 1H, H-5′, J ) 1.3), 7.28 (s,
1H, H-5), 4.32 (s, 3H, CH3O), 2.26 (d, 3H, CH3, J ) 1.3), 2.25 (s,
3H, CH3CO); IR 3329, 3110, 2954, 1711, 1677, 1594, 1531, 1389,
1142. Anal.(C15H13NO5) C, H, N.
3-Amino-8-methoxy-4′-methylfuro[3,2-g]coumarin (20). A
mixture of 19 (500 mg, 1.74 mmol), methyl alcohol (150 mL), and
3 M HCl (75 mL) was heated under reflux for 1 h. The mixture
was allowed to cool and was neutralized with NaHCO3 and
concentrated to 30% of its initial volume. The precipitate was
filtered off to give 20: yield 409 mg (96%); mp 139-140 °C; 1H
NMR (CDCl3) 7.43 (d, 1H, H-5′, J ) 1.3), 7.04 (s, 1H, H-5), 6.81
(s, 1H, H-4), 4.28 (s, 3H, CH3O), 2.24 (d, 3H, CH3, J ) 1.3); IR
3324, 3110, 2947, 1707, 1638, 1577, 1352, 1169, 1087, 766. Anal.
(C13H11NO4) C, H, N.
3-Bromoacetylamino-4′-methylfuro[3,2-g]coumarin (22). To
a solution of 3-amino-4′-methylfuro[3,2-g]coumarin44 (21, 500 mg,
2.32 mmol) in toluene (200 mL) was added pyridine (0.2 mL, 2.47
mmol) and bromoacetyl chloride (1 mL, 12.10 mmol). The mixture
was stirred for 1 h at room temperature. The resulting precipitate
was filtered off and the solution evaporated under vacuum to give
22: yield 632 mg (80%); mp 215-217 °C; 1H NMR (CDCl3) 8.90
(bs, 1H, HN), 8.81 (s, 1H, H-4), 7.64 (s, 1H, H-5), 7.49 (d, 1H,
H-5′, J ) 1.3), 7.45 (s, 1H, H-8), 4.04 (s, 2H, CH2), 2.28 (d, 3H,
CH3, J ) 1.3); IR 3309, 1707, 1673, 1542, 1362, 1161. Anal.
(C14H10BrNO4) C, H, N.
3-Phthalylacetamido-4′-methylfuro[3,2-g]coumarin (23). To
a solution of potassium phthalimide (350 mg, 1.90 mmol) in
dimethylacetamide (20 mL) was added bromo derivative 22 (500
mg, 1.48 mmol). The mixture was stirred for 4 h at room
temperature, acetic acid (2 mL) was added, and the mixture was
left overnight at 0 °C. The resulting precipitate was filtered off
and washed with toluene to give 23: yield 512 mg (85.6%); mp
312 °C; 1H NMR (DMSO-d6) 10.37 (s, 1H, HN), 8.69 (s, 1H, H-4),
8.00-7.85 (m, 6H, H-5 + H-5′ + phthalimide), 7.70 (s, 1H, H-8),
4.63 (s, 2H, CH2), 2.22 (s, 3H, CH3); IR 3341, 1725, 1710, 1688,
1532, 1420, 1200, 1147. Anal. (C22H14N2O6) C, H, N.
3-Aminoacetamido-4′-methylfuro[3,2-g]coumarin (24). A mix-
ture of phthalimido derivative 23 (500 mg, 1.24 mmol), methyl
hydrazine (0.5 mL, 9.35 mmol), HCl3 (12 mL), and ethyl alcohol
(12 mL) was stirred for 6 h at room temperature. The solvent was
evaporated under vacuum and the residue washed with hexane and
recrystallized from ethyl alcohol to give 24: yield 225 mg (66.5%);
mp 291-293 °C; 1H NMR (DMSO-d6) 8.75 (s, 1H, H-4), 8.01 (s,
1H, H-5), 7.88 (d, 1H, H-5′, J ) 1.1), 7.69 (s, 1H, H-8), 3.91 (s,
2H, CH2), 2.26 (d, 3H, CH3, J ) 1.1); IR 3245 1711, 1663, 1626,
1540, 1458, 1447, 1141. Anal. (C14H12N2O4) C, H, N.
Animals and Brain Mitochondria Preparations. Adult male
Sprague-Dawley rats weighing 200-250 g were used. The rats
were anesthetized with carbon dioxide and killed by decapitation.
