The Se^…S/N interactions as a possible mechanism of δ-aminolevulinic acid dehydratase enzyme inhibition by organoselenium compounds: A computational study

Graphical abstract

Highlights

• •DPDS and PSA interacts with cysteine residues from AlaD active site.
• •The Se…S interactions could be involved in the δ-AlaD inhibition.
• •δ-AlaD from Cucumis sativus does not present cysteine residues in the active site.
• •Se…N interactions could be involved in the organoselenium action.

Abstract

Organoselenium compounds present many pharmacological properties and are promising drugs. However, toxicological effects associated with inhibition of thiol-containing enzymes, such as the δ-aminolevulinic acid dehydratase (δ-AlaD), have been described. The molecular mechanism(s) by which they inhibit thiol-containing enzymes at the atomic level, is still not well known. The use of computational methods to understand the physical–chemical properties and biological activity of chemicals is essential to the rational design of new drugs. In this work, we propose an in silico study to understand the δ-AlaD inhibition mechanism by diphenyl diselenide (DPDS) and its putative metabolite, phenylseleninic acid (PSA), using δ-AlaD enzymes from Homo sapiens (Hsδ-AlaD), Drosophila melanogaster (Dmδ-AlaD) and Cucumis sativus (Csδ-AlaD). Protein modeling homology, molecular docking, and DFT calculations are combined in this study. According to the molecular docking, DPDS and PSA might bind in the Hsδ-AlaD and Dmδ-AlaD active sites interacting with the cysteine residues by Se^…S interactions. On the other hand, the DPDS does not access the active site of the Csδ-AlaD (a non-thiol protein), while the PSA interacts with the amino acids residues from the active site, such as the Lys291. These interactions might lead to the formation of a covalent bond, and consequently, to the enzyme inhibition. In fact, DFT calculations (mPW1PW91/def2TZVP) demonstrated that the selenylamide bond formation is energetically favored. The in silico data showed here are in accordance with previous experimental studies, and help us to understand the reactivity and biological activity of organoselenium compounds.

1. Introduction

The utilization of selenium (Se) in organic synthesis has been producing a vast number of organoselenium compounds since the second half of the 19th century. For instance, Ebselen (EBS) was synthesized in 1924, and nowadays is the most investigated of the organoselenium compounds (Fig. 1 A) [1]. Diphenyl diselenide (DPDS) is the simplest diaryl diselenide and has been tested as a pharmacological agent [2]. The organoselenium derivatives present many pharmacological properties, such as anti-inflammatory, cardioprotective, neuroprotective, and antioxidant, this last one due to their ability to reduce hydrogen peroxide (H₂O₂) to water (H₂O). Therefore, these compounds are considered mimetics of the glutathione peroxidase (GPx) enzyme and are promising drugs [3], [4], [5], [6].

Fig. 1.

(A) The structural formula of some organoselenium compounds, (B) the 5-aminolevulinic acid (5-Ala) substrate and porphobilinogen (PBG) product of the δ-AlaD.

In addition, EBS and DPDS can oxidize thiol groups of proteins [3], [4], [7] as observed in the mammalian enzyme δ-aminolevulinic acid dehydratase (mδ-AlaD) or porphobilinogen synthase (PBGS) (EC 4.2.1.24). Since the δ-AlaD is an important enzyme involved in the porphyrins’ synthesis, its inhibition can have toxicological consequences [8], [9], [10], [11]. The δ-AlaD catalyzes the asymmetric condensation of two molecules of 5-aminolevulinic acid (δ-aminolevulinic acid – 5-Ala), forming the porphobilinogen (PBG), which is the precursor of porphyrins’ synthesis (Fig. 1B). In the enzyme active site, each substrate binds at two different subsites (A and P), leading to the regioselective product PBG. The acetic acid and propanoic acid side-chains of PBG originate from the subsites A and P, respectively [12], [13], [14]. Porphyrins are essential to living beings, particularly to the aerobic life, due to the heme prosthetic group, which is involved in the transport of oxygen (hemoglobin and myoglobin), xenobiotic metabolism (cytochrome P450), protection against peroxides (peroxidases and catalases), and chlorophyll synthesis [13], [15], [16], [17]. There are two major classes of δ-AlaD: the Zn-dependent enzymes (that are present in mammals, fungi and some bacteria, such as Escherichia coli) [15], [18], [19], and the Mg-dependent enzymes, that are found mainly in plants, protozoa and other bacteria [13], [20], [21], [22].

