Structural analysis and molecular docking of trypanocidal aryloxy-quinones in trypanothione and glutathione reductases: a comparison with biochemical data

A set of aryloxy-quinones, previously synthesized and evaluated against Trypanosoma cruzi epimastigotes cultures, were found more potent and selective than nifurtimox. One of the possible mechanisms of the trypanocidal activity of these quinones could be inhibition of trypanothione reductase (TR). Considering that glutathione reductase (GR) is the equivalent of TR in humans, biochemical, kinetic, and molecular docking studies in TR and GR were envisaged and compared with the trypanocidal and cytotoxic data of a set of aryloxy-quinones. Biochemical assays indicated that three naphthoquinones (Nq-h, Nq-g, and Nq-d) selectively inhibit TR and the TR kinetic analyses indicated that Nq-h inhibit TR in a noncompetitive mechanism. Molecular dockings were performed in TR and GR in the following three putative binding sites: the catalytic site, the dimer interface, and the nicotinamide adenine dinucleotide phosphate-binding site. In TR and GR, the aryloxy-quinones were found to exhibit high affinity for a site near it cognate-binding site in a place in which the noncompetitive kinetics could be justified. Taking as examples the three compounds with TR specificity (TRS) (Nq-h, Nq-g, and Nq-d), the presence of a network of contacts with the quinonic ring sustained by the triad of Lys62, Met400′, Ser464′ residues, seems to contribute hardly to the TRS. Compound Nq-b, a naphthoquinone with nitrophenoxy substituent, proved to be the best scaffold for the design of trypanocidal compounds with low toxicity. However, the compound displayed only a poor and non-selective effect toward TR indicating that TR inhibition is not the main reason for the antiparasitic activity of the aryloxy-quinones.


Introduction
Parasitic diseases are a major obstacle to human health and economic development in many parts of the world and cause high rates of mortality and morbidity (WHO, 2011). Current therapies against these diseases are unsatisfactory, with treatment failure being common due to widespread resistance and severe side effects (Bern, 2011;Bern et al., 2007;Rassi, Rassi, & Marin-Neto, 2010;Viotti et al., 2006;Yun et al., 2009). Thus, there is a need for the development of new, efficient, and safe drug (Carraro, Iribarne, & Paulino, 2016;Hoelz et al., 2015;Manoel-Caetano Fda & Silva, 2007;Molina et al., 2014;Woodcock & Woosley, 2008). In recent years, naphthoquinones and heterocyclic derivatives are considered privileged structures in the development of novel drugs against parasitic diseases such as trypanosomiasis (Prati et al., 2015;Salas, Faundez, Morello, Maya, & Tapia, 2011).
There are several proposed mechanisms of action for the trypanocidal activity of quinone derivatives. One of them is the well-known ability of quinones for generating reactive oxygen species (ROS) through redox cycling with molecular oxygen and consequently oxidative stress and cell death (Benites et al., 2010;Henry & Wallace, 1996;Kovacic, 2007;O'Brien, 1991;Paulino et al., 2008).
The metabolic differences between the parasite and host cells to avoid the damage mediated by ROS are considerable (Gutteridge & Halliwell, 2000;Salas et al., 2011). Trypanothione reductase (TR) constitutes a major component of the general oxidative-stress defence in Trypanosoma cruzi (Turrens, 2004) and is absent in mammals. The main function of TR is to maintain a reducing intracellular environment by keeping trypanothione in the dithiol state (Gonzalez-Chavez, Olin-Sandoval, Rodriguez-Zavala, Moreno-Sanchez, & Saavedra, 2015). TR is essential in all trypanosomatids. Parasites with lowered TR activity display an increased sensitivity toward hydrogen peroxide (Fairlamb, Blackburn, Ulrich, Chait, & Cerami, 1985;Gonzalez-Chavez et al., 2015). The counterpart of the parasite TR in the mammalian host is glutathione reductase (GR). A comparison with the available crystal structures of GR (Berkholz, Faber, Savvides, & Karplus, 2008;Karplus & Schulz, 1987Pai, Karplus, & Schulz, 1988;Pai & Schulz, 1983) reveals that the overall structure of the two enzymes is highly conserved. There is an overall 40% sequence identify and both enzymes are homodimers. TR may also replace thioredoxin reductase, although trypanosomatids do possess a conventional thioredoxin.
The key feature of TR and GR is the mutually exclusive specificity for their cognate disulfide substrate. Thus, in principle, it should be possible to inhibit selectively the parasite enzyme without affecting the mammalian one. A detailed analysis of the enzymes (Iribarne, Paulino, Aguilera, Murphy & Tapia 2002) indicates that in both TR and GR, several alternative sites exist in addition to the binding sites for the disulfide substrate and nicotinamide adenine dinucleotide phosphate (NADPH) and the dimer interface.
For a set of aryloxy-quinones previously synthetized and tested for their trypanocidal and cytotoxic activities (Vazquez et al., 2015), it was proposed that their selective antiparasitic activities may correlate with the inhibition of TR. To prove this hypothesis, here, we report on the biochemical and kinetic analysis of recombinant T. cruzi TR and human GR inhibition. In parallel, in silico molecular docking studies of TR/GR complexes with aryloxy-quinones in all putative binding sites aimed at a deeper understanding of the atomic interactions between these compounds and TR and GR, respectively. As a result, quinone derivatives with good and selective TR inhibitor capacity and with low GR affinity can be proposed as candidates to develop more effective drugs against Chagas and other diseases caused by trypanosomatids.

