5-Nitro-2-furfuriliden derivatives as potential anti-Trypanosoma cruzi agents: Design, synthesis, bioactivity evaluation, cytotoxicity and exploratory data analysis

Article abstract of DOI:10.1016/j.bmc.2013.06.017

The anti-Trypanosoma cruzi activity of 5-nitro-2-furfuriliden derivatives as well as the cytotoxicity of these compounds on J774 macrophages cell line and FN1 human fibroblast cells were investigated in this study. The most active compounds of series I and II were 4-butyl-[N′-(5-nitrofuran-2-yl) methylene] benzidrazide (3g; IC₅₀ = 1.05 μM ± 0.07) and 3-acetyl-5-(4-butylphenyl)-2-(5-nitrofuran-2-yl)-2,3-dihydro,1,3,4-oxadiazole (4g; IC₅₀ = 8.27 μM ± 0.42), respectively. Also, compound 3g was more active than the standard drugs, benznidazole (IC₅₀ = 22.69 μM ± 1.96) and nifurtimox (IC₅₀ = 3.78 μM ± 0.10). Regarding the cytotoxicity assay, the 3g compound presented IC ₅₀ value of 28.05 μM (SI = 26.71) against J774 cells. For the FN1 fibroblast assay, 3g showed IC₅₀ value of 98 μM (SI = 93.33). On the other hand, compound 4g presented a cytotoxicity value on J774 cells higher than 400 μM (SI >48), and for the FN1 cells its IC₅₀ value was 186 μM (SI = 22.49). Moreover, an exploratory data analysis, which comprises hierarchical cluster (HCA) and principal component analysis (PCA), was carried out and the findings were complementary. The molecular properties that most influenced the compounds' grouping were C log P and total dipole moment, pointing out the need of a lipophilic/hydrophilic balance in the designing of novel potential anti-T. cruzi molecules.

Full text of DOI:10.1016/j.bmc.2013.06.017

Bioorganic & Medicinal Chemistry 21 (2013) 5395–5406
Contents lists available at SciVerse ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
5-Nitro-2-furfuriliden derivatives as potential anti-Trypanosoma
cruzi agents: Design, synthesis, bioactivity evaluation, cytotoxicity
and exploratory data analysis
a
b
Fanny Palace-Berl a, , Salomão Dória Jorge , Kerly Fernanda Mesquita Pasqualoto ,
Adilson Kleber Ferreira b,c, Durvanei Augusto Maria c, Rodrigo Rocha Zorzi a, Leandro de Sá Bortolozzo a,
José Ângelo Lauletta Lindoso d, Leoberto Costa Tavares a
⇑
a Department of Biochemical and Pharmaceutical Technology, Faculty of Pharmacy, University of São Paulo, Av. Prof. Lineu Prestes, 580, São Paulo, SP 05508-000, Brazil
b Biochemistry and Biophysics Laboratory, Butantan Institute, SP, Brazil
c Experimental Physiopathology, Faculty of Medicine, University of Sao Paulo, Sao Paulo, Brazil
d Laboratory of Seroepidemiology, LIM 38-HC FMUSP, Institute of Tropical Medicine and Institute of Infectology Emílio Ribas, SP, Brazil
a r t i c l e i n f o
a b s t r a c t
Article history:
The anti-Trypanosoma cruzi activity of 5-nitro-2-furfuriliden derivatives as well as the cytotoxicity of
these compounds on J774 macrophages cell line and FN1 human fibroblast cells were investigated in this
study. The most active compounds of series I and II were 4-butyl-[N0-(5-nitrofuran-2-yl) methylene] ben-
Received 25 March 2013
Revised 29 May 2013
Accepted 6 June 2013
Available online 15 June 2013
zidrazide (3g; IC50 = 1.05
dro,1,3,4-oxadiazole (4g; IC50 = 8.27
the standard drugs, benznidazole (IC50 = 22.69
ing the cytotoxicity assay, the 3g compound presented IC50 value of 28.05
cells. For the FN1 fibroblast assay, 3g showed IC50 value of 98 M (SI = 93.33). On the other hand, com-
pound 4g presented a cytotoxicity value on J774 cells higher than 400 M (SI >48), and for the FN1 cells
its IC50 value was 186 M (SI = 22.49). Moreover, an exploratory data analysis, which comprises hierar-
lM
0.07) and 3-acetyl-5-(4-butylphenyl)-2-(5-nitrofuran-2-yl)-2,3-dihy-
0.42), respectively. Also, compound 3g was more active than
1.96) and nifurtimox (IC50 = 3.78 0.10). Regard-
M (SI = 26.71) against J774
l
M
l
M
lM
Keywords:
l
Trypanosoma cruzi
5-Nitro-2-furfuriliden
Nifuroxazide
Azomethine derivatives
Oxadiazoline derivatives
Molecular modeling
l
l
l
chical cluster (HCA) and principal component analysis (PCA), was carried out and the findings were com-
plementary. The molecular properties that most influenced the compounds’ grouping were ClogP and
total dipole moment, pointing out the need of a lipophilic/hydrophilic balance in the designing of novel
potential anti-T. cruzi molecules.
Ó 2013 Elsevier Ltd. All rights reserved.
1. Introduction
expectative of cure decreases to 20–60%. Moreover, due to the side
effects associated with BZ and NFX, patients under treatment re-
Chagas disease (CD) is a chronic and debilitating parasitic infec-
tion caused by the intracellular protozoan Trypanosoma cruzi. CD,
also called American Trypanosomiasis, affects around 8–10 million
people worldwide, mainly in Latin America.1 However, as a result
of the influx of immigrants, for instance, CD is increasingly being
reported in several non-endemic areas of Europe and North Amer-
ica as well, becoming a serious global problem.1,2
Currently there are only two drugs available for the treatment
of CD, benznidazole (BZ) and nifurtimox (NFX) (Fig. 1A), which
are effective in the acute phase and initial chronic phase of the
infection, but not in its late stage.2–4 Chemotherapy can shorten
the acute phase of disease and may achieve the parasitological cure
in 50–80% cases. However, for the initial chronic phase the
quire a careful monitoring.2
Considering the global incidence of CD and the lack of novel
therapeutic alternatives, it is still necessary the search for new
anti-T. cruzi agents to improve the cure rate and reduce the adverse
effects on chagasic patients. Many classes of compounds are being
investigated as anti-trypanocidal, and among them are the nitro-
heterocyclic compounds.4,5 In this regard, our research group
focused the investigation on nifuroxazide (NF), 4-hydroxy-N0-(5-
nitro-2-furfuryliden) benzhydrazide (Fig. 1B), as a lead compound,
since it has shown quite promising biological responses by consid-
ering molecular modifications on its chemical structure.6,7 The NF
mechanism of action is not fully elucidated, but previous studies
have suggested that its antimicrobial activity would be related to
the reduction of the nitro group and formation of free radical toxic
species.8,9 Then, in this study, two sets (series I and II, Fig. 1C) of NF
analogues were designed envisaging the identification of the most
active compound against T. cruzi.
⇑
Corresponding author. Tel.: +55 11 30913693x38156386, mobile: +55 11
998673742.
E-mail address: palaceberlf@usp.br (F. Palace-Berl).
0968-0896/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.bmc.2013.06.017