Studies have demonstrated that the DPDS can inhibit the δ-AlaD enzyme from human (Hsδ-AlaD) and rodents [10], [11], [23], [24], [25], [26], [27], [28]. The δ-AlaD from Drosophila melanogaster (Dmδ-AlaD) can also be inhibited by DPDS [29]. In contrast, DPDS do not inhibit δ-AlaD from cucumber, Cucumis sativus (Csδ-AlaD); nevertheless, its putative metabolite, the phenylseleninic acid (PSA), can inhibit the Csδ-AlaD [30]. In fact, the toxicity of organoselenium compounds could be associated with their metabolic oxidation by flavin-containing monooxygenases [4], [31], [32]. However, the inhibition mechanism(s) involved in these cases has not been established yet.

To complement and better understand the in vivo and in vitro data, in silico methods have been used to analyze, simulate, and predict the pharmacology and toxicity of chemicals [33], [34], [35], [36], [37]. There are many types of computational methods, where the molecular docking stands out by simulating the interactions between macromolecules (proteins and DNA) and ligands (substrate, inhibitor, and agonist). This method consists in predicting the binding mode of the ligand at the binding site of a given target, in addition to the estimation of affinity for the receptor, by predicting binding free energy (ΔG) [38], [39], [40], [41]. Quantum mechanical methods, such as the density functional theory (DFT) approach, are frequently used in the study of structures, reactions, and molecular properties [42], [43], [44], but are strictly limited to systems of few hundreds of atoms. In addition, the protein homology modeling has been successfully employed to predict the 3D protein structure, which is essential in many cases when the tertiary or quaternary structure must be studied [45], [46], [47], [48], [49].

Different in silico methods have been adopted to predict the reactivity, toxicity, and pharmacology of organoselenium compounds and selenoproteins [44], [50], [51], [52], [53], [54], [55], [56], [57], [58]. Here, to better understand the toxicological effects of organoselenium molecules, and how they interact with target proteins, we propose an in silico approach combining protein homology modeling, molecular docking simulations, and DFT calculations (Scheme 1 ). Based on the difference of DPDS and PSA inhibition behavior on δ-AlaD enzymes, this study aims to compare the intermolecular interactions between the Hsδ-AlaD, Dmδ-AlaD and Csδ-AlaD enzymes and the DPDS and PSA organoselenium compounds, to gain insight into their mechanisms of inhibition.

Scheme 1.

Overview of all the steps involved in this study.

2. Materials and methods

2.1. Protein homology modeling

First, the homology analysis of the primary structure of δ-AlaD enzymes from cucumber (Cucumis sativus), fruit fly (Drosophila melanogaster), human (Homo sapiens), mouse (Mus musculus), zebrafish (Danio rerio), cockroach (Blattella germanica), protozoa (Toxoplasma gondii), yeast (Saccharomyces cerevisiae), archaeon (Pyrobaculum calidifontis) and bacteria (Chlorobaculum parvum, Escherichia coli, Pseudomonas aeruginosa, Staphylococcus aureus and Wolbachia) were performed. The FASTA amino acid sequences for δ-AlaD enzymes were obtained from the the National Center for Biotechnology Information – NCBI (https://www.ncbi.nlm.nih.gov/pubmed/), UniProt (http://www.uniprot.org/) and Protein Data Bank – PDB (http://www.rcsb.org/pdb), according to the respective codes: Blattella germanica: UniProt (A0A2P8XHW3_BLAGE); Chlorobaculum parvum: PDB (2C1H); Cucumis sativus: UniProt (A0A0A0LQK9_CUCSA); Danio rerio: NCBI (NP_001017645.1); Drosophila melanogaster: UniProt (Q9VTV9_DROME); Escherichia coli: PDB (1L6S); Homo sapiens: PDB (1E51); Mus musculus: NCBI (NP_001263375.1); Pseudomonas aeruginosa: PDB (1GZG); Pyrobaculum calidifontis: PDB (5LZL); Saccharomyces cerevisiae: PDB (1H7N); Staphylococcus aureus: UniProt (HEM2_STAAR); Toxoplasma gondii: PDB (3OBK); Wolbachia: NCBI (WP_041571452.1). Regarding of FASTA from PDB, it was used the FASTA associated with the corresponding PDB file on the database (we do not extract the FASTA from the PDB file). The Clustal Omega server (http://www.ebi.ac.uk/Tools/msa/clustalo) was used to make the multiple sequence alignment, and the similarity between the δ-AlaD sequences was calculated from the Geneious program (https://www.geneious.com) (Fig. 2 , S1, S2 and Table S1).