The brains were removed in an ice-cold isolation medium containing
0.32 M sucrose, 10 mM Tris-HCl buffer, pH 7.4. Blood vessels
and pial membranes were removed and mitochondria were then
prepared according to a previously published method.45
Figure 3. (A) Inhibition of MAO-A by compound 8. Each point rep-
resents mean ( SEM from triplicate determinations. (B) Lineweaver-
Burk plots and secondary replot of MAO-A inhibition by 8 showing a
linear competitive mechanism. All values represent mean ( SEM from
triplicate determinations.
Experimental Section
Chemistry. Melting points were determined using a Reichert
Kofler thermopan or in capillary tubes on a Bu¨chi 510 apparatus
and are uncorrected. IR spectra were recorded on a Perkin-Elmer
1640FT spectrophotometer. 1H and 13C NMR spectra were recorded
on a Bruker AMX spectrometer at 300 and 75.47 MHz, respectively,
using TMS as internal standard (chemical shifts in δ values, J in
hertz). Mass spectra were obtained using a Hewlett-Packard 5988A
spectrometer. Elemental analyses were performed using a Perkin-
Elmer 240B microanalyzer and were within (0.4% of calculated
values in all cases. Silica gel (Merck 60, 230-00 mesh) was used
for flash chromatography (FC). Analytical thin-layer chromatog-
raphy (TLC) was performed on plates precoated with silica gel
(Merck 60 F254, 0.25 mm).
3-Acetylamino-7-hydroxy-8-methoxycoumarin (17). To a solu-
tion of compound 1643(2 g, 9.66 mmol) in 30% acetic acid (250
mL) at 0 °C was added acetic anhydride (5 mL, 52.94 mmol). The
resulting precipitate was filtered off to give 17: yield 2.04 g (85%);
mp 228 °C; 1H NMR (DMSO-d6) 10.25 (bs, 1H, HO), 9.68 (s, 1H,
HN), 8.50 (s, 1H, H-4), 7.26 (d, 1H, H-5, J ) 8.5), 6.86 (d, 1H,
H-6, J ) 8.5), 3.83 (s, 3H, CH3O), 2.13 (s, 3H, CH3CO); IR 3324,
2990, 2975, 1708, 1670, 1603, 1544, 1237. Anal. (C12H11NO5) C,
H, N.
7-Acetonyloxy-3-acetylamino-8-methoxycoumarin (18). To a
solution of 17 (1 g, 4.01 mmol) in dry acetone (300 mL) were
added K2CO3 (4 g) and chloroacetone (1 mL, 12.6 mmol). The
Determination of MAO-A Activity. MAO-A activity was
determined spectrophotometrically according to a previously re-
ported method.41 Mitochondrial incubations (final protein concen-

QSAR Model for Screening of MAO-A Inhibitors
Journal of Medicinal Chemistry, 2006, Vol. 49, No. 3 1155
(12) Carrieri, A.; Carotti, A.; Barreca, M. L.; Altomare, C. Binding models
of reversible inhibitors to type-B monoamine oxidase. J. Comput.
Aid.- Mol. Des. 2002, 16, 769-778.
tration 1 mg/mL) were performed in 100 mM potassium phosphate
buffer (pH 7.4) at 37 °C. The mitochondria were preincubated at
37 °C for 5 min with the selective irreversible MAO-A inhibitor
clorgyline (250 nM). After a second preincubation with DMSO
(control) or the potential inhibitor dissolved in DMSO, the
nonselective substrate kynuramine was added at a concentration
equal to its KM (90 µM for MAO-A). The final DMSO concentration
was 5% (v/v). The formation of 4-hydroxyquinoline was continu-
ously monitored at 314 nm.
For each inhibitor, IC50 values were determined from MAO-A
inhibition/-log concentrations plots, using the graph package Origin
v. 6.0 (Microcal Software Inc., Northampton, MA). Analysis of
the corresponding Lineweaver-Burk plots enabled the mechanism
of the inhibition to be assessed.
The reversibility of the inhibition was assessed by dialysis, as
reported previously.46 Briefly, the procedure involved incubation
of mitochondria preparations (1 mg/mL) containing clorgyline (250
nM) at 37 °C for 15 min in the absence (control) or presence of
the inhibitor concentration equal to its IC50. These mixtures were
then dialyzed using a Biodialyzer (Sigma Chemical Co.) with an
ultrafiltration membrane of nominal molecular weight limit 10 000.
