Optimization of 1,4-Naphthoquinone Hit Compound: A Computational, Phenotypic, and In Vivo Screening against Trypanosoma cruzi

Chagas disease (CD) still represents a serious public health problem in Latin America, even after more than 100 years of its discovery. Clinical treatments (nifurtimox and benznidazole) are considered inadequate, especially because of undesirable side effects and low efficacy in the chronic stages of the disease, highlighting the urgency for discovering new effective and safe drugs. A small library of compounds (1a–i and 2a–j) was designed based on the structural optimization of a Hit compound derived from 1,4-naphthoquinones (C2) previously identified. The biological activity, structure-activity relationship (SAR), and the in silico physicochemical profiles of the naphthoquinone derivatives were analyzed. Most modifications resulted in increased trypanocidal activity but some substitutions also increased toxicity. The data reinforce the importance of the chlorine atom in the thiophenol benzene ring for trypanocidal activity, highlighting 1g, which exhibit a drug-likeness profile, as a promising compound against Trypanosoma cruzi. SAR analysis also revealed 1g as cliff generator in the structure-activity similarity map (SAS maps). However, compounds C2 and 1g were unable to reduce parasite load, and did not prevent mouse mortality in T. cruzi acute infection. Phenotypic screening and computational analysis have provided relevant information to advance the optimization and design of new 1,4-naphthoquinone derivatives with a better pharmacological profile.


Introduction
Chagas disease (CD), a serious parasitic disease caused by T. cruzi, continues to be a neglected disease with a high social and economic impact in Latin America, despite the control strategies of the World Health Organization's programs launched to combat this disease [1]. Unlike other neglected tropical diseases (NTDs), little progress has been achieved in the treatment and diagnosis of CD, with the pediatric formulation of benznidazole (Bz) [2], the reference drug, and the recent approval of Bz by the United States Food and Drug Administration [3] being the major advances to date.
Despite improved control of vector-borne transmission in wide areas of Southern Cone countries, outbreaks of oral infection and the appearance of secondary peridomestic vectors have resulted in the re-emergence of acute infection in many endemic countries [4][5][6]. Non-endemic countries have also faced this silent disease because of the emigration of asymptomatic infected individuals, who are unaware that they harbor T. cruzi, making as described in lapachone-derived naphthoimidazoles and naphthofuranquinones compared to parental compounds [30,31]. The presence of methylthio group (SCH 3 ) has been associated with the redox potential of several naphthoquinones, since these compounds can participate in oxidation reactions forming reactive sulfur species [32]. Other substituents, including fluoro (F), nitro (-NO 2 ), propyl (C 3 H 7 ), and naphthyl (C 10 H 8 ) groups were added in order to improve the therapeutic properties.
In this study, a small library of 1,4-naphthoquinone derivatives was synthesized and screened against T. cruzi. Also, a computational multi-parameter analysis was performed to predict physicochemical properties, analyze structure-activity relationship and systematically characterize the activity landscape of 1,4-naphthoquinone derivatives against T. cruzi, allowing us to determine structural changes associated with activity differences, in order to identify a promising candidate for treatment of Chagas disease.

Hit Compound Optimization
The series of 2-hydroxy-3-phenylsulfanylmethyl-[1,4]-naphthoquinone derivatives (1a-f, 1h and 1i) was designed considering the results reported in previous studies of our research group, where we observed that the introduction of chlorine atoms in R1, combined with benzene in R2, promoted good anti-T. cruzi activity [29]. Inspired by these results, we started this study trying to improve the anti-T. cruzi activity by exploring the introduction of electron-withdrawing groups like chlorine in R1 (compounds 1a-d) and/or electron-releasing groups like methoxy (OCH 3 ; 1g-i) in R2. Later, in order to better correlate anti-T. cruzi activity and the electronic profile, we designed a new series of compounds (2a-j) to study the effect of different substituents in R1 (CH 3 , NO 2 , F, OCH 3 , SCH 3 , C 3 H 7 , C 10 H 8 ) in the presence of methoxy group, an electron-donating group, in the benzene ring (R2) [33]. Finally, we decided to observe if the introduction of a triazole in R2 (1-(3,5-dichlorophenyl)-1H-1,2,3-triazole, 1e), a heterocycle extensively present in antiproliferative compounds [27], or presence of hydroxyl group in the quinonic aromatic ring (R3; 1f), usually associated with increased ROS formation [34], could improve anti-T. cruzi activity.