Fig. 2.

Multiple alignments of the δ-AlaD amino acids sequence of different organisms. Only a fragment from the active site of the proteins are shown. The residues from the active site are highlighted: Cys (yellow); residues that remain conserved (cyan), residues that are not conserved when compared to the human enzyme (green and pink). The complete alignment is shown in Fig. S1. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Since there is no available three-dimensional structure of Dmδ-AlaD and Csδ-AlaD, the Swiss-Model (https://swissmodel.expasy.org) [59], Phyre2 [60], and Geno3D servers [61] were used to obtain their structures, using the amino acid sequence of the Cucumis sativus and Drosophila melanogaster δ-AlaD, taken from UniProt with the codes A0A0A0LQK9_CUCSA and Q9VTV9_DROME, respectively. The 3D structures of the Chlorobaculum parvum (PDB: 2C1H [62]), Pseudomonas aeruginosa (PDB: 1GZG [63]), and Toxoplasma gondii (PDB: 3OBK [21]) where used as template to build the Csδ-AlaD models, while the Escherichia coli (PDB: 1L6S [64]), Pyrobaculum calidifontis (PDB: 5LZL [18]), and Saccharomyces cerevisiae (PDB: 1H7N [19]) structures where used as template to build the Dmδ-AlaD models. The validation of the protein models were carried out by the programs: Verify 3D [65], [66], ProSA [67], PROCHECK [68], [69], and ERRAT [70]. The Ramachandran plot was made by the PDBsum server (www.ebi.ac.uk/pdbsum/) [71]. More details can be found in the Supporting information.

2.2. Molecular docking

To carry out the docking simulations, the Hsδ-AlaD was obtained from PDB with the code 1E51 [72], and the Dmδ-AlaD-1L6S and Csδ-AlaD-3OBK models were obtained from protein homology modeling by the Swiss-model program (as described above). The CHIMERA 1.8 program [73] was used to add the hydrogen atoms to the proteins. The Lys199/195/291 and Lys252/248/344 residues were considered neutral (deprotonated) [14], which was confirmed by H++ analysis (http://biophysics.cs.vt.edu). The ligands (PBG and the organoselenium compounds) were built in the Avogadro 1.1.1 software [74], followed by a geometric optimization using the MOPAC program (http://openmopac.net/MOPAC2012.html) with the semi-empirical method PM6 (with the water dielectric constant) [75]. The PSA was considered deprotonated (pka = 4.79) [76] during the docking simulations. The protein and ligands were converted to the pdbqt format with the AutoDockTools [77], with the Gasteiger and MOPAC charges, respectively. The partial charge (0.302) of the Zn²⁺ ion from Hsδ-AlaD and Dmδ-AlaD were obtained from a previous study [51].

The AutoDockVina 1.1.1 software [78] was used for the docking simulations, with exhaustiveness of 100. The best docking protocol was obtained using the ligands and the side chain of Arg209 and Lys252 residues from Hsδ-AlaD (Arg205 and Lys248 from Dmδ-AlaD-1L6S, and Arg301 and Lys344 from Csδ-AlaD-3OBK) flexible. The grid boxes (with spacing of 1 Å) were centered in the active site of the enzymes Hsδ-AlaD (coordinates: x = 31.63; y = 73.65; z = 57.08), Dmδ-AlaD-1L6S (coordinates: x = 19.72; y = 83.35; z = 52.14), and Csδ-AlaD-3OBK (coordinates: x = -64.60; y = -77.40; z = 28.05), with a size of 25 × 25 × 25 Å, in both cases. The Discovery Studio Visualizer 17.2.0. (DSV) program (https://www.3dsbiovia.com/) was used to analyze the results, where the conformers of lowest binding free energy (ΔG) were selected as the best model. The molecular docking protocols were validated by the RMSD (root-mean-square deviation) values from the PBG molecules, which give the relationship between the experimental and the theoretical data in a receptor-ligand complex. RMSD values lower than 2.0 Å indicate good quality of data reproduction (Fig. S4) [41], [79], [80] (details can be found in the Supporting Information).