The dialysis was preformed at 4 °C using 250 mL of outer buffer
(100 mM potassium phosphate buffer, pH 7.4). The outer buffer
was replaced with fresh buffer every 30 min for a period of 2 h.
Dialyzed mixtures were then assayed for MAO-A activity.
Computer Software. The calculation of the molecular descrip-
tors was implemented in the in-house software MARCH-INSIDE.
This software has a graphical interface that makes it user-friendly
for medicinal chemists.
(13) 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. 2005, 102, 12684-12689.
(14) Binda, C.; Hubalek, F.; Li, M.; Herzig, Y.; Sterling, J.; Edmondson,
D. E.; Mattevi, A. Crystal structures of monoamine oxidase B in
complex with four inhibitors of the N-propargylaminoindane class.
J. Med. Chem. 2004, 47, 1767-1774.
(15) Mevedev, A. E.; Veselovsky, A. V.; Shvedov, V. I.; Tikhonova, T.
A.; Moskvitinia, T. A.; Fedotova, O. A.; Axenova, L. N.;
Kamyshanskaya, A. Z.; Kirkel, A. Z. and Ivanov, A. Inhibition of
monoamino oxidase by pirlindole analogues: 3D-QSAR and CoMFA
analysis. J. Chem. Inf. Comput. Sci. 1998, 38, 1137-1144.
(16) Nunez, M. B.; Maguna, F. P.; Okulik, N. B.; Castro, E. A. QSAR
modeling of the MAO inhibitory activity of xanthone derivatives.
Bioorg. Med. Chem. Lett. 2004, 14, 5611-5617.
(17) Todeschini, R and Consonni, V. Handbook of Molecular Descriptors;
Wiley VCH: Weinheim, Germany, 2000.
(18) Kier, L. B.; Hall, L. H. Topological Indices and Related Descriptors
in QSAR and QSPR; Gordon and Breach Science Publishers:
Amsterdam, 1999; pp 445-489.
(19) Estrada E.; Uriarte, E. Recent advances on the role of topological
indices in drug design discovery research. Curr. Med. Chem. 2001,
8, 1573-1588.
(20) Estrada, E.; Uriarte, E.; Montero, A.; Teijeira, M.; Santana, L.; De
Clercq, E. A novel approach for the virtual screening and rational
design of anticancer compounds. J. Med. Chem. 2000, 43, 1975-
85.
(21) Karelson, M.; Lobanov, V. S.; Katritzky, A. R. Quantum-chemical
descriptors in QSAR/QSPR studies. Chem. ReV. 1996, 96, 1027-
1043.
(22) Bultinck, P.; Langenaeker, W.; Carbo-Dorca, R.; Tollenaere, J. P.
Fast calculation of quantum chemical molecular descriptors from the
electronegativity equalization methodology. J. Chem. Inf. Comput.
Sci. 2003, 43, 422-428.
Acknowledgment. We are grateful to the Xunta de Galicia
(PGIDIT02BTF20301PR and PGIDIT05BTF20302PR2) for
partial financial support. The authors would like to acknowledge
the referees for useful comments, which have undoubtedly
improved the final presentation of the manuscript.
(23) Sanderson, R. T. Electronegativity and bond energy. J. Am. Chem.
Soc. 1983, 105, 2259-2261.
(24) Mortier, W. J.; Ghosh, S. K.; Shankar, S. Electronegativity-
equalization method for the calculation of atomic charges in
molecules. J. Am. Chem. Soc. 1986, 108, 4315-4320.
(25) Freund, J. A.; Poschel, T., Eds. Stochastic processes in physics,
chemistry, and biology. In Lecture Notes in Physics; Springer-
Verlag: Berlin, Germany. 2000.
(26) Gonza´lez-D´ıaz, H.; Agu¨ero, G.; Cabrera, M. A.; Molina, R.; Santana,
L.; Uriarte, E.; Delogu, G.; Castan˜edo, N. Unified Markov thermo-
dynamics based on stochastic forms to classify drugs considering
molecular structure, partition system, and biologic species. Distribu-
tion of the antimicrobial G1 on rat tissues. Bioorg. Med. Chem. Lett.
2005, 15, 551-557.