Methodology
A set of 28 aryloxy-quinones ( Figure 1) were previously synthetized and tested (Vazquez et al., 2015). The set of quinones was divided in three subsets based on three main scaffolds as naphthoquinones, quinolinequinones, and furanoquinones. Moreover, each main scaffold has different aryloxy substituents. To understand the manner in which these molecules function in the parasite, as well as in the couple TR/GR trypanosomatids and mammalian enzymes, in a first step of this work twenty of these molecules were selected to make TR/GR inhibition and kinetics studies. In a second step an in silico, strategy was applied to the entire set of quinones.

TR assays
The activity of TR was measured in 40 mM Hepes, 1 mM EDTA, pH 7.5 at 25°C. The assay mixture contained, in a total volume of 1 mL, 100 μM NADPH, 5-10 mU of TcTR, and 50 μL DMSO as control or the same volume with the tested compound dissolved at a final concentration of 100 μM when the quinone inhibition was tested. The reaction was started by the addition of 100 μM of TS 2 and NADPH consumption was measured by following the absorbance decrease at 340 nm at 25°C (ε = 6.22 mM −1 cm −1 ) (Persch et al., 2014). From these data, the percentage of inhibition was calculated. Many compounds proved to be insoluble under the assay conditions and thus could only be measured at rather low concentrations. One representative compound of each subset of compounds (Nq-h, Fq-c, and Qq-b) was subjected to a detailed kinetic analysis.
2.1.2. GR assays GR was purified from human erythrocytes by the method previously reported (Worthington & Rosemeyer, 1974). The enzyme was stored at a concentration of 5 mg/mL in potassium phosphate buffer, pH 7.0, I .1 M, containing .2 M KCI, 1 mM EDTA, and .1% (v/v) 2-mercaptoethanol.
The activity of human GR was measured in 200 mM KCl, 60 mM K 2 HPO 4 , and 1 mM EDTA, pH 6.9 at 25°C. The assay mixture contained in a total volume of 1 mL, 100 μM NADPH, and 5-20 mU of hGR and 5 mM of inhibitor on DMSO.
After checking for nonspecific NADPH oxidation for two minutes, 1.1 mM of GSSG was added to start the enzymatic reaction. The decrease in NADPH concentration was monitored spectrophotometrically. Residual enzyme activity in the presence of inhibitor was determined relative to controls containing DMSO according to Nordhoff et al. (Fernandez-Blanco, Font, & Ruiz, 2016;Maran, Fernandez, Barbieri, Font, & Ruiz, 2009;Nordhoff, Bucheler, Werner, & Schirmer, 1993).

In silico assays
All calculations were done running the Molecular Operating Environment (MOE 2015.10, Chemical Computing Group Inc., http://www.chemcomp.com) suite on Linux, on a workstation with a quad-core processor hyperthread equipped.
A molecular docking of 28 carbo and heterocyclic derivatives of aryloxy-quinones was done to study the interactions with the putative enzymatic receptors TR and GR. This procedure implies a ligand and a receptor preparation, the development of a validated docking procedure and the molecular docking itself.