2.3. Density functional theory calculations

All quantum chemistry calculations have been performed using density functional theory (DFT) approach as implemented in Gaussian 09 rev. E.01 program [81]. mPW1PW91 (Perdew-Wang hybrid functional) [82] was used, in combination with the def2TZVP (Triple zeta quality with polarization functions) basis set for all the atoms [83], [84].Full geometry optimizations were carried out in gas phase; solvation (water) effects were taken into account in subsequent single point calculations at the same level of theory using PCM approximation [85].

3. Results and discussion

3.1. Protein sequence comparison and homology modeling

Considering that DPDS inhibits the Hsδ-AlaD [11] and Dmδ-AlaD [29] and does not inhibit Csδ-AlaD [30], we initially compared the primary structure of the δ-AlaD enzymes (including other different species) through multiple sequence alignment (Fig. 2, Fig. S1/S2 and Table S1). The analysis of the sequence alignment data demonstrated that are two groups of proteins, i.e. Group A, which includes the species that present Cys residues in the active site (Saccharomyces cerevisiae, Drosophila melanogaster, Danio rerio, Homo sapiens, Mus musculus, Escherichia coli, Pyrobaculum calidifontis and Staphylococcus aureus), and Group B, which includes the species that have Asp residues (Toxoplasma gondii, Cucumis sativus, Wolbachia, Pseudomonas aeruginosa and Chlorobaculum parvum) (Fig. S2). Interestingly, the Blattella germanica δ-AlaD is small when compared to the other species (146 vs ~ 330 residues) and has not the Cys region of the active site; however, the catalytic Lys residues are conserved (Fig. S1). According to the phylogenetic tree (Fig. S2) it belongs to Group A.

In general, the three cysteine residues from the active site of Group A δ-AlaD were replaced by two aspartate residues and one alanine residue in Group B (Fig. 2 and Fig. S1/S2) indicating a significant change in the nature of the active site. In addition, the Arg221 (in the human protein) were replaced by a Lys residue in the δ-AlaD from Group B (Lys313 in Csδ-AlaD). As the Lys and Arg are basic and positively charged residues, practically, the same physical–chemical properties are conserved. These observations are in accordance with previous studies of Kervinen et al. (2001) [86] where five δ-AlaD enzymes (from Pisum sativum, Pseudomonas aeruginosa, Bradyrhizobium japonicum, Escherichia coli, and H. sapiens) sequences were analysed, and the metal-binding region determined.

Here, based on these observations, we can suppose that DPDS does not inhibit the Csδ-AlaD because this enzyme does not present Cys residues in its active site. However, it does not explain why PSA inhibits the Csδ-AlaD. For a better understanding of the interactions between inhibitors and enzymes, the molecular docking simulations were performed. Taking into account that there are no Csδ-AlaD and Dmδ-AlaD structures available, the 3D model of these enzymes were built using protein homology modelling.

Homology modeling is the most accurate method to build protein structure models [87], [88], [89]. Among the different programs developed for this purpose, in this study we have chosen the Swiss-Model [59], Phyre2 [60], and Geno3D [61] to create the Dmδ-AlaD and Csδ-AlaD structures. Taking in account the primary structure similarity between the δ-AlaD enzymes (Fig. 2, Figs. S1–S2 and Table S1), three templates were selected for Dmδ-AlaD (PDB ID: 1H7N, 1L6S and 5LZL) and three for Csδ-AlaD (PDB ID: 1GZG, 2C1H, and 3OBK). Each template was used in the protein homology modeling with the three programs above cited, to find the best protein model. The 3D structure models of Dmδ-AlaD and Csδ-AlaD built were validated using the programs: Verify 3D [65], [66], ProSA [67], PROCHECK [68], [69], and ERRAT [70] (Tables S1–S2).