5396
F. Palace-Berl et al. / Bioorg. Med. Chem. 21 (2013) 5395–5406
Furthermore, cytotoxicity assays on mammalian cells, using
J774 murine macrophages,14,15 and on FN1 human fibroblast
cells16 were performed to evaluate the investigated compounds.
Molecular modeling methods were also applied, herein, to se-
lect an energetically favourable conformation of each compound
and, then, calculate the molecular properties of different nature
(electronic, steric, hydrophobic, topological, geometric), which
were used to carry out the exploratory data analysis (hierarchical
cluster analysis, HCA, and principal components analysis,
PCA).17,18 Exploratory analysis allows the investigation of inters-
amples relationships through either similarity indices, or linear
combinations from the original data. The findings can be used to
establish some qualitative structure–activity relationships, since
the calculated molecular properties are directly related to the com-
pounds’ chemical structures.
2. Results and discussion
2.1. Chemistry
The two designed sets of 5-nitro-2-furfuriliden derivatives were
obtained as shown in Scheme 1. The substituent groups were se-
lected according to the influence of their physicochemical proper-
Figure 1. (A) Chemical structures of the drugs used in the CD treatment
(benznidazole and nifurtimox). (B) Chemical structure of lead compound (nifurox-
azide). (C) Chemical structures of the azomethine (series I) and 3-acetyl-2,5-
disubstituted-2,3-dihydro-1,3,4-oxadiazole derivatives (series II), pointing out the
regions where chemical modifications were carried out in the present study.
ties (electronic (r) and hydrophobic (p) character) employing the
Craig diagram.10
The benzhydrazides (2a–j) were obtained from benzoic acids
through consecutive reaction of esterification and hydrazinolysis,
without isolation of the corresponding methyl esters.19 Further-
more, the azomethine derivatives of series I (3a–j) were obtained
by nucleophilic addition of the derivatives 2a–j with 5-nitro-2-fur-
aldehyde diacetate.20 The 3-acetyl-2,5-disubstituted-2,3-dihydro-
1,3,4-oxadiazole derivatives of series II (4a–j) were obtained by
oxidative cyclization of azomethine derivatives (3a–j) with acetic
anhydride.6,11,21
The hydrazinolysis reactions provided satisfactory yields for all
benzhydrazides (values between 81% and 97%). The synthetic pro-
cedure was modified to obtain 4-nitrobenzhydrazide, 2f. The reac-
tions were performed in a controlled temperature (ꢀ3 to 2 °C) to
avoid the reduction of the compounds by the hydrazine hydrate,
which acts as a reducing agent.19,22
The synthesis of the azomethynic series (3a–j) presented yield
values between 89% and 99%, as already reported in previous stud-
ies.20 The identification of compounds 3a–j was carried out
through the comparison of the experimental melting point values
to those previously reported in the literature (see Supplementary
data section) as well as through the 1H and 13C NMR spectroscopy
and elemental analysis (CNH). The 3-acetyl-2,5-disubstituted-
1,3,4-oxadiazoline derivatives (4a–j) obtained yields between
The structure of NF was maintained to design the series I, and
molecular modifications were performed at position 4 of the ben-
zene ring, and consequently the phenolic hydroxyl group was re-
placed by different substituents. The same kind of substitution
was also performed to the compounds of series II. The substituents’
selection was based upon the Craig’s diagram, a rational method
where the physicochemical properties (hydrophobicity and elec-
tronic) are interrelated.10
The series II was planned primarily considering the cyclization
of the central portion of the compounds from series I (azomethynic
group) to form the 3-acetyl-2,5-disubstituted-2,3-dihydro-1,3,4-
oxadiazoline ring. Besides the improvement of pharmacokinetics
efficacy due to the ability to establish hydrogen bonding interac-
tions, the oxadiazoline ring can provide important structural fea-
tures to the molecules, as conformational changes, for instance.
Recently, several 3-acetyl-2,5-disubstituted-2,3-dihydro-1,3,4-
oxadiazoline ring systems have been associated with diverse bio-
logical activities, that is, antibacterial, antifungal,6,11 anti-T. cruzi,6
anticancer.12 They have also been reported as important ring sys-
tems for Parkinson disease regarding the inhibitory activity on
MAO-B.13
Scheme 1. Synthetic route for series I (3a–j) and II (4a–j). Reaction conditions: (I): CH3OH, H2SO4/reflux/4 h; N2H4 80% aq/reflux/30 min; (II) 5-nitro-2-furaldehyde diacetate,
H2O, H2SO4, CH3COOH, C2H5OH/reflux/1 h; (III) (CH3CO)2O/110–120 °C/12 h.

F. Palace-Berl et al. / Bioorg. Med. Chem. 21 (2013) 5395–5406
5397
Figure 2. 1H NMR spectra (DMSO-d6, 300 MHz) of compounds 3g and 4g. Chemical shifts (d) highlighting the differences between the azomethynic group and oxadiazolinic
ring: (A) compound 3g—H8 (12.17 ppm) and H6 (8.44 ppm); (B) compound 4g—H2 (7.37 ppm) and H22 (2.30 ppm). NMR attributions of compounds were done according to
the numbered chemical structure.
52% and 86%. During the synthesis procedure, thin layer chroma-
tography (TLC) analysis was performed to verify the formation of
3-acetyl-2,5-disubstituted-1,3,4-oxadiazolines.6
Figure 2B. The 1H NMR spectrum of compound 4g presented one
signal around d 7.4 ppm, evidencing the formation of oxadiazoline
ring, and another singlet signal near d 2.3 ppm, which confirms the
three protons of the acetyl group (see Fig. 2B; H2 and H22, respec-
tively). Analyzing the 13C NMR spectra of compounds from series I,
the signals were near 162 and 135 ppm, respectively, indicating
the presence of the azomethine group. The compounds of series
II displayed a characteristic peak around 21 ppm related to the
presence of the acetyl group in the oxadiazoline ring. Moreover,
elemental analysis (CNH) was carried out, and taken together with
1H and 13C NMR, the data were in agreement with the proposed
structures (see Supplementary data, p. S2 and S4).
The structural elucidation of these derivatives was confirmed
through 1H and 13C NMR spectra analysis, considering chemical
shifts (d) related to either the residual solvent peak, or the internal
standard. Compounds of the series I showed characteristic NMR
spectra, as illustrated in Figure 2A (1H NMR for the 3g compound).
The presence of two singlet signals were observed around d
12 ppm and d 8 ppm, respectively, indicating the protons of the
azomethine group (see Fig. 2A; H8 and H6). For the compounds
of series II, two characteristic peaks were found, as displayed in