Computational Analysis: Physicochemical Properties, SAR, and SAS Maps
The high structure similarity and high activity difference between molecules of the series led us to analyze structure-activity relationship (SAR), using the concept of activity landscape, and the physicochemical properties of all derivatives. In this light, structureactivity similarity (SAS) maps were analyzed with emphasis on the activity cliff that clusters structurally similar compound pairs with large activity differences. The SAS maps for the 19 derivatives against trypomastigotes and intracellular amastigotes revealed 172 data points with each one corresponding to a pairwise comparison (Figure 1). Data points were additionally differentiated by the potency (pIC50) using a color scale from low (blue) to high (red). Most of the compound pairs generated for trypomastigotes and intracellular amastigotes were identified in similarity cliff (region III; R3) and smooth SAR (region IV; R4), showing small structural similarity with high activity similarity and high structural and activity similarity, respectively ( Figure 1). The quantitative analysis revealed that 86.5% and 77.7% of compound pairs were distributed in the R3 and R4 regions of SAS maps for both amastigotes and trypomastigotes, respectively ( Figure 1). In contrast, the compound pairs' lowest proportion of data points were in the activity cliff (region II; R2). A higher proportion of activity cliff pairs was identified for intracellular amastigotes (7%) compared to trypomastigotes (3.5%) (Figure 1). The pairs 1g_2j, 1g_2c, 1g_2a, 1g_2b, 1h_2j, and 1h_2c illustrate the activity cliffs in the SAS map of trypomastigotes. Data points 1g_2h, 1g_2j, 1h_2j, 1i_2j, 2a_2h, 2a_2j, 2b_2j, 2c_2j, 2d_2j, 2e_2j, 2e_2h, and 2i_2j represent the activity cliffs of intracellular amastigotes (Figure 1). Interesting, two compound pairs, 1g_2j and 1h_2j, were identified in the activity cliffs in the SAS map of both trypomastigotes and intracellular amastigotes. Although small structural changes were evident among these molecules, 1g is among the most active compounds while 2j is the least effective compound in the series. Table 2. Trypanocidal activity and structure activity relationship (SAR) of 1,4-naphthoquinone derivatives. The effect of changes in R-groups against intracellular amastigotes of T. cruzi (Dm28c-Luc clone). derivatives (1a, 1b, 1g, 2a, 2e, and 2f) ( Table 2). Although these derivatives have IC50 values ≤6.9 µM, they showed lower activity than Bz (IC50 = 1.4 ± 0.4 µM). However, 1g (IC50 = 6.7 ± 1.8 µM) was the only derivative with SI > 10 among the most active compounds (Table 2).  Thus, we addressed the question of how the structural changes in the series affected the physicochemical properties of the derivatives and resulted in drastic changes in compound activity. Then, we assessed the molecular properties typically analyzed in drug discovery software, including molecular mass (MM), lipophilicity (LogP), number of hydrogen donors (HBDs) and acceptors (HBAs), polar surface area (PSA), rotatable bonds, pIC50 and drug-likeness, using Datawarrior software. These properties are important to determine compound absorption, distribution, metabolism, and excretion (ADME) profile. The series 1(a-i) has the highest MW (418.46-542.82) compared to the series 2(a-j), which achieved maximal of 452.42 ( Figure 2). Most of the compounds have high lipophilicity (LogP > 4.27), except for 2f with moderate lipophilicity (LogP = 3.9), but they do not violate Lipinski's rule of five (Ro5) (Figure 2). Hydrogen bonds, which increase aqueous solubility, also fit the Ro5 with HBD < 5 and HBA < 10. The PSA values, associated with permeability and oral absorption prediction, are distinct between the series. The maximum PSA value reached 134.72 Å 2 (2f), and three groups of compounds were identified with PSA values of 79.67 Å 2 (1a-d), 88.9 Å 2 (1g, 2a-e and 2g-h), and 99.9 to 114.2 Å 2 (1e-f, 1h, 2i-j) (Figure 2). The number of rotatable bonds, a parameter involved in the molecule flexibility, was also measured in both series. Four (21%), five (47.4%), and six (31.6%) rotatable bonds were observed ( Figure 2). Thus, reduced molecule flexibility (≤10 rotatable bonds) and low PSA (≤140 Å 2 ) point to good oral bioavailability of the compounds. Two main clusters of compounds share similar physicochemical properties. As expected, because of similarity in chemical structure, compounds 1a-d overlap in the same chemical space, with C2 positioned close to this group. High values of HBA and PSA detached compounds 1e, 1h 1i, 2f, 2i, and 2j from the selective compounds (2d, 1g, and C2) in the chemical space. Data dimension reduction was obtained by principal component analysis (PCA) of six relevant physicochemical Two main clusters of compounds share similar physicochemical properties. As expected, because of similarity in chemical structure, compounds 1a-d overlap in the same chemical space, with C2 positioned close to this group. High values of HBA and PSA detached compounds 1e, 1h 1i, 2f, 2i, and 2j from the selective compounds (2d, 1g, and C2) in the chemical space. Data dimension reduction was obtained by principal component analysis (PCA) of six relevant physicochemical properties related to Veber and Lipinski rules. Compounds 2a-e, 2h, 1f, and 1g have a set of similar physicochemical properties and clustered in PCA analysis. Chemical space revealed that the first two PCs account for 75% of variance. HBA account for most (0.54) PC1 impact, followed by PSA (0.51) and rotatable bonds (0.48). HBD had the largest contribution to PC2 (0.24) (Figure 2). Most of modifications in these compound series did not alter significantly the activity against parasite, however the variability of physicochemical properties in the analyzed series by PCA revealed a high influence of hydrogen bond descriptors and PSA, parameters related to compound permeability and polarity, which could impact the activity against intracellular amastigote forms.

Trypanocidal Activity in Mouse Model of Acute Infection
Based on the promising in vitro trypanocidal effects of compound C2, previously identified as a hit compound [29], and the trypanocidal activity of 1g, we proceed with these compounds for in vivo assay using a murine model of T. cruzi acute infection. First, we subjected the mice to five cumulative oral (o.p.) and intraperitoneal (i.p.) doses, 1 h apart, at concentrations of 50 mg/kg and 100 mg/kg of compounds C2 and 1g for acute toxicity assessment (NOAEL). No toxic side effects, according to the OECD guidelines, were noted up to 48 h post-treatment at the total cumulative dose of 250 mg/kg for both compounds analyzed (data not shown). Additionally, changes in mouse behavioral characteristics were not observed. However, one animal died in the intraperitoneal treatment regimen of 1g at the cumulative dose of 500 mg/kg.
Next, the efficacy of C2 and 1g was then evaluated in Swiss Webster male mice infected with 10 4 bloodstream trypomastigotes (Y strain) followed by i.p. treatment with C2 and 1g for 5 consecutive days after detection of positive parasitemia (5 dpi). Untreated and vehicle-only groups had high levels of parasitemia, as expected for this acute model of T. cruzi infection, achieving parasitemia peak at 8 dpi (Figure 3). C2 and 1g treatment was not efficient in eliminating the parasites. The parasitic load of the C2 and 1g treated groups remained similar or even slightly higher than the untreated or vehicle-treated control groups (Figure 3). In contrast, the Bz-treated group had undetectable parasitemia by microscopic analysis. Besides not reducing the parasite burden, the compounds did not bring benefits to the animals survival. Treatment with 1,4-naphthoquinones-derived compounds was not able to prevent mortality, as evidenced by Bz treatment (Figure 3). Most animals, untreated and treated with naphthoquinone derivatives (C2 and 1g) died between 14 and 16 dpi, except for animals treated with 100 mg/kg C2 which died earlier, between 8 and 10 dpi, reaching 100% at the end of treatment (Figure 3). Bz-treated groups survived during all period analyzed (26 days).

Discussion
The discovery and development of new drugs is a time-consuming (average 10-15 years) and costly ($800 million to $1 billion) process, which makes it a great challenge for the efficient and safe treatment of various diseases [35]. Despite the increased success rate of molecular entity approval in recent years (2017-2018) [36], only 1.65% of the new products for treatment of neglected tropical diseases (NTDs) entered phase I clinical trial [37]. Among the 20 NTDs proposed for control or elimination by 2020, Chagas disease (CD), discovered 110 years ago, has made little progress in finding new drugs effective in treating the disease. The recently released results of the BENDITA study showed the efficacy of Bz in reduced doses (150 mg/kg) and shorter treatment regimen (2 weeks of treatment) in Bolivian patients [17], produced fewer side effects, bringing new hope for individuals with this silent disease. However, screening for new safe drugs are still a priority, because of natural resistance of T. cruzi strains to reference drug and its inability to prevent cardiomyopathy, encouraging the search for new hit and lead compounds for CD treatment. In this study, we invested in the optimization of a previously identified hit compound (2-hydroxy-3-phenylsulfanylmethyl-[1,4]-naphthoquinone) and analyzed a small library of 1,4-naphthoquinone analogues for their trypanocidal activity using in vitro and in vivo preclinical assays and computational approaches.
Different strategies, based on C-ring, redox center, and A-ring modifications, have been utilized to develop bioactive naphthoquinoidal derivatives with antiplasmodial, trypanocidal, and leishmanicidal activity [38,39]. Herein, we exploited the synthesis of several lapachol analogues containing [1,2,3]-triazole, thiophenol, alkyl, and naphthalene nucleus. Our findings demonstrated that optimization of compound C2, by addition of chlorine atom in the thiophenol benzene ring, increased the lipophilicity (1a-d) improving trypanocidal activity compared to C2 and Bz, and also enhancing mammalian cell toxicity. The high lipophilic character of naphthoquinone derivatives containing furane moiety, methoxy group, and aliphatic side chain has been proposed to benefit the trypanocidal activity by improving the compound permeability through the plasma membrane of the parasite [40]. However, the increased toxicity level has been reported to be related to the promiscuity of highly lipophilic compounds (logP >5) which bind with high affinity to nonspecific hydrophobic targets [41]. Alternatively, the insertion of the heterocyclic ring [1,2,3]-triazole into 1,4-naphthoquinone (1e) enhanced the activity against trypomastigotes, but not amastigotes, compared to Bz. Optimization of 1,4-naphthoquinones activity by the addition of triazole has generated either promising or completely inactive derivatives against T. cruzi [27], suggesting that the position of triazole insertion into quinones or its association with other substitutes modulates its biological activity. High trypanocidal potency has been proposed to be associated with orthoand para-quinoidal moieties of [1,2,3]-triazole-coupled naphthoquinones and their electrophilic properties, probably related to high ROS induction [42].
Among the 1,4-naphthoquinone derivatives we highlight compound 1g in this small library as a promising compound with a better trypanocidal effect. Although the structural changes were minor, only the inclusion of the methoxy group in the benzene ring of the hit compound (C2) has improved the physicochemical properties, including surface polar area, rotatable bonds and hydrogen bond acceptors, making compound 1g slightly more effective than C2 against trypomastigotes and intracellular amastigotes. It is important to note that merely replacing the chlorine atom with other substituents in series 1(a-i), including halogens and nitro, propyl, naphthyl groups, affecting the physicochemical properties of the derivatives, induced a reduction in trypanocidal activity and increased toxicity.
SAS map indicated the total number of pairwise matches for all compounds, revealed few compounds with similar structures and different activity in activity in the cliff region. Compound 1g, the most active compound evidenced by SAR analysis, was identified as a cliff generator on SAS maps (both trypomastigotes and amastigotes) by pairing with at least five distinct derivatives (2a, 2b, 2c, 2h, and 2j), corroborating the relevance of 4-chlorine in the benzene ring (R1). The change of the chlorine atom in R1 by the 4-methyl (2c), 5-methyl (2b), or thiol group (2j) reduced the trypanocidal activity two-to ten-fold. Comparing with C2, the introduction of the methoxy group to benzene ring in R2 improved the lipophilicity and trypanocidal activity of 1g. In contrast, the other structural changes in R2, by introduction of [1,2,3]-triazole, propyl and naphthyl groups, did not contribute to the improvement of bioactivity. Unfortunately, both compounds C2 and 1g failed to reduce parasite burden or ensure mouse survival, suggesting that physicochemical parameters still need to be improved for better compound bioavailability and effectiveness. Although numerous naphthoquinone derivatives have been analyzed against T. cruzi, few analogues have evolved into preclinical in vivo assays. The 2,3diphenyl-1,4-naphthoquinone (DPNQ), for instance, has been reported as a potential chemotherapeutic agent against T. cruzi because of its trypanocidal activity in phenotypic screening and in murine T. cruzi experimental infection [43]. Treatment of C3H/HeN female infected mice with DPNQ reduced two-fold the parasite load and ensured 60% animal survival up to 70 dpi, stimulating compound optimization in an attempt to improve efficacy. In fact, the translational interface between in vitro and in vivo assays as well as preclinical and clinical trials is still a major gap in the development of new drugs, demonstrating the importance of physicochemical and pharmacokinetic properties and the choice of experimental models.
In conclusion, the optimization of compound C2, by introduction of methoxy group in benzene ring (1g), moderately improved trypanocidal activity in vitro but highlighted 1g as cliff generator in SAS map. However, C2 and 1g were unable to reduce parasite load, and did not prevent mouse mortality. Design and synthesis of a new library of 1,4-naphthoquinones may contribute to the identification of high potent and low toxic drugs for Chagas disease treatment.

Synthetic Compounds
Compounds 1a-i were obtained by multicomponent reaction of lawsone with the appropriate aldehyde to generate the intermediate o-quinone methide in situ, followed by nucleophilic addition of a substituted thiol, as previously described by our group [44].
The reactions were carried out by microwave irradiation (150 • C, 20 min). The results are presented in Figure 4. Naphtoquinones 2a-j were synthetized as previously described [33]. All products were purified by column chromatography using silica gel and were fully characterized by spectroscopic analysis (Supplementary Material).

General Procedure for Preparing 1a-i and 2a-j
A 10 mL microwave tube was loaded with naphthoquinone (2.9 mmol), aldehyde (5.8 mmol), arylthiol (5.8 mmol), and ethanol (5 mL). The mixture was irradiated for 20 min at 150 • C, and the solvent was then evaporated under reduced pressure. The residual was purified by column chromatography on silica gel and eluted with an increasing polarity gradient of hexane and ethyl acetate.   13

Parasites
T. cruzi Dm28c-Luc clone, genetically modified to express firefly luciferase, and Y strain were used in the drug screening assay. Vero cells were infected with T. cruzi, Dm28c-Luc or Y strain, in a 10:1 parasites/host cell ratio and maintained at 37 • C in humidified atmosphere of 5% CO 2 . Trypomastigotes were harvested from the infected culture supernatant on the 4th day post-infection (dpi) followed by quantification of the number of parasites/mL in Neubauer chamber.

Cytotoxicity In Vitro Assay
To evaluate the toxic effects of the 1,4-naphthoquinone derivatives on mammalian cells, Vero cells were seeded at a density of 1.5 × 10 4 cells/well in 96-well white culture plates. Twenty-four hours later, the cell cultures were incubated for 72 h at 37 • C with 1,4naphthoquinone derivatives (series 1(a-i) and 2(a-j)) and Bz (500-1.9 µM) diluted in RPMI medium supplemented with 10% FBS. After incubation, the cell viability, measured by ATP level, was assessed by adding 20 µL/well of the CellTiter Glo (Promega Corporation, Madison, WI, EUA) solution. The luminescent signal was read on the FlexStation 3 reader (Molecular Devices, Sunnyvale, CA, USA). The concentration of compound that reduces 50% of mammalian cell viability (CC 50 ) was determined by linear regression. All treatments and controls were performed at low concentration (≤1%) of dimethyl sulfoxide (DMSO). At least three independent assays were performed in duplicate.