According to the data in Tables S2 and S3, the best Dmδ-AlaD and Csδ-AlaD models were obtained from the PDB ID 1L6S and PDB ID 3OBK templates, respectively, using the Swiss-Model program, which turned out to be the most performant program for this task. Dmδ-AlaD-1L6S and Csδ-AlaD-3OBK models showed a satisfactory protein structure, because the validation parameters are in the range of native protein structure (see the Supporting information), and they were used for the molecular docking simulations.

Despite the differences in the primary structure between the Homo sapiens δ-AlaD, Dmδ-AlaD and Csδ-AlaD, the comparison of the tertiary structure of the three enzymes exhibited a very similar organization of the residues, with the simulated PBG binding pose presenting practically the same conformation and interactions (Fig. 3 ). Here, we highlighted the major difference in the active site of both enzymes. As shown in Fig. 3AB, in Hsδ-AlaD and Dmδ-AlaD the thiolates of the Cys residues are coordinated to a zinc ion (Zn²⁺), where this metal nucleus acts as a Lewis acid and Zn^…N coordination with the amino moiety (Lewis base) from PBG is formed. This Zn^…N interaction is essential to the catalysis of the δ-AlaD, because it specifically guides one molecule of 5-Ala substrate in subsite A, before the cyclization to pyrrole ring [86], [90]. In fact, the Cys mutations cause a dramatic reduction in the enzyme activity [12]. On the other hand, according to the docking simulation between the PBG and Csδ-AlaD, the orientation of one molecule of 5-Ala substrate is likely driven by the H-bonds between the amino moiety from 5-Ala and the carboxylate groups of Asp217 and Asp225 residues (Fig. 3C). The PBG binding pose obtained by the docking in Csδ-AlaD is very similar to the crystallographic data collected from T. gondii δ-AlaD [21]. Interestingly, Asp217 and Asp225 are the residues that correspond to Cys124 and Cys132 residues in the human enzyme, respectively (Fig. 2, Fig. 3).

Fig. 3.

Comparison between the 3D structures of δ-AlaD from Homo sapiens (A), Drosophila melanogaster (B) and Cucumis sativus (C). The active site is highlighted, and the carbon atoms of PBG are represented in pink color. (A) Human δ-AlaD structure from the crystal PDB ID 1E51 [72], and (B) Dmδ-AlaD and (C) Csδ-AlaD from protein homology modeling, with the PBG binding pose from the molecular docking. The hydrogen atoms were omitted for clarity. H-bonds, electrostatic (charge-charge) and hydrophobic (π–π) interactions, besides the zinc coordination, are represented by green, orange, purple, and blue dotted lines, respectively; all distances are in Å. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

In addition, in the case of Csδ-AlaD, the Mg²⁺ ion is not present in the active site (Fig. 3C), and does not participate directly in the catalysis. However, the Mg²⁺ is essential to enzyme function, as observed in E. coli, Bradyrhizobium japonicum, Pseudomonas aeruginosa, and P. sativum, due to the H-bonding network around this metal ion maintaining the quaternary structure of δ-AlaD [13], [14], [15], [86]. This difference in the active site of δ-AlaD from different species must be taken into account in the design of selective inhibitors with useful applications, such as in the case of δ-AlaD from Wolbachia [91], [92], [93] and Staphylococcus aureus [94]. Moreover, due to the similarity of the active site from δ-AlaD of the group B, the use of plant δ-AlaD (such as cucumber) can provide a simple, practical and cheap in vitro assay to find new selective inhibitors.

3.2. Organoselenium molecular docking study

Molecular docking simulations were carried out to understand the δ-AlaD inhibition by DPDS and PSA. According to the docking between the Hsδ-AlaD and DPDS, this latter interacts with the enzyme active site mainly by hydrophobic interactions (π-π stacking with Phe79, Tyr205 and Phe208 residues and alkyl-π with Pro125). The selenium atoms of DPDS interact with the carboxylic group of Asp120 and with the Zn²⁺ ion, besides the thiolate group from Cys124 (Fig. 4 A). The putative DPDS metabolite, PSA, also interacts in the Hsδ-AlaD active site, by π-π stacking with Tyr205 and Phe208, H-bond with Ser168, and interactions with Tyr196 (anion-π interaction between the seleninate and the phenyl moieties), Asp120 (repulsive electrostatic interaction between the seleninate and carboxyl groups), and zinc ion (coordination). In addition, Se^…S interaction with Cys124 is observed (Fig. 4B).

Fig. 4.