5398
F. Palace-Berl et al. / Bioorg. Med. Chem. 21 (2013) 5395–5406
Table 1
T. cruzi (NF, IC50 = 120.46
lar modifications performed in the phenyl and azomethynic group
generated satisfactory results.
l
M
4.06), evidencing that the molecu-
Anti-T. cruzi activity of 5-nitro-2-furfuriliden derivatives, 3a–j and 4a–j, benznidazole,
nifurtimox and nifuroxazide
a
a
Compounds (R)
IC50
(
lM)
Compounds (R)
IC50 (lM)
2.3. Cytotoxicity assays
3a
3b
3c
3d
3e
3f
3g
3h
3i
H
4.44 0.48
3.71 0.14
12.51 0.94
3.94 0.36
4.03 0.23
9.45 0.54
1.05 0.07
4.42 0.14
3.74 0.15
3.21 0.19
4a
4b
4c
4d
4e
4f
4g
4h
4i
H
10.22 0.41
9.84 0.59
13.44 0.80
9.26 1.02
11.46 0.81
13.91 0.95
8.27 0.42
11.14 1.07
9.59 0.31
11.52 0.99
CH3
CN
OCH3
Cl
NO2
C4H9
CF3
Br
CH3
CN
OCH3
Cl
NO2
C4H9
CF3
Br
Compounds presenting high, moderate and low biological activ-
ity against epimastigote forms were elected for the cytotoxicity
tests using mice macrophages and human fibroblast cells, and
the findings are presented in Table 2.
Regarding series I, compounds 3g and 4g displayed higher
selectivity indices (SI). Also, the results pointed out the compounds
from series I as more cytotoxic than those from series II. The J774
mice macrophages were chosen as a model to obtain an unspecific
mammalian cytotoxicity, since the macrophages are host cells for
the parasite. Moreover, it was performed cytotoxicity assays using
FN1 human fibroblast cells to determine the in vitro toxicity of the
synthesized compounds against human normal cells (not infected).
Compounds 3e, 3g and 4f showed lower cytotoxicity for human
fibroblasts than for mice macrophages cells. Compound 3g also
presented an interesting SI value regarding both cell lines (SI for
J774 cells = 27.58; SI for FN1 cells = 93.33). The SI value for com-
pounds 3e, 4e and 4f was not exactly determined because the max-
imum concentration considered in the assay for those compounds
was not toxic.
3j
I
4j
I
Benznidazole
Nifurtimox
Nifuroxazide
22.69 1.96
3.59 0.46
120.46 4.06
IC50: half maximal inhibitory concentration.
a
Values corresponding to the average of triplicates.
It is noteworthy that, among the twenty compounds synthe-
sized and identified, herein, there are five new compounds, and
all from series II (4c, 4g–4j) (see 1H and 13C NMR spectra in Supple-
mentary data, p. S6).
2.2. Anti-T. cruzi activity assays
2.4. Molecular modeling approach, calculation and selection of
descriptors
The biological activity of the designed compounds was evalu-
ated through the percentage of growth inhibition of the epimasti-
gote forms of T. cruzi (Y strains), according to the experimental
protocols. Results were expressed as IC50 values, which represent
the concentration of compound necessary to inhibit fifty percent
of the parasite growth. Parasites in the stationary phase were incu-
bated, during 5 days, increasing the concentration of the investi-
gated compounds from series I and II, reference drugs (NFX and
BZN) and lead compound (NF). All assays were performed using
DMSO as control, at the same concentration employed to solubilise
the compounds and reference drugs, and which did not exceeded
1.0%.14,23 No effect on the epimastigotes’ growth was observed in
the culture media by the presence of up to 1.0% DMSO. The findings
are presented in Table 1 (see also Supplementary data, p. S16).
Accordingly, the most active compounds of each series were the
4-C4H9 (3g and 4g) substituted derivatives while the lowest activ-
ity was observed for the 4-CN (3c and 4c) and 4-NO2 (4f and 5f)
substituted derivatives. Comparing the two series, compounds of
series I were more active than those from series II. Even so, all
compounds showed higher activity in comparison to the reference
drug BZ. The compound 3g presented threefold higher activity than
The three-dimensional (3D) molecular models of compounds
were built up and energy-minimized as reported in Section 4.
The lowest-energy conformer of each reference template, NF (ser-
ies I) and compound 4a (series II), was selected from the equilib-
rium region of the conformational ensemble profile (CEP) from
molecular dynamics (MD) simulations (see Section 4.4 and Supple-
mentary data p. S18, Figs. 1S and 2S). The root mean square devi-
ation (RMSD) values between the atomic positions of the selected
conformers from CEP and of the energy-minimized structures were
computed as a criterion to verify whether the structural integrity
was maintained after simulations. Higher RMSD values indicate
loss of structural integrity during MD simulations if there is con-
sidered a crystallographic structure as template or reference.24
The RMSD value found between the lowest-energy conformer
and the initial energy-minimized structure was 0.07 and 0.14 Å
for NF and compound 4a, respectively. The higher RMSD value
for compound 4a was already expected since the molecular model
of this compound was built up using different crystallographic
fragments not an entire experimental structure (X-ray diffraction
or RMN). However, regarding the main portion of molecular struc-
ture (5-nitrofuran moiety), for instance, the RMSD value was
NFX (3g, IC50 = 1.05
l
M
0.07; NFX, IC50 = 3.78
lM
0.10). The
lead compound, NF, did not show significant activity against
Table 2
Cytotoxicity of 5-nitro-2-furfuriliden derivatives (3e, 3f, 3g, 4e, 4f and 4g), benznidazole and nifurtimox against J774 mice macrophages, FN1 human fibroblast cells and their
respective selectivity index
a
a
a
Compounds (R)
T. cruzi IC50
(
lM)
J774 cells IC50
(
lM)
SIb
FN1 cells IC50
>200
(lM)
SIb
3e
3f
3g
4e
4f
4g
Cl
4.03 0.23
9.45 0.54
1.05 0.07
11.46 0.81
13.91 0.95
8.27 0.43
22.69 1.96
3.78 0.10
57.11 1.29
79.09 7.89
28.96 1.05
>400
114.27 2.05
>400
14.17
8.37
27.58
>34
8.21
>48
>49.63
18.83
93.33
>17
>14
22.49
>9
NO2
C4H9
Cl
NO2
C4H9
17,800 11.93
9800 4.72
>200
>200
18,600 9.50
>200
Benznidazole
Nifurtimox
>400
280.48 2.44
>18
74.2
>200
>53
IC50: half maximal inhibitory concentration.
a
Values corresponding to the average of triplicates.
SI—selectivity index (SI = IC50 of cells line/IC50 of T. cruzi).
b

F. Palace-Berl et al. / Bioorg. Med. Chem. 21 (2013) 5395–5406
5399
0.07 Å, reinforcing the structural integrity at the pharmacophore
group region. Additionally, the selected conformer for NF and for
compound 4a presented total potential energy value of 19.90 and
30.86 kcal/mol, respectively, considering the intramolecular en-
ergy contribution of solvation (Esolv) and of hydrogen bonding
(EHb) (see Supplementary data, p. S18, Table 1S).
data (see equation in 4.5.1 Hierarchical cluster analysis sec-
tion).17,18,35 Herein, HCA analysis was performed for (i) a matrix
considering the biological activity as dependent variable allowing
the grouping of compounds or samples (see Fig. 3), and (ii) another
matrix considering the biological activity also as independent var-
iable to verify intervariables relationships (see Fig. 4).
Then, forty descriptors or molecular properties of different nat-
ure were calculated (thermodynamic, hydrophobic, electronic,
topological, geometric, and steric) for each compound. A matrix
composed of 21 rows (samples or compounds plus NF) and 41 col-
umns (independent variables or descriptors or molecular proper-
ties plus biological activity) was generated and used as input for
the exploratory data analysis (HCA and PCA). All calculated
descriptors and their methods are shown in Supplementary data
(p. S19, Table 2S; p. S21, Table 3S).
The biological data (anti-T. cruzi activity) was transformed into
potency (log1/IC50 or—logIC50 or pIC50) and corresponds to the
dependent variable. Compounds of the two series were distributed
according to a gradient of biological activity as follows: less, mod-
erately, and highly active compounds. It was considered each ser-
ies separately to emphasize the substituent groups’ contribution
during the analysis. For series II, 4c and 4f were considered as less
active compounds (pIC50 values of 4.86 and 4.87); 4a, 4b, 4d, 4e,
4g–j as moderately active compounds (pIC50 values from 4.87 to
5.03), and 4g was the most active compound of series II (pIC50 va-
lue of 5.08). Similar distribution of biological activity was observed
for series I, where 3c and 3f were considered less active com-
pounds (pIC50 values of 4.90 and 5.02); 3a, 3b,3d, 3e, 3g–j as mod-
erately active compounds (pIC50 values from 5.35 to 5.49); and 3g
as the most active compound (pIC50 value of 5.98).
Regarding the dendrogram of samples (Fig. 3), the compounds
were split into three main clusters: A (45.2% similarity), B (52.0%
similarity), and C (53.9% similarity), using a similarity cursor of
0.4 (dashed line) as reference. Clusters A and B share 35% of simi-
larity whereas cluster C has no similarity with those two.
The compounds were grouped, in general, accordingly to their
substituent groups and, then, subdivided in each series (see cluster
A and B). Considering the gradient of biological activity [less (1),
moderately (2) and highly (3) active compounds of each series],
cluster A is contained of moderately active compounds from series
II and I. Also, it presents more lipophilic compounds and has two
subclusters, regarding each series: A’ (4i -Br, 4j -I, 4e -Cl, 4b –
CH3, and 4h –CF3) and A’’ (3j -I, 3i -Br, 3e -Cl, 3h –CF3, 3b –CH3,
and 3d –OCH3), which share 71.7% and 59.1% of similarity, respec-
tively. The values of Hansch constant (p) for the substituent groups
in subclusters A’ and A’’ range from 0.56 to 1.12 and from -0.02 to
1.12, respectively, evidencing a more hydrophobic contribution to
those compounds. Cluster B have grouped moderately and less ac-
tive compounds and it was subdivided into subclusters B’ (4f –NO2,
4d –OCH3, 4c –CN, and 4a -H) and B’’ (3f –NO2, 3c –CN, 3a –H, and
NF), which displayed similarity indices of 0.58 and 0.64, respec-
tively. The NO2 and CN substituent groups present more negative
values for the Hansch constant,
p equal to -0.28 and -0.57, meaning
a more hydrophilic contribution. The lead compound, NF, was put
together into cluster B’’ along with the non-substituted moderately
active compound, 3a, probably due to the structural resemble with
compounds from series I as well as the character more hydrophilic
Before proceeding the exploratory data analysis, a preliminary
variable selection was carried out using as criteria (i) the Pearson’s
linear correlation coefficient between descriptors and inhibitory
activities (cut off value R = 0.2), and (ii) a visual inspection of scat-
ter plots for the biological activity (pIC50 values) versus each calcu-
lated molecular property or descriptor.17 These findings are
presented in Table 4S (Supplementary data, p. S23). Ten calculated
descriptors showed uniform distribution/dispersion with linear
tendency, and they were: the intramolecular energy of solvation
(Esolv); the electrostatic potential atomic charge (ESP or CHELPG)
of nitrogen in the N@C bond for series I (N7) and nitrogen in
oxadiazolinic ring for series II (N4) (CHELPG_N7,4), using the HF/
of its substituent group (-OH,
pounds [3g and 4g, n-butyl as substituent group,
p
= -0.67). The most active com-
= 2.13], regard-
p
ing the both series, were grouped into cluster C and share a
similarity index of 0.54.
The dendrogram of variables or descriptors (see Fig. 4) showed
that the electronic molecular properties,
more close to the biological activity and grouped in the same clus-
ter. The cluster shares 37.0% of similarity, and CHELPG_N7,4 and
display a similarity index of 0.74. Dipole moment, , is a global
l and CHELPG_N7,4, are
l
l
3-21G⁄ method;25,26 total dipole moment (
l); the solvent accessi-
molecular property, which is related to the molecule’s polarity.
ble surface area of all hydrophobic atoms (ASA_h);27 the STERIMOL
parameter L defined as the length of a substituent along the axis of
the bond between the first atom of the substituent and the parent
molecule;28 the calculated n-octanol/water partition coefficient
applying the Ghose et al., 1998-SLN method29 (ClogPSG-SLN), the
weighted method using Viswanadhan et al.,1989,30 Klopman
et al., 199431 and PHYSPROPÓ databases32 (ClogPWM), and the
Viswanadhan et al., 198930 method (ClogPSV); molar refractivity
Based upon this property, the compounds of series I are more polar
(hydrophilic) than those from series II. The average value of
l
found for series I (including NF) was 12.68 whereas for series II
was 6.75. ESP or CHELPG of nitrogen in the N=C bond for series I
(N7) ranged from -0.3297 to -0.3932 whereas for the corresponding
nitrogen in series II (N4 in the oxadiazolinic ring) varied from -
0.5099 to -0.5971.
The unsupervised principal components analysis (PCA) aimed
to correlate the descriptors to find other uncorrelated variables,
the principal components (PCs), which can describe the informa-
tion from the original data set.17,18
The PCA findings are showed in Fig. 5. According to the factors
selection, the two first factors or PCs discriminated more than sev-
enty percent (76.68%) of total variance from the original data. Also,
regarding the scores’ plot (Fig. 5), complementary findings were
obtained for the both methods, PCA and HCA, since the compounds
were split in the same previous groups (A’, blue rectangle; A’’, cyan
rectangle; B’, light pink square; B’’, dark pink square; and C, yellow
oval circle). Interestingly, PC2 or factor 2 (26.37%) seems to be
responsible for split the compounds into the two series, I and II
(horizontal gray line in the scores’ plot). Additionally, the most ac-
tive compounds of each series, 3g and 4g, were a little bit more dis-
tant from the rest of data set and from one another, evidencing the
(MR);33 and the hydrophobic Hansch constant ( ).34 The final ma-
p
trix was used as input for the exploratory data analysis. The data
matrix was autoscaled because the calculated independent vari-
ables presented different magnitude orders.
2.5. Exploratory data analysis
As already mentioned, exploratory data analysis is a multivari-
ate method of classification, which comprises hierarchical cluster
analysis (HCA) and principal components analysis (PCA). HCA is
an unsupervised pattern recognition analysis, where the distance
between similar samples is smaller than for unrelated samples,
and classifies the results according to similarity indices that range
from zero to one. The similarity matrix generates a dendrogram,
which is a tree-diagram constructed from the Euclidean distances

