Pharmacophore based discovery of potential antimalarial agent targeting haem detoxification pathway

Article abstract of DOI:10.1016/j.ejmech.2008.02.005

Pharmacophore hypotheses were generated from molecules having putative antimalarial activities targeting haem detoxification pathway of malarial parasite. A training set consisting of 33 compounds, whose activities were evaluated for haem polymerization inhibition and against chloroquine resistant (K1) strain of Plasmodium falciparum, was optimized to generate hypotheses. The hypothesis showing optimum correlation between actual and estimated activities was validated by Fischer's randomization test at 95% confidence level and used as a model to screen our in-house compound database. Nicotinic acid [trans-3-(4-ethoxy-3-methoxy-phenyl)-1-(4-hydroxy-phenyl)-allylidene]-hydrazide (ALH5) was obtained as a hit. The compound was synthesized and evaluated against chloroquine sensitive (MRC-02) and resistant (RKL9) strains of malarial parasite P. falciparum. The compound showed antimalarial activity in nanomolar range and was found comparable with chloroquine.

Full text of DOI:10.1016/j.ejmech.2008.02.005

Available online at www.sciencedirect.com
European Journal of Medicinal Chemistry 43 (2008) 2840e2852
http://www.elsevier.com/locate/ejmech
Original article
Pharmacophore based discovery of potential antimalarial agent
targeting haem detoxification pathway
*
Badri Narayan Acharya, Deepika Saraswat, Mahabir Parshad Kaushik
Discovery Centre, Defence Research and Development Establishment, Jhansi Road, Gwalior, Madhya Pradesh 474002, India
Received 27 November 2007; received in revised form 5 February 2008; accepted 7 February 2008
Available online 4 March 2008
Abstract
Pharmacophore hypotheses were generated from molecules having putative antimalarial activities targeting haem detoxification pathway of
malarial parasite. A training set consisting of 33 compounds, whose activities were evaluated for haem polymerization inhibition and against
chloroquine resistant (K1) strain of Plasmodium falciparum, was optimized to generate hypotheses. The hypothesis showing optimum correla-
tion between actual and estimated activities was validated by Fischer’s randomization test at 95% confidence level and used as a model to screen
our in-house compound database. Nicotinic acid [trans-3-(4-ethoxy-3-methoxy-phenyl)-1-(4-hydroxy-phenyl)-allylidene]-hydrazide (ALH5)
was obtained as a hit. The compound was synthesized and evaluated against chloroquine sensitive (MRC-02) and resistant (RKL9) strains of
malarial parasite P. falciparum. The compound showed antimalarial activity in nanomolar range and was found comparable with chloroquine.
Ó 2008 Elsevier Masson SAS. All rights reserved.
Keywords: Malaria; Plasmodium falciparum; Haem detoxification; Pharmacophore; Predictive model; Hydrazide
1. Introduction
Discovery and development of novel antimalarials with bet-
ter activities targeting haem detoxification pathway will re-
quire an understanding of both the structural and the
functional characteristics (pharmacophoric features) of the
known potential antimalarial molecules. Pharmacophore mod-
eling is a drug-designing tool, which exploits the wealth of in-
formation available, including the structural and biological
activity data of known active molecules. A pharmacophore
model represents the three-dimensional (3D) arrangements
of chemical features in a molecule that may be important for
its bioactivity [6]. Qualified pharmacophore models may be
used as a query for searching chemical databases to find
new chemical entities. Kurosawa et al. have described the
use of a 3D pharmacophore model as a query for virtual
screening of database to discover antimalarial molecules [7].
Besides the application as query for database searching, 3D
pharmacophore models may be used for predicting the activity
of novel molecules [8e13].
Malaria is a major health problem in today’s world, with
200e500 million clinical cases and two million deaths occur-
ring each year [1]. The development of drug resistance in ma-
laria causing parasites imposes the need for the discovery of
new potential antimalarial molecules either from natural sour-
ces or by synthesis [1]. Haem detoxification is an important
biochemical process for the survival of malarial parasite [2].
Drug discovery methods that focus on the inhibition of haem
detoxification can identify antimalarials, which are both potent
and likely to remain useful for a longer period of time [3]. One
excellent example relates to quinoline antimalarials, which
acts by interfering with the haem detoxification, allowing the
intraparasitic build-up of toxic haem [4]. This has defined in-
hibition of haem detoxification as an attractive target for de-
signing new antimalarial drugs [5].
We have reported a pharmacophore based predictive model
[12] to discover potential antimalarial agents targeting haem
detoxification pathway, which has been developed on the basis
* Corresponding author. Tel.: þ91 751 2343972; fax: þ91 751 2341148.
E-mail address: mpkaushik@rediffmail.com (M.Parshad Kaushik).
0223-5234/$ - see front matter Ó 2008 Elsevier Masson SAS. All rights reserved.
doi:10.1016/j.ejmech.2008.02.005