Molecular docking of organoselenium compounds with Hsδ-AlaD (A, B), Dmδ-AlaD (C, D) and Csδ-AlaD (E, F). (A, C, and E) DPDS binding pose in Hsδ-AlaD, Dmδ-AlaD, and Csδ-AlaD enzymes, respectively. (B, D, and F) PSA binding pose in Hsδ-AlaD, Dmδ-AlaD, and Csδ-AlaD, respectively. H-bonds (green), hydrophobic (π-π, alkyl-π) (purple), cation-π, anion-π, electrostatic interactions (orange), and zinc coordination (blue), are represented by dotted lines; all the distances are in Å. The ligands and the amino acids lateral chains are represented by ball and stick, and stick models, respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

The simulation of DPDS with the Dmδ-AlaD demonstrated that this organoselenium compound could access the active site making hydrophobic interactions with Arg205, Pro212 (alkyl and phenyl groups), Phe204 and Tyr201 (phenyl and phenyl moieties), besides interacting with Arg217 via H-bond (selenyl and guanidinyl groups [95], [96]) (Fig. 4C). In addition, the DPDS showed a Se^…S interaction with Cys122. The PSA molecule also binds in the Dmδ-AlaD active site, through hydrophobic π-π stacking with Phe204 and Tyr201 (phenyl and phenyl moieties), through H-bonds with Ser165, Lys195, and Gln221 (seleninate and OH, NH and C=O groups, respectively), and Zn^…O coordination. Similarly to DPDS, the PSA also showed Se^…S interaction with Cys122 (Fig. 4D).

On the other hand, the docking simulations between the Csδ-AlaD and DPDS demonstrated that it does not enter in the Csδ-AlaD active site. In fact, DPDS binds in the superficial region of the enzyme, close to the entrance of the active site, interacting with the Lys313 (phenyl and carbon chain) and presenting an intramolecular π-π stacking (phenyl- phenyl) (Fig. 4E). In contrast, PSA can access the active site of Csδ-AlaD (Fig. 4F), making H-bonds with Arg301 and Lys291 residues, stabilized by electrostatic interaction with Asp217, and π-π stacking with Phe330 (phenyl and phenyl moieties).

Finally, we simulated the interactions of other putative oxidized organoselenium forms [97] (Fig. S5) to verify if these molecules are able to interact with the δ-AlaD enzymes, and its binding partner. For Hsδ-AlaD, all organoselenium molecules show Se^…S interaction (3.1–5 Å) with the Cys124 residue (Fig. S6), except R,R-DPDS(O). Conversely, for Dmδ-AlaD, only S,R-DPDS(O), R-DPDS(O) and PhSeOH show Se^…S interaction (4–4.4 Å) (Fig. S7). In relation of the Csδ-AlaD, we verified that all the selenoxide forms of DPDS do not bind in the active site (Fig. S8), as observed with DPDS. However, like PSA, PhSeOH enters in the active site and interacts with Lys291. These data suggest that for Csδ-AlaD small organoselenium electrophilic moieties can indeed inhibit the enzyme. In addition, the stereochemistry of the compounds play an essential role in the binding mode in the enzyme.

The predicted binding free energy (ΔG_bind) for the Hsδ-AlaD indicates that the interaction of DPDS with the enzyme is energetically more favored than the interaction PSA-enzyme (Table 1 ). In contrast, ΔG_bind for Dmδ-AlaD suggests a more favorable PSA-enzyme than DPDS enzyme binding. Similarly, in Csδ-AlaD, PSA showed (negatively) larger binding energy than DPDS. Finally, the presence of oxygen atoms in the oxidized forms of DPDS enabled the formation of H-bonds facilitating thermodynamically the binding.

Table 1.

Predicted ΔG_bind (kcal/mol) from molecular docking.