5400
F. Palace-Berl et al. / Bioorg. Med. Chem. 21 (2013) 5395–5406
Figure 3. HCA of samples: considering a similarity cursor of 0.4 (dashed line), the compounds were grouped into three main clusters: A (moderately active compounds; 45.2%
similarity), B (less and moderately active compounds; 52.0% similarity), and C (highly active compounds (53.9% similarity); n-butyl substituent group). Numbers 1, 2, or 3
were added after the name of each compound to define its gradient of biological activity as less, moderately, or highly active, respectively.
Figure 4. HCA of variables: the electronic descriptors grouped in the same cluster of biological activity were CHELPG_N7,4 and l. They are highlighted in a light gray dashed
rectangle.
two series classification. Regarding the HCA of samples, these two
compounds share 54% of similarity.
The loadings’ table, also presented in Fig. 5, provides informa-
tion about which descriptors or molecular properties most influ-
enced the separation procedure of samples/compounds. In PC1,
or factor 1 (50.31%), hydrophobic (ClogPSG-SLN, ClogPSV, ClogPWM
), steric/hydrophobic (MR), steric/geometric (ASA_h) and steric
(L) molecular properties displayed higher loading values (absolute
,
p

F. Palace-Berl et al. / Bioorg. Med. Chem. 21 (2013) 5395–5406
5401
Figure 5. PCA findings: factor selection, scores’ plot and loadings’ table. In the scores’ plot, the clusters observed in HCA are highlighted as blue rectangle (A0), cyan rectangle
(A00), light pink square (B0), dark pink square (B00), and yellow oval circle (C).
Table 3
change the molecular properties of a compound, and consequently,
Calculated values found for the two molecular properties that displayed the highest
loading value in each PC or factor
can influence its activity.36 This kind of influence was observed in
studies based on the physicochemical properties of NFX analogues
Compounds
R
ClogPWM (PC1)
l (Debye) (PC2)
aiming to explore substituent groups in the opposite side of the ni-
tro-furan group. The property was directly related to the anti-T.
NF
OH
H
CH3
CN
OCH3
Cl
NO2
C4H9
CF3
Br
1.75
2.05
2.57
1.91
1.89
2.66
1.99
3.90
2.93
2.82
2.98
2.36
2.87
2.21
2.20
2.96
2.30
4.21
3.24
3.13
3.29
13.19
14.52
15.06
10.99
13.77
12.76
9.80
15.15
11.07
11.46
11.74
5.57
6.72
7.22
6.20
6.40
9.34
6.75
6.61
6.30
l
3a-2
3b-2
3c-1
3d-2
3e-2
3f-1
3g-3
3h-2
3i-2
3j-2
4a-2
4b-2
4c-1
4d-2
4e-2
4f-1
4g-3
4h-2
4i-2
4j-2
cruzi activity, meaning more reactive molecules had larger values
of total dipole moment.37 As we know, hydrophobic properties,
such as ClogP, are quite important to the diffusion of compounds
through the biological system. Additionally, the environment of a
binding site in enzymes or receptors is generally hydrophobic
and the partition complementarity with the respective ligand is
crucial for the biological response. In this regard, the compounds
of series I and II with higher ClogP values presented also higher
activity values. Therefore, the hydrophobic/hydrophilic balance
and electronic density distribution were pointed out as important
molecular properties for the biological activity in some studies
with similar compounds, which had the mechanism of action by
reduction.37–39
The outliers’ diagnosis, implemented in Pirouette 3.11 soft-
ware,40 was also performed through the Mahalanobis distance,41
as can be seen in Fig. 6. The sample residual threshold was based
upon a ninety-five percent of confidence level interval set inter-
nally in Pirouette 3.11.40 The compounds did not exceed the con-
sidered threshold, meaning the calculated properties were
sufficient to describe the structural features of the both data sets
(see Fig. 6).
The lipophilic and electrostatic potential maps (MLP and MEP)
were calculated onto the molecular surface of the most active com-
pound of each set to visualize the molecular properties responsible
for the compounds’ discrimination, and are shown in Fig. 7.
MLP (Fig. 7A) and MEP (Fig. 7C) can be interpreted through a
color range scheme, which varies depending on the software used
for calculation. It is well-known that molecular surfaces can trans-
late the shape of any molecular system, and the most active com-
pound from series II, 4g, is shorter and more bent than the most
active compound from series I, 3g. Furthermore, the total dipole
moment vector was also displayed to the most active compounds
(blue arrows in Fig. 7B).
The MLP property gives information of the molecular hydropho-
bic/hydrophilic balance, where the blue color corresponds to
hydrophilic regions and the brown color to hydrophobic regions.42
The lipophilicity can be numerically expressed by the calculated n-
octanol/water partition coefficient (ClogP), applying Ghose et al.,—
SLN, Sybyl Line Notation.29,42 The brown color is mainly concen-
trated onto the substituent group region (–C4H9) for the both
I
H
CH3
CN
OCH3
Cl
NO2
C4H9
CF3
Br
I
6.41
The calculated partition coefficient (n-octanol/water) ClogPWM was obtained
applying the weighted method using Viswanadhan et al., 1989, Klopman et al., 1994
and PHYSPROPÓ databases.30-32
The values of total dipole moment (
l
) were calculated employing the HF/3.21G⁄
method.25,26
Numbers 1, 2, or 3 were added after the name of each compound to define its
gradient of biological activity as less (1), moderately (2), or highly active (3),
respectively. Clog P and dipole moment of the most active 5-nitro-2-furfuriliden
derivatives are in bold.
values). ClogPWM and p, for instance, presented the highest loading
values, 0.42 and 0.40, respectively, reinforcing the discrimination
pattern obtained, which was closely the same observed in HCA.
In PC2,
l and CHEPLG_N7,4, which are electronic molecular proper-
ties, showed the highest loading values, 0.56 and 0.55, respectively.
Regarding the investigated compounds, the calculated values for
the two molecular properties presenting the highest loading values
in each PC, or factor, are shown in Table 3.
It is noteworthy that the ClogP and
active compound of each set, 3g and 4g, could be considered as the
optimum values to the anti-T. cruzi activity for these investigated
sets, for instance. The
atoms with different electronegativity are interconnected. Consid-
ering that, substituent groups with different electronegativity can
l values found for the most
l measures the delocalized charge when