B.N. Acharya et al. / European Journal of Medicinal Chemistry 43 (2008) 2840e2852
2841
of quinoline antimalarials [4]. In the present study we report
2.2. Selection of training set and test set molecules
on the development of a pharmacophore model from structur-
ally diversified molecules with similar mode of action, i.e. in-
hibition of haem detoxification process of malarial parasites.
The molecules were taken from literature [4,14e21] along
with their antimalarial activities against chloroquine resistant
(K1) strain of malarial parasite Plasmodium falciparum. Num-
ber of training set molecules and input pharmacophoric fea-
tures were optimized to develop a quantitative model having
good correlation between actual and predicted biological
activities. The optimized model was validated by Fischer’s
randomization test at 95% confidence level. These models
were further evaluated as an activity prediction model by esti-
mating the antimalarial activities of a test set consisting of 44
molecules taken from literature [4,14e21]. The best model
was used to screen our in-house database and a compound
nicotinic acid [trans-3-(4-ethoxy-3-methoxy-phenyl)-1-(4-
hydroxy-phenyl)-allylidene]-hydrazide (ALH5) was obtained
as a hit. The molecule was synthesized and evaluated against
chloroquine sensitive (MRC-02) and resistant (RKL9) strains
of P. falciparum and found comparable with chloroquine
diphosphate (CQ).
Selection of training set molecules is the most important as-
pect of hypothesis generation. The selection followed some
basic requirements, such as a minimum of 16 structurally di-
verse compounds must be selected to avoid any chance of cor-
relation. The activity data should have a range of 3e4 orders
of magnitude. Several active molecules were included in the
training set so that they could provide critical requirement
for pharmacophore generation. Several moderately active
and inactive molecules were included to spread the activity
range. By following the above criteria, 39 molecules (C1e
C39) were chosen from the constructed database and put in
the training set (Fig. 1). Here, biological activity data are rep-
resented as IC50 in nM for antimalarial activity against chloro-
quine resistant (K1) strain of P. falciparum. The activities
against malarial parasite (in nM) have been classified some-
what arbitrarily as follows: highly active (<100), moderately
active (101e1000) and poorly active (>1000). Rest of the
molecules (T1eT44) were put in the test set (Fig. 2).
2.3. Generation of pharmacophore model
2. Development of pharmacophore model
In the current work, HypoGen module of Catalyst was used
to generate pharmacophore models from the training set mol-
ecules. The compounds were built using Catalyst 2De3D
sketcher. Conformational models of all training set molecules
were generated using ‘‘best quality’’ conformational search
option in Catalyst using a constraint of a 20 kcal molꢀ1 energy
threshold above the global energy minimum and CHARMM
force field parameters. A maximum of 250 conformations
were generated to cover maximum conformational space. In-
stead of using just the lowest energy conformation of each
compound, all conformational models for molecules in the
training set were used in Catalyst HypoGen for pharmaco-
phore hypothesis generation.
Five pharmacophoric features were used to generate phar-
macophore hypotheses. The minimum and maximum counts
for all the five features were kept 0 and 5, respectively. The
values for uncertainty, minimum points and minimum subset
points were kept the default values of Catalyst. The HypoGen
module performs three important cost calculations in the units
of bits that determine the success of any pharmacophore hy-
potheses. ‘‘Fixed cost’’, which represents the simplest model
fits all data perfectly. ‘‘Null cost’’ represents the highest cost
of a pharmacophore with no features and estimates the activity
to be the mean of the activity data of the training set mole-
cules. Its absolute value is equal to the maximum occurring er-
ror cost. A meaningful pharmacophore hypothesis may result
when the difference between these two values (null cost -
ꢀ fixed cost) is more than at least 20. ‘‘Total cost’’ is the
sum of three cost factors: a weight, an error and a configuration
cost. The weight cost increases if the weight factor for the
chemical features deviates from default value of 2. The error
cost is solely dependent upon the root mean square (rms) dif-
ferences between the estimated and actual activities of the
training set molecules. The rms value represents the quality
All modeling works were performed on a PIV @ 2 GHz
computer with 1 GB RAM, Nvidia Quadro4 980XGL graphics
card, running on Redhat enterprise Linux 2.1 WS. Catalyst
4.10 software [22] was used to generate pharmacophore
models.
2.1. Building of database
Pharmacophore modeling requires structure and activity
data of several active and inactive molecules for hypothesis
generation. A particular molecule may show different activi-
ties in terms of IC50 values against different strains of P. fal-
ciparum. Hence in this work, these molecules were
considered whose antiparasitic activity against CQ resistant
K1 strain of P. falciparum has been reported. Structures of
haem polymerization inhibitors with their antiparasitic activity
against K1 strain from various medicinal chemistry as well as
life science journals [4,14e21] were collected and developed
as a unique database. The database comprises of 83 structur-
ally diversified compounds with a wide range of experimental
in vitro antimalarial activities against K1 strain of P. falcipa-
rum. Fig. 1 shows 39 molecules (C1eC39) and Fig. 2 shows
44 molecules (T1eT44). The molecules belong to different
chemical classes such as quinolines (C28eC36 and T32e
T40) [4], pyridines (C27 and T41) [4], bis-quinolines (C1,
C2 and T1) [14], cryptolepines (C5eC14 and T2eT17)
[15,16], acridines (C15eC20 and T18eT27) [17], phenothia-
zines (C21eC23 and T28eT30) [18], tetrahydro acridine dio-
nes (C3, C4 and T31) [19] and quinoxalines (C24eC26, T43
and T44) [20]. Some carbocyclic aromatic compounds having
very good antimalarial activity such as C37 [4], C38, C39 and
T42 [21] were also included in the database.

2842
B.N. Acharya et al. / European Journal of Medicinal Chemistry 43 (2008) 2840e2852
Fig. 1. Structures of 39 training set molecules.
of correlation between actual and estimated activity data. The
configuration cost is represented as log2 P, where P is the
number of initial hypotheses created in the constructive phase
and that survived in the subtractive phase. It should not be
more than 17.
2.4. Validation of pharmacophore model
To evaluate the statistical relevance of the models, Fisher’s
randomization test was applied. The purpose of the test is to
randomize the activity data associated with the training set.

B.N. Acharya et al. / European Journal of Medicinal Chemistry 43 (2008) 2840e2852
2843
Fig. 2. Structures of 44 test set molecules.
The randomized training sets are used to generate hypotheses
using the same features and parameters. If the randomized data
sets result in the generation of pharmacophore with similar or
better cost values, rms and correlation, then the original hy-
pothesis is considered to have been generated by chance.
The statistical significance is given by the equation of signifi-
cance ¼ [1 ꢀ (1 þ x)/y], where ‘x’ is total number of hypothe-
ses having total cost lower than the most significant hypothesis
and ‘y’ is the number of initial HypoGen runs þ random
runs. With the aid of CatScramble program available in
Catalyst/HypoGen module, the experimental activities of
training set molecules were scrambled randomly and resulting
training sets were used for HypoGen runs. Thereby all param-
eters were kept as per the initial HypoGen calculation. For
a 95% confidence level, CatScramble created 19 spreadsheets.
2.5. Virtual screening
A small compound library of 90 molecules was created us-
ing Catalyst 2De3D sketcher. Conformational models of all

2844
B.N. Acharya et al. / European Journal of Medicinal Chemistry 43 (2008) 2840e2852
the molecules were generated using ‘‘best quality’’ conforma-
tional search option in Catalyst using a constraint of
a 20 kcal molꢀ1 energy threshold above the global energy min-
imum and CHARMM force field parameters. A maximum of
250 conformations were generated to cover a broad conforma-
tional space. The library was virtually screened by a statisti-
cally valid pharmacophore model using Catalyst Compare/
Fit module. The hit molecule was synthesized and analyzed
against chloroquine sensitive (MRC-02) and resistant
(RKL9) strains of P. falciparum.
human AB (þ) serum and 0.2% sodium bicarbonate (Sigma)
and maintained at 5% CO2. Cultures containing early ring
stages were synchronized by addition of 5% D-sorbitol
(Sigma) lysis [24], used for testing. Initial culture was main-
tained in small vials with 10% haematocrit, i.e. 10 ml eryth-
rocytes containing 1.0% ring stage parasite in 100 ml
complete media. The culture volume per well of the assay
was 100 ml. Number of parasites for the assay was adjusted
from 1% to 1.5% by diluting with fresh human B (þ)
RBC. Assay was done in 96-well microtitre flat-bottomed
tissue culture plates incubated at 37 ꢁC for 24 h in the pres-
ence of twofold serial dilutions of compounds and chloro-
quine diphosphate for their effect on schizont maturation.
Compounds were dissolved in ethanol and further diluted
with RPMI-1640 medium (the final ethanol concentration
did not exceed 0.5% which did not affect parasite growth).
Chloroquine diphosphate was dissolved in aqueous medium.
Test was done in duplicate wells for each dose of the drugs.
Solvent control culture containing the same concentrations
of the solvent as present in the test wells was prepared
with RPMI-1640 containing 10% AB (þ) serum. Parasite
growth was found unaffected at the solvent concentrations.
Growth of the parasites from duplicate wells of each concen-
tration was monitored by Giemsa stained blood smears by
counting the number of schizonts per 100 asexual parasites.
Percent schizont maturation inhibition was calculated by the
formula: (1 ꢀ Nt/Nc) ꢂ 100, where Nt and Nc represent the
number of schizonts in the test and control wells,
respectively.
3. Chemistry
Synthesis of nicotinic acid [trans-3-(4-ethoxy-3-methoxy-
phenyl)-1-(4-hydroxy-phenyl)-allylidene]-hydrazide (ALH5)
involved two simple steps (Scheme 1). In the first step synthe-
sis of a,b-unsaturated ketone, trans-3-(3-ethoxy-4-methoxy-
phenyl)-1-(4-hydroxy-phenyl)-propenone (3) was carried out
by the well-known ClaiseneSchmidt condensation reaction
and the product was purified by re-crystallization from meth-
anol. In the second step, condensation reaction between 3 and
nicotinic acid hydrazide (4) was carried out in methanol at
65 ꢁC. After the reaction, methanol was removed at reduced
pressure and ALH5 was purified by preparative HPLC (Knuer,
Germany).
4. Antimalarial evaluation
P. falciparum strains MRC-02 (CQ sensitive) and RKL9
(CQ resistant) were obtained from National Institute of Ma-
laria Research, New Delhi, India and maintained in a contin-
uous culture using the standard method described by Trager
and Jensen [23]. Parasites were cultured in human B (þ)
erythrocytes in RPMI-1640 media (GIBCO-BRL, Paisely,
Scotland) supplemented with 25 mM HEPES buffer, 10%
5. Determination of pKa
Acid dissociation constants (pKas) were determined spec-
trophotometrically as previously described [25] by using
a Unicam UV 300 spectrometer of Thermo Spectronic.
Scheme 1. Synthesis of nicotinic acid [trans-3-(4-ethoxy-3-methoxy phenyl)-1-(4-hydroxy-phenyl)allylidene]-hydrazide (ALH5). Reagents and conditions: (i)
NaOMe, MeOH, r.t., 24 h; (ii) methanol, stirred for 72 h at 65 ꢁC.

