Unraveling Potential Glyoxalase-I Inhibitors Utilizing Structure-Based Drug Design Techniques

Introduction

By dividing into new cells, human cells develop and multiply as required by the body. When old cells pass away due to aging or damage, new cells take their place. Cancer develops when the host cells grow out of control and spread to other parts of the body.¹ It is an out-of-control growth that is caused by the accumulation of genetic and epigenetic alterations in healthy cells.²

To allow normal cell survival, the glyoxalase system detoxifies ketoaldehydes, such as methylglyoxal, to innocuous metabolites, such as D-lactic acid, in a GSH-catalyzing manner.^3,4 The detailed detoxification process is shown in Figure 1. The isomerization mechanism of the glyoxalase system starts with proton abstraction from the C1 of hemithioacetal substrate by Glu172 that is produced non-enzymatically by the reaction of methylglyoxal with glutathione (GSH) to the produce enediol intermediate. Subsequently, Glu172 covey the proton to C2 of the substrate resulting in the formation of the S,D-lactoylglutathione (SLG), and this represents the substrate for Glo-II that converts the thioester into non-toxic D-lactic acid by hydrolysis.

Figure 1 Methylglyoxal detoxification pathway by glyoxalase system. (GSH; glutathione).

Many metalloenzymes in living organisms have been considered as pivotal targets for cancer treatment. The histone deacetylase family (HDAC), carbonic anhydrase, bacterial methionine aminopeptidase and matrix metalloproteinase 3 (MMP-3) were subjected to extensive research to find potential inhibitors. The main strategy was to find a suitable zinc-binding moiety that is considered the main determinant for enzyme inhibition.^5–8

A series of studies have reported Glo-I overexpression in several cancer types, including pancreatic cancerous tissues,⁹ breast cancer proteomics,¹⁰ prostate cancer cells,¹¹ and melanoma,¹² while these tumor cells have increased glycolytic activity, which increases the cellular levels of hazardous substances, such as methylglyoxal.^13–15

Inhibition of this enzyme leads to the accumulation of toxic byproducts which cause tumor cell death.^16,17 Glo-I is a potential target for the design and discovery of new Glo-I inhibitors with various chemical scaffolds, because it is a rate-limiting step in this system. Glo-I overexpression has been linked to multidrug resistance in various malignancies.^18–20

Over the past decades, there had been extensive research on the design and development of Glo-I inhibitors with potential anticancer activity.^21,22 The earliest inhibitors that emerged were either substrates or transition-state analogues, which are known for being metabolically deactivated by glutamyl transpeptidase (GGT). Due to the obstacles related to metabolic instability of GSH-based inhibitors, and to the violations of drug-like properties, some researchers have created non-GSH-based Glo-I inhibitors.^23,24 Natural flavonoids and anthocyanidins were investigated by Takasawa et al, as potential Glo-I enzyme inhibitors.²⁵ In order to identify crucial structural elements of flavonoids to serve as Glo-I inhibitors, computational and experimental approaches were performed by Al-Balas et al, where the structure-activity relationship (SAR) of a panel of 24 compounds belong to the flavonoid class of natural products, was explored. It was found that a C2/C3 double bond, along with a di- or tri-hydroxylation of the benzene rings, particularly ring A, are pivotal for enzyme inhibition (primarily as zinc-binding features). Additionally, the inhibitory effect of these compounds does not entirely depend on the ketol system. Scutellarein is one of the flavonoids that had been studied and have been found the most effective inhibitor.^26,27

Structurally, Glo-I has a homodimeric structure with two similar active sites, each monomer weighing 42kDa and containing 184 amino acids. The active sites could be easily identified due to the presence of zinc atom as a catalytic cofactor. Each active site had three primary areas: a positively charged entrance (Arg37, Arg122, and Lys156), a structurally central zinc atom – crucial for stabilizing the enediolate intermediate, and a deep hydrophobic pocket (Phe62, Leu69, Leu92, and Met179) (Figure 2).^28–31 Our research team has worked on this target over the past period.^32–35 This work is a continuation of previous efforts, and it is mainly based on the investigation of potential inhibitors, hinged on structure-based (SBDD) techniques, by utilizing the ASINEX^® commercial database and testing the potential candidates by in vitro assays.