Ligand preparation: modeling and conformational analysis
The three-dimensional (3D) structures of the 28 molecules under study (Vazquez et al., 2015) (Figure 1) were submitted to energy minimization by means of MMFF94x force field (Halgren, 1996a(Halgren, , 1996b and the program suite MOE. To use all flexibility capacity of the program, a conformational analysis was performed for all naphthoquinones, quinolinequinones, and furanoquinones using the low-mode molecular dynamics (LowModeMD) approach (Labute, 2010) to search minimum energy conformers. The resulting conformations are saved to the output database provided they meet the energetic and geometric criteria.
An energy minimization was done and terminated when the root mean square (RMS) gradient test fall below .005 kcal/mol. The maximum number of energy minimization iterations was selected as 500. Two conformations are judged equal if the optimal heavy atom RMS superposition distance is less than the specified tolerance of .25 Å. All conformations with an energy greater than the sum of the global minimum energy plus a cutoff of 7 kcal/mol were discarded.
2.2.2. Crystal structure of the putative targets 2.2.2.1. Trypanothione reductase. The catalytic function of TR is the reduction of its cognate substrate TS 2 to the dithiol form T(SH) 2 . TR is active as a homodimer with a subunit mass of about 52 kD. The two identical catalytic sites are composed of residues of both subunits and comprise at least two regions, namely the NADPH site and the disulfide substrate site, which are separated by the flavin ring (Paulino et al., 2005).
The crystal structure of T. cruzi TR complexed with its physiological substrate TS 2 was obtained from the Protein Data Base (PDB 1BZL), 2.4 Å resolution (Bond et al., 1999). The homodimeric protein is composed of two peptide chains each containing 486 amino acids, two molecules of the cofactor flavin adenine dinucleotide (FAD) and two molecules of ligand, the bis(gammaglutamyl-cysteinyl-glycinyl)spermidine (TS 2 ). Both TR cocrystallized ligands are posed each one in both catalytic sites.
2.2.2.2. Glutathione reductase. The crystal structure of human GR was obtained from the PDB base (4GR1) with 2.4 Å resolution (Janes & Schulz, 1990). GR is a homodimer with two chains of 478 amino acids and a molecule of the cofactor FAD each. It is cocrystallized with a molecule of 4-N-malonyl cysteinyl-2,4diaminobutyrate disulfide (RGS).

Quinones putative binding sites
To identify possible quinones-binding site in the crystal structures of GR and TR, the Site Finder module of the MOE 2015.10 was used. Forty-four putative sites in TR and fifty-five55 in GR were detected. The site finder shows the findings as red (polar) or grey (apolar) spheres and a number that is the total count of spheres represents the size of a site. The sites have been analyzed in their size as well their structure and amino acid composition. The greater detected sites in both enzymes were the catalytic sites described earlier in the crystallographic structures. The Site Finder detected too the NADPH-binding site in both enzymes. Finally, for both enzymes, a site localized in the interface surface between both monomers conforming the dimer was identified (Figure 2). Several crystal structures of GR-inhibitor complexes have shown binding of the ligands at this cavity (e.g. Bilzer et al., 1984;Savvides & Karplus, 1996).
When the TS 2 -binding site in TR is described based on a list of contacting amino acids at a 4.5 Å distance of TS 2 , it could observed that amino acids from both A and B chains participate in the binding (Bond et al., 1999).
The overall structure of the active site of GR is similar to that of TR. However, in contrast to the negatively charged TS 2 binding site in TR, the GSSG-binding site in GR displays an overall positive charge.
A hydrophobic region named "Z-site" was defined in 1991 by el-Waer, Douglas, Smith, and Fairlamb (1991). They suggested the Z-site as a relevant binding pocket for TR inhibitors although Glu466 and Glu467, the residues mainly responsible for the electrostatic interactions with the inhibitors, are conserved in GR and TR (de Paula da Silva, Bernardes, da Silva, Zani, & Carvalho, 2012). Thus, it is difficult to assume a specific binding of TR inhibitors at this site. Indeed, although docking simulations yielded the Z-site as favored binding site, the crystal structure revealed the inhibitor at the hydrophobic wall lining the TS 2 -binding site (Persch et al., 2014). Here we will demonstrate that for aryloxyquinones, a mixture of both catalytic and Z-site residues, in its neighborhood, could confer selectivity.

Alternative putative binding sites.
In each active site of the homodimer, the cofactor FAD is found, together with the dithiol/disulfide bridge, essential for catalysis. During catalysis, reducing equivalents flow from NADPH to the disulfide substrate (e.g. trypanothione) with FAD and the disulfide bridge as intermediates. Two proton relays, one at each site, modulate the transfer. This complexity justifies the proposal of the NADPH site as a putative-binding site, in the same way by which the endogenous substrate binding site was proposed before.
At the subunit interface, both enzymes have a cavity that was marked in the Site Finder strategy, as a likely binding site. This is in accordance with previous crystal structures of GR inhibitor complexes showing ligand binding at this site (e.g. Bilzer et al., 1984;Savvides & Karplus, 1996). In comparison to GR, the cavity of TR is more extended and narrow, contacting both catalytic sites in neighborhood of the above-described Z-site ( Figure 2). This topology like a bridge between both catalytic sites caused us to think that this site may bind molecules. For this reason, in the docking approaches, we compared this region with other sites.

Molecular docking
Separate configurational databases in which each one of naphthoquinones, quinolinequinones, and furanoquinones configurations were used. As ligand, it was used a configurational database of molecules.
Four sites are assayed separately. The catalytic binding sites were defined from a 4.5 Å sphere around the cocrystallized ligand of TR and GR, respectively. The interface site definition was made through the Site Finder procedure above described. In the case of NADPH, a 4.5 Å sphere around the FAD adenine moiety defined the site.
Docking validation tests were performed with three set of conformational databases for naphthoquinones, quinolinequinones, and furanoquinones and the Catalytic Site 1. Two placement methods: Triangle Matcher and Alpha PMI (Udatha, Sugaya, Olsson, & Panagiotou, 2012) were assayed. Affinity dG (AdG) (Halgren, 1996a(Halgren, , 1996b and London dG were assayed for scoring and re-scoring. Finally, MMFF94x force field was used to energy minimize the resulting structures.