B.N. Acharya et al. / European Journal of Medicinal Chemistry 43 (2008) 2840e2852
2845
Absorbances of each compound (0.025 mg mlꢀ1) dissolved
7. Results and discussion
in buffers of different pH ranging from 5.8 to 11.2 as
well as in 0.1 M NaOH and 0.1 M HCl were determined.
Buffers from pH 5.8 to 8.0 were prepared from 0.1 M solu-
tions of monobasic potassium phosphate (KH2PO4) and di-
basic potassium phosphate (K2HPO4) and from pH 8.4 to
11.2 were prepared from 0.1 M glycine and 0.1 M NaOH.
Analytical wavelengths (212 nm, 240 nm, 270 nm and
340 nm) were chosen at which significant difference be-
tween the absorbances of molecules in 0.1 M HCl and
0.1 M NaOH was observed. To reduce error, pKa was deter-
mined in duplication at two different wavelengths for each
compound.
HypoGen develops quantitative pharmacophore hypothesis
in three steps. The first step is the constructive phase, in which
HypoGen identifies and stores all the possible pharmacophoric
features of the active molecules. Second step is the subtractive
phase, in which pharmacophoric features present in the inac-
tive molecules are identified and subtracted from the pharma-
cophores stored in the constructive phase. Third step is the
optimization phase in which small perturbations are applied
to those features, which survived in the subtractive phase.
But the quality of pharmacophoric models depends upon the
input parameters like the number and diversity (structure and
activity) of training set molecules and the input pharmaco-
phoric features, which HypoGen can’t decide. Hence some
preliminary experiments must be carried out with HypoGen
to optimize the number of training set molecules and pharma-
cophoric features as input parameters. Optimization was car-
ried out in three steps. In the first step HypoGen was used
for the selection of pharmacophoric features. In the second
step, number of training set molecules was optimized and in
the third step pharmacophoric features were optimized with
respect to design the best possible models in order of their
predictive ability.
6. b-Hematin inhibition activity
The quantitative BHIA (b-hematin inhibitory activity) as-
say is based on the differential solubility of hematin and b-
hematin in DMSO and NaOH solution, respectively
[26,27]. The method determines a 50% inhibitory concentra-
tion for b-hematin formation inhibition (BHIA50) of the
compound. In a microcentrifuge tube, 100 ml of 6.4 mM so-
lution of hematin freshly dissolved in 0.2 N NaOH solution,
50 ml of compound dissolved in ethanol (chloroquine diphos-
phate was dissolved in water), 200 ml of 3 M solution sodium
acetate trihydrate (C2H3NaO2$3H2O), 50 ml of glacial acetic
acid were incubated at 37 ꢁC for 24 h. Concentration of com-
pounds was varied from 6 mM to 0.3 mM range. Water and
ethanol were taken as controls for water soluble and ethanol
soluble compounds, respectively. Test was done in duplicates
for each dose of drug. After incubation, the tubes were
centrifuged for 15 min at 3300g. The supernatant was dis-
carded and the pellet was reconstituted in DMSO and again
centrifuged for 15 min at 3300g. The supernatant was dis-
carded and b-hematin was obtained as a pellet. Following
centrifugation to isolate DMSO insoluble b-hematin, the pel-
let was dissolved in 0.1 N NaOH and absorbance was re-
corded at 405 nm to calculate BHIA50 as previously
described [27].
7.1. Selection of pharmacophoric features
The training set molecules used for pharmacophore gener-
ation are highly diversified. By observing the structures of all
the molecules, it was seen that seven pharmacophoric feature
types such as hydrogen bond acceptor (HBA), hydrogen bond
donor (HBD), hydrophobic aliphatic (HYALI), hydrophobic
aromatic (HYAR), positive charge (PC), positive ionizable
(PI) and ring aromatic (RA) could effectively map critical
chemical features of all molecules in the training and test
sets. But HypoGen can consider only five features at a time
for hypothesis generation. Therefore eight pharmacophore
models (hypo 1 to hypo 8) were generated with different com-
bination of five input pharmacophoric features. The input and
output parameters like pharmacophore features, cost values,
Table 1
Input and output parameters of pharmacophore modelsa (hypo 1 to hypo 8) generated from 39 training set moleculesb
Model no. Input featuresc
Output featuresc
Total cost Fixed cost Null cost Conf. cost Weight cost rms
rd
hypo 1
hypo 2
hypo 3
hypo 4
hypo 5
hypo 6
hypo 7
hypo 8
a
HBA, RA, HYAR, HYALI, PI
HBA, HBD, HYAR, RA, HYALI HBD, HYALI, HYALI, RA 197.209
HBA, HYALI, PI, RA
196.210
148.319
149.209
148.407
148.561
147.457
146.914
148.319
145.277
240.362
240.362
240.362
240.362
240.362
240.362
240.362
240.362
16.013
16.903
16.101
16.255
15.151
14.635
16.013
12.971
1.244
1.301
1.580
1.422
1.824
1.911
1.417
1.128
1.565 0.750
1.571 0.747
1.549 0.756
1.605 0.734
1.574 0.746
1.513 0.769
1.382 0.811
1.573 0.746
HBD, PI, RA, HYALI, PC
HBA, RA, HYALI, HYAR, PI
HYAR, PC, RA, PI, HYALI
HBD, PI, RA, HYALI, PC
HBA, HBD, HYALI, PI, RA
HBA, HBD, HYAR, PI, RA
HBD, HYALI, PC, RA
HBA, HYALI, RA
HYALI, PI, RA, RA
HBD, RA HYALI, PI
HBD, HYALI, PI, RA
HBD, PI, RA
195.656
199.122
196.501
192.366
185.862
193.575
Null cost ¼ 240.362 for all models.
b
c
Molecules present in Fig. 1.
HBA, hydrogen bond acceptor; HYAR, hydrophobic aromatic; HYALI, hydrophobic aliphatic; PI, positive ionizable; RA, ring aromatic; HBD, hydrogen bond
donor; PC, positive charge.
d
Correlation coefficient. All costs are in bits.