Figure 2 The active site of Glo-I enzyme; (A) Positively ionized mouth (blue), hydrophobic pocket (brown) amino acids, (B) Zinc atom (grey sphere) with major amino acids bound to it (Glu99, Gln33, his126, Glu172).

Materials and Methods

The workflow of this effort is illustrated in Scheme 1.

Scheme 1 Detailed scheme of the main steps that were conducted within this project.

3D Ligand Pharmacophore Mapping

The first step was extracting ligands that possess zinc-binding groups, such as thiol; phosphate; carboxylic acid; five-membered aromatic ring possessing N-N atoms, such as (imidazole, tetrazole, and triazole); hydroxamic acid from ASINEX^® database by using “3D Ligand Pharmacophore Mapping” protocol in Discovery Studio software from Biovia^®. The ASINEX^® database Elite Library and the Synergy Library contained 91,001 compounds, and these were screened in this process.

In silico Docking

The mapped compounds from the previous step were docked into the Glo-I active site using the CDOCKER protocol in the Discovery Studio software. This is a validated software, and it uses a CHARMm-based molecular dynamics (MD) scheme to dock ligands into a receptor binding site. Random ligand conformations are generated using high-temperature MD. The conformations are then translated into the binding site. Candidate poses are thereafter created using random rigid-body rotations, followed by simulated annealing. A final minimization is then employed to refine the ligand poses. Glo-I was retrieved from the Protein Databank PDB code 3W0T with a 1.35 Å resolution. The protein was prepared and minimized using default parameters before docking, and the zinc atom had a +2 charge, and was assigned an octahedral geometry. In addition, all mapped ligands were prepared using a preparation ligand protocol with default parameters, with the exception that the generated isomers were false (the geometry of the commercial compounds was conserved). The CDOCKER protocol was applied for the docking analysis. The parameters of the CDOCKER protocol were set as default; the number of top hits was 10, number of random conformations also 10, simulated annealing was set to true, the force field was CHARMm, the full potential was set to obtain more accurate results and Ligand Partial Charge Method was assigned to be Momany-Rone which uses a set of bond-charge-increment rules for assigning atom partial charges. Finally, this docking protocol was conducted on Glo-I enzyme active sites with a 10 Å sphere to ensure that all three major areas (a positively charged entrance, central zinc atom, and a deep hydrophobic pocket) were covered, and the docked compounds could be in contact with them. The second active site was assigned to a 9.95 Å sphere as a technical step to differentiate between the two active sites in the final stage of analysis.

Selection of Potential Inhibitors

The docking analysis was based on the CDOCKER-INTERACTION energy, and the best positions between the two active sites with the highest CDOCKER-INTERACTION energy were chosen. The next step was to select compounds with the highest CDOCKER-INTERACTION energy at the two active sites. This process yielded 21 poses that represent the highest pose from each compound was selected. Finally, 14 compounds were selected by avoiding repetition of the same functional groups and ensuring diversity in the zinc-binding functional groups.

In-Situ Minimization and Calculation of Binding Energy

The 14 selected compounds were checked for their suitability for purchase as they should have acceptable Total Binding Energy values (negative). In situ minimization was conducted as a preparatory step before calculating the Total Binding Energy. The poses that were elected for the in situ minimization were the top 14 poses from the previous docking step. Regarding the parameters, the, Adopted Basis NR was used as a minimization algorithm as it gives more accurate results in which better convergence to the local minima. Regarding RMSD gradient, it is a tolerance applied to the average gradient during a cycle of minimization (kcal/(mol-Å), it was set 10⁹ to give more accurate results. Subsequently, the binding energy protocol was used to calculate the Total Binding Energy. With regard to parameters, Poisson Boltzmann with non-polar Surface Area (PBSA) was utilized as an implicit solvent model in which PB equation is fundamental for the most precise continuum models of solvation effects. In situ minimization was set False because this step was conducted previously as a separated step the Ligand Conformational Entropy set to True in order to determine the total binding energy. As the ligand flexibility is based on this method, the run time can be greatly extended. The negative total binding energy number means that the ligand bound spontaneously without using any energy, whereas a positive value means that the binding required energy to take place and could only do so if the energy was available.