5402
F. Palace-Berl et al. / Bioorg. Med. Chem. 21 (2013) 5395–5406
Figure 6. Plot of sample residual versus Mahalanobis distance. The sample residual threshold (red line) is based upon a 95% probability limit (set internally in Pirouette
program).
compounds, 3g and 4g (see Fig. 7A). Regarding the MEP property,
regions with higher electronic density distribution are displayed
in the red color while regions with lower electronic density distri-
bution are in blue.43,44 The nearest neighbor moiety to the –C4H9
substituent group is less negatively charged (green/light blue col-
or) in compound 4g than in compound 3g (yellow/green color).
Thus, the findings indicate that the hydrophobic/hydrophilic bal-
ance and electronic density distribution seem to play an important
role for the anti-T. cruzi activity.
(0.01 mol) of sulphuric acid. The reaction mixture was cooled
down to room temperature and the hydrazine hydrate 80% (v/v)
(10.0 mL, 0.13 mol) was added. The system was maintained by vig-
orously stirring for more 30 min in reflux. In the case of compound
with 4-nitro substituent group attached to the benzene moiety,
after the addition of hydrazine hydrate 80% (v/v) at room temper-
ature, the reaction mixture was cooled down in ice bath and main-
tained into stirring during 1 h. After this period, the mixture was
maintained at low temperature to give 2.19,22
4.1.2. General procedure for the synthesis of azomethine
derivatives (3a–j)
3. Conclusion
Compounds of series I were synthesized by refluxing 5-nitro-2-
furaldehyde diacetate 98% (5 mmol) and benzhydrazides (3)
(5 mmol) in water, sulphuric acid, acetic acid, and methanol
(8:7:8:20 v/v) for 1 h. After cooling, the mixture was poured into
cold water to precipitate the azomethine derivatives20 (see struc-
tural elucidation of the compounds of series I in Supplementary
data, p. S2).
The findings indicated that the anti-T. cruzi activity of the inves-
tigated compounds regarding the two series is strongly influenced
by the physicochemical properties of the substituent group at-
tached to the benzene ring.
The cytotoxicity assay shown that some compounds have pre-
sented interesting selectivity indices, highlighting especially com-
pound 3g, which had SI value between parasites and cell lines of
27.58 for J774 cells, and of 93.33 for FN1 cells.
The complementary findings generated in the HCA and PCA ap-
proach suggested primarily the hydrophobic and electronic prop-
erties as the responsible for the compounds grouping and,
consequently, for the anti-T. cruzi activity of the 5-nitro-2-furfuril-
iden derivatives.
4.1.3. General procedure for the synthesis of 1,3,4-oxadiazoline
derivatives (4a–j)
Compounds of series II (1 mmol) were dissolved in acetic anhy-
dride (10 mL, 0.10 mol), and the mixture was heated between 110
and 120 °C for 12 h under stirring and nitrogen gas atmosphere.
After this period, 10 mL of distilled water was added to the system
under constant stirring until its precipitate6,21,43 (see structural
elucidation of the compounds of series II in Supplementary data,
p. S4). Identification of five new chemical entities, 4c and 4g–4j,
which belong to series II, are described as follows (see numbered
structure in Fig. 2 and 1H and 13C NMR spectra in Supplementary
data, p. S6).
4. Experimental sections
4.1. Chemistry
NMR spectra were recorded on a Bruker ADPX Advanced
(300 MHz) spectrometer employing DMSO-d6 solutions with tetra-
methylsilane as internal standard. Melting points were determined
using Büchi M-560 apparatus and elemental analysis was per-
formed on a Perkin–Elmer 24013 CHN Elemental Analyzer.
4.1.3.1. 3-Acetyl-5-(4-ciano-phenyl)-2-(5-nitrofuran-2-yl)-2,3-
dihydro-1,3,4-oxadiazole (4c).
Light yellow solid (55%); mp
177.0–178.0 °C. 1H NMR (DMSO-d6, 300 MHz): d (ppm): 7.99 (d,
2H, H15, H19, J = 8.7 Hz), 7.96 (d, 2H, H16, H18, J = 8.7 Hz), 7.65
(d, 1H, H8, J = 3.8 Hz), 7.41 (s, 1H, H2), 7.20 (d, 1H, H7, J = 3.8 Hz),
2.32 (s,3H, H22); 13C NMR {H} (DMSO-d6, 75 MHz): d (ppm):
167.4 (C20), 153.3 (C6), 151.8 (C5), 150.3 (C9), 132.9 (C16, C18),
4.1.1. General procedure for the synthesis of benzhydrazides
(2a–j)
Each substituted benzoic acid (1) (0.02 mol) was refluxed for
4 h in 20.0 mL (0.49 mol) of anhydrous methanol and 0.5 mL