Trypanocidal Activity
Screening of the compounds was performed against trypomastigotes and intracellular amastigote forms. Trypomastigotes (1 × 10 6 parasites/well), Dm28c-Luc clone and Y strain, were incubated for 24 h at 37 • C with the 1,4 naphthoquinone derivatives and Bz (0.41-100 µM) and their viability determined by the activity of the luciferase enzyme, after addition of luciferin substrate (300 µg/mL), or CellTiter Glo. The luminescent signal was detected on the FlexStation 3 reader.
The effect of the compounds against intracellular amastigotes was screened in 72 h T. cruzi-infected Vero cells. Briefly, Vero cells seeded (1.5 × 10 4 cells/well) on 96-well white plate infected with trypomastigotes (Dm28c-Luc) at a 10:1 multiplicity of parasites/cell. After 24 h of infection, the cultures were washed with PBS and then, treated for 72 h at 37 • C with naphthoquinone derivatives and Bz (0.41-100 µM). After treatment, luciferin (300 µg/mL) was added to the culture and the viability of the parasites was assessed by reading the luminescent signal using the FlexStation 3 reader. Controls were also performed in no-toxic concentrations of DMSO (≤1%). The concentration that reduces the number of viable parasites by 50% (IC 50 ) or 90% (IC 90 ) was calculated by linear regression.

In Silico Analysis
Molecular structure of compounds (SMILES code) and pIC50 (-log IC 50 ) values were inserted in Datawarrior software version 5.0 to explore the chemical space, physicochemical properties, and structure activity-relationship (SAR) [45]. For principal component analysis (PCA), six properties related to Veber and Lipinski rules were applied for each compound. The activity landscape and the structure-activity similarity (SAS) maps were analyzed using Activity Landscape Plotter, an online open web platform (https://www.difacquim. com/d-tools/). MACCS molecular fingerprints were used in SAS map analysis and the threshold for structural similarity were set to 0.8.

Mouse Acute Toxicity
To establish the no-observed-adverse-effect level (NOAEL), increasing doses of the 1,4 naphthoquinone derivatives were orally (o.p.) and intraperitoneally (i.p.) administered into male Swiss mice (21 to 24 g; n = 2 for each group). Treated animals were separated into two groups: (1) 5 consecutive doses of 50 mg/kg and (2) 100 mg/kg every 1h, totaling a cumulative dose of 250 mg/kg and 500 mg/kg, respectively. Untreated and vehicle treated animals (PBS with 3% Tween-80) were used as control groups. Mice were inspected for toxic and sub-toxic symptoms according to Organization for Economic Co-operation and Development (OECD) guidelines [46].

In Vivo Experimental T. cruzi Infection
Male Swiss Webster mice (18 g) were obtained from the Institute of Science and Technology in Bio-models of the Oswaldo Cruz Foundation, Rio de Janeiro, Brazil. Mice were housed in groups of 5 animals per cage in a ventilated cabinet under controlled temperature (22 ± 1 • C), 55% ± 5% humidity and a 12 h light-dark light cycle. The animals received water and an ad libitum feeding regimen. Mice were intraperitoneally (i.p.) infected with bloodstream trypomastigotes of T. cruzi Y strain (10 4 parasites/animal) and parasitemia was daily evaluated from the 4th dpi. Animals were distributed into ten groups: (1) uninfected and untreated, (2) uninfected and treated with C2 (50 mg/kg); (3) uninfected and treated with C2 (100 mg/kg); (4) uninfected and treated with 1g (50 mg/kg), (5) T. cruziinfected and untreated, (6) T. cruzi-infected and treated i.p. with vehicle, (7) T. cruziinfected and treated i.p. with C2 50 mg/kg/day, (8) T. cruzi infected and treated i.p. with C2 100 mg/kg/day, (9) T. cruzi-infected and treated i.p. with 1g 50 mg/kg/day, and (10) T. cruzi-infected and treated orally with Bz 100 mg/kg/day. Daily treatment (0.1 mL i.p. dose) started at 5 days post-infection (dpi) with positive parasitemia and followed up to 10 dpi.
Parasitemia was estimated by Pizzy-Brener method [47]. Briefly, animals parasitemia were determined by microscopic quantification of fresh blood (5 µL) collected from animals' caudal vein. Mortality was daily evaluated up to 26 days and expressed in survival rates. All animal procedures were previously approved by the Oswaldo Cruz Institute Ethical Committee for the Use of Animals (License L15/17).