Ligand-based design, synthesis, and experimental evaluation of novel benzofuroxan derivatives as anti-Trypanosoma cruzi agents

Article abstract of DOI:10.1016/j.ejmech.2013.03.053

A set of substituted-[N′-(benzofuroxan-5-yl)methylene]benzohydrazides (4a-t), previously designed and synthesized, was experimentally assayed against Trypanosoma cruzi, the etiological agent of Chagas' disease, one of the most neglected tropical diseases. Exploratory data analysis, Hansch approach and VolSurf formalism were applied to aid the ligand-based design of novel anti-T. cruzi agents. The best 2D-QSAR model showed suitable statistical measures [n = 18; s = 0.11; F = 42.19; R² = 0.90 and Q² = 0.77 (SDEP = 0.15)], and according to the optimum 3D-QSAR model [R² = 0.98, Q ² = 0.93 (SDEP = 0.08)], three latent variables explained 62% of the total variance from original data. Steric and hydrophobic properties were pointed out as the key for biological activity. Based upon the findings, six novel benzofuroxan derivatives (4u-z) were designed, synthesized, and in vitro assayed to perform the QSAR external prediction. Then, the predictability for the both models, 2D-QSAR (R_pred² = 0.91) and 3D-QSAR (R_pred² = 0.77), was experimentally validated, and compound 4u was identified as the most active anti-T. cruzi hit (IC₅₀ = 3.04 μM).

Full text of DOI:10.1016/j.ejmech.2013.03.053

Accepted Manuscript
Ligand-based design, synthesis, and experimental evaluation of novel benzofuroxan
derivatives as anti-Trypanosoma cruzi agents
Salomã o Dória Jorge, Fanny Palace-Berl, Kerly Fernanda Mesquita Pasqualoto,
Marina Ishii, Adilson Kleber Ferreira, Carolina Maria Berra, Rosemary Viola Bosch,
Durvanei Augusto Maria, Leoberto Costa Tavares
PII:
S0223-5234(13)00213-4
10.1016/j.ejmech.2013.03.053
EJMECH 6091
DOI:
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 28 January 2013
Revised Date: 19 March 2013
Accepted Date: 24 March 2013
Please cite this article as: S.D. Jorge, F. Palace-Berl, K.F. Mesquita Pasqualoto, M. Ishii, A.K. Ferreira,
C.M. Berra, R.V. Bosch, D.A. Maria, L.C. Tavares, Ligand-based design, synthesis, and experimental
evaluation of novel benzofuroxan derivatives as anti-Trypanosoma cruzi agents, European Journal of
Medicinal Chemistry (2013), doi: 10.1016/j.ejmech.2013.03.053.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT
Graphical abstract

ACCEPTED MANUSCRIPT
Highlights
►
►
►
►
►
A set of benzofuroxan derivatives as anti-Trypanosoma cruzi agents is presented.
Exploratory data analysis, Hansch approach and VolSurf formalism were applied.
The predictability for the both QSAR models was experimentally validated.
Cytotoxicity assays were performed with human fibroblasts cells.
The most active compound of the entire set was identified.

ACCEPTED MANUSCRIPT
1
Ligand-based design, synthesis, and experimental evaluation of novel
benzofuroxan derivatives as anti-Trypanosoma cruzi agents
a,∗
a
b
Salomão Dória Jorge, Fanny Palace-Berl, Kerly Fernanda Mesquita Pasqualoto,
a
b
b
b
Marina Ishii, Adilson Kleber Ferreira, Carolina Maria Berra, Rosemary Viola Bosch,
b
a
Durvanei Augusto Maria, Leoberto Costa Tavares
a
Laboratório de Planejamento e Desenvolvimento de Fármacos, Departamento de Tecnologia
Bioquímico Farmacêutica, Faculdade de Ciências Farmacêuticas, Universidade de São Paulo,
b
São Paulo, SP 05508-900, Brasil; Laboratório de Bioquímica e Biofísica, Instituto Butantan,
São Paulo, SP 05503-900, Brasil.
Abstract
A
set of substituted-[N’-(benzofuroxan-5-yl)methylene]benzohydrazides (4a-t),
previously designed and synthesized, was experimentally assayed against Trypanosoma
cruzi, the etiological agent of Chagas’ disease, one of the most neglected tropical
diseases. Exploratory data analysis, Hansch approach and VolSurf formalism were
applied to aid the ligand-based design of novel anti-T. cruzi agents. The best 2D-QSAR
2
model showed suitable statistical measures [n=18; s=0.11; F=42.19; R =0.90 and
2
2
Q =0.77 (SDEP=0.15)], and according to the optimum 3D-QSAR model [R =0.98,
2
Q =0.93 (SDEP=0.08)], three latent variables explained 62% of the total variance from
original data. Steric and hydrophobic properties were pointed out as the key for
biological activity. Based upon the findings, six novel benzofuroxan derivatives (4u-z)
were designed, synthesized, and in vitro assayed to perform the QSAR external
2
prediction. Then, the predictability for the both models, 2D-QSAR (R pred=0.91) and
∗
Corresponding author address: Av. Prof Lineu Prestes, 580, São Paulo, SP 05508-900, Brasil. Phone:
+
55 11 30913693; fax: +55 11 38156386; e-mail: sdjorge@usp.br; sdjorge@gmail.com.

ACCEPTED MANUSCRIPT
2
2
3
D-QSAR (R pred=0.77), was experimentally validated, and compound 4u was
identified as the most active anti-T.cruzi hit (IC =3.04 ꢀM).
5
0
Keywords: benzofuroxan derivatives, Trypanosoma cruzi, exploratory data analysis,
QSAR approaches, external prediction.
1
. Introduction
Neglected Tropical Diseases (NTDs) constitute one of the most serious public
health burdens, affecting primarily people living on extreme poverty. They are endemic
in Africa, Asia, and Americas. The major impact of NTDs on public health systems is
the lifelong disability and morbidity [1]. Chagas’ disease (CD) or American
trypanosomiasis, for instance, caused by the protozoal hemoflagellate Trypanosoma
cruzi, is considered one of the most neglected diseases. The fifth annual G-FINDER
report demonstrated that the global funding for CD R&D investment in 2011 was
$
17.4m (0.57% of total funding for NTDs R&D) [2]._ENREF_2
The World Health Organization (WHO) recommends preventive chemotherapy
as one approach against NTDs although the currently available medicines, nifurtimox
NFX) and benznidazole (BZ) used to control CD, were effective mainly during the
(
early phase of infection [3]._ENREF_3 Those drugs, however, exhibit variable efficacy in
different regions of high endemicity because the high biological, biochemical, and
genetic diversity of T. cruzi strains, which have been grouped into six discrete typing
units (DTUs), TcI to TcVI [4]._ENREF_4 Furthermore, the natural resistance of T. cruzi
has been suggested as an important factor to explain the low rates of cure detected in
chagasic patients [5, 6]. _ENREF_5_ENREF_6The complexity of the protozoal life cycle
and the pathologic mechanisms involved in CD has also significantly impaired the fight