F. Palace-Berl et al. / Bioorg. Med. Chem. 21 (2013) 5395–5406
5403
114.6 (C7), 113.1 (C8), 84.2 (C2), 34,7 (CH2), 32,6 (CH2), 21,6
(CH2), 21,0 (C22), 13,6 (CH3); Anal. Calcd for (C18H19N3O5): C,
60.50; H, 5.36; N, 11.76. Found: C, 60.11; H, 5.47; N, 11.58.
4.1.3.3. 3-Acetyl-5-(4-trifluoromethyl-phenyl)-2-(5-nitrofuran-
2-yl)-2,3-dihydro-1,3,4-oxadiazole (4h).
Light yellow solid
(86%); mp 166.0–167.0 °C. 1H NMR (DMSO-d6, 300 MHz):
d
(ppm): 8.06 (d, 2H, H15, H19, J = 8.1 Hz), 7.92 (d, 2H, H16, H18,
J = 8.4 Hz), 7.73 (d, 1H, H8, J = 3.8 Hz), 7.45 (s, 1H, H2), 7.28 (d,
1H, H7, J = 3.8 Hz), 2.33 (s,3H, H22); 13C NMR {H} (DMSO-d6,
75 MHz): d (ppm): 172.6 (C20), 158.6 (C6), 157.0 (C5), 155.5
(C9), 137.1 (C14), 133.7 (C17), 132.6 (C15, C19), 131,3 (C16),
131,4 (C18), 127.0 (CF3), 119.9 (C7), 118.3 (C8), 90.2 (C2), 26.3
(C22); Anal. Calcd for (C15H10F3N3O5): C, 48.79; H, 2.73; N, 11.38.
Found: C, 48.80; H, 2.66; N, 11.25.
4.1.3.4. 3-Acetyl-5-(4-bromo-phenyl)-2-(5-nitrofuran-2-yl)-2,3-
dihydro-1,3,4-oxadiazole (4i).
Light yellow solid (67%); mp
160.0–161.0 °C. 1H NMR (DMSO-d6, 300 MHz): d (ppm): 7.76 (d,
2H, H15, H19), 7.76 (d, 2H, H16, H18), 7.70 (d, 1H, H8, J = 3.7 Hz),
7.38 (s, 1H, H2), 7.23 (d, 1H, H7, J = 3.7 Hz), 2.28 (s,3H, H22); 13C
NMR {H} (DMSO-d6, 75 MHz): d (ppm): 167.2 (C20), 153.8 (C6),
151.8 (C5), 150.3 (C9), 132.3 (C16, C18), 128.5 (C15, C19), 125,8
(C14), 122.6 (C17), 114.2 (C7), 112.9 (C8), 84.7 (C2), 21.1 (C22);
Anal. Calcd for (C14H10BrN3O5): C, 44.23; H, 2.65; N, 11.05. Found:
C, 44.13; H, 2.55; N, 10.73.
4.1.3.5.
3-Acetyl-5-(4-iodo-phenyl)-2-(5-nitrofuran-2-yl)-2,3-
Light yellow solid (78%); mp
dihydro-1,3,4-oxadiazole (4j).
175.0–176.0 °C. 1H NMR (DMSO-d6, 300 MHz): d (ppm): 7.91 (d,
2H, H16, H18, J = 8.2 Hz), 7.70 (d, 1H, H8, J = 3.8 Hz), 7.58 (d, 2H,
H15, H19, J = 8.2 Hz), 7.38 (s, 1H, H2), 7.23 (d, 1H, H7, J = 3.8 Hz),
2.29 (s,3H, H22); 13C NMR {H} (DMSO-d6, 75 MHz): d (ppm):
167.2 (C20), 154.1 (C6), 151.8 (C5), 150.4 (C9), 138.0 (C16, C18),
128.2 (C15, C19), 122.8 (C14), 114.5 (C7), 113.1 (C8), 99.7 (C17),
84.6 (C2), 21.1 (C22); Anal. Calcd for (C14H10IN3O5): C, 39.37; H,
2.36; N, 9.84. Found: C, 39.47; H, 2.29; N, 9.68.
4.2. Anti-T. cruzi activity assay
In vitro anti-T. cruzi activity assays were performed against the
epimastigote forms of Y strain of T. cruzi isolated by Silva and Nus-
senzweig.45 The epimastigote form, at the stationary phase, and
from a culture incubated for 10 days at 28 °C, was diluted into liver
infusion tryptose medium (LIT) with 10% of fetal calf serum to ob-
tain a final concentration of 2.0 ꢁ 106 parasites per mL. The nitro
derivatives, the standard drugs, BZ and NFX, and the lead com-
pound, NF, were separately dissolved in DMSO and diluted into
LIT culture medium to obtain a concentration from 0.25 to
Figure 7. Visualization of lipophilic and electronic molecular properties for the
most active 5-nitro-2-furfuriliden derivatives, 3g (–C4H9) and 4g (–C4H9): (A) the
MLP property and ClogP values were calculated using the SYBYL 8.0 package,42 and
the color range scheme shows hydrophobic regions in brown (0.13) and hydrophilic
regions in blue color (ꢀ0.13). The compounds are displayed in stick capped model
(carbon atoms are in light gray color, oxygen in red, nitrogen in blue, and hydrogen
in white). (B) Total dipole moment vector (l) (blue arrows) was calculated using the
HF/3-21G⁄ method.25,26 (C) The MEP property was calculated using GAUSSIAN 03W43
and the visualization was generated employing GaussView 3.0.44 The color range
scheme indicates positive regions (0.099) in blue color and negative regions
(ꢀ0.099) in red. The molecules are displayed in stick capped model (carbon atoms
are in light gray color, oxygen in red, nitrogen in blue, and hydrogen in white).
180.0 lM, depending on the biological activity of each compound.
An aliquot of the inocula containing 2.0 ꢁ 106 viable parasites per
mL was added to the compounds being evaluated and incubated at
28 °C for 5 days. After this period, the amount of viable parasites
was determined using a spectrophotometer (T80 plus, PG Instru-
ments, United Kingdom) at 580 nm (10 mm light path length).
The assays were performed in triplicate and the anti-T. cruzi activ-
ity was determined by the relation: PGI = {1ꢀ[(Ap ꢀ A0p)/
(Ac ꢀ A0c)]} ꢁ 100, where: PGI = percentage of growth inhibition;
Ap = A580 of the culture containing the drug at day 5; A0p = A580 of
the culture containing the compound after the addition of the
inoculums at day 0; Ac = A580 of the culture in the absence of any
compound at day 5; A0c = A580 of the culture in the absence of
127.7 (C14), 127.2 (C15, C19), 119.0 (CN), 117.8 (C17), 114.3 (C7),
112.9 (C8), 85.1 (C2), 20.9 (C22); Anal. Calcd for (C15H10N4O5): C,
55.22; H, 3.09; N, 17.17. Found: C, 54.88; H, 2.92; N, 17.18.
4.1.3.2. 3-Acetyl-5-(4-buthyl-phenyl)-2-(5-nitrofuran-2-yl)-2,3-
dihydro-1,3,4-oxadiazole (4g).
Light yellow solid (52%); mp
89.0–90.0 °C. 1H NMR (DMSO-d6, 300 MHz): d (ppm): 7.76 (d, 2H,
H15, H19, J = 8.2 Hz), 7.72 (d, 1H, H8, J = 3.8 Hz), 7.37 (s, 1H, H2),
7.36 (d, 2H, H16, H18, J = 8.0 Hz), 7.23 (d, 1H, H7, J = 3.8 Hz), 2.65
(t, 2H, CH2), 2.30 (s, 3H, H22), 1.56 (t, 2H, CH2), 1.31 (m, 2H,
CH2), 0.89 (s, 3H, CH3). 13C NMR {H} (DMSO-d6, 75 MHz): d
(ppm): 167.2 (C20), 154.6 (C6), 151.7 (C5), 150.6 (C9), 147.0
(C17), 129.3 (C14), 128.9 (C15), 128,3 (C19), 126,6 (C16, C18),
any compound at day 0. The inhibitory concentration 50%, IC50
,
was calculated from the percentage of inhibition growth using
nonlinear model growth/sigmoid dose–response for each case,
employing at least, five concentrations.6,23