Biological Assay

Once the selected compounds had been purchased, their Glo-I inhibitory activity was evaluated in vitro against human recombinant Glo-I, as described in other studies.^32,33 All chemicals and bulk solvents were obtained from Acros (Thermo Fisher Scientific, New Jersey, US) and Sigma-Aldrich Co., respectively, and had the highest purity throughout the country. Following the manufacturer’s instructions (R&D Systems^® Corporation, USA), the biological activity of the selected compounds was assessed based on their in vitro inhibitory activity against human recombinant Glo-I. The enzyme was reconstituted in our laboratory, kept at −70° C, which will be thawed prior to use on the day of the test. Regarding the assay buffer, it was prepared by utilizing 0.5 M sodium phosphate monobasic and 0.5 M sodium phosphate dibasic. DMSO was used to dissolve the test compounds to prepare a 10mM stock solution, and the absorbance of each well was measured for 200 s. at λ_max 240 nm using a UV microplate reader at 25° C. The substrate solution was prepared by adding 25 mL buffer with 706 µL methylglyoxal and 706 µL glutathione, which was prepared previously, and then vortexed for 15s to incubate them in a water bath at 37° C for 15 min. A blank was prepared by mixing the substrate mixture with the assay buffer; the human recombinant Glo-I was then added. Three duplicates of each test were performed, and the average was calculated. To determine the top-hit compounds, IC₅₀ was calculated using the Prism software. Myricetin was used as a positive control with an IC₅₀ of 3.38 ± 0.41 µM.

Results and Discussion

3D Pharmacophore Mapping

The presence of zinc-binding groups is the main criterion for the discovery of potent Glo-I inhibitors. Therefore, the 3D pharmacophore mapping was conducted to extract the “Zinc Binding Groups” containing compounds, by building a zinc query file that contains the major groups; these were reported with good zinc-binding ability such as thiol, phosphate, carboxylic acid, and five-membered aromatic ring possessing N-N atoms, such as (imidazole, tetrazole, and triazole), and hydroxamic acid. “Elite and Synergy Libraries” that contain a panel of early ADMET filters have been used to screen these compounds and ensure that screening hits can be quickly optimized from hit to lead, and that they are free of ADMET issues. A total of 1809 zinc-binding group-containing compounds were obtained from the 91,001 compounds.

In silico Docking

Molecular docking was performed using the CDOCKER protocol, a grid-based molecular-docking algorithm. The two binding sites were defined using the Define and Edit Binding site tool within the DS, with a sphere of 10 Å and 9.95 Å radius, which is based on a co-crystallized ligand-protein complex. The CDOCKER protocols were employed with default parameters, with the exception of full potential to obtain more accurate results. The resulting pose ranking was based on the CDOCKER interaction energy, which included the interaction energy, plus ligand strain with the interaction energy alone (Table 1). This protocol gives 10 poses for each docked ligand with its score and in our work, the highest scored poses were adopted for the selection purposes.

Table 1 The Values of -CDOCKER-Interaction at Two Active Sites

Selection of Potential Inhibitors

The selection criteria were based on the highest CDOCKER interaction energy score, while their investigation manually is intended to avoid any inconsistencies that could happen during the docking process. The next step was to identify ligands with the highest CDOCKER interaction energy shared by the two active sites. ADM 22644975 scored the highest, with a value of 72.6 for one of the active sites, and 64.3 for the second. This has resulted in the identification of 21 common compounds that have high scores on both active sites (Table 1). Fourteen compounds were chosen for purchase, and particular attention was given to avoid the repetition of similar ligands (Figure 3).

Figure 3 The chemical structures of the finally selected and purchased chemicals from the commercial database.

In-Situ Minimization and Calculation of Binding Energy

For the chosen compounds, these two stages serve as confirmation steps prior to the purchase. Estimating the Total Binding Energy was conducted after the in situ minimization had taken place. The Adopted Basis Newton Raphson NR was used as a minimization algorithm, whereas the remaining parameters were left at their default settings. Given that we need to calculate the total binding energy rather than the binding energy, which is the sum of the binding energy and the conformational entropy of the ligand, the Poisson Boltzmann with non-polar Surface Area (PBSA) model was used as the implicit solvent model. All compounds showed total binding energy values within the range of −122.439 to −2.725 kcal/mol (Table 2). This has confirmed that the selected compounds were suitable for purchase.