Protein-ligand interaction fingerprints (PLIF)
The PLIF descriptors implemented in the MOE were used as a benchmark with respect to interaction fingerprints. Interactions are classified as hydrogen bonds, ionic interactions, and surface contacts according to the residues. The PLIF descriptors for all protein-bound ligands were generated with the default parameter set in MOE and presented in Population Display and the Barcode display.
The Population Display is a histogram showing the number of ligands (Y-axis) with which each residue (plotted in the X-axis) interacts. In the Barcode representation of binding interactions, they are horizontal lines (rows) that correspond to each of the studied compounds.
Additionally, two other analysis of PLIF results were performed and PLIF-ContactMap and PLIF-HeatMap were generated.

H-Bond-ligand interactions
The ligand interactions application provides a tool to visualize an active site of a complex in diagrammatic form. A selection of interacting entities, which includes hydrogen-bonded residues, close but nonbonded residues, solvent molecules and ions are drawn about the ligand, their positions in 2D being chosen to be representative of the observed 3D distances, as well as taking into account aesthetic considerations.

Results and discussion
3.1. Enzymatic inhibition and kinetics 3.1.1. TR and GR inhibition screening Table 1 summarizes the experimental data obtained for all aryloxy-quinones investigated in this study. The trypanocidal activity toward T. cruzi, cytotoxicity vs. J-774 cells (Vazquez et al., 2015), as well as the in vitro TR/ GR inhibitor activities are given.
Considering the three subsets of aryloxy-quinones studied, a naphthoquinone (Nq-h) proved to be the most active inhibitor of TR. However, no correlation between TR inhibition and the antiparasitic activity efficiency was observed. Whereas compound Nq-b showed the highest trypanocidal activity, it was a comparably weak inhibitor of TR and was even more active against GR. Thus, the compound is a good example that other mechanisms are responsible for the cellular mode of action of the quinones.
Nq-b is a naphthoquinone with a nitro substituent in its aryloxy moiety which may be the reason for its excellent trypanocidal activity (IC 50 T. cruzi epimastigote = .02 μM). It is even more potent than nifurtimox (IC 50 T. cruzi epimastigote = 7 μM) and possibly acts by a similar mechanism.
Nq-h was the most effective and selective inhibitor of TR. Its trypanocidal activity was good; although ten fold lower than that of Nq-b. The halogen (Br) on the naphthoquinone and the O-naphthyl substituent suggest that the electronegativity as well as the planar aromaticity may confer selectivity for TR vs. GR inhibition.

TR kinetic assays
Nq-h, Qq-b, and Fq-c were selected as representatives for the naphthoquinone, quinolinquinone, and furanquinone subsets and subjected to a detailed kinetic analysis. The type of inhibition was derived from Lineweaver-Burk plots (Supplementary Material Figure I). The inhibitor constants K i were calculated from direct plots. The data obtained are shown in Table 2. All three compounds inhibited TR with a noncompetitive type of inhibition indicating that their binding site does not coincide with that for TS 2 . With a K i -value of 1.1 μM, Nq-h was clearly the best TR inhibitor.
3.2. The 3D structure of the assayed quinones database and it Structure-Activity Relationships (SAR) Table 3 summarizes the main structural characteristics and results to allow a SAR analysis of the trypanocidal and TR inhibitory activities.

SAR with respect to the trypanocidal activity
Being that the aim of this search for a trypanocidal nontoxic drug is to improve the design of nifurtimox (better growth inhibition together with a lesser toxicity), it was been considered that a quinone with a good trypanocidal activity must inhibit T. cruzi growth in a concentration under .17 μM, a medium activity will correspond to concentrations between .17 and 7 μM, and quinones with inhibitory concentrations over 7 μM (that of nifurtimox) will be considered with low activity.
As we concluded in a previous paper (Vazquez et al., 2015), when comparing the naphthoquinones, furanoquinones, and quinolinequinones, the naphthoquinone moiety proved to be the best substructure for trypanocidal activity and lower toxicity. This is because the naphthoquinones were specially analyzed here. Inside the naphthoquinone sample, we could observe many variants including the addition of nitro, bromine, phenyl, or naphthyl moieties.
The presence of two putative redox active moieties: the quinonic ring and the nitrophenyloxy group (Nq-b) is an example of a very good improvement giving to the molecule very good (the best) activity in T. cruzi. With respect to the toxicity (J-774 cells), a comparison with nifurtimox seems put in evidence that the presence of the quinonic planar structure together with a phenoxy group, gave to the molecule selectivity in favor to the parasite inhibition (low toxicity than nifurtimox). Finally, the addition of a halogen when a nitro moiety is present (like in Nq-e) seems to lower the T. cruzi selective growth inhibition.
The presence of a halogen or a dimethyl-phenyloxy (comparing Nq-a with Nq-d or Nq-a with Nq-c) did not helped to improve the T. cruzi growth inhibition either to lower the toxicity.
The results of molecules with an alpha or beta naphtyloxy moieties, with or without a bromine atom in the structure, seems indicate that in which refers the T. cruzi growth inhibition these moieties have quite similar impact. In some cases (comparing Nq-i with Nq-j), the bromination contributes to lower the toxicity when a beta naphthyloxy is present in the structure. The halogenation of an alpha naphtyloxy moiety as in the case of the Nq-h seems not to give to the molecule a special either selective activity (compared with Nq-g).