5404
F. Palace-Berl et al. / Bioorg. Med. Chem. 21 (2013) 5395–5406
4.3. Cytotoxicity assays
The energy minimization of the 3D molecular models (4a and
NF) was also performed employing the MOLSIM 3.2 program55
by applying the steepest descent and, subsequently, the conjugated
gradient method, using a convergence criterion of 0.01 kcal/mol.
The energy-minimized structures were the input to the MD simu-
lations of 1 ns (1,000,000 steps; step size of 1 fs) at 301 K (28 °C),
the same temperature of the biological assay. Dielectric constant
of 3.5 was used to simulate the environment of the biological
membranes and enzyme or receptor binding sites. It was assigned
a fictitious atomic mass of 5000 u.m.a. to some atoms position in
order to maintain the structural integrity of the molecular models
during simulation. Trajectory files were recorded every 20 simula-
tion steps to generate 50,000 conformations. The lowest-energy
conformation was selected from the equilibrium region of confor-
mational ensemble profile (CEP) and compared to the initial en-
ergy-minimized molecular model to verify whether the structural
integrity was maintained after simulation. The RMSD value was
used as criterion to verify the maintenance of structural integrity
employing the Hyperchem 8.0 program.49
4.3.1. Cell culture
Cell culture of murine macrophage J774 cell line and FN1 hu-
man fibroblast cells were maintained in RPMI 1640 containing
10% FBS, 100 UI penicillin and 100
95% air and 5% CO2.
lg/ml gentamycin at 37 °C in
4.3.2. J-774 murine macrophage-like cells
It were seeded (1.0 ꢁ 104 cells/well) in 96 well microplates with
200 mL of RPMI 1640 medium supplemented with 5% heat-inacti-
vated fetal calf serum. Cells were allowed to attach for 24 h in a
humidified 5% CO2/95% air atmosphere at 37 °C. Then, cells were
exposed to compounds 3c, 3f, 3g, 4c, 4f, 4g, the standard drugs,
BZ and NFX, and the lead compound, NF, in concentrations be-
tween 0.8 and 400.0 lM for 48 h. Afterwards, cell viability was as-
sessed by measuring the mitochondrial-dependent reduction of
MTT (Sigma) to formazan. For that purpose, MTT was added to cells
to a final concentration 0.4 mg/mL and cells were incubated at
37 °C for 3 h. After removing the media, formazan crystals were
dissolved in DMSO (180 mL), and the absorbance at 595 nm was
read using a microplate reader (Titertek Uniscience, Multiskan
MCC/340, USA). Results are expressed as IC50 (compound concen-
tration that reduce 50% control absorbance at 595 nm). Every
IC50 is the average of six different experiments and was calculated
from the percentage of inhibition growth using nonlinear model
growth/sigmoid dose–response for each case using, at least, five
concentration values. Cytotoxicity percentages (%C) were determined
as following: %C = 100ꢀ[(ODd ꢀ ODdm)/(ODc ꢀ ODcm) ꢁ 100], where
ODd is the mean of OD595 of wells with macrophages and different
concentrations of the compounds; ODdm is the mean of OD595 of
wells with different compounds concentration in the medium;
ODc is the growth control and ODcm is the mean of OD595 of wells
with medium only.14,46,47
The selected lowest-energy conformation for NF and compound
4a were used as template for building up the molecular models of
compounds from series I and II, respectively.49 The geometry of
each model was optimized employing the MM+ force field,53 and
the partial atomic charges were calculated using the AM1 semiem-
pirical method.54 The MOLSIM 3.2 program55 was also employed to
energy-minimization procedure, applying the steepest descent and
conjugated gradient methods. The convergence criterion was
0.01 kcal/mol. At this time, some thermodynamic properties were
obtained as follows: the total potential energy from minimization
step (Etotal); the intramolecular energy contribution of solvation
(Esolv), applying the hydration shell model proposed by Forsythe
and Hopfinger;56 and, the intramolecular energy contribution of
hydrogen bonding (EHb). The resulting 3D molecular model for
each compound was the input structure to the calculation of phys-
icochemical and structural properties, which were explored in this
study.
4.3.3. FN1 human fibroblast cells
It were plated in triplicate in 96-well plates at a density of
The electronic properties as Charges from Electrostatic Poten-
tials using a Grid based method (CHELPG),26 maps of electrostatic
potential (MEP), dipole moment (total and x, y, z), and frontier
molecular orbital energy (EHOMO and ELUMO) were calculated using
the HF/3-21G⁄ method,25,26 implemented in the Gaussian 03W
program.43
The maps of lipophilic potential (MPL) and the respective values
of the calculated n-octanol/water partition coefficient (ClogP)
were obtained using the Sybyl 8.0 package.42 The maps were calcu-
lated onto Connolly molecular surfaces, using a probe with 1.4 Å
radii (water molecule). Two methods available in the Sybyl 8.0 pro-
1 ꢁ 104 cells/well into flat microtiter plates and incubated for
overnight at 37 °C in
a humidified incubator containing 5%
CO2. Then, the cells were treated with the compounds from
12.5 to 200 M in a final volume of 100 L. After 24 h of treat-
l
l
ment, cells were exposed to 5 mg/mL of MTT for 2 h, and the
precipitated formazan was dissolved in 0.1 N HCl in isopropanol
and measured at 540 nm with a microplate reader Thermo Plate
(Thermoplate TP Reader, Tokay Hit, Japan) and the IC50 values
determined.27
4.4. Molecular modeling approach, calculation and selection of
descriptors
gram were employed to compute ClogP: Ghose et al., 1998-SLN
42
(ClogPSG-SLN
)
and Viswanadhan et al. (ClogPSV).30
The Marvin 5.2.1_1 program57 was used to calculate steric,
topological/geometric, and lipophilic properties. ClogPWM was cal-
culated by the weighted method, assigning equal weight for each
method.30–32
The 3D molecular models of the compounds from series I were
built up using the Cartesian coordinates of crystallized structure of
the lead compound NF (LEQTAC code, Rfactor = 0.11)48 retrieved
from the Cambridge Structural Database (CSD) as reference geom-
etry.49 For compounds from series II, the 3D molecular models
were constructed from the crystallized structure of NF and a sim-
ilar fragment, 2-[4-acetyl-5-(biphenyl-4-yl)-4,5-dihydro-1 3,4-
oxadiazol-2-yl]-phenyl acetate, containing the oxadiazolinic ring
(Rfactor = 0.035),50 giving the 3D molecular model of compound
4a.49 For the reference drugs, the Cartesian coordinates of crystal-
lized structure of BZ (Rfactor = 0.025)51 and NFX analogue (Rfac
tor = 0.042)52 were used as template.49
Properties related to the substituent groups were retrieved
from the literature.33 Further detailed information regarding all
descriptors, methods, and respective software used to perform
these calculations are listed in Supplementary data, p. S21,
Table 3S.
The independent variables or descriptors or molecular proper-
ties of different nature were calculated and generated a matrix
containing 41 columns (independent variables or descriptors plus
biological activity (log1/C)) and 21 rows corresponding to the
number of compounds (11 from series I, including NF, and 10 from
series II). The compounds were distributed in gradients regarding
the biological activity (less, moderately and highly active
compounds).
Each molecular model was energy-minimized in the MM+ force
field,53 without any constrains. The partial atomic charges were
calculated employing the AM1 (Austin Model 1) semiempirical
method.54

F. Palace-Berl et al. / Bioorg. Med. Chem. 21 (2013) 5395–5406
5405
Procedure to select the most representative molecular proper-
Acknowledgements
ties: two filters were used as criteria to previously select the inde-
pendent variables: (i) the Pearson correlation coefficient and (ii) a
visual inspection of scatter plots for biological activity versus each
variable or descriptor or molecular property.17 These data are pre-
sented as Supplementary data (p. S23, Table 4S). The Pearson cor-
relation coefficient value of 0.22 was formerly used as cut off for
selecting the calculated independent variables. Regarding the vi-
sual inspection, only variables presenting uniform distribution
and linear tendency with the biological data were selected to com-
pose the final matrix.
The authors thank Professor Maria Júlia Manso Alves of the Lab-
oratory of Biochemical Parasitology, Department of Biochemistry,
Institute of Chemistry, University of São Paulo (USP) for providing
the epimastigote form of T. cruzi strain. Prof. Elizabeth Igne Ferreira
of the LAPEN laboratory, Faculty of Pharmacy, University of São
Paulo (USP), which had kindly provided access to MOLSIM 3.2 pro-
gram. The authors thank Brazilian scientific committees, FAPESP,
CAPES and CNPq for the financial support.
The final matrix was used as input for the exploratory data anal-
ysis. Due to the distinct magnitude orders among the calculated
variables, the autoscaling procedure was applied as a preprocess-
ing method.35
Supplementary data
Supplementary data (structure elucidation of 5-nitro-2-furfuril-
iden derivatives (series I); structure elucidation of 3-acetyl-2,5-
disubstituted-2,3-dihydro-1,3,4-oxadiazole derivatives (series II);
1H and 13C NMR spectra of new compounds synthesized (4c and
4g–j); dose–response curve of anti-T. cruzi activity assay; confor-
mational ensemble profile of nifuroxazide NF; conformational
ensemble profile of compound 4a (series II); values of the thermo-
dynamic parameters obtained for the lowest-energy conformer of
nifuroxazide (NF) and of 4a compound, selected in CEP from MD
simulations; Independent variables obtained from literature and
from calculation for nifuroxazide (NF), azomethine (3a–g) and
oxadiazoline derivatives (4a–g); descriptors, methods and soft-
ware used to perform calculation for the two series of 5-nitro-2-
furfuriliden derivatives; Pearson’s correlation coefficient and
scatter plots of the descriptors versus biological activity) associ-
ated with this article can be found, in the online version, at
http://dx.doi.org/10.1016/j.bmc.2013.06.017.
4.5. Exploratory data analysis17,18
4.5.1. Hierarchical cluster analysis, HCA
HCA was performed using the Pirouette 3.11 program,40
employing the complete linkage method and Euclidean distance.
In HCA, distances between pairs of samples (or variables) are calcu-
lated and compared. When distances between samples are rela-
tively small, this implies that the samples are similar. The
calculated distances between samples were set on a similarity ma-
trix whose elements are called similarity indices, ranging between
0 and 1, where 1 is equivalent to a maximum similarity. The results
are usually presented as
a two-dimensional chart, called
dendrogram.40
The multivariate distance dab between two samples vectors, a
and b, was determined by computing differences at each of the
m variables:
References and notes
"
#
1=M
m
X
1. Schmunis, G. Lancet Infect. Dis. 13, 2013, 283.
2. Rassi, A.; Marin-Neto, J. A. Lancet 2010, 375, 1388.
3. Astelbauer, F.; Walochnik, J. Int. J. Antimicrob. Agents 2011, 38, 118.
4. Coura, J. R. Mem. Inst. Oswaldo Cruz 2009, 104, 549.
5. Cerecetto, H.; González, M. Pharmaceuticals 2010, 3, 810.
6. Ishii, M.; Jorge, S. D.; Oliveira, A. A.; Palace-Berl, F.; Sonehara, I. Y.; Pasqualoto,
K. F. M.; Tavares, L. C. Bioorg. Med. Chem. 2011, 19, 6292.
7. Jorge, S. D.; Ishii, M.; Palace-Berl, F.; Ferreira, A. K.; De-Sá-Júnior, P. L.; Oliveira,
A. A.; Sonehara, I.; Pasqualoto, K.; Tavares, L. Med. Chem. Commun. 2012, 3, 824.
8. Paula, F. R.; Serrano, S. H. P.; Tavares, L. C. Quím. Nova 2009, 32, 1013.
9. Viodé, C.; Bettache, N.; Cenas, N.; Krauth-Siegel, R. L.; Chauvière, G.; Bakalara,
N.; Périé, J. Biochem. Pharmacol. 1999, 57, 549.
M
dab
¼
ðXaj ꢀ XbjÞ
i
M is the order of the distance, and here corresponds to the
Euclidean distance (M = 2).40 Because of inter-sample distances
can vary with the type and number of measurements, it is cus-
tomary to transform them onto a somewhat more standard
scale of similarity, where dmax is the largest distance in the data
set:
10. Craig, P. N. J. Med. Chem. 1971, 14, 680.
dab
similarityab ¼ 1 ꢀ
dmax
11. Rollas, S.; Gulerman, N.; Erdeniz, H. IL Farmaco 2002, 57, 171.
12. Lee, L.; Robb, L. M.; Lee, M.; Davis, R.; Mackay, H.; Chavda, S.; Babu, B.; O’Brien,
E. L.; Risinger, A. L.; Mooberry, S. L. J. Med. Chem. 2009, 53, 325.
13. Maccioni, E.; Alcaro, S.; Cirilli, R.; Vigo, S.; Cardia, M. C.; Sanna, M. L.; Meleddu,
R.; Yanez, M.; Costa, G.; Casu, L.; Matyus, P.; Distinto, S. J. Med. Chem. 2011, 54,
6394.
14. Romeiro, N. C.; Aguirre, G.; Hernández, P.; González, M.; Cerecetto, H.; Aldana,
I.; Pérez-Silanes, S.; Monge, A.; Barreiro, E. J.; Lima, L. M. Bioorg. Med. Chem.
2009, 17, 641.
HCA was carried out for samples and variables. In the first case,
biological activity was considered as dependent variable. In the
second option, the biological activity is considered as independent
variable.17,18,40
4.5.2. Principal components analysis, PCA
15. Ralph, P.; Nakoinz, I. Nature 1975, 257, 393.
16. Ferreira, A. K.; Meneguelo, R.; Pereira, A.; Mendonça, F. O. R.; Chierice, G. O.;
Maria, D. A. Anticancer Res. 2012, 32, 95.
17. Ferreira, M. J. Braz. Chem. Soc. 2002, 13, 742.
18. Beebe, K. R.; Pell, R. J.; Seasholtz, M. B. Chemometrics: A Practical Guide; Wiley-
Interscience: New York, 1998.
The exploratory analysis of PCA was also carried out employing
the Pirouette 3.11 software.40 For PCA, the biological activity was
considered as the dependent variable. The data were decomposed
into two matrices, one of scores related to the samples, and an-
other of loadings, related to the variables.17,18
19. Jorge, S. D.; Palace-Berl, F.; Masunari, A.; Cechinel, C. A.; Ishii, M.; Pasqualoto, K.
F. M.; Tavares, L. C. Bioorg. Med. Chem. 2011, 19, 5031.
The new set of axes generates the PCs, or factors, into which are
the information related to the original descriptors. Thus, the num-
ber of PCs that explain most of the variability in the data set can be
determined, considering that these PCs are uncorrelated and mutu-
ally orthogonal variables built up as simple linear combinations
from the original data.17,18
20. Tavares, L. C.; Penna, T. C.; Amaral, A. T. Boll. Chim. Farm. 1997, 136, 244.
21. Somogyi, L. J. Heterocycl. Chem. 2007, 44, 1235.
22. Li, C.-K.; Ma, Y.-J.; Cao, L.-H. J. Chin. Chem. Soc. 2009, 56, 182.
23. Cerecetto, H.; Di Maio, R.; González, M.; Risso, M.; Saenz, P.; Seoane, G.;
Denicola, A.; Peluffo, G.; Quijano, C.; Olea-Azar, C. J. Med. Chem. 1999, 42, 1941.
24. Baber, J. C.; Thompson, D. C.; Cross, J. B.; Humblet, C. J. Chem. Inf. Model. 2009,
49, 1889.
25. Fock, V. Z. Phys. A: Hadron Nucl. 1932, 75, 622.
In this exploratory data analysis, PCA was run up to ten fac-
tors or PCs. The outliers’ diagnosis, implemented in Pirouette
3.11 software,40 was also performed through the Mahalanobis
distance.41
26. Brenneman, C. M.; Wiberg, K. B. J. Comput. Chem. 2004, 11, 361.
27. Ferrara, P.; Apostolakis, J.; Caflisch, A. Proteins 2002, 46, 24.
28. Verloop, A. The STERIMOL Approach to Drug Design; Marcel Dekker: New York,
1987.