Supporting Information Available: Elemental analyses data
for synthesized compounds; bibliographic references for database
compounds; Table 3, a complete list of compounds used in training
sets with the indication of the compound number in the corre-
sponding bibliographic reference as well as their observed clas-
sification, predicted classification, training, and subsequent vali-
dation probabilities; and Table 4, a list of values of the molecular
descriptors contained in the model. This material is available free
of charge via the Internet at http://pubs.acs.org.
(27) Gonza´lez-D´ıaz, H.; Cruz-Monteagudo, M.; Molina, R.; Tenorio, E.;
Uriarte, E. Predicting multiple drugs side effects with a general drug-
target interaction thermodynamic Markov model. Bioorg. Med. Chem.
2005, 13, 1119-1129.
(28) Gonza´lez-D´ıaz, H.; Torres-Go´mez, L. A.; Guevara, Y.; Almeida, M.
S.; Molina, R.; Castan˜edo, N.; Santana, L.; Uriarte, E. Markovian
chemicals “in silico” design (MARCH-INSIDE), a promising ap-
proach for computer-aided molecular design III: 2.5D indices for
the discovery of antibacterials. J. Mol. Mod. 2005, 11, 116-123.
(29) Van Waterbeemd, H. Discriminant analysis for activity prediction.
In Chemometric Methods in Molecular Design. Method and Prin-
ciples in Medicinal Chemistry; Manhnhold, R., Krogsgaard-Larsen,
P., Timmerman, H., Eds.; VCH: Weinhiem, 1995; Vol. 2, pp 265-
282.
References
(1) Dostert, P.; Strolin Benedetti, M.; Jafre, M. Monoamine Oxidase:
Basic and Clinical Frontiers; Elsevier: Amsterdan, 1982.
(2) Singer, T. P. In Chemistry and Biochemistry of FlaVoenzymes; Muller,
F., Ed.; CRC Press: Boca Raton, FL, 1991; Vol. 2, p 437.
(3) Zeller, E. Classification and Nomenclature of Monoamine Oxidase
and Other Amine Oxidases: Structure, Function and Altered Func-
tions; Academic Press: New York. 1979.
(4) Wouters, J. Structural aspects of monoamine oxidase and its reversible
inhibition. Curr. Med. Chem. 1998, 5, 137-162.
(5) Yamada, M.; Yasuhara, H. Clinical pharmacology of MAO inhibi-
tors: Safety and future. Neurotoxicology 2004, 25, 11-20.
(6) Rudorfer, M. V.; Potter, V. Z. Antidepressants. A comparative review
of the clinical pharmacology and therapeutic use of the “newer”
versus the “older” drugs. Drugs 1989, 37, 713-738.
(7) Gerlach, M.; Double, K.; Reichmann, H.; Riederer, P. Arguments
for the use of dopamine receptor agonists in clinical and preclinical
Parkinson’s diseases. J. Neural Transm. Suppl. 2003, 167-193.
(8) Riederer, P.; Danielczyk, W.; Grunblatt, E. Monoamine oxidase-B
inhibition in Alzheimer’s desease. Neurotoxicology 2004, 25, 271-
277.
(9) Cesura, A. M.; Pletscher, A. The new generation of monoamine
oxidase inhibitors. Prog. Drug Res. 1992, 38, 171-297.
(10) Youdim, M. B. The advent of selective monoamine oxidase A
inhibitor antidepressants devoid of the cheese reaction. Acta Psychiatr.
Scand. Suppl. 1995, 386, 5-7.
(11) Edmondson, D. E.; Mattevi, A.; Binda, C.; Li, M.; Hubalek, F.
Structure and mechanism of monoamine oxidase. Curr. Med. Chem.
2004, 11, 1983-1993.
(30) Kleeman, A.; Engel, J.; Kutscher, B.; Reichert, D. Pharmaceutical
Substances, 4th ed.; George Thieme Verlag: Stuttgart, 2001.
(31) Negwer, M. Organic Chemical Drugs and Their Synonyms;
Akademie-Verlag: Berlin, 1987.
(32) Santana, L.; Uriarte, E.; Roleira, F.; Mihazes, N.; Borges, F.
Furocoumarins in medicinal chemistry. Synthesis, natural occurrence
and biological activity. Curr. Med. Chem. 2004, 11, 3239-3261.