2846
B.N. Acharya et al. / European Journal of Medicinal Chemistry 43 (2008) 2840e2852
root mean square deviations (rms) and correlation coefficients
(r) are listed in Table 1.
mean square deviations (rms) and correlation coefficients (r)
are listed in Table 2. Model hypo 12 showed r ¼ 0.900, which
is the best among the five hypotheses. The output pharmaco-
phoric features of hypo 9e13 were without HBA. Hence
hypo 14e18 were generated without giving HBA in the input
features. But no significant improvement in correlation was
observed. From this set of experiments it was observed that
the training set (Fig. 1 without C7, C11, C12, C19, C23 and
C33 molecules) can give a quantitative model with maximum
ability of antimalarial activity prediction.
The fixed cost values of the above eight hypotheses were
found close to each other because all the models were gener-
ated from same training set molecules. The weight cost and
configuration cost of all the molecules were within the permis-
sible limits. The pharmacophoric features returned in hypo 7
(Table 1) were HBD, HYALI, PI and RA with ‘rms’ and ‘r’
values of 1.382 and 0.811, respectively. For other models the
above two values were found lower than hypo 7. HypoGen re-
turned RA and PI features in all the models. HBD was re-
turned in all the five models (hypo 2, 3, 6e8) where it was
given as an input parameter. HYALI was returned in four
out of five models and HBA was returned in two out of four
models. HYAR and PC features were not returned in any of
the models. Hence further developments of pharmacophore
models were carried out with HBA, HBD, HYALI, PI and
RA as input features.
7.3. Optimization of pharmacophoric features
In the previous section, 10 models (hypo 9e18) were gen-
erated containing either three or four pharmacophoric features
(Table 2). The most predictive hypo 12 was a four featured
model. But five features were given to HypoGen in the input
parameters to generate all the hypotheses. Hence another
five models (hypo 19e23) were generated by giving four fea-
tures in input parameters to find out the pharmacophoric fea-
tures which can map the optimized training set molecules
effectively. The input and output parameters of hypo 19e23
are listed in Table 3. It was observed that when any of the fea-
tures from RA (hypo 22), PI (hypo 19) and HYALI (hypo 23)
was dropped, the correlation coefficient was found below 0.9.
The reason may be, without any of the above three features,
the model generated could not map the training set molecules
effectively and leading to wrong activity prediction. Hence
RA, PI and HYALI are the three most essential features. Be-
tween the rest two features HBA and HBD, the latter was pre-
ferred by HypoGen, which can be clearly observed in the case
of hypo 19. Hence HBD, HYALI, PI and RA are the four im-
portant pharmacophoric features, which can map the opti-
mized training set molecules (Fig. 1 without C7, C11, C12,
C19, C23 and C33 molecules) effectively. After optimization
hypo 21 was found showing best correlation (0.906) between
7.2. Optimization of training set
Correlation coefficient signifies the closeness of the pre-
dicted biological activity with the actual activity. The better
the prediction the closer is the ‘r’ value to 1. For hypo 7,
the ‘r’ value was 0.811. By studying the HypoGen output
file of hypo 7, 11 molecules (C1, C2, C7eC9, C11, C12,
C19, C23 and C33) in the training set were found out whose
predicted activities by hypo 7 were deviating >5 times from
their actual values. C1 and C2 were deviating 5e7.5 times
of their actual activities. C8 and C9 were found deviating be-
tween 7.5 and 10 times and others >10 times of their actual
activities. In order to improve the ‘r’ value, some of the mol-
ecules were dropped from the training set and five models
(hypo 9e13) were generated. The input pharmacophoric fea-
tures were kept HBA, HBD, HYALI, PI and RA which have
been optimized in the previous section. The dropped mole-
cules, output pharmacophoric features, cost values, root
Table 2
Input and output parameters of pharmacophore models (hypo 9 to hypo 18)
Model no. Molecules dropped from
the training seta
Output featuresb
Total cost Fixed cost Null cost Conf. cost Weight cost rms
rc
hypo 9
C1, C2
C8, C9
HBD, PI, RA
186.005
172.274
148.817
143.869
135.605
137.890
141.592
130.124
128.137
120.033
228.812
200.310
201.054
197.976
156.109
12.311
16.013
14.363
16.013
14.636
1.131
1.847
1.881
1.532
2.096
1.612 0.734
1.272 0.778
1.027 0.890
0.963 0.900
0.970 0.848
hypo 10
hypo 11
hypo 12
hypo 13
HBD, HYALI, PI, RA
HBD, HYALI, PI, RA
HBD, HYALI, PI, RA
C7, C11, C12, C19, C23
C7, C11, C12, C19, C23, C33
C1, C2, C7, C8, C9, C11, C12, HBD, HYALI, PI, RA
C19, C23, C33
hypo 14
hypo 15
hypo 16
hypo 17
hypo 18
C1, C2
C8, C9
HBD, PI, RA
HBD, HYALI, RA
HBD, HYALI, PI, RA
185.973
206.760
155.338
138.816
138.532
131.501
132.392
127.337
228.812
200.310
201.054
197.976
156.109
13.238
12.953
16.013
16.904
15.212
1.620
2.301
1.837
1.174
1.626
1.587 0.744
1.903 0.598
1.165 0.856
1.310 0.901
0.994 0.804
C7, C11, C12, C19, C23
C7, C11, C12, C19, C23, C33
HBD, HYALI, HYALI, PI 161.656
144.171
C1, C2, C7, C8, C9, C11, C12, HBD, HYALI, PI, RA
C19, C23, C33
a
Molecules present in Fig. 1.
HBD, hydrogen bond donor; HBA, hydrogen bond acceptor; PI, positive ionizable; RA, ring aromatic; HYALI, hydrophobic aliphatic.
Correlation coefficient. All costs are in bits.
b
c