SAR with respect to the TR/GR inhibitory activity
The nitro moiety, so good in the design of a trypanocidal molecule, is not related with a good and selective TR inhibition. This is the case of Nq-b and Nq-e, which are not selective TR inhibitors even if they are nontoxic and trypanocidal molecules.
A naphthoquinone scaffold accompanied by a phenyl nitro compound like that of Nq-b is depicted as the ideal scaffold to reach higher T. cruzi inhibition together with very low toxicity, even if the action mechanism will not be the selective inhibition of TR.
The phenoxy group is good when it has a bromination in the naphthyl moiety (Nq-d).
The alpha naphthyloxy group as in Nq-h and Nq-g, seems to be related to a selective TR inhibition. When the alpha naphthyloxy group has an halogen (Nq-h), the selectivity is as good than when it is absent (Nq-g).
They are no TR inhibition data to analyze the influence of a beta naphthyl (as it is the case of Nq-j of Nq-i) even if both have low activity in GR.
In the last columns of the Table 3, main questions referring these observations were YES/NO answered. In consequence, those molecules with both YES answers (T. cruzi and TR-selective inhibition): Nq-h, Nq-d, and Nq-g, were selected to further discussion.
Nq-h, Nq-d, and Nq-g are good and selective T. cruzi inhibitors and at the same time of TR. For these   Table 3. Summary of the structural characteristics for the assayed sets of naphtoquinones (Columns 2 to 6). Columns 7 and 8: color-coded T. cruzi inhibition growth and J-774 measurements (Vazquez et al., 2015): T. cruzi growth inhibition was considered TRYPANOCIDAL (GREEN) for compound concentrations below .17 μM; MEDIUM for concentrations between .17 and 7 μM and LOW (RED) those equal or higher than 7 μM (that of Nfx). Columns 9 and 10: color coded TR and GR inhibition results. TR inhibition was considered good (GREEN) when inhibitions of more than 43% were detected with concentrations below 5 μM. All other cases were considered low (RED); GR inhibition was considered low (RED) when a 50% inhibition was obtained for concentrations above 100 μM; all other cases were considered a low inhibition (GREEN). Column 11: Index of selectivity (IS) evaluated from data of Table 1   three cases, the hypothesis of a selective inhibition of TR associated with a good and selective inhibition of T. cruzi growth is true. Considering that the Nq-h was the best, we could say that for the set of studied aryloxy-quinones, a naphthoquinone moiety, accompanied with an alpha naphthyloxy substituent is the scaffold to lead the design of a trypanocidal, nontoxic and TR-selective inhibitor drug.

Molecular docking
The previous observations were limited by the lacking of all necessary experimental results. This is because in this work we had the proposal of doing all in silico evaluation of the complete set of the quinones under study.
The in silico will be in this case, complementary to all the discussion envisaged in the Section 3.2.

Docking validation
The validation step aimed into obtaining the broader range of energies to assure a good conformation sampling. The full set of 10 naphthoquinones, nine quinolinequinones, and nine furanquinones was used to generate three conformational databases of 63, 59, and 64 conformers, respectively, and they were docked in the Catalytic Site 1 of the TR (1BZL). The results are presented in the Table 4. The best results were obtained for the Alpha PMI and the London ⊗G methods.
As the second criterion of validation, we look for a maximum overlap between the cocrystallyzed and docked TS 2. The Alpha PMI and the London ⊗G methods were used to dock the TS 2 in the catalytic Site 1. The overlapping between the cocrystallyzed and docked TS 2 was measured and resulted an root mean square deviation (RMSD) of 2.02 Å.
In consequence, the procedure was declared validated.