Enzyme	Hsδ-AlaD	Dmδ-AlaD	Csδ-AlaD
DPDS	−6.2	−5.9	−5.2
PSA	−5.1	−6.1	−5.9
R,R-DPDS(O)	−7.0	−8.0	−5.1
S,R-DPDS(O)	−7.0	−7.9	−6.7
S,S-DPDS(O)	−6.1	−6.1	−4.8
R-DPDS(O)	−7.0	−6.5	−5.2
S-DPDS(O)	−6.3	−6.0	−5.1
PhSeOH	−4.6	−5.8	−4.8

In the Hsδ-AlaD and Dmδ-AlaD enzymes, both PSA and DPDS presented similar binding pose, interacting with amino acid residues from the active site. Notably, Cys124 and Cys122 (Hsδ-AlaD and Dmδ-AlaD, respectively), stabilization occurs via Se^…S interaction (Fig. 4A–D). However, for Csδ-AlaD, only PSA binds in the active site, and no Se^…S interaction is present because the Csδ-AlaD does not have Cys residues in the active site (Fig. 4E–F). These outcomes strongly suggest that organoselenium compounds binding in the active sites could prevent the entrance of the substrate 5-Ala, thus inhibiting the enzymes.

The previous in vitro assays have indicated that the mechanism of Hsδ-AlaD (or mammalian δ-AlaD) and Dmδ-AlaD inhibition by organoselenium compounds involves Cys oxidation because dithiothreitol (DTT_red) could reactivate the enzyme from these sources [10], [11], [24], [29], [30], [98]. The Se^…S interaction could lead to the formation of the selenenyl sulfide bond (Se–S) [99], [100], an adduct between the protein and the selenium compound, by means of a nucleophilic attack of the thiolate moiety of Cys124(122) to the Se atom of either DPDS or PSA. In fact, previous experimental as well as theoretical studies have indicated that Se-S bound can be easily formed between reduced thiol-containing molecules and diselenide- (Se-Se) and seleninic acid (R-SeO₂H)-containing molecules [99], [101], [110], [102], [103], [104], [105], [106], [107], [108], [109].

In the next step, a vicinal thiol group – from Cys122(120) and/or Cys132(130) – could perform a nucleophilic attack to the electrophilic S atom of the Se–S bond, leading to the disulfide bridge (S–S) formation, i.e., thiol oxidation, and the release of zinc ion [7], [99], [108], [111], [112], [113]. In fact, the distances between the S atoms are 3.7–4.6 Å for both Hsδ-AlaD and Dmδ-AlaD. Previous studies suggested that the cysteine oxidation (S–S) in the Hsδ-AlaD active site involves Cys124 and Cys132 residues. The Cys124 residue is the first thiolate that reacts with diselenides or selenides/selenoxides, forming the Se–S intermediate; then, Cys132 reacts with this intermediate leading the disulfide bridge, denaturing the active site [51], [114].

Csδ-AlaD has no Cys residues in the active site and consequently, the Cys oxidation mechanism is not possible. PSA, likely due to its polarity, has a better affinity for the active site of Csδ-AlaD (where polar and basic residues are present). PSA has a highly electrophilic Se atom [30], [115], [116]. Its Hirshfeld partial charge is higher than the one computed for Se in DPDS and in the other selenium compounds of this study, indicating a deficiency of electrons (Table S4). In addition, due to the short distance between the Se atom and the amino group from Lys291 (Se^…N = 3.8 Å, Fig. 4F), a nucleophilic attack from the Lys291 on PSA could occur, forming a seleninamide moiety (Ph-Se(O)NH-Lys), i.e., an adduct between the enzyme and the organoselenium moiety, which might inhibit the Csδ-AlaD. The seleninamide formation from seleninic acid has already been reported in the literature [115], [117]. The formation of seleninamide could prevent the reaction between the Lys291 residue and the 5-Ala substrate (the Schiff base formation, which is an essential step in the δ-AlaD catalytic cycle [14], [15]). The in vitro study of Farina et al. (2002) [30] showed that in the presence of DTT_red the Csδ-AlaD is not inhibited. A possible explanation is that the sulfur atom from DTT could react with the seleninamide adduct, forming a thioseleninate moiety (Ph-Se(O)S-DTT) releasing the free Lys291 and consequently reactivating the enzyme (Ph-Se(O)NH-Lys + DTT-SH → Ph-Se(O)S-DTT + Lys-NH₂). In fact, the thioseleninate intermediate can be formed via a reaction between seleninamide and thiol molecules [5], [100], [118], [119].