ACCEPTED MANUSCRIPT
3
against this NTD, however many efforts have been made to develop more efficient and
safer drugs to be used in shorter treatment regimens.
In this regard, the development of drugs based upon the structural optimization
of compounds with known biological activities is advantageous, considering time and
efforts consumption, even though the emergence of parasite resistance were developed
in a short period of time. Nifuroxazide (NF), for instance, a 5-nitrofuran compound,
presents a chemical structure that favors chemical modifications, but its mechanism of
action is not fully elucidated. In general, nitroheterocyclic compounds, including NFX,
have not yet their mode of action completely clear. Previous studies suggested that the
biological activity of this kind of compounds is related to the reduction of nitro group,
which implies the formation of free radical toxic species [7-10]._ENREF_7 Non-selective
bioreduction of these compounds is probably the reason of its toxic effects in the
mammalian host.
Concerning to the mechanistic aspect, the use of benzofuroxan system has been
reported as a promising alternative in the search for new anti-chagasic drugs_ENREF_11
less toxic than BZ and NFX, which contain a nitroaromatic system [11-14]._ENREF_13
-
The presence of =N(
ꢀ
O)O moiety in benzofuroxan derivatives may confer them the
electron-accepting properties similar those of aromatic nitrocompounds or N-oxides,
which are also used as bioreductible activated antitumor and/or cytotoxic agents. Then,
the redox properties and cytotoxicity of benzofuroxans bear certain similarities
(
reactions with single-electron-transferring flavoenzymes) and differences
electrochemical reduction, cytotoxicity) from nitroaromatic compounds. Moreover, the
(
reactions of benzofuroxans with thiol groups have been considered as a parallel
cytotoxicity mechanism [15],_ENREF_15 reinforcing, for instance, the use of these
derivatives as potential trypanothione reductase inhibitors.

ACCEPTED MANUSCRIPT
4
All trypanosomes and leishmanias possess a unique thiol metabolism in which
the ubiquitous glutathione/glutathione reductase (GR) system is replaced by
trypanothione and trypanothione reductase (TR) [16]._ENREF_16 The dithiol
trypanothione [bis(glutathionyl)spermidine] is the donor of reducing equivalents for the
synthesis of DNA precursors, the homeostasis of ascorbate, the detoxification of
hydroperoxides, and the sequestration/export of thiol conjugates [17]._ENREF_17
Reduction of trypanothione disulfide is crucial for the maintenance of a reducing
intracellular milieu which renders the parasite specific TR an attractive drug target
molecule [18-21]._ENREF_18 TR and human GR have similar catalytic mechanisms; 14
of the 19 amino acid residues close to the binding site are conserved. However, they are
specific to their respective disulfide substrates, and this specificity is mainly due to the
exchange of five residues, Ala34, Arg37, Ile113, Asn117 and Arg347 in human GR that
are replaced by Glu18, Trp21, Ser109, Met113 and Ala343 in T. cruzi TR, for instance.
GR has a hydrophilic, positively charged region in its active site that interacts with the
glycine carboxylates of glutathione disulfide, while TR has a larger binding site, with a
hydrophobic and negatively charged region with which the spermidine moiety of
2
trypanothione disulfide (T[S] ) binds. The absence of TR from the mammalian host and
the sensitivity of trypanosomatids to oxidative stress make TR an attractive target for
trypanosomiasis therapeutics [22-24]._ENREF_22
The primary amino acids sequences of GR and T. cruzi TR enzymes obtained
from National Center for Biotechnology Information (NCBI) (GenBank accession
number AAH69244.1 and CAA78360.1, respectively) were aligned [25],_ENREF_25
revealing 36% overall amino acid sequence identity. Moreover, multiple protein
sequence alignment was also done to find local sequence similarities between the two
sequences [26],_ENREF_26 and these regions are delimited by boxes in Figure 1. The

ACCEPTED MANUSCRIPT
5
predicted ligands binding regions for GR and TR, using the Sequence Annotated by
Structure (SAS) tool from European Bioinformatics Institute website
(http://www.ebi.ac.uk/thornton-srv/databases/sas/) [PDB IDs: 1GRE and 1BZL] [27,
2
8],_ENREF_27_ENREF_27 are presented as dark and light gray highlights, respectively. It
is noteworthy that these regions are quite similar in both sequences, and little
differences are mainly found in C-terminus region of TR enzyme. Furthermore, the
amino acid residues highlighted as black dots in TR sequence are probable ligand
binding sites. Also, they correspond to non-ligand binding sites for GR, and are not
placed within the identical regions regarding both proteins.
FIGURE 1
Before to explore more the TR enzyme as a target, herein, we carried out a
ligand-based drug design approach regarding a set of N’-(benzofuroxan-5-il)methylene
benzohydrazide derivatives (4a-t). Previously, we reported the synthesis, antibacterial
activity and preliminary testing against epimastigote forms of T. cruzi of these
compounds [14, 29, 30]. In continuation of our research to identify the physicochemical
and molecular properties that have most influence on anti-T.cruzi activity of
benzofuroxan derivatives (4a-t), which are NF functional analogues (figure 2), we
firstly performed an exploratory data analysis, which comprises the principal
component analysis (PCA) and hierarchical cluster analysis (HCA) [31]._ENREF_29
Then, to generate quantitative structure-activity prediction models, a classical 2D-
QSAR [32, 33]_ENREF_30 approach (also known as Hansch analysis) and 3D-QSAR
VolSurf [34-36]_ENREF_32 based method were applied, and regarding the QSAR
findings, six novel benzofuroxan derivatives (4u-z) were designed, synthesized, and

ACCEPTED MANUSCRIPT
6
assayed to perform the QSAR external validation. The biological in vitro activity was
tested against epimastigote forms of T. cruzi Y strain and cell viability in fibroblasts
was also carried out to verify the compounds’ toxicity effect on mammalian cells.
FIGURE 2
2
. Results and Discussion
Anti-Trypanosoma cruzi activity: the biological activity of benzofuroxan derivatives
4a-t) and of the lead compound, NF, were tested in vitro against the epimastigote form
(
of T. cruzi Y strain, isolated by Silva and Nussenzweig in 1953 [37]._ENREF_38 The use
of T. cruzi epimastigote forms is relevant in a preliminary evaluation of the
antiproliferative effect of potential antichagasic agents, since those forms are present in
mammalian’s cells [38, 39]._ENREF_39 The Y strain is classified as TcII, originally from
São Paulo/Brazil and isolated from a human host [4],_ENREF_4 showing a percentage of
cure induced by NFX and BZ of 47%, for each drug, in groups of mice [6]._ENREF_6
Epimastigote forms were grown in axenic media as described in the Experimental
6
protocols. Parasites in the stationary growth phase (from an initial culture with 2.0 × 10
epimastigotes/mL) were incubated with increasing concentrations of the compounds 4a-
t (range from 1.0 to 160.0 ꢀM) in DMSO (0.5 and 1.0% final concentration) for 5
days.[40, 41]_ENREF_41 BZ and NFX were used as the reference anti-T. cruzi drugs. No
effect on epimastigotes’ growth was observed in the presence of up to 1% (v/v) DMSO
in the culture media. Parasite viability and growth response were determined by
absorbance readings at 580 nm (A580). The percentage of grown inhibition (table 1) was
calculated as follows: PGI = {1 – [(A – A )/(A – A )]}× 100, where A = A of the
p
0p
c
0c
p
580
culture containing the drug at day 5; A0p = A580 of the culture containing the drug right

ACCEPTED MANUSCRIPT
7
after the inocula addition (day 0); A = A of the culture in the absence of any drug
c
580
(
control) at day 5; A = A in the absence of any drug at day 0 [42-45]._ENREF_43 The
0c 580
biological endpoint was determined as the concentration which causes a fifty percent of
parasite growth inhibition (PGI50). The values of log1/PGI50 (or pPGI50) were
considered as the biological parameter or the dependent-variable for the theoretical
analyses carried out.
The two most active compounds were the 4-I (4q) and 4-Br (4o) substituted
derivatives having PGI50 values of 6.11 ꢀM (approximately 2.50 ꢀg/mL) and 7.38 ꢀM
(
2.80 ꢀg/mL), respectively. Otherwise, the lowest activity was observed for the 4-NH
4c) and 4-SO NH (4p) substituted derivatives, which presented PGI50 values of 116.96
M (34.74 ꢀg/mL) and 60.38 ꢀM (21.80 ꢀg/mL), respectively. The 4-OH (4d) and 4-
N(CH ) (4k) substituted derivatives did not show any anti-T. cruzi activity at the
2
(
2
2
ꢀ
3
2
concentrations tested.
TABLE 1
Calculation and selection of descriptors: the independent variables or descriptors
related to steric, electronic, hydrophobic, thermodynamic, topological and geometric
contributions were obtained from literature [46, 47]_ENREF_47 and from computational
approaches [48-50] as reported in the Experimental protocols. The initial data matrix
contains eighteen rows (number of compounds) and thirty-four columns, including
biological activity (log1/PGI50 values). A list of all descriptors used in the exploratory
analysis and 2D-QSAR approach is presented in Table 2, and the values of each
independent variable or descriptor obtained for the investigated compounds are
properly reported in Table 1S (Supplementary Information section).

ACCEPTED MANUSCRIPT
8
TABLE 2
The correlation between the biological data (log1/PGI50) and the independent
variables was visualized through the respective scatter plots (log1/PGI50 versus each
descriptor) as a criterion for the selection of descriptors [51]._ENREF_53 The thirty-three
descriptors were evaluated and those which did not show
a
proper
distribution/dispersion regarding the biological data were excluded in this step. The
findings are presented in table 2S (Supplementary Information section) and the
discarded descriptors are marked in red. The Pearson correlation coefficient was also
employed as a filter for the variables selection, but just for the exploratory data analysis,
since the Pirouette 3.11 program (Infometrix, Inc., 2003)[52] uses only linear functions.
Variables presenting uniform dispersion/distribution and non-linear correlation
tendency, regarding the biological data, were also considered in the 2D-QSAR studies,
since non-linear functions can be used to build up the prediction models.
Exploratory data analysis: the input data matrix was composed of eighteen
compounds (4a-t, excepted 4d and 4k, which did not have its PGI50 values properly
2 3 4
determined) and nine descriptors (CHELPG_C , CHELPG_C , CHELPG_C ,
CHELPG_N19, CHELPG_O22, ClogPSLN, ClogPWM, ASA_h and π), which were
preliminary selected using the scatter plots and Pearson correlation coefficients (cutoff
value = 0.3) as filtering.
PCA was run up to nine factors and the findings are presented in figure 3a. In
PCA multivariate method the PCs or factors contain most of the variability in the data
set, albeit in a much lower dimensional space. PC1, or factor 1, is defined in the
direction of maximum variance of the whole data set whereas PC2 or factor 2 is the

ACCEPTED MANUSCRIPT
9
direction that describes the maximum variance in the orthogonal subspace to PC1. The
subsequent components are taken orthogonal to those previously chosen and describe
the maximum of the remaining variance _ENREF_54[31, 52].
According to the factors selection, in fig. 3a, PC1 or factor1 contains 46.28% of
total variance from the original data whereas PC2 or factor 2 describes 24.09% of total
variance. Both PCs can be considered as representative for analysing the set of
benzofuroxan derivatives because together they discriminate 70.37% of total variance
from the original descriptors or molecular properties. Regarding the scores plot (fig. 3a),
the two first factors classified the benzofuroxan derivatives into four groups: A
[
compounds with amine and nitro substituent; 4c (R
CF )], B [highly active compounds; 4g, 4i, 4n, 4o, 4q, 4r and 4s], C [moderately active
compounds (4b, 4e and 4h), except to 4l (R = OC H ), which is more active], and D
1 2 1 2 2
= NH ) and 4t (R = NO , R =
3
1
2
5
[
4a, 4f, 4m, 4p and 4j, which are less active compounds]. In general, the more active
compounds are most commonly found on the right upper side of PC1 (positive values).
The loadings table, also presented in fig. 3a, provides information about which
descriptors or molecular properties influenced more for the samples/compounds
classification. In PC1 or factor 1, electronic (CHELPG_C , CHEPG_C , CHELPG_N ,
2
3
19
CHELPG_O22) and steric/geometric (ASA_h) properties showed higher loading values
in module). ASA_h corresponds to the solvent accessible surface area of all
hydrophobic atoms. In PC2, although CHELPG_C presented a high loading value,
(
4
primarily hydrophobic properties (ClogPWM, π) influenced the samples classification.
Moreover, the plot of sample residual versus Mahalanobis distance [53] (also in
fig. 3a) indicated there were no outliers. The sample residual threshold (light green line)
is based upon a ninety-five percent of confidence level interval set internally in
Pirouette 3.11 [52]._ENREF_55 The compounds did not exceed a threshold of 95%,

ACCEPTED MANUSCRIPT
1
0
meaning the calculated properties were sufficient to describe the structural features of
the entire data set.
In HCA,_ENREF_53 distances between pairs of samples or variables are
calculated and compared. When the distance between samples is relatively small, this
implies that samples are similar. Conversely, samples separated by relatively large
distances are mostly dissimilar. Complementary findings were obtained for the both
methods, PCA and HCA, as can be seen in the dendrogram of samples (fig. 3b). Four
clusters were formed according to the samples’ similarity indices: A with 50.5%
similarity [4c (R
n, 4o, 4q, 4r and 4s; log1/PGI50 values from 5.00 to 5.21], C with 58.8% similarity
4b, 4e, 4h and 4l; log1/PGI50 values from 4.82 to 4.93, except to 4l which has
1 2 1 2 2 3
= NH ) and 4t (R = NO , R = CF )], B with 51.2% similarity [4g, 4i,
4
[
log1/PGI50 value of 5.02], and D with 51.2% similarity [4a, 4f, 4m, 4p and 4j;
log1/PGI50 values lower than 4.80]. Additionally, the dendrogram of variables (fig. 3b),
which considers the anti-T. cruzi activity also as an independent variable, indicated the
hydrophobic properties (π, ClogPWM and ClogPSLN) as the descriptors most closely
related to the biological data (y), since they are in the same cluster sharing 53.3%
similarity.
FIGURE 3
The exploratory data analysis pointed out the hydrophobic (ClogP and π),
electronic (CHELPG) and steric/geometric (ASA_h) properties as important for the
anti-T. cruzi activity, regarding the set of benzofuroxan derivatives investigated and
these findings are going to support the construction and interpretation of 2D/3D-QSAR
models.

ACCEPTED MANUSCRIPT
1
1
To visualize the properties highlighted as relevant to the anti-T. cruzi activity,
the maps of lipophilic and electrostatic potential (MLP and MEP) were calculated onto
the molecular surface of less [4c; R
1
1 ;
= NH2; log1/PGI50 = 3.93) and more [4q; R = I
log1/PGI50 = 5.21)] active compounds and are presented in fig. 4A and B, respectively.
Also, the partial atomic electrostatic potential charges (ESP, CHELPG) by color and the
solvent accessible surface area of all hydrophobic atoms (ASA_h) for compounds 4c
and 4q are showed in fig. 4C and D as well. Since MEP, MLP, and ASA_h are
calculated onto the molecular surface, they all reflect molecular shape.
The MLP property (fig. 4A) [49] gives the information of molecular
hydrophilic/hydrophobic balance and can be explained using a color scheme, where
blue color corresponds to hydrophilic regions and brown color to hydrophobic regions.
The lipophilicity can be numerically expressed by the calculated n-octanol/water
partition coefficient (ClogP). In this regard, the 4c compound presented a more
2 1
hydrophilic character than 4q. The contribution of NH group (R ) to hydrophobicity is
negative (π = -1.23), reflecting in the global ClogP value (ClogPSLN = 0.11 and
ClogPWM = 1.23) of compound 4c.
According to the MEP property (fig. 4B) [54], the regions with higher electronic
density distribution (as carbonyl oxygen and N-oxide) are displayed in red to yellow
color while regions with lower electronic density distribution are in blue to green color.
2
Electron donor substituent groups (as NH ) attached to the benzene ring increase the
electron density in the ring (green to yellow color in the ring center) region whereas
electron-withdrawing substituents, as iodine, reduce the electron density in the benzene
ring (only green color in the ring center). Moreover, ESP charges are represented by a
color range from red (negative; -1.11) to green (positive; 1.11) (see fig. 4B). For
compound 4c, the carbon attached to the substituent group is in bright green color and

ACCEPTED MANUSCRIPT
1
2
its charge value is positive (0.761) while the carbon attached to iodine (compound 4q) is
in dark color and its charge value is negative (-0.033). Additionally, the carbonyl carbon
is slightly more positive (0.796; bright green) for compound 4c than for compound 4q
(0.750; bright green).
The ASA_h property [55], in fig. 4C, indicates that the solvent accessible
surface area of all hydrophobic (non-polar) atoms is bigger for compound 4q (more
2
active; 390.50 Å ) than 4c, and it is mostly concentrated onto iodine and benzene ring
region. This finding is in agreement with the MLP of compound 4q (brown color
visualized onto iodine and benzene ring moiety).
FIGURE 4
Hansch analysis: considering the data presented in Table 1, the best 2D-QSAR model
was constructed employing multiple linear regression (MLR) (BuidQSAR 2.1.0
software [56])._ENREF_58 A systematic search using three independent variables
resulted in a linear model or Eq. 1, which is related to the hydrophobic (Hansch‐Fujita
hydrophobicity constant, π), electronic (total dipole moment, ꢀ) and steric/geomeric
(solvent accessible surface of all hydrophobic atoms, ASA_h) properties. The inter-
correlation between descriptors was less than 0.50 (R < 0.5), which means each
descriptor provides original information to the model. The correlation matrix obtained
for the selected descriptors are given in Table 3. Eq. 1a was obtained from autoscaled
data.
pPGI = + 0.23 (± 0.07) π – 0.0585 (± 0.03) ꢀ + 0.0035 (± 0.00) ASA_h + 4.08 (± 0.51)
5
0
(
Eq. 1)
2
2
n = 18 ; R = 0.90 ; s = 0.11 ; F = 42.13 ; Q = 0.77
pPGI50 = 0.21 (± 0.06) π – 0.12 (± 0.07) ꢀ + 0.15 (± 0.07) ASA_h + 4.83 (± 0.06)
Eq. 1a)
(

ACCEPTED MANUSCRIPT
1
3
The Hansch‐Fujita hydrophobicity descriptor is a constant group that measures
the individual contribution of a given substituent group for the partition coefficient of a
molecule [33]._ENREF_31 Hydrophobicity is a physicochemical property, which is
related to the ability of a compound diffuses through the biological membranes and
reaches its site of action. The positive regression coefficient of π, in Eq. 1, suggests that
the compounds’ activity would increase as the hydrophobicity of the substituent groups
also increases.
The total dipole moment (ꢀ) is a property that measures the delocalized charge
when atoms with different electronegativity are interconnected. Substituent groups with
different electronegativity can change the molecular properties of a compound, such as
acidity and basicity. In this regard, the dipole moment could help the understanding of
some aspects related to the reactivity of compounds [57]._ENREF_59 The negative
coefficient of ꢀ, in Eq. 1, indicates that the compounds’ activity increases as the
compounds’ polarity decreases (lower dipole moment values). The same behavior was
reported by Boiani and co-workers [58],_ENREF_60 showing the importance of polarity
for anti-T. cruzi activity in molecules bearing the benzofuroxan system.
The ASA_h descriptor is dependent of conformation and partial charges of a
molecule and, as already mentioned, it corresponds to the solvent accessible surface
area of all hydrophobic atoms (| qi | < 0.125, where | qi | is the absolute value of the
partial atomic charge), calculated using the radius of the solvent molecule (1.4 Å for the
water molecule, for instance) [50]._ENREF_51 It is noteworthy that steric/geometric
properties are related to the molecular shape, which together with the electrostatic field
can be considered as the major requirements for the molecular recognition process.
Also, the hydrophobic character, besides to confer the ability to passively cross

ACCEPTED MANUSCRIPT
1
4
biological membranes, can be related to the complementarity with the environment of a
binding site in enzymes or receptors, which can be essentially lipophilic.
A comparison between observed and predicted activity (log1/PGI50) for the
benzofuroxan derivatives used in the construction of 2D-QSAR model, or Eq. 1, is
shown in fig. 5A. In this study, the compounds whose differences in
experimental/observed and predicted/calculated activities exceeded 2.0 standard
deviations (95% confidence level) from the mean of a model would be considered as
outliers. The column plot presented in fig. 5B shows the absence of outliers for the best
2
D-QSAR model.
As already mentioned, the selected descriptors in Eq. 1 were not inter-correlated
(
(
R < 0.5), providing unique information to the QSAR model. The fraction of variance
2
expressed by r ) was 0.90, indicating the 2D-QSAR model explains 90% of the
2
variance of data. According to the literature, a well-fitted QSAR model must have r ≥
0
.6 [31, 33]. Moreover, the Eq. 1 showed a small standard deviation value (s = 0.11) as
well as a large value of Fischer statistics (F), which is the ratio between explained and
unexplained variance for a given number of freedom degrees [59]._ENREF_61 The F-
value was 44.86, which is statistically significant at the 95% confidence level.
2
The leave-one-out (LOO) crossvalidation correlation coefficient (Q LOO) was
2
0
.77. Regarding literature data, a predictive QSAR model must have Q ≥ 0.5
[
60]._ENREF_62 Multiple leave-N-out (LNO) crossvalidation procedure [60]_ENREF_62
was also performed to confirm the robustness of the 2D-QSAR model or Eq. (1), and
the critical N found was 5 (28% out) (fig. 5C) [61]. The critical N is the maximum value
2
2
of N at which Q LNO is still stable, similar to the original model (Q LOO), and do not
oscillate significantly for N up to critical N, in case of multiple LNO. Regarding
multiple LNO, QSAR models should be considered robust when two standard

ACCEPTED MANUSCRIPT
1
5
deviations should not be greater than 0.1 for N = 2, 3, etc., including the critical value of
N. _ENREF_62[60]
FIGURE 5
Thus, 2D-QSAR model or Eq. 1 can be indeed considered a good QSAR model
for the investigated set of compounds, since it is robust (passed through LNO
procedure), its statistical measures are suitable, each descriptor provides unique
information to the model, and the model can be interpreted regarding a chemical and/or
biological point of view, as well. Then, it could be used to predict/design novel
potentially more active anti-T. cruzi benzofuroxan derivatives.
3
D-QSAR/VolSurf analysis: molecular interaction fields (MIF) were computed using
GRID force field in VolSurf employing the water (OH2), hydrophobic (DRY), carbonyl
oxygen (O) and amide (N1) probes.[62]_ENREF_64 Subsequently, a set of 104
descriptors was computed from these four probes. Moreover, six GRID-independent
descriptors were included resulting a total of 110 descriptors. The 110 descriptors
encode information about molecular size and shape, hydrophilic and hydrophobic
regions and the respective hydrophilic/hydrophobic balance, molecular diffusion, ClogP
(calculated partition coefficient n-octanol/water), ClogD (calculated distribution
coefficient), the ‘charge state’ descriptors, 3D pharmacophoric descriptors, and some
relevant ADME properties [63]._ENREF_65
All descriptors were normalized through the autoscaling preprocessing, and
subjected to a multivariate statistical analysis applying the partial least square regression
(PLS) method. From initial models encompassing all 110 descriptors, a set of 29

ACCEPTED MANUSCRIPT
1
6
informative descriptors was selected using the fractional factorial design (FFD)
approach implemented in VolSurf _ENREF_64[62].
The PLS model found for 29 VolSurf descriptors and discriminated by three
relevant PLS components or latent variables (LV) or factors presented the following
2
2
2
statistical measures: R = 0.98; Q LOO = 0.93 (SDEP = 0.08) and Q L2O = 0.93 (SDEP =
0
.08). This model better correlated the log1/PGI50 values to the 3D molecular structures
of the set of benzofuroxan derivatives. A comparison between observed and predicted
log1/PGI50 values of the investigated compounds is shown in Fig. 6A. The multiple
leave-N-out (LNO) crossvalidation method was also applied to PLS model and showed
a critical N value equal to 7 (39% out) (fig. 6B) [61].
FIGURE 6
According to the selected PLS model, the three first factors discriminated 62.1%
of total variance from the original data (factor 1 = 37.2%; factor 2 = 17.4%; and, factor
3
= 7.5%). The VolSurf descriptors that showed a stronger impact on log1/PGI are
50
highlighted in the PLS coefficients plot (see Figure 7). The biological activity is directly
influenced by the following descriptors: molecular size and shape (VOH2, SOH2 and
ROH2), integy moment from the hydrophilic probe (IW2OH2, IW3OH2 and IW7OH2),
lipophilic regions (D1DRY and D3DRY) and the best volume from the lipophilic probe
(BV21DRY and BV22DRY), and calculated partition coefficient (ClogP). Otherwise,
hydrophilic region (W2OH2), capacity factors (CW1OH2, CW2OH2 and CW3OH2),
and hydrophilic local interaction energy minima distance from the hydrophilic probe
(D13OH2) inversely influenced the biological activity. Also, the best volumes from the
hydrophobic probe, BV11DRY and BV22DRY, hydrophilic regions (W2-6O and W3-

ACCEPTED MANUSCRIPT
1
7
4
N1), and hydrogen bonding from the oxygen and amide probes (HB1O, HB2O, HB6O,
HB8N1) exert unfavorable impact on anti-T. cruzi activity of the benzofuroxan
derivatives.
FIGURE 7
The VolSurf descriptors obtained from PLS analysis showed that steric and
hydrophobic properties seem play the key role on anti-T. cruzi activity regarding the set
of compounds investigated. Then, the findings are in agreement with those formerly
described for the 2D-QSAR approach. Furthermore, the hydrophobic properties were
also identified in exploratory analysis among the descriptors that most influenced the
compounds classification, either by linear combination of original variables (PCA), or
similarity indices (HCA).
The VOH2, SOH2 and ROH2 descriptors, which mean volume, shape and
rugosity (ratio volume/surface), respectively, are related to the shape and size of the
compounds, and positive values of these descriptors would favorable contribute to the
activity. Moreover, higher values of hydrophilic integy moment (IW*OH2) can indicate
that the hydrophilic regions would be concentrated in a particular area of molecular
surface regarding the most active compounds whereas negative values of hydrophilic
capacity factors (Cw*OH2) can indicate that these regions are diffused onto the
molecular surface. Once the investigated compounds are congeners, the differences in
hydrophilic regions are given by the molecular characteristics of the substituent groups
attached to the benzene ring. So, the inclusion of hydrophilic groups would increase the
values of the capacity factor (due to the increasing of the hydrophilic regions onto the
molecular surface), and, therefore, the hydrophilic integy moment values would

ACCEPTED MANUSCRIPT
1
8
decrease, since these descriptors measure the unbalance between the center of mass of
the compounds and the barycenter of their hydrophilic region. The same behavior can
be observed for descriptors of hydrophobic regions (D*DRY), which are related to the
distribution of hydrophobic regions and the best volumes (BV*DRY), indicating the
highest volume of hydrophobic regions into the molecular structure. Also, there is a
positive contribution of ClogP, which means its value would increase with the addition
of hydrophobic groups attached to the aromatic ring.
In order to show the explanatory potency regarding the VolSurf descriptors, a
comparative profile using a highly [4q; R
NH ; log1/PGI50 = 3.93] active compound is presented in Figure 8. Higher values for
shape and size descriptors (VOH2, SOH2 and ROH2), hydrophilic integy moments
IW2OH2, IW3OH2 and IW7OH2), lipophilic regions (D1DRY and D3DRY) and the
1 1
= I; log1/PGI50 = 5.21] and weakly [4c; R =
2
(
best volume from the lipophilic probe (BV21DRY and BV22DRY) as well as for the
calculated partition coefficient (ClogP) could explain the high potency of compound 4q.
The weakly active compound 4c exhibits lower, or even null, values for the same
descriptors.
The findings obtained in this approach also provided the understanding of the
properties that most influenced the anti-T. cruzi activity regarding the investigated
benzofuroxan derivatives. Thus, based on that, the proposition of novel potentially more
active derivatives could be carried out.
FIGURE 8
External Prediction: a test set (n = 6; 4u-z) based upon the findings from 2D and 3D-
QSAR models was firstly designed, then synthesized and, subsequently, in vitro assayed
to be used as external validation. It is important to remember that one of the main goals

ACCEPTED MANUSCRIPT
1
9
of constructing QSAR models is the prediction of biological activity of novel
compounds outside the training set considering the relevant physicochemical properties,
which appear in the selected QSAR model and define the biological response [33, 64,
6
5]. Then, a test set was designed aiming to find compounds with improved anti-T.
cruzi activity in comparison to the set of compounds used to build the models (training
set).
The designed compounds of test set were obtained as shown in Figure 9.
Benzhydrazides (2u-z) were obtained directly from substituted benzoic acids (1u-z) in a
single synthetic step, without the need of isolating their respective intermediate esters
[
30]._ENREF_37 Further, benzofuroxan derivatives (4u-z) were obtained by reaction of
the derivatives 2u-z with 5-formylbenzofuroxan (3). The aldehyde 3, which was used in
the final stage of the synthesis of compounds 4u-z, was synthesized from 4-chloro-3-
nitrobenzaldehyde as reported in reference [66]._ENREF_68
FIGURE 9
The results of the external validation for both QSAR models (2D and 3D) are
displayed in table 4, and the graphic results for the experimental versus predicted
activity for both, training and test set, are shown in figure 10.
2
As can be seen through the high values of predictive R measure (2D-QSAR
2
2
model: R pred = 0.91; 3D-QSAR model: R pred = 0.77), in table 4, the predicted values
found were in good agreement with the experimental log1/PGI values, confirming the
5
0
external predictability of both QSAR models. According to the literature, a predictive
2
QSAR model should present R pred value higher than 0.5 _ENREF_69[67].

ACCEPTED MANUSCRIPT
2
0
The prediction power of 2D-QSAR model was 67%, which means four of six
compounds were well predicted by the model. These four compounds presented its
residual values (in module) equal to or lower than 1.0 SD (99% confidence level) from
the mean of residuals (0.17). For the 3D-QSAR model, the prediction power was 50 %.
TABLE 4
FIGURE 10
Except to compound 4v (PGI = 7.12 ꢀM (2.30 ꢀg/mL), the designed test set
5
0
3 7
showed better activity than training set. The most active compound was the 4-n-C H
substituted derivative (4u), which presented PGI50 value of 3.04 ꢀM (0.99 ꢀg/mL),
approximately seven-fold more potent than the reference drug BZ (PGI50 = 20.84 ꢀM;
5
.42 ꢀg/mL), and sharing the same activity with NFX (PGI50 = 3.78 ꢀM; 1.03 ꢀg/mL).
Cytotoxicity assay: cytotoxicity of the ten most active derivatives and two weakly
active compounds against human fibroblasts cells (FN1) was evaluated in vitro at range
from 12.5 to 200.0 ꢀM. The assays were carried out in triplicate, and the biological
endpoint was determined as the concentration which causes a fifty percent of inhibition
of cell proliferation (IC50-FN1) determined by the MTT (3-[4,5 dimethylthiazol-2-yl]-
2
,5diphenyltetrazolium bromide) colorimetric assay. NF, BZ and NFX were used as the
reference drugs [68]. The cytotoxic activity (IC50-FN1) was related to the anti-T. cruzi
activity (PGI50) in order to determine the corresponding selectivity index (SI= IC50-FN1
PGI50) (see Table 5).
/
TABLE 5

ACCEPTED MANUSCRIPT
2
1
The results demonstrated that compounds 4u, 4x, 4y, 4w, 4z, 4q, 4v, 4o and 4c
were non-toxic at the highest assayed doses. They also presented selectivity indices
higher than 25 (more active compounds). Compound 4c, however, showed low SI value
(> 1.71) probably because it can be defined as a weakly anti-T. cruzi active compound.
BZ and NFX also showed non-toxicity effect, but just NFX had a selectivity index
higher than 52.91 and, comparatively, compounds 4u and 4x presented a better profile
than the NFX.
3
. Conclusions
The chemometric studies (HCA, PCA, classical QSAR and 3D-QSAR)
performed herein for a set of benzofuroxan derivatives with activity against
epimastigotes forms of T. cruzi have revealed interesting points regarding structure-
property-activity relationships.
The complementary findings generated by each approach suggested that
hydrophobicity and steric properties are the most crucial molecular features for the anti-
T. cruzi activity of the investigated benzofuroxan derivatives, and they play a positive
contribution on biological activity. Additionally, the electronic effect was also pointed
out as important physicochemical property, and it was showed that compounds with
withdrawing-electron substituent groups can increase the benzofuroxan derivatives'
activity. It is important to highlight that the classical or 2D-QSAR approach generated a
QSAR model simple (a few terms; only linear functions), robust, ease of interpretation,
and with suitable external predictability.

ACCEPTED MANUSCRIPT
2
2
Regarding the designed external prediction set, novel compounds having new
and unpublished chemical structures were synthesized and assayed. Also, it is
noteworthy that the most active compound of the entire set (hit; IC50=3.04 ꢀM) was
identified and it was more active than the lead compound (NF) and reference drugs (BZ
and NFX), experimentally validating the model obtained with the training set.
The cytotoxicity assay demonstrated absence of toxic effects on human
fibroblasts for some of the tested analogues. Along with the anti-T.cruzi activity and
allied to preliminary non-mutagenic results of benzofuroxan derivatives [14],_ENREF_35
the investigated set of compounds has been revealed as promising hit-compounds to be
explored in the rational drug design of novel anti-T. cruzi agents.
4
4
. Experimental
.1. Chemistry
NMR spectra were recorded on a Bruker ADPX Advanced (300 MHz)
spectrometer employing DMSO-d solutions with tetramethylsilane as internal standard.
6
1
13
H and C NMR spectra of compounds 4u-z are reported in figures 2S-13S
(Supplementary Information section). Melting points were determined using Micro-
Química MQAPF-301 apparatus. Elemental analysis was performed on a Perkin-Elmer
4013 CHN Elemental Analyzer and used as procedure to confirms the purity of
2
compounds. The deviation of elemental analysis results from the calculated was < 0.4%.
The synthetic procedure of benzydrazides (2u-z) and benzofuroxans derivatives (4u-z)
has been described previously [30]. 5-formylbenzofuroxan 3, was synthesized from 4-
chloro-3-nitrobenzaldehyde as reported in reference [66]._ENREF_68
4
.1.1. General procedure for the preparation of benzhydrazides (2u-z): each
substituted benzoic acid (0.01 mol) was refluxed during 4 hours in 20.0 mL (0.50 mol)

ACCEPTED MANUSCRIPT
2
3
of anhydrous methanol and 0.5 mL (1.0 mmol) of sulfuric acid. The reaction mixture
was cooled down to room temperature and the hydrazine hydrate 80% (v/v) (10.0 mL,
0
.11 mol) was added. The system was maintained into vigorously stirring for more 30
minutes. After this period, the mixture was maintained at cold temperature to give 2.
4
.1.2. General procedure for the preparation of benzofuroxans derivatives (4u-z): a
mixture of 5-formylbenzofuroxan (1.0 mmol) and benzhydrazides (1.0 mmol) in water,
sulfuric acid, acetic acid and methanol (8:7:8:20 v/v) was heated under reflux during 1
h. After cooling down, the mixture was poured into cold water to give 4. The
recrystallization was performed in boiling acetonitrile to give the compounds as pale
yellow needles.
FIGURE 11
4
.1.2.1. 4-n-Propyl-(benzofuroxan-5-yl)methylene benzhydrazide (4u): yellow pale
1
needle (93%); mp 107.0 – 108.0 ºC. H NMR (DMSO-d
6
, 300 MHz): δ (ppm): 11.88 (s,
1
H, H10), 8.50 (s, 1H, H8), 7.93 (d, 1H, J = 9.2 Hz, H6), 7.84 (d, 2H, J = 8.3 Hz,
H13,H17), 7.79 (s, 1H, H4), 7.69 (d,1H, J = 9.5 Hz, H7), 7.33 (d, 2H, J = 8.4 Hz,
1
3
H14,H16), 2.65 (t, 2H, CH
2
), 1.65 (q, 2H, CH
2 3
), 0.92 (t, 3H, CH ); C NMR {H}
(
DMSO-d , 75 MHz): δ (ppm): 163.9 (C11), 146.4 (C8), 144.4 (C15), 137.7 (C5), 135.3
6
(C12), 132.1 (C7a), 131.3 (C3a), 129.1 (C6), 128.2 (C14,C16), 127.9 (C13,C17), 116.1
(
C7), 114.4 (C4), 37.1 (CH ), 23.5 (CH ), 13.3 (CH ); Anal. Calcd for (C H N O ): C,
2
2
3
17 16
4
3
6
2.95%; H, 4.97%; N, 17.27%. Found: C, 63.29%; H, 4.96%; N, 17.37%.
4
.1.2.2. 4-Isopropyl-(benzofuroxan-5-yl)methylene benzhydrazide (4v): yellow pale
1
needle (95%); mp 235.0 – 236.0 ºC. H NMR (DMSO-d
6
, 300 MHz): δ (ppm): 11.81 (s,

ACCEPTED MANUSCRIPT
2
4
1
H, H10), 8.52 (s, 1H, H8), 8.30 (d, 1H, J4,6 = 1.9 Hz, H4), 8.00 (dd,1H, J6,4 = 1.9 Hz,
J6,7 = 8.4 Hz, H6), 7.86 (d, 2H, J = 8.2 Hz, H13,H17), 7.78 (d, 1H, J7,6 = 8.4 Hz, H4),
1
3
7
3
.38 (d, 2H, J = 8.2 Hz, H14,H16) 2.98 (m, 1H, CH); 1.25 (d, 6H, CH ); C NMR {H}
(
DMSO-d , 75 MHz): δ (ppm): 163.9 (C11), 152.6 (C8), 148.0 (C15), 143.8 (C5), 135.3
6
(C12), 132.1 (C7a), 131.3 (C3a), 130.9 (C6), 127.9 (C14,C16), 126.2 (C13,C17), 125.6
(
C7), 123.2 (C4), 33.3 (CH), 23.4 (CH ); Anal. Calcd for (C H N O ): C, 62.95%; H,
3
17 16
4
3
4
.97%; N, 17.27%. Found: C, 63.09%; H, 5.15%; N, 17.44%.
4
.1.2.3. 4-tert-Butyl-(benzofuroxan-5-yl)methylene benzhydrazide (4w): yellow pale
1
needle (91%); mp 237.0 – 238.0 ºC. H NMR (DMSO-d
6
, 300 MHz): δ (ppm): 11.88 (s,
1
H, H10), 8.51 (s, 1H, H8), 8.30 (d, 1H, J4,6 = 1.5 Hz, H4), 8.01 (dd,1H, J6,4 = 1.7 Hz,
J
6,7 = 8.4 Hz, H6), 7.86 (d, 2H, J = 8.5 Hz, H13,H17), 7.79 (d, 1H, J7,6 = 8.2 Hz, H4),
13
7
3 6
.53 (d, 2H, J = 8.5 Hz, H14,H16), 1.33 (s, 9H, CH ); C NMR {H} (DMSO-d , 75
MHz): δ (ppm): 163.9 (C11), 154.9 (C8), 148.0 (C15), 144.4 (C5), 135.3 (C12), 132.1
(
C7a), 131.3 (C3a), 130.4 (C6), 127.7 (C14,C16), 125.6 (C13,C17), 125.0 (C7), 123.2
C4), 34.6 (C), 30.8 (CH ); Anal. Calcd for (C18 ): C, 63.89; H, 5.36%; N,
6.56%. Found: C, 63.71%; H, 5.30%; N, 16,47%.
(
3
18 4 3
H N O
1
4
.1.2.4. 4-n-Butyl-(benzofuroxan-5-yl)methylene benzhydrazide (4x): yellow pale
1
needle (87%); mp 95.0 – 96.0 ºC. H NMR (DMSO-d
6
, 300 MHz): δ (ppm): 11.87 (s,
1
H, H10), 8.50 (s, 1H, H8), 7.93 (d, 1H, J = 9.5 Hz, H6), 7.84 (d, 2H, J = 8.2 Hz,
H13,H17), 7.80 (s, 1H, H4), 7.69 (d,1H, J = 9.5 Hz, H7), 7.33 (d, 2H, J = 8.0 Hz,
H14,H16), 2.67 (t, 2H, CH ), 1.61 (m, 2H, CH ), 1.33 (m, 2H, CH ), 0.91 (t, 3H, CH );
, 75 MHz): δ (ppm): 164.0 (C11), 148.0 (C8), 146.6 (C15),
44.4 (C5), 135.3 (C12), 132.1 (C7a), 131.3 (C3a), 130.6 (C6), 128.2 (C14,C16), 127.9
2
2
2
3
1
3
C NMR {H} (DMSO-d
6
1

ACCEPTED MANUSCRIPT
2
5
(
C13,C17), 116.1 (C7), 114.4 (C4), 34.6 (CH ), 32.5 (CH ), 21.5 (CH ), 13.4 (CH );
2 2 2 3
Anal. Calcd for (C H N O ): C, 63.89; H, 5.36%; N, 16.56%. Found: C, 64.15%; H,
18
18
4
3
5
.13%; N, 16,90%.
4
.1.2.5. 4-Propoxy-(benzofuroxan-5-yl)methylene benzhydrazide (4y): yellow pale
1
needle (85%); mp 204.0 – 205.0 ºC. H NMR (DMSO-d
6
, 300 MHz): δ (ppm): 11.74 (s,
1
H, H10), 8.50 (s, 1H, H8), 8.29 (s, 1H, H4),7.98 (d, 1H, J = 8.4 Hz, H6), 7.90 (d, 2H, J
=
8.7 Hz, H13,H17), 7.78 (d,1H, J = 8.2 Hz, H7), 7.03 (d, 2H, J = 8.6 Hz, H14,H16),
13
4
.03 (t, 2H, CH
2 2 3 6
), 1.76 (m, 2H, CH ), 0,99 (t, 3H, CH ); C NMR {H} (DMSO-d , 75
MHz): δ (ppm): 163.6 (C15), 161.7 (C11), 148.0 (C8), 144.0 (C5), 143.3 (C7a), 132.1
(
C13,C17), 131.3 (C3a), 129.8 (C6), 125.2 (C12), 123.1 (C7), 116.1 (C4), 114.1
C14,C16), 69.4 (CH ), 21.9 (CH ), 10.1 (CH ); Anal. Calcd for (C17 ): C,
9.99%; H, 4.74%; N, 16.48%. Found: C, 60.15%; H, 4.64%; N, 16.59%.
(
2
2
3
16 4 4
H N O
5
4
.1.2.6. 4-Butoxy-(benzofuroxan-5-yl)methylene benzhydrazide (4z): yellow pale
1
needle (92%); mp 191.0 – 192.0 ºC. H NMR (DMSO-d
6
, 300 MHz): δ (ppm): 11.85 (s,
1
H, H10), 8.51 (s, 1H, H8), 7.98 (d, 1H, J = 9.5 Hz, H6), 7.93 (d, 2H, J = 8.6 Hz,
H13,H17), 7.80 (s, 1H, H4), 7.71 (d,1H, J = 9.2 Hz, H7), 7.05 (d, 2H, J = 8.7 Hz,
H14,H16), 4.01 (t, 2H, CH ), 1.75 (m, 2H, CH ), 1.47 (m, 2H, CH ), 0.98 (t, 3H, CH );
, 75 MHz): δ (ppm): 163.5 (C15), 161.7 (C11), 144.0 (C8),
35.4 (C5), 132.1 (C7a), 129.9 (C13,C17), 129.0 (C3a), 125.5 (C6), 125.1 (C12), 123.1
C7), 116.1 (C4), 114.2 (C14,C16), 67.6 (CH ), 30.6 (CH ), 18.6 (CH ), 13.5 (CH );
): C, 61.01%; H, 5.12%; N, 15.81%. Found: C, 60,94%; H,
2
2
2
3
1
3
C NMR {H} (DMSO-d
6
1
(
2
2
2
3
18 4 4
Anal. Calcd for (C18H N O
5
,14%; N, 15,51%.

ACCEPTED MANUSCRIPT
2
6
4
.2. Biological activity
In vitro assays were performed against the epimastigote form of T. cruzi Y strain
cultured at 28 °C in liver infusion tryptose medium (LIT), supplemented with 10% heat-
inactivated fetal calf serum. Cells from a 10-day-old culture (stationary phase) were
inoculated into 50 mL of fresh culture medium to give an initial concentration of
6
4
.0x10 cells/mL [40, 41]._ENREF_41 The benzofuroxan derivatives, NF, BZ, and NFX
were separately dissolved in dimethylsulphoxide (DMSO) and diluted into LIT culture
medium to obtain a concentration of compounds from 2.00 to 320.0 µM (1.0 and 2.0%
DMSO final concentration). Aliquots of 0.5 mL of stock solutions were added into 0.5
6
mL of fresh culture medium (to give 2.0x10 cells/mL initial concentration). The final
concentration of DMSO in the culture media never exceeded 1.0% and the control was
run in the presence of 0.5 and 1.0% DMSO and in the absence of any drug. The
solutions were incubated at 28 °C for 5 days. After this period, the amount of viable
parasites was determined using a spectrophotometer (T80 plus, PG Instruments, United
Kingdom) at 580 nm (10 mm light path length). The assays were performed in
triplicate. No effect on epimastigote growth was observed in the presence of up to 1%
DMSO in the culture media. The concentration of each compound responsible for fifty
percent of parasite growth inhibition, PGI50, was calculated from the percentage of
inhibition growth, using non-linear dose response regression for each case and
employing at least six concentration values.
4
.3. Molecular modeling approach
The three-dimensional structure of ligand 4d (4-hydroxy-[N'-(benzofuroxan-5-
yl)methylene]benzhydrazide), in its neutral form, was constructed employing the

ACCEPTED MANUSCRIPT
2
7
HyperChem 7 software [69]._ENREF_70 The Cartesian coordinates of crystallized
structures of NF were retrieve from Cambridge Structural Database
(CSD)[70]_ENREF_71 (entry code LEQTAC (R-factor 0.11) [71]_ENREF_72 and used as
geometry references in the building up of ligand 4d. Crystallographic information of
benzofurazan 1-oxide was retrieved from the file tk2442.cif (R-factor 0.035)
[
72]_ENREF_73 and also used as a reference particularly in the construction of the
benzofuroxan system moiety. The energy minimization was carried out employing the
HyperChem 7 MM+ force field without any constrains. The MOLSIM 3.2 program
[
73]_ENREF_74 was also used for geometry optimization. Partial atomic charges were
calculated using the AM1 semiempirical method [74],_ENREF_75 also implemented in
the HyperChem program [69].
The molecular structure modeled as described above was used as input in
molecular dynamics (MD) simulations [75],_ENREF_76 which are employed to perform
the conformational ensemble profile (CEP) of the ligand 4d (MOLSIM 3.2). The MD
simulation protocol included 1,000,000 steps, a step size of 0.001 ps (1fs), and a
simulation temperature of 310 K. Trajectory files were recorded every 20 simulation
steps to generate 50,000 conformations. The lowest-energy conformation was selected
from the equilibrium region of CEP (Supplementary Information, figure 1S) and used to
constructed the three-dimensional molecular structure of each ligand (4a - 4z). CHarges
from Electrostatic Potentials using a Grid based method (CHELPG) [76]_ENREF_77
were computed using the HF/3-21G* method, implemented in the Gaussian 03W
program [48]._ENREF_49
The independent variables or descriptors or molecular properties of different
nature were calculated employing MarvinView [50],_ENREF_51 Gaussian 03W
[
48],_ENREF_49 HyperChem 7 [69]_ENREF_70 and Sybyl 8.0 [49]_ENREF_57_ENREF_50