Docking in the catalytic sites of TR and GR
The three conformational databases were used as "ligand" in the docking.
As "Receptor," they were settled all atoms of TR or GR. As "Site," it was defined a sphere of 4.5 Å around the crystallographic ligand TS 2 or RGS, respectively. After selection, TS 2 (or RGS) molecules were deleted and FAD atoms remained as unselected atoms.
With the aim of comparing within the experimental results shown in the Table 1, in the Supplementary Material Table A, were detailed the ΔG of the best docked conformation for each set of aryloxy-quinones. All poses were classified in two ranges of high and low energies. As it could be observed inside the left bottom frame of the Figure 2(A), two populations of naphthoquinones are detected in TR. The red colored poses, that is, the more energetic bindings, are in a site in the borderline between the Z site and the catalytic site.
When the furane and quinoline quinones were docked against the catalytic site 1 of TR, similar results were obtained being the more energetic bindings near to the catalytic site in the borderline with the Z site (Supplementary Material Figures II(a) and III(a)).
To describe graphically, this new aryloxy-quinones site, an electrostatic map was obtained for TR and GR (Figure 2 The best energies in the docking against GR were observed in a subsite of the catalytic site 1 similar to that detected for TR (left bottom frame in Figure 2(B) and Supplementary Material Figures II(b) and III(b)).

Docking in the alternative sites of TR and GR
The docking for the three conformational databases in the interface site of TR (1BZL) and y GR (4GR1) shown that all best scored poses were placed in the middle of the interface but with lesser free energies than in the other assayed sites (data not shown). As an example, in the Supplementary Material, Figure IV shows the naphthoquinones docked in the interface site of TR and GR.
The results of the docking in the NADPH sites in TR and GR are shown in the Supplementary Material Table A  Two list of contacts emerges and they are, in the case of TR is: Lys62, Thr66, Leu399′, Met400′, His401′, Lys402′, Asp432′, Asn433, His461′, Thr463′, and Ser464′.
Those contacts that have counterpart in TR/GR are highlighted in bold: Lys62 in TR is Lys67 in GR; Thr66 in TR is Asn71 in GR; Leu399′ in TR is Met406′ in GR (a nonconservative change); His461′ in TR is the His467′ in GR; Thr463′ in TR is Thr469′ in GR and Ser464′ in TR is Ser470′ in GR.  All other contacts of docked quinones in TR does not have a counterpart in GR and vice versa.
In the Figure 4(a) and (b), the Y-axis shows the relative counts for the bits, directly related with the number of contacts. For example, the Lys62 has the three higher bars and that means that this amino acid makes three kinds of contacts with the most of quinones. On the contrary, Asn433 has three bars with very low Y-value meaning that it forms three different and nonfrequent contacts.
Lys62 is of most importance for the TR binding and is evidenced by the three columns in the Figure 3(a). In the case of GR, Lys67 forms hydrogen bonds with all the analyzed quinones evidenced by the two columns if the Figure 3(b). Even if the mammalian enzyme offers a more electrostatic region to binding and that the Lys62 is conserved in GR, some selectivity for TR could be evidenced by an increased number of contacts.
Leu399′ in TR is a contact detected for quinones that is a nonconserved residue in GR (Met406′). This contact seems to be important to the pose of quinones in the TR. However, Met406′ makes more contacts in GR than Leu399′ in TR, allowing to conclude that both residues are of similar important in the TR or GR binding.
The Met400′ in TR appears, as a surface contact in the PLIF analysis for Nq-h, Qq-d, and Fq-b and it is absent as contact in GR. For this reason, we could propose this contact as important for the TR specificity (TRS). This residue deserved special consideration in the interaction analysis of quinones by de Molfetta, de Freitas, da Silva, and Montanari (2009).
Referring the contact with Ser464′ in TR (Ser470′ in GR), it appears more frequently in TR than in GR as an important H-bonding interaction in the PLIF analysis.
In GR (Figure 4(b)), there is a differential contact found in the Glu473′ (not present in TR) giving to this environment the electrostatic differential characteristic pointing to a selective binding in favor of TR.
3.4.1.2. PLIF-ContactMap and PLIF-HeatMap. The PLIF-ContactMap and the PLIF-HeatMap were created from the PLIF analysis as an ad hoc designed tool to observe all docked conformations by means of a colorful image of binding in TR and GR that confirmed and completed the previous observations. They are presented in the Table 5 and in the Supplementary Material  Table B.
In the case of PLIF-ContactMap, to each given contact, it was assigned a number representing the percentage of conformers of a quinone making a contact with a given amino acid in TR or GR. For data visualization, the numbers of contacts were highlighted by means of a color gradient from dark red (100%) to light blue (4%). The result is shown in the Table 5 (PLIF-ContactMap). Alternatively, the percentage of contacts was associated with the intensity of the color from dark to light red (PLIF-HeatMap).
Lys62 in TR makes two kinds of H-bond acceptor and surface contacts with high percentage of conformations making contacts (e.g. 64% for Nq-h). Met400′ has unique and colorful annotations in TR representing the contacts made by a high percentage of quinones conformations (e.g. 55% for Nq-h). Finally, higher percentage of conformations make four kind of contacts (two H-bond side-chain acceptor and two backbone H-bond acceptor) with Ser 464′ in TR (e.g. 69% for Nq-d). The PLIF-HeatMap (Supplementary Material Table B) presents a similar picture.