Table 2 The Values of in-situ Minimization, Binding Energy and Total Binding Energy

Biological Evaluation

The inhibitory activities of the 14 compounds were conducted in vitro against human recombinant Glo-I. The slope for each compound was calculated by plotting absorbance versus time, and the inhibition percentage was computed as well. Myricetin was used as a positive control. Fourteen compounds were screened at a concentration of 50 µM, and the percentage of Glo-I inhibition was calculated and compared to inhibition of the positive control (myricetin).

Pursuant to the screening process, two compounds, SYN 22881895 and SYN 25285236, have showed inhibition percentages greater than 50%. The remaining compounds have showed a low percentage of inhibition. The average IC₅₀ for SYN 22881895 and SYN 25285236 compounds was 48.77 µM and 48.18 µM, respectively (Table 3).

Table 3 The Average Percent of Inhibition and IC50 for the Tested Compounds

Analysis of Docked Poses of the Active Compounds

Figure 4 shows the binding patterns of SYN22881895 and SYN25285236. Regarding compound SYN22881895, the benzene ring performed good hydrophobic interactions with amino acids Phe62 and Met179 in the hydrophobic pocket. The nitrogen atom of the isoxazole ring coordinates with the zinc atom via a metal-acceptor bond. The tetrazole ring showed ionic interactions with the positive mouth basic amino acids, Arg37 and Arg122. Regarding SYN25285236, the benzene ring showed hydrophobic interactions with the hydrophobic pocket of amino acids Met179, Leu92, and Phe62. These represent tetrazole-chelated zinc atoms via attractive charges. Additionally, the ether oxygen forms a hydrogen bond with the basic amino acids.

Figure 4 An interaction representation of the most active compounds depicted as 3D and 2D within the active site of Glo-I of (A) SYN22881895 with its most important interactions both 3D and 2D and (B) SYN25285236 with most important interactions both 3D and 2D.

The tetrazole ring in the most active compound has performed optimal interaction with the zinc atom. This is due to the fact that the adjacent phenyl ring has positioned itself optimally inside the hydrophobic pocket next to the zinc atom. Meanwhile, the phenoxy group has showed ion dipole interaction with basic amino acid Arg37. On the other hand, in compound SYN 22881895, the zinc chelator was the isoxazole ring instead of the tetrazole ring. The para-fluoro methoxy moiety has perfect fit inside the hydrophobic pocket, while the negatively ionized tetrazole ring favored to have ionic interaction with the active site Arg 122 and Arg 37.

It is worth noting that the two active compounds were ranked as the top 21 of the entire database. This is a good indication of the ability of the software to choose actives. For example, the most active compound SYN25285236 ranked the eighth at one active site and the seventh at the other. SYN22881895 ranked the ninth at one active site and the eleventh at the other.

Conclusion

Two active lead hits were identified in this work, which possessed a tetrazole ring. The most active compound SYN25285236 offered the tetrazole ring as the zinc-chelating moiety. On the other hand, the SYN22881895 compound revealed that the tetrazole ring favored ionic interactions with positively ionized entrance, while the isoxazole was the zinc chelator. The most active compound had an IC₅₀ of 48.18µM. This work is based on selecting the crystal structure of Glo-I with excellent resolution (1.35 Å), which was then prepared and minimized for docking purposes. The CDOCKER protocol was used as the main determinant to prioritize the compounds in the commercial ASINEX^® database for the future purchase. The docking process was conducted on the two active sites of this protein to add more credit to the results retrieved from this step. The finally selected compounds were investigated as a confirmatory step by calculating their total binding energies which also returned satisfactory. An in vitro assay was conducted against this enzyme. Generally, acceptable correlation has been detected between the in silico and the in vitro results as those compounds were ranked as the top 21 of the in silico results and showed promising biological activity. In the future, these two compounds will be lead-guiding for deciphering more active compounds.