B.N. Acharya et al. / European Journal of Medicinal Chemistry 43 (2008) 2840e2852
2847
Table 3
Input and output parameters of pharmacophore models (hypo 19 to hypo 22) with modified training seta
Model no.
Input featuresb
Output featuresb
Total cost
Fixed cost
Conf. cost
Weight cost
rms
rc
hypo 19
hypo 20
hypo 21
hypo 22
hypo 23
a
HBA, HBD, HYALI, RA
HBA, HYALI, PI, RA
HBD, HYALI, PI, RA
HBA, HBD, HYALI, PI
HBA, HBD, PI, RA
HBD, HYALI, RA, RA
HYALI, PI, RA, RA
HBD, HYALI, PI, RA
HBD, HYALI, HYALI, PI
HBD, PI, RA
142.002
135.541
142.255
148.497
144.246
123.175
123.175
126.760
128.227
124.198
11.051
11.051
14.636
16.103
12.073
1.392
1.133
1.129
1.130
1.238
1.067
0.865
0.968
1.108
1.099
0.885
0.826
0.906
0.875
0.878
Molecules present in Fig. 1 except C7, C11, C12, C19, C23 and C33.
b
c
HBD, hydrogen bond donor; HBA, hydrogen bond acceptor; PI, positive ionizable; RA, ring aromatic; HYALI, hydrophobic aliphatic.
Correlation coefficient. Null cost is 197.976 for hypo 19e23. All costs are in bits.
actual and predicted antimalarial activities of training set
molecules.
generated on the basis of 19 scrambled spreadsheets. The re-
sults of CatScramble run are shown in Table 5. The data
clearly show that none of the generated hypotheses after ran-
domization had a cost value better than hypo 21. This con-
firmed the statistical validity of the model.
7.4. Validation of pharmacophore model
Fig. 3 shows regression between actual and estimated activ-
ities of the 33 training set molecules. Table 4 shows the actual
and estimated activities of training set molecules in which 24
molecules out of 33 molecules were predicted correctly.
All the highly active molecules (IC50 < 100 nM) were pre-
dicted correctly. Two moderately active molecules
(101 nM < IC50 < 1000 nM) were predicted as inactive while
four inactive molecules were predicted as moderately active.
CatScramble was used to carry out Fisher’s randomization
test. The statistical validity of hypo 21 was tested at 95% con-
fidence limit. In return pharmacophore hypotheses were
7.5. Activity prediction of test set molecules
The purpose of the hypotheses is not just to predict the ac-
tivity of the training set molecules accurately but also to pre-
dict the activity of the test set molecules and classify them
correctly as active or inactive. The actual and predicted ac-
tivities of 44 test set molecules are given in Table 6. Most
of the highly active molecules (17 out of 18) were predicted
correctly. In the case of moderately active molecules 8 out of
14 were predicted correctly. T26, which is a moderately
Fig. 3. Correlation (r ¼ 0.906) of actual vs predicted activities by hypo 21 for the training set molecules.

2848
B.N. Acharya et al. / European Journal of Medicinal Chemistry 43 (2008) 2840e2852
Table 4
Actual and estimated activities of training set molecules calculated on the basis of hypo 21
a
Sl no
Code
Fit
IC50 (nM)
Actual
Activity scaleb
Actual
Error
References
Estimated
Estimated
1
C37
C38
C15
C36
C3
5.85
5.38
5.79
5.50
5.69
5.24
5.31
5.15
5.48
5.47
5.71
5.55
5.19
5.80
4.79
3.75
3.82
5.37
3.85
4.50
4.18
3.83
3.94
3.75
3.85
3.94
3.87
3.91
3.75
3.85
3.75
2.13
2.15
2.8
5.6
9.1
27
þþþ
þþþ
þþþ
þþþ
þþþ
þþþ
þþþ
þþþ
þþþ
þþþ
þþþ
þþþ
þþþ
þþþ
þþþ
þþ
þþþ
þþþ
þþþ
þþþ
þþþ
þþþ
þþþ
þþþ
þþþ
þþþ
þþþ
þþþ
þþþ
þþþ
þþþ
þ
þ3.3
þ4.8
þ1.2
þ2.2
þ1.4
þ2.9
þ2
[4]
2
[21]
[4]
3
9.1
9.4
11
21
4
[4]
5
9.5
13
37
[19]
[4]
6
C29
C2
13
16
7
31
45
[14]
[4]
8
C28
C1
18
25
þ2.4
ꢀ1.2
ꢀ2
9
21
22
[14]
[4]
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
a
C20
C25
C35
C30
C26
C34
C6
43
50
12
18
ꢀ4
[20]
[4]
51
69
ꢀ2.8
ꢀ1.6
ꢀ7.8
þ1.3
þ7.7
þ5.2
þ1.5
þ3.7
ꢀ1.2
þ1.5
þ2.4
þ1.7
þ2.6
þ1.1
ꢀ1.4
ꢀ1.4
ꢀ1.7
ꢀ2
42
10
[4]
80
81
[20]
[4]
100
1100
960
280
880
210
420
950
730
1100
900
740
870
780
1100
880
1100
48 000
45 000
150
190
190
240
250
280
400
420
440
800
1000
1200
1300
2300
4700
6300
100 000
100 000
[15]
[21]
[4]
C39
C32
C10
C16
C24
C4
þþ
þþ
þþ
þþ
þþ
þþ
[16]
[17]
[20]
[19]
[18]
[15]
[17]
[18]
[17]
[4]
þþ
þþ
þþ
þþ
þþ
þþ
C22
C5
þþ
þþ
þþ
þ
C17
C21
C18
C27
C31
C13
C14
C8
þþ
þþ
þ
þþ
þ
þþ
þ
þþ
þ
þ
[4]
þ
þþ
ꢀ5.4
ꢀ5.5
ꢀ2.1
ꢀ2.2
[16]
[16]
[15]
[15]
þ
þ
þ
þ
C9
þ
þ
Data for activities for chloroquine resistant (K1) strain of P. falciparum are from Refs. [4,14e21].
Activity scale: þþþ (0e100 nM, highly active), þþ (101e1000 nM, moderately active), þ (>1000 nM, poorly active).
b
active compound, was predicted highly active and rest five
moderately active molecules were predicted inactive. For in-
active molecules four were predicted correctly. Rest six mol-
ecules were predicted moderately active. Overall 29 out of
44 molecules (65%) of the test set were predicted correctly.
The crossover in predictability was found mostly between
moderately active and inactive molecules rather than be-
tween active to moderately active or active to inactive
molecules. Hence this model may be used as a tool for
new antimalarial discovery targeting haem detoxification
pathway.
pharmacophore mappings of ALH5 are shown in Fig. 4. Three
out of four features of hypo 21 were found fitting well with
ALH5. The only feature, which was not mapping was PI
and due to this ALH5 was predicted as moderately active.
7.7. Synthesis and antimalarial evaluation
According to our literature survey, no report has been ap-
peared dealing with the synthesis and antimalarial evaluation
of ALH5 or similar molecules. Recently Posner et al. have re-
ported on a novel artemisinin derivative [28] which contains
nicotinic acid hydrazide side chain like ALH5. Hence the
compound was synthesized and evaluated against malaria
parasite. Synthesis of ALH5 was carried out by condensing
a,b-unsaturated ketone with nicotinic acid hydrazide. Conden-
sation between a,b-unsaturated ketone and hydrazides are well
reported in literature [29]. In the presence of acid or base or
at higher temperature the reaction didn’t stop after condensa-
tion but proceeds for cyclization by formation of bond be-
tween nitrogen and b-carbon. Hence substituted pyrazole
7.6. Virtual screening and pharmacophore mapping
Pharmacophore model hypo 21 was used as a virtual
screening tool for our in-house compound library and com-
pound ALH5 (nicotinic acid [trans-3-(4-ethoxy-3-methoxy-
phenyl)-1-(4-hydroxy-phenyl)-allylidene]-hydrazide) was ob-
tained as a hit with estimated activity of 160 nM (moderately
active). The two-dimensional (2D) and three-dimensional (3D)

B.N. Acharya et al. / European Journal of Medicinal Chemistry 43 (2008) 2840e2852
2849
Table 5
Results from cross-validation run using CatScramble for hypo 21
CY5 showed characteristic double doublet (dd) pattern at
d values 3.14 ppm (H4a), 3.85 ppm (H4b) and 5.67 ppm (H5).
ALH5 did not show above pattern in the 1H NMR, which con-
firmed the non-cyclic structure. In deuterium exchange NMR,
two protons of ALH5 at d values 9.857 ppm and 11.366 ppm
were found exchanged. This observation further validated the
formation of ALH5.
ALH5 was evaluated against both CQ sensitive (MRC-02)
and CQ resistant (RKL9) strains of P. falciparum. The results
are shown in Table 7. Vacuolar accumulation ratios (VARs) of
ALH5 and chloroquine diphosphate (CQ) were theoretically
calculated from experimentally determined pKa values using
Eq. (1) [30]. In Eq. (1), VAR represents the ratio of drug inside
and outside of the parasite food vacuole. pHv ¼ pH inside food
vacuole (assumed to be pH 5.5) and pHe ¼ pH externally (as-
sumed to be pH 7.4).
Hypothesis Total
no.
Fixed
cost
Conf. cost Weight rms
cost
Correlation
coefficient (r)
cost
Trial 1
Trial 2
196.379 113.124 1.000
202.718 114.709 2.584
190.219 121.628 9.503
197.976 110.999 0.000
181.640 125.173 13.049
211.232 120.925 8.800
186.555 122.408 10.284
188.507 124.143 12.019
185.389 121.965 9.840
183.274 124.906 12.782
175.321 122.807 10.683
184.640 122.111 9.987
198.522 121.171 9.047
191.439 120.066 7.942
197.976 110.999 0.000
190.100 121.573 9.449
196.238 121.251 9.126
196.505 121.251 9.126
202.253 117.294 5.169
142.255 126.760 14.636
3.549
5.533
1.124
0.000
1.124
5.327
1.148
1.150
2.238
1.129
1.418
1.557
3.708
1.229
0.000
1.144
2.287
2.282
4.915
1.129
2.213 0.276
2.250 0.220
0.000 0.000
2.295 0.000
1.849 0.592
2.284 0.152
1.971 0.512
1.974 0.510
1.943 0.537
1.880 0.573
1.779 0.632
1.939 0.536
2.128 0.380
2.078 0.425
2.295 0.000
2.037 0.460
2.115 0.392
2.119 0.388
2.217 0.272
0.968 0.906
Trial 3
Trial 4
Trial 5
Trial 6
Trial 7
Trial 8
Trial 9
Trial 10
Trial 11
Trial 12
Trial 13
Trial 14
Trial 15
Trial 16
Trial 17
Trial 18
Trial 19
hypo 21
8
9
ꢀ
ꢀ
ꢁ
ꢀ
ꢁ
2
v
Hþ
Hþ
>
>
>
>
<
>
>
>
>
=
v
1 þ
1 þ
þ
þ
Qv
Qe
Ka2
Hþ
Ka1Ka2
Hþ
VAR ¼
¼
ð1Þ
ꢁ
ꢀ
ꢁ
2
>
>
>
>
>
:
>
>
>
;
Null cost ¼ 197.976. All costs are in bits.
e
e
Ka2
Ka1Ka2
[5-(4-ethoxy-3-methoxy-phenyl)-3-(4-hydroxy-phenyl)-4,5-
dihydro-pyrazol-1-yl]-pyridin-3-yl-methanone (CY5) was ob-
tained as product. To overcome this problem, condensation
was carried out at lower temperature (65 ꢁC) for a longer pe-
riod of time (72 h). The overall yield of ALH5 was 20% after
purification by preparative HPLC. ALH5 and CY5 are iso-
mers and were easily differentiated by 1H NMR spectroscopy.
ALH5 might have three pKa values. pKa1 and pKa2 due to
exchangeable protons of hydrazone moiety and phenolic hy-
droxyl group, respectively, have been confirmed in deuterium
exchange proton NMR. The third pKa for protonable nitrogen
of pyridine might have value below 5 [31]. Its contribution for
VAR would be negligible and hence not determined in the cur-
rent study. The two pKa values of CQ were determined
Table 6
Actual and estimated activities of test set molecules estimated by hypo 21
Code
Activity
Actual
Error
References
Code
Activity
Actual
Error
References
Estimated
Estimated
T1
T2
17
166
448
1210
419
950
63
34
1200
1200
1200
120
2
7
[14]
[15]
[15]
[15]
[15]
[15]
[15]
[15]
[15]
[15]
[15]
[16]
[16]
[16]
[16]
[16]
[16]
[17]
[17]
[17]
[17]
[17]
T23
T24
T25
T26
T27
T28
T29
T30
T31
T32
T33
T34
T35
T36
T37
T38
T39
T40
T41
T42
T43
T44
1100
1000
350
900
860
800
65
ꢀ1.2
ꢀ1.2
2.3
[17]
[17]
[17]
[17]
[17]
[18]
[18]
[18]
[19]
[4]
T3
T4
2.6
ꢀ1
ꢀ3.5
ꢀ1.1
1.6
2.8
3.9
2.6
2
620
34
ꢀ9.5
2.5
T5
T6
86
870
100
2000
7000
1800
16
590
4100
740
12
ꢀ3.4
ꢀ1.7
ꢀ2.4
ꢀ1.4
1.1
T7
T8
37
100
T9
260
45
1000
120
T10
T11
T12
T13
T14
T15
T16
T17
T18
T19
T20
T21
T22
38.8
166
42
45
49
100
ꢀ3.7
3.8
[4]
[4]
260
650
650
520
7180
7620
350
40
1000
1100
1200
1200
2000
880
3.9
1.8
1.8
2.2
ꢀ3.6
ꢀ8.6
2.6
2
11.5
9.1
44
39
4.3
[4]
[4]
1338
89.6
69.3
56.2
90.6
15 119
27
200
26
ꢀ6.7
ꢀ3.4
ꢀ2.1
1.4
[4]
[4]
33
76
[4]
[4]
900
80
87
1.1
7500
18
ꢀ2
[4]
30
50
25
75
ꢀ1.2
1.5
5.4
ꢀ1.5
ꢀ2.3
ꢀ5.3
[21]
[20]
[20]
1780
790
790
150
150
810
Activity scale: 0e100 nM, highly active; 101e1000 nM, moderately active; >1000 nM, poorly active.