Analysis of docking in the alternatives binding sites
When the interface site was selected in the TR (Supplementary Material Table A), all quinones finally were posed near the catalytic site, indicating a tendency to bind in this subsite (Supplementary Material Figure IV). However, when the same strategy was used for GR, it is noticeable that all quinones were placed in the middle of the interface site (Supplementary Material Figure IV). Earlier studies in GR  shown that a putative site with allosteric properties placed in the dimer interface should be considered as a putativebinding site. Tridimensional and molecular dynamics analysis (Hikichi et al., 1995) show that the GR dimer interface is quit adequate to lodge planar molecules as they are too the aryloxy-quinones here studied. However, if we take into account the scoring obtained for all of assayed sites, the dimer interface seems not to be the best for binding.
When the NADPH was used to center a putative binding site (Supplementary Material Table A), in the case of TR the scores of all quinones ranked in lower values than in the aryloxy-quinones site near the catalytic one described in the 3.4.1 Section. In principle, then, the NADPH site could be not considered as a secondary one for binding and for inhibition, remaining as more relevant the binding near the catalytic site. On the contrary, in the GR enzyme, it seems to be a site with similar "appealing" for quinones than the active site, with similar scores for both sites. Any case, the NADPH site in GR seems to be a good alternative for binding. This result is in agreement with previous observations . Then, for a molecule with a high binding energy either in the catalytic or the NADPH site in GR, it will be predictable a nonselective in TR inhibition.
3.4.3. 3D Graphical analysis and modeling 3.4.3.1. Putative modeling of selective TR inhibitors. Taking the observations related to the catalytic and the NADPH sites together, it could be suggested that if a quinone must be designed as a noncompetitive TR inhibitor, it must be bound near the catalytic site for TR but allowing the nonproducing posing of TS 2 . In addition, this compound must have low binding energies in the catalytic either the NADPH GR sites.
In the Figure 5, an hypothetical complex of a TR with Nq-h posed near the catalytic site (as described in the 3.4.1 Section) with the cocrystallized TS 2 in the crystallographic pose is shown. The TR-TS 2 -Nq-h energy minimized model was used as putative virtual complex to show that there is enough room inside the catalytic site to lodge the TS 2 and the inhibitors. Furthermore, this model could agree with a noncompetitive inhibition.
Based again in our results, to add selectivity against GR, it will be necessary that the quinone, even if it will be posed in the NADPH site or in the catalytic site, the bound will be so weak to no interfere with the GR activity. This is being sustained, for example in the case of Nq-h, by a low GR score, concomitantly with a very low GR inhibition capacity.
Moreover, if the aim of an anti-Chagasic drug should be to design optimal and selective TR inhibitors with maximal trypanocidal activity, structures like that of Nq-h could be used as leads to the improvement of the design.
3.4.3.2. Graphical analysis of best and selective TR inhibitors and comparison with the best trypanocidal assayed quinonic compounds. In the Table 3, we add three more columns containing the score in TR and GR, and finally, the ratio between both measurements (TR/GR) as a theoretical determination of the TRS.
If we take as good specificity a ratio over 1.1, the prediction of the TRS is that Nq-h and Nq-g are selective, and all other naphthoquinones are nonselective. We were successful into predict two of three TR selective inhibitors, and all (7) nonselective TR inhibitors. It was just only one failure into predict the Nq-d as nonselective (being selective). In view of this result, we could be confident that the in silico methods have a very good capacity to predict if a molecule will be a selective TR inhibitor and to give an atomic description of the interaction. Table 5. PLIF-ContactMap: Contact percentage detected by PLIF between a given aminoacid and all conformers of each quinone in TR and GR. X-Axis: TR (grey) and GR (yellow) contact codes are detailed: Side chain H-acceptor (ChAcc), Backbone H-acceptor (BkAcc), and Surface contact (Surf). Blocks are colored by means of a color gradient from blue (low percentage of contact) to red (high percentage of contact). Red blocks indicate a highly frequent contacts. Black frames were used to groupe similar residues in TR/GR. Y-axis: aryloxy-quinones set.
Nq-g, Nq-d, and Nq-h were further analyzed and shown in the Figures 6 and 7 (TR and GR, respectively). To do that, further energy minimization was done for the six complexes of those molecules docked in the TR and GR. The fact thatas it is clear in the Figure 6 the place occupied by quinones in the docking is in the binding site of one of the carboxylate moieties of trypanothione, could have the meaning that the anchorage of the TS 2 in the presence of the quinones is prevented, generating a bad and nonproductive linkage.
The Figure 6(d) shows clearly that the Lys62, Met400′, and Ser464′ are the most frequent and intense contacts for a TR-specific binding. This conclusion reinforce the massive observations (PLIF analysis) made in the Section 3.4.1.
To compare with another interesting molecule with low score free energy to TR, we analyzed the Nq-b, the best trypanocidal molecule with low toxicity in the J-774 cells. In it best pose in TR, the nitro phenyl group is placed in the same region that the quinone ring in the Nq-h being the Nq-b quinone ring redirected toward the interface. The low score obtained for this Nq-b pose indicated that the nitro group, with its charge localized in the nitrogen and oxygen atoms, is not adequate to bound this region in which the Lys62 and the Ser464′ are sustaining the quinone rings of three best and specific TR ligand quinones above-mentioned (Nq-h, Nq-g, and Nq-d). The difference in the charge distribution (very concentrate in the nitro group and polarized in the oxygens of quinone rings) must be the key to understand the difference in their TR activity. Then, the same key moiety, which gave best trypanocidal activity to the quinones (a naphthoquinone with a nitrophenyl group), seems to be the responsible for a weak union to TR, making this design not corresponding to a selective TR inhibitor. In summary, we are showing that the action mechanism of naphthoquinones with a nitro group in it structure, making these molecules the best design for a nontoxic and trypanocidal compounds is not associated with the TR-selective inhibition. Moreover, the lack of TRS could be explained too taking into account the global electrostatic charge of both TR and GR catalytic sites. The GR-catalytic site has a global negative charge, and it will be very appealing to a negative-charged nitro compound. However, the very good specific trypanocidal properties of the Nq-b compound is related to it good reactivity in the ROS mechanisms and some capacity to lower the toxicity with respect to nifurtimox. To associate the previous consideration with the role of different residues in the TR-selective inhibition, for the case of Nq-h, Nq-g, and Nq-d, we could consider for one hand (Figure 6), the presence of a network of contacts with the quinonic ring sustained by the Lys62, Ser464′, and Leu399′ residues. The alpha naphthyl moieties (Nq-h and Nq-g) contribute to the posing and are oriented to the interface (Nq-h) or to the catalytic site (Nq-g). In the case of Nq-d, its phenyl moiety accomplishes with the same role posing the quinonic ring near to the Lys62, Ser464′, and Leu399′. On the other hand, the Met400, in the neighborhood of the bromine atoms (Nq-h and Nq-g) makes important surface contacts ( Figure 6) giving better proper specificity to the TR anchorage. In the case of the nonbrominated Nq-d, with its quinonic ring pointing to the polar region of Met400, a backbone donor H-bonding is giving the differential contact needed to confer TRS.
Once again, it was demonstrated for one hand that a molecule containing a naphthoquinone ring together to a brominated alfa naphthoxy or phenyloxy moiety seems to be a very good model to lead the design of a trypanocidal and nontoxic antitrypanosomatid (e.g. anti-Chagas) drug. On the other hand, this pharmacological property was associated with a selective TR inhibition capacity mediated by a great number of contacts with the triad Lys62, Ser464′, and Met400, with the Leu399′ accomplishing an important and subtle role in the spatial orientation inside the site.