5406
F. Palace-Berl et al. / Bioorg. Med. Chem. 21 (2013) 5395–5406
29. Ghose, A. K.; Viswanadhan, V. N.; Wendoloski, J. J. J. Phys. Chem. 1998, 102,
43. GAUSSIAN 03W-revision B.02 for Windows, version 6; Gaussian Inc.: Pittsburgh,
PA, 1995–2003.
3762.
30. Viswanadhan, V. N.; Ghose, A. K.; Revankar, G. R.; Robins, R. K. J. Chem. Inf.
Comput. Sci. 1989, 29, 163.
31. Klopman, G.; Li, J.-Y.; Wang, S.; Dimayuga, M. J. Chem. Inf. Comput. Sci. 1994, 34,
44. GAUSSVIEW 3.0; Gaussian Inc.: Pittsburgh, PA, 2000–2003.
45. Silva, L. H. P.; Nussenzweig, V. Folia Clín. Biol. 1953, 20, 191.
46. Mosmann, T. J. Immunol. Methods 1983, 65, 55.
752.
32. PHYSPROP
47. Denizot, F.; Lang, R. J. Immunol. Methods 1986, 89, 271.
48. Pniewska, B.; Januchowski, M. Pol. J. Chem. 1998, 72, 2629.
49. HYPERCHEM Program for Windows; Hypercube, Inc.: Gainesville, FL, 2008.
50. Yehye, W. A.; Ariffin, A.; Rahman, N. A.; Ng, S. W. Acta Crystallogr. Sect. E: Struct.
Rep. Online 2010, 66, o878.
51. Soares-Sobrinho, J. L.; Cunha-Filho, M. S.; Rolim Neto, P.; Torres-Labandeira, J.
J.; Dacunha-Marinho, B. Acta Crystallogr. Sect. E: Struct. Rep. Online 2008, 64
(Pt. 3), o634.
52. Caracelli, I.; Stamato, F.; Mester, B.; Paulino, M.; Cerecetto, H. Acta Crystallogr. C
1996, 52, 1281.
53. Allinger, N. L. J. Am. Chem. Soc. 1977, 99, 8127.
54. Dewar, M. J.; Zoebisch, E. G.; Healy, E. F.; Stewart, J. J. J. Am. Chem. Soc. 1985,
107, 3902.
database.
http://www.syrres.com/what-we-do/databaseforms.
aspx?id=386.
33. Hansch, C.; Leo, A.; Hoekman, D. Exploring QSAR: Hydrophobic, Electronic, and
Steric Constants; American Chemical Society: Washington, 1995.
34. Hansch, C.; Fujita, T. J. Am. Chem. Soc. 1964, 86, 1616.
35. Ferreira, M.; Antunes, A. M.; Melgo, M. S.; Volpe, P. L. O. Quim. Nova 1999, 22, 724.
36. Arroio, A.; Honório, K. M.; da Silva, A. B. F. Quim. Nova 2010, 33, 694.
37. Soriano-Correa, C.; Raya, A.; Esquivel, R. O. Int. J. Quantum Chem. 2008, 108, 1369.
38. Jorge, S. D.; Palace-Berl, F.; Pasqualoto, K. F. M.; Ishii, M.; Ferreira, A. K.; Berra,
C. M.; Bosch, R. V.; Maria, D. A.; Tavares, L. C. Eur. J. Med. Chem. 2013, 64, 200.
39. Boiani, M.; Cerecetto, H.; Gonzalez, M.; Gasteiger, J. J. Chem. Inf. Model. 2008, 48,
213.
40. PIROUETTE 3.11, Infometrix Inc.: Woodinville, WA, 1990–2003.
41. Mahalanobis, P. C. J. Assist. Soc. Bengal 1930, 26, 541.
42. SYBYL 8.0 for Linux; Tripos, Inc.: 2007. 1699 South Hanley Rd., St. Louis, MO
63144–2917, USA.
55. Doherty, D. MOLSIM: Molecular Mechanics and Dynamics Simulation Software-
User’s Guide, version 3.2; The Chem21 Group Inc.: Lake Forest, 1997.
56. Forsythe, K.; Hopfinger, A. Macromolecules 1973, 6, 423.
57. Marvin Beans, version 5.2.1_1; ChemAxon Ldt.: 1998–2010, Budapest, Hungary.