2850
B.N. Acharya et al. / European Journal of Medicinal Chemistry 43 (2008) 2840e2852
Fig. 4. Pharmacophore mapping of ALH5 with hypo 21. (a) Two-dimensional mapping. (b) Three-dimensional mapping; the blue contour represents the HYALI
feature, the orange contour represents AR feature, the magenta contour represents HBD feature and the red contour represents PI feature. Distances between fea-
tures are in angstrom unit. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
experimentally and were found close to the previously re-
ported values [30]. The VAR and BHIA50 values of ALH5
were found lower than CQ. Due to this CQ was showing bet-
ter antimalarial activity against chloroquine sensitive MRC-02
strain. But in the case of resistant strain, the actual VAR of CQ
is much lower than the calculated value of Eq. (1). As RKL9 is
not resistant to ALH5, its VAR would be comparable with the
VAR of sensitive strain. Hence activity of ALH5 was found
better than CQ in the case of resistant strain RKL9. Activity
against RKL9 strain was found closer to the estimated value
by pharmacophore model hypo 21, which further validated
the practical applicability of the model to discover new anti-
malarial compound targeting haem detoxification pathway.
of training set molecules and pharmacophoric features were
optimized through a series of pharmacophore hypothesis gen-
eration. The statistically valid model was used as a virtual
screening tool to discover a novel antimalarial agent nicotinic
acid [trans-3-(4-ethoxy-3-methoxyphenyl)-1-(4-hydroxy-phe-
nyl)-allylidene]-hydrazide (ALH5). The molecule was synthe-
sized and evaluated against malarial parasite P. falciparum. It
showed better activity than chloroquine against CQ resistant
(RKL9) strain of P. falciparum. SAR studies of this class of
compounds are in progress and will be communicated
separately.
9. Experimental protocol
8. Conclusion
9.1. Chemistry
In this work, we have demonstrated the generation and ap-
plication of pharmacophore model (hypo 21) to discover po-
Nicotinic acid hydrazide, 2-hydroxy acetophenone and 3-
ethoxy-4-methoxy benzaldehyde were obtained from Sig-
maeAldrich. Mass spectra were acquired on a Micromass
Q-ToF high-resolution mass spectrometer equipped with elec-
trospray ionization (ESI) on Masslynx 4.0 data acquisition
tential antimalarial agent ALH5.
pharmacophore model was generated and validated targeting
A
quantitative
haem detoxification pathway of P. falciparum. The number
Table 7
In vitro antiplasmodial activities of nicotinic acid [trans-3-(4-ethoxy-3-methoxyphenyl)-1-(4-hydroxy-phenyl)-allylidene]-hydrazide (ALH5) and chloroquine
diphosphate (CQ)
e
BHIA50 (mM)
Compound
Predicted antimalarial
activitya (nM)
Activity vs MRC-02
strainb (nM)
Activity vs RKL9
strainc (nM)
pKa2, pKa1
VARd
ALH5
CQ
a
160
280
75 ꢃ 35
21 ꢃ 3
129 ꢃ 12
177 ꢃ 4
9.05, 6.87
9.61, 8.55
1472.04
5889.85
2.35 ꢃ 0.016
1.89 ꢃ 0.009
Predicted activity by pharmacophore model (hypo 21) against chloroquine resistant (K1) strain of P. falciparum, IC50, nM.
b
c
d
e
Antiplasmodial activity against chloroquine sensitive (MRC-02) strain of P. falciparum, IC50, nM ꢃ SD, results of two separate determinations.
Antiplasmodial activity against chloroquine resistant (RKL9) strain of P. falciparum, IC50, nM ꢃ SD, results of two separate determinations.
Vacuolar accumulation ratio (VAR) calculated using Eq. (1) and assuming vacuolar pH of 5.5 and external pH of 7.4.
b-Hematin inhibitory activity, BHIA50, mM ꢃ SD, results of two separate determinations.