(33) Gnerre, C.; Catto, M.; Francesco, L.; Weber, P.; Carrupt, P.-A.;
Altomare, C.; Carotti, A.; Testa, B. Inhibition of monoamine oxidase
by functionalized coumarin derivatives: Biological activities, QSAR,
and 3D-QSARs. J. Med. Chem. 2000, 43, 4747-4758.
(34) Chimenti, F.; Secci, D.; Bolasco, A.; Chimenti, P.; Granese, A.;
Befani, O.; Turini, P.; Alcaro, S.; Ortuso, F. Inhibition of monoamine
oxidases by coumarin-3-acyl derivatives: Biological activity and
computational study. Bioorg. Med. Chem. Lett. 2004, 14, 3697-
3703.

1156 Journal of Medicinal Chemistry, 2006, Vol. 49, No. 3
Santana et al.
(35) Antonello, C.; Marciani-Magno, S.; Gia, O.; Carlassare, F.;
Baccichetti, F.; Bordin, F. Diethylaminoalkyloxycoumarin and
-furocoumarin derivatives. Il Farmaco, Ed. Sci. 1979, 34, 139-156.
(36) Hansen, J. B.; Bjerring, P.; Buchardt, O.; Ebbesen, P.; Kanstrup, A.;
Karup, G.; Knudsen, P. H.; Nielsen, P. E.; Norden, B.; Ygge, B.
Psoralenamines. 3. Synthesis, pharmacological behavior, and DNA
binding of 5-(aminomethyl)-8-methoxy-, 5-[[(3-aminopropyl)oxy]
methyl]- and 8-[(3-aminopropyl)oxy]psoralen derivatives. J. Med.
Chem. 1985, 28, 1001-1010.
(37) Nakamura, Y. Photocrosslinking and the cleavage of DNA by iron
complex-substituted psoralen. Chem. Lett. 1989, 1, 51-54.
(38) Gia, O.; Tonconi, M. T.; Uriarte, E. Benzo and tetrahydrobenzo-
furocoumarins: New potential photochemotherapeutic drugs. Il
Farmaco 1997, 52, 289-294.
(39) Corda, L. Delogu, G.; Favretto, D.; Maccioni, E. Podda, G.; Santana,
L.; Tomaselli, P.; Traldi, P.; Uriarte, E. Synthesis and mass
spectrometric investigation of some new coumarin derivatives. Rapid.
Comm. Mass Spectrom. 1998, 12, 2041-2046.
(40) Dalla Via, L.; Gia, O.; Viola, G.; Bertoloni, G.; Santana, L.; Uriarte,
E. Photobiological studies of new cyclopentenepsoralens. Il Farmaco
1998, 12, 638-644.
(41) Tera´n, C.; Miranda, R.; Santana, L.; Teijeira, M.; Uriarte, E. Synthesis
of angular and linear benzofurocoumarins. Synthesis 1997, 12, 1384-
1386.
(42) Dalla Via, L.; Gia, O.; Marciani-Magno, S.; Teijeira, M.; Santana,
L.; Uriarte, E. New tetracyclic analogues of photochemothera-
peutic drugs 5-MOP and 8-MOP: Synthesis, DNA interacction
and antiproliferative activity. J. Med. Chem. 1999, 42, 4405-
4413.
(43) Rodighiero, G.; Antonello, C. Sintesi di alcuni derivati 3-ammino-
cumarinici e prime notizie sulle loro propieta` antibatteriche. Bolletino
Chim. Farm. 1958, 97, 592-601.
(44) Antonello, C.; Marciani-Magno, S.; Baessato, O.; Palumbo, M.
3-Amino- and 3-oxy-derivatives of psoralen: Preparation and
interactions with DNA. Il Farmaco, Ed. Sci. 1981, 36, 565-
584.
(45) Me´ndez-AÄ lvarez, E.; Soto-Otero, R.; Sa´nchez-Sellero, I.; Lo´pez-
Rivadulla, L. M. Inhibiton of brain monoamine oxidase by adducts
of 1,2,3,4-tetrahiisoquinoline with components of cigarette smoke.
Life Sci. 1997, 60, 1719-1727.
(46) Me´ndez-AÄ lvarez, E.; Soto-Otero, R.; Sa´nchez-Sellero, I.; Lo´pez-
Rivadulla, L. M. In Vitro inhibition of catalase activity by cigarette
smoke: Relevance for oxidative stress. J. Appl. Toxicol. 1998, 18,
443-448.
JM0509849