Conclusions
The aryloxy-quinones presented a tendency to have good docking scores in a new subsite inside the catalytic one, configuring a good model for the noncompetitive type of inhibition obtained experimentally. In this putative binding of quinones, there is enough room for the substrate trypanothione that would be anchored but in a way that would interfere with the overall catalysis.
In the TR aryloxy-quinones-binding site, the main contacts were detected with Lys62, Met400′, and Ser464′. In GR, a similar sub site was occupied by quinones, but there is a differential contact found in the Glu473′ giving to this environment the electrostatic differential characteristic pointing to a selective binding in favor of TR.
For most of the compounds studied, a relationship between trypanocidal activity and TR inhibition could not be observed. However, three of the naphthoquinones, namely Nq-h, Nq-g, and Nq-d, proved the hypothesis of a relationship between a good and nontoxic activity in T. cruzi, accompanied with low toxicity in mammalian cells associated with a specific TR inhibition.
Overall, our studies provided a more thorough understanding of the experimental trypanocidal results and revealed for selected compounds TR/GR as key enzymes in the action mechanisms of some aryloxy-quinones.
Interaction with other cellular targets cannot be discarded. Evidence for this is the very good and non-toxic activity of Nq-b, a nitro compound. A promising approach to optimize these compounds are in silico techniques such as for example the reverse docking that allows to find other targets that could contribute to the aryloxy-quinones trypanocidal action. A strategy involving the cycles of biochemical, parasitological and in silico assays, appears to be the best strategy for compound optimization and required to reach the final goal of a new anti-Chagas nontoxic drugs.