B.N. Acharya et al. / European Journal of Medicinal Chemistry 43 (2008) 2840e2852
2851
system. ESI was used in þve ionization mode. Vibrational
spectra were recorded on a Bruker Tensor 27 FTIR spectrom-
eter. 1H NMR spectra were acquired on a Bruker DPX 400 FT
NMR spectrometer at 400 MHz. 13C NMR spectra were ac-
quired on Bruker AV 400 FT NMR at 100 MHz. All the
NMR spectra were acquired in DMSO-d6 at 27 ꢁC. To check
the purity, liquid chromatographyemass spectrometry (LCe
ESI-MS) was carried out for all the molecules with Waters-
Micromass LC-ESI-MS system equipped with Waters 1525 bi-
nary gradient pump and Micromass Q-ToF micromass spec-
trometer, ionized by þve ESI mode (capillary 3000 V,
sample cone 30 V, extraction cone 1 V). Separations were car-
ried out on Waters ExterraÔ C8 MS column (2.1 ꢂ 30 mm,
3.5 m) with binary linear gradient of 5% methanol in water
to 100% methanol in 20 min.
J ¼ 8.2 Hz, 13-H, 17-H), 8.23 (1H, d, J ¼ 6.6 Hz, 24-H),
8.72 (1H, s, 22-H), 9.03 (1H, s, 20-H), 10.03 (1H, s, OH).
Acknowledgement
We thank Dr. R. Vijayaraghavan, Director, DRDE, Gwalior
for his keen interest and encouragement in the present work.
We especially thank Dr. Shri Prakash and Dr. M.M Parida
for their helpful advises and critical comments, Mr. Avik Ma-
zumdar and Mr. Basant Lal for NMR analysis and Mr. A.K.
Srivastava for FTIR and ESI-MS analyses.
References
[1] J. Sachs, P. Malaney, Nature 415 (2002) 680e685.
[2] D.E. Goldberg, A.F.G. Slater, Parasitol. Today 8 (1992) 280e283.
[3] R.G. Ridley, W. Hofheinz, H. Matile, C. Jaquet, A. Dorn, R. Masciadri,
S. Jolidon, W.F. Richter, A. Guenzi, M.A. Girometta, H. Urwyler,
W. Huber, S. Thaithong, W. Peters, Antimicrob. Agents Chemother. 40
(1996) 1846e1854.
9.2. Procedure for synthesis of nicotinic acid [trans-3-
(4-ethoxy-3-methoxy-phenyl)-1-(4-hydroxy-phenyl)-
allylidene]-hydrazide (ALH5) and [5-(4-ethoxy-3-
methoxy-phenyl)-3-(4-hydroxy-phenyl)-4,5-dihydro-
pyrazol-1-yl]-pyridin-3-yl-methanone (CY5)
[4] S.R. Hawley, P.G. Bray, M. Mungthin, J.D. Atkinson, P.M. O’Neill,
S.A. Ward, Antimicrob. Agents Chemother. 42 (1998) 682e686.
[5] T.J. Egan, H.M. Marques, Coord. Chem. Rev. 190e192 (1999) 493e517.
[6] O.F. Guner, Pharmacophore Perception, Development and Use in Drug
Design, International University Line, La Jolla, CA, 2000.
[7] Y. Kurosawa, A. Dorn, M. Kitsuji-Shirane, H. Shimada, T. Satoh,
H. Matile, W. Hofheinz, R. Masciadri, M. Kansy, R.G. Ridley, Antimi-
crob. Agents Chemother. 44 (2000) 2638e2644.
To a solution of 10 mmol (1360 mg) of 4-hydroxy aceto-
phenone (1) in 10 ml methanol kept in an ice bath, 30 ml of
2 N methanolic NaOH solution was added followed by
10 mmol (1800 mg) of 4-ethoxy-3-methoxy benzaldehyde
(2) in 10 ml methanol. The reaction mixture was stirred at
30 ꢁC for 24 h and neutralized with hydrochloric acid. trans-
3-(4-Ethoxy-3-methoxy-phenyl)-1-(4-hydroxy-phenyl)-prope-
none (3) appeared as a light yellow precipitate was separated
by filtration and re-crystallized from methanol (2050 mg,
65% yield). Equal amount of 3 (4 mmol, 1192 mg) and 4 (nic-
otinic acid hydrazide, 4 mmol, 548 mg) was stirred in metha-
nol (10 ml) at 65 ꢁC for 72 h. Preparative HPLC was run to
separate and purify the products. ALH5 (353 mg, 20% yield),
yellow amorphous solid; mp: 94e96 ꢁC; LCeMS Rt (min):
10.73; ESI-MS (M þ H): C24H24N3O4, 418.1667 (calculated),
418.1785 (observed); IR (KBr, cmꢀ1): 3414 (OH), 3048 (NH),
[8] A. Palomer, J. Pascual, F. Cabre, M.L. Garcia, D.J. Mauleon, J. Med.
Chem. 43 (2000) 392e400.
[9] M. Chopra, A.K. Mishra, J. Chem. Inf. Model. 45 (2005) 1934e1942.
[10] M.O. Taha, J. Zhang, K. Yu, W. Zhu, H. Jiang, Bioorg. Med. Chem. Lett.
16 (2006) 3009e3014.
[11] M.-Y. Li, K.-C. Tsai, L. Xia, Bioorg. Med. Chem. Lett. 15 (2005) 657e
664.
[12] B.N. Acharya, M.P. Kaushik, Med. Chem. Res. 16 (2007) 213e229.
[13] N. Bharatham, K. Bharatham, K.W. Lee, J. Mol. Graphics Model. 25
(2007) 813e823.
[14] F. Ayad, L. Tilley, L.W. Deady, Bioorg. Med. Chem. Lett. 11 (2001)
2075e2077.
[15] O. Onyeibor, S.L. Croft, H.I. Dodson, M. Feiz-Haddad, H. Kendrick,
N.J. Millington, S. Parapini, R.M. Phillips, S. Seville, S.D. Shnyder,
D. Taramelli, C.W. Wright, J. Med. Chem. 48 (2005) 2701e2709.
[16] C.W. Wright, J. Addae-Kyereme, A.G. Breen, J.E. Brown, M.F. Cox,
S.L. Croft, Y. Gokcek, H. Kendrick, R.M. Phillips, P.L. Pollet, J. Med.
Chem. 44 (2000) 3187e3194.
1
1640 (C]O), 1601 (C]N), 1265 (CeN); H NMR (DMSO-
d6, 400 MHz) d/ppm: 1.31 (3H, t, J ¼ 6.84, CH3), 3.79 (3H, s,
OCH3), 4.02 (2H, q, J ¼ 6.8 Hz, OCH2), 6.75e7.49 (m, Ar,
10H), 8.24 (1H, m, 22-H), 8.71 (1H, s, 21-H), 9.04 (1H, s,
19-H), 9.93 (1H, s, OH), 11.39 (1H, s, CONH); 13C NMR
(DMSO-d6, 100 MHz) d/ppm: 14.72, 55.75, 63.79, 111.29,
112.57, 115.07, 115.15, 116.19, 117.58, 121.49, 123.43,
128.11, 128.89, 130.09, 130.59, 135.88, 140.08, 149.01,
149.38, 151.91, 157.19, 158.66, 158.90, 162.81. CY5
(750 mg, 43% yield), bright yellow crystalline solid; mp:
98e100 ꢁC; LCeMS Rt (min): 11.78; ESI-MS (M þ H):
C24H24N3O4, 417.1667 (calculated), 418.2295 (observed); IR
(KBr, cmꢀ1): 3066, 2977, 1595 (C]O), 1515, 1442, 1258,
1138, 1029, 836; 1H NMR (DMSO-d6, 400 MHz) d/ppm:
1.30 (3H, t, J ¼ 6.84, CH3), 3.14 (1H, dd, J ¼ 13.5, 4.3 Hz,
4-Ha), 3.75 (3H, s, OCH3), 3.85 (1H, m, 4-Hb), 3.96 (2H, q,
J ¼ 6.8 Hz, OCH2), 5.67 (1H, dd, J ¼ 11.2, 4.2 Hz, 5-H),
6.82 (2H, d, J ¼ 8.2 Hz, 14-H, 16-H), 6.78 (1H, m, 7-H),
6.89 (2H, m, 10-H, 11-H), 7.51 (1H, m, 23-H), 7.56 (2H, d,
[17] S. Auparakkitanon, W. Noonpakdee, R.K. Ralph, W.A. Denney,
P. Wilairat, Antimicrob. Agents Chemother. 47 (2003) 3708e3712.
[18] M. Kalkanidis, N. Klonis, L. Tilley, L.W. Deady, Biochem. Pharmacol.
63 (2002) 833e842.
[19] A. Dorn, J.P. Scovill, W.Y. Ellis, H. Matile, R.G. Ridley,
J.L. Vennerstrom, Am. J. Trop. Med. Hyg. 65 (2001) 19e20.
[20] J. Guillon, P. Grellier, M. Labaied, P. Sonnet, J.-M. Leger, R. Deprez-
Poulain, I. Forfar-Bares, P. Dallemagne, N. Lemaitre, F. Pehoureq,
J. Rochette, C. Sergheraert, C. Jarry, J. Med. Chem. 47 (2004) 1997e
2009.
[21] A.M.W. Stead, P.G. Bray, I.G. Edwards, H.P. Dekoning, B.C. Elford,
P.A. Stocks, S.A. Ward, Mol. Pharmacol. 59 (2001) 1298e1306.
[22] CATALYST 4.10, Accelrys Inc., San Diego, CA, 2005.<http://www.
accelrys.com/>.
[23] W. Trager, J.B. Jensen, Science 193 (1976) 673e675.
[24] C. Lambros, J.P. Vanderberg, J. Parasitol. 65 (1979) 418e420.
[25] A. Albert, E.P. Serjeant, The Determination of Ionization Constants,
Chapman and Hall Ltd., London, 1971, pp. 44e64.

2852
B.N. Acharya et al. / European Journal of Medicinal Chemistry 43 (2008) 2840e2852
[26] N. Basilico, E. Pagini, D. Monti, P. Olliaro, D.J. Taramelli, Antimicrob.
Agents Chemother. 42 (1998) 55e60.
[29] J.-T. Li, X.-H. Jhang, Z.-P. Lin, Beil. J. Org. Chem. 3 (2007) 13.
doi:10.1186/1860-5397-3-13.
[27] R. Baelamans, E. Deharo, G. Bourdy, V. Munoz, C. Quenevo,
M. Sauvain, H. Ginsburg, J. Ethnopharmacol. 73 (2000) 271e275.
[28] G.H. Posner, I.H. Paik, W. Chang, K. Borstnik, S. Sinishataj,
A.S. Rosenthal, T.A. Shapiro, J. Med. Chem. 50 (2007) 2516e2519.
[30] C.H. Kaschual, T.J. Egan, R. Hunter, N. Basilico, S. Parapini,
D. Taramelli, E. Pasini, D. Monti, J. Med. Chem. 45 (2002) 3531e3539.
[31] A. Albert, E.P. Serjeant, The Determination of Ionization Constants,
Chapman and Hall Ltd., London, 1971, pp. 94.