The reaction between the PSA and the active site in Csδ-AlaD was investigated by means of DFT calculations. For this purpose, we set up a model reaction, using EtNH₂ as a model of the Lys residue and PSA in the protonated form (PhSeOOH), as it should be due to its proximity to Arg301 (Fig. 4F) and because water is a better leaving group than hydroxyl anion. Our results (Fig. 5 ) indicate that the seleninamide formation is energetically favored, both in the gas and water phase. The reactant complex (PhSeOOH·EtNH₂) is characterized by an H-bond between the hydroxyl and amino groups and by a short distance Se^…N (3.8 Å), promoting the release of a water molecule and the formation of the Se–N bond in the product complex (PhSeONHEt·H₂O) (Fig. 5A).

Fig. 5.

The reaction of PSA with Lys residue model (EtNH₂) modeled at (PCM)-mPWPW91/def2PVTZ level of theory. (A) Intermolecular interactions of reagent and product complexes; the distances are in Å. (B) Energetics (kcal/mol) of the reaction in the gas and water phase. PhSeOOH + EtNH₂ and PhSeONHEt + H₂O correspond to the reagents and products, while PhSeOOH·EtNH₂ and PhSeONHEt·H₂O are the reactant and product complexes, respectively. All energy values are relative to the free reactants.

The proximity between electrophilic forms of organoselenium molecules and nucleophilic moieties from critical amino acids residues (in this case Se^…S/N interactions from Cys124, Cys122, and Lys291, from Hsδ-AlaD, Dmδ-AlaD, and Csδ-AlaD, respectively) could lead to covalent bonds formation, and consequently, these adducts can impair the functions of enzymes, inhibiting them. This mechanism could justify the toxicity of some organoselenium compounds.

The understanding of the mechanism of organoselenium compounds toxicity will be crucial in the designing of new molecules less toxic and more selective in relation to pharmacological targets. In this sense, the potential role of metabolites of a given drug can also be informative, as suggested by our present study. Organoselenium molecules have promising biological activity, and Ebselen is under clinical trials as potential lithium mimetic for bipolar disorder [120]. Of particular importance, Ebselen has been recently used against SARS-CoV-2 in vitro and presented antiviral activity possibly by inhibiting the main protease (Mpro) enzyme from the virus of COVID-19 [121]. Selenothymidines (selenium-containing AZT derivatives) are potential pharmacological agents against cancer [122]. DPDS presents many therapeutics properties (anxiolytic, antidepressant-like, anticancer, neuroprotective, and others) and its mechanism of action involves the modulation of the cellular redox status [123]. DPDS could modulate any protein having reactive thiol groups due to the lack of specific molecular targets. In this way, new DPDS derivatives with higher selectivity for specific protein targets still need to be developed.

4. Conclusion

The present work, entirely performed in silico and combining multiscale approaches, provides an efficient explanation to experimental in vitro data, giving evidence that DPDS inhibits Hsδ-AlaD and Dmδ-AlaD enzymes, but does not inhibit Csδ-AlaD [11], [29], [30]. The molecular docking simulations between the selected organoselenium molecules and δ-AlaD could provide a possible explanation for this observation. The homology modeling showed that Csδ-AlaD does not present Cys residues in the active site, and consequently, DPDS has not a substrate to oxidize. On the other hand, the putative metabolite PSA could access the active site, interacting with the Lys291 residue (Se^…N), preventing the entrance of the 5-Ala substrate, and consequently inhibiting the Csδ-AlaD. By DFT calculations, we have demonstrated that the reaction between PSA and Lys is indeed energetically favored. In Hsδ-AlaD and Dmδ-AlaD enzymes, both DPDS and PSA can access the active site, interacting with Cys124 (122), by Se^…S interaction, which could lead to Cys oxidation, and, consequently, protein denaturation and enzyme inhibition. This type of study is essential to understand the reactivity and selectivity of organoselenium compounds in biological systems and can lead to better rational drug design. On the basis of these promising computational results, further studies are prompted. In addition, due to its protein similarity and organoselenium binding pose, Dmδ-AlaD rather than Hsδ-AlaD could be used as a model to test the toxicity of new organoselenium molecules.

CRediT authorship contribution statement

Pablo Andrei Nogara: Conceptualization, Methodology, Investigation, Validation, Writing - original draft. Laura Orian: Methodology, Resources, Writing - review & editing, Supervision, Funding acquisition. João Batista Teixeira Rocha: Methodology, Resources, Writing - review & editing, Supervision, Funding acquisition.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article: