Synthesis, in vitro inhibitor screening, structure–activity relationship, and molecular dynamic simulation studies of novel thioquinoline derivatives as potent α-glucosidase inhibitors

New series of thioquinoline structures bearing phenylacetamide 9a–p were designed, synthesized and the structure of all derivatives was confirmed using different spectroscopic techniques including FTIR, 1H-NMR, 13C-NMR, ESI–MS and elemental analysis. Next, the α-glucosidase inhibitory activities of derivatives were also determined and all the synthesized compounds (IC50 = 14.0 ± 0.6–373.85 ± 0.8 μM) were more potent than standard inhibitors acarbose (IC50 = 752.0 ± 2.0 μM) against α-glucosidase. Structure–activity relationships (SARs) were rationalized by analyzing the substituents effects and it was shown that mostly, electron-donating groups at the R position are more favorable compared to the electron-withdrawing group. Kinetic studies of the most potent derivative, 9m, carrying 2,6-dimethylphenyl exhibited a competitive mode of inhibition with Ki value of 18.0 µM. Furthermore, based on the molecular dynamic studies, compound 9m depicted noticeable interactions with the α-glucosidase active site via several H-bound, hydrophobic and hydrophilic interactions. These interactions cause interfering catalytic potential which significantly decreased the α-glucosidase activity.

α-Glucosidase inhibitory activity. To develop new glucosidase enzyme inhibitory agents, all synthesized 9a-p derivatives were screened to assess their potential α-glucosidase inhibitory activities. As presented in Table 1, all synthetic compounds exhibited better α-glucosidase inhibitory activity with IC 50 values in the range  As can be seen in Table 1 (entry 1) the compound (9a, R = phenyl) exhibited an IC 50 value of 88.50 µM with approximately nine-fold improvement in inhibitory potency compared to the positive control. Compound 9b bearing fluorophenyl substitution (as an electron-withdrawing group) at ortho position showed less inhibitory effect than 9a. Changing the position from ortho to para revealed the most potent electron withdrawing entry in this set of compounds 9c with an IC 50 value of 46.00 µM. However, the chloro and bromo substituent derivatives (9d and 9e respectively) were not successful to improve the inhibitory potency compared to their corresponding fluorine derivatives. The other potent agent-bearing electron-withdrawing group came back to 9g bearing 4-nitrophenyl with an IC 50 value of 93.74 ± 1.5 µM. However, this entry was still less effective compared to 9c as an unsubstituted one. The reason can be ascribed to the differences in electronegativity of the mentioned substitutions as well as their size. It seems that the presence of a small electron withdrawing group at the para position of the phenyl ring improved the activities and with the increase in their size at the para position the potencies reduced.
Methyl substitution as a small electron-donating group did not improve the potency compared to 9a as an unsubstituted one although the ortho position (9h) demonstrated better activity compared to para one (9i). Interestingly replacement of 4-methyl with 4-methoxy moiety significantly reduced the IC 50 value to 24.70 µM. Assessments on 9k (R = 4-ethyl phenyl) reveal that the increase of bulkiness of the electron-donating groups at the para-position of phenyl is in favor of inhibition. Similarly, multi substitution is favorable so that 9l bearing 2,6-dimethylphenyl recorded the best inhibitory activity (14.0 ± 0.6).
Further, the effect of ring replacements was also evaluated. Results disclosed that bulky ring substitutions such as naphthyl (9n) improved the inhibitory activity significantly compared to the phenyl counterpart (9a).
As can be seen in benzyl derivatives, elongation of the linker deteriorated the potency. This trend can easily be seen in 9p versus 9c.
The summary of SAR was presented in Fig. 3. The highest potency to inhibit α-glucosidase was observed in compounds bearing 2,6-dimethylphenyl followed by naphthyl and 4-methoxyphenyl with IC 50 values of 14.0, 18.42 and 24.70 µM, respectively (Table 1). Generally, electron-donating groups at R are more favorable compared to the electron-withdrawing groups. Also, in the case of electron donating groups bulk substituent at R is more favorable. On the other hand, the small electron withdrawing group at the para position improved the activity.
Enzyme kinetic studies. According to Fig. 4a, the Lineweaver-Burk plot showed that the K m gradually increased and V max remained unchanged with increasing inhibitor concentration indicating a competitive inhibition. The results showed 9m binds to the active site on the enzyme and competed with the substrate for binding to the active site. Furthermore, the plot of the K m versus different concentrations of inhibitor gave an estimate of the inhibition constant, K i of 18.0 µM (Fig. 4b).
Homology modeling and molecular docking study. It should be noted that the in vitro assay was conducted by using the α-glucosidase enzyme (EC. 3. 2. 1. 20) of Saccharomyces cerevisiae. Since the 3-D crystallographic structure of α-glucosidase of Saccharomyces cerevisiae is not available homology modeling structure method was applied via the protein sequence obtained from uniport.org by using isomaltose (PDB: 3A47) of S. cerevisiae. It was shown that the isomaltose template had high sequence similarity (85% similarity) with the α-glucosidase Saccharomyces cerevisiae. The sequence alignment was exposed in Fig. S17 (see supplementary information). Additionally, tthe Ramachandran plot estimating the homology-modeled protein structure and the conformation of amino acids in the protein was shown in Fig. S18 in the supplementary file. Ramachandran plot distributions indicated that most of the residues are in the favored and allowed regions. Next, to determine the binding sites of the modeled α-glucosidase enzyme, the site mapping tool was applied. Five potential active sites were identified, based on the site map score and overall surface area of active sites. As demonstrated in Fig. 5, the chosen active sites contains a plausible surface area of H-bond acceptor/donor and hydrophobic sites.
The results of the molecular docking study for compound 9m as the most potent derivative are displayed in Fig. 5 and the following interactions were observed between 9m and the active site pocket residues of the enzyme; H-bond between the carbonyl of amide group and His239, dual pi-pi stacking interactions between the phenyl www.nature.com/scientificreports/ ring of benzohydrazone moiety, Tyr71, and Phe177 residues plus many hydrophobic interactions with Phe157, Phe158, Phe300, Val303, Phe311, and Tyr313 residues.

Molecular dynamic simulations.
The comparison between the stability of the enzyme-inhibitor complex and enzyme was assessed using the backbone root mean square deviation (RMSD) during the 100 ns molecular dynamic (MD) simulation (Fig. 6). The RMSD value of α-glucosidase enzyme stabilized after 5 ns in the average value of (3 Å) and remained on the same situation with fewer fluctuations till the 40 ns then the RMSD value had a major rise and had further rising trend until the end of simulation with the average RMSD value of (4.5 Å). The RMSD plot of the α-glucosidase-9m complex was shown in Fig. 6. The complex stabilized after 2 ns at (1.8 Å) and then leaped to 2.75 Å at 20 ns and remained at the same interval with a slightly decreasing trend for the rest of simulation time. The overall RMSD values of systems had a significant difference which can be interpreted as the stabilizing effect of 9m on the enzyme as a potent inhibitor.
The root means square fluctuations (RMSF) of Cα atoms from both systems revealed the detailed mechanism of the ligand interactions with the enzyme. Upon the binding of the ligand to the α-glucosidase, residues movement decreased as a result of non-bonding interactions between the ligand and the enzyme. The most difference among the fluctuations of the system was observed between (amino acids: 250-300) and (amino acids: www.nature.com/scientificreports/ 390-450). As it's showed in Fig. 7, these sequences correspond to the active site's nearby α-helix, β-sheet, and double α-helix regions respectively. The interactions of 9m with the active site pocket of the enzyme which was present in more than 30% of simulations duration are demonstrated in Fig. 8. The interactions briefly consisted of (1) H-bound interaction between the carbonyl of the benzohydrazone group and Arg312, (2) direct and water bridged H-bond between the hydrazone group and Asp349, (3) water bridged H-bond between the quinoline system's nitrogen and the Glu304, (4) pi-cation interaction between the quinoline system and the Arg312 and (7) multiple hydrophobic interactions with Phe300, Phe157, and Lue218.
Next, contributing energy component of non-covalent interactions in the simulation duration is demonstrated in Fig. 9. As in the x-axis there are the interacting residues of the active site with the ligand and in the y-axis, there is the time fraction interaction of the simulation and the stacked bar charts are normalized throughout the trajectory. As it is shown in Fig. 9, Phe157, Phe177, Ala278, and Phe300 exhibited hydrophobic interactions with   www.nature.com/scientificreports/

Conclusion
In this study, a series of n a series of thioquinoline-benzohydrazide linked to different phenylacetamides were designed, synthesized, and evaluated as possible α-glucosidase inhibitors. All synthesized derivatives displayed increased inhibitory activity with IC 50 values in the range of 14.0 ± 0.6 to 373.85 ± 0.8 µM compared to acarbose as the positive control. SARs exhibited the favorable role of balk and spacious electron-donating groups at the para position of the phenyl ring compared to the electron-withdrawing group. The most potent candidate in this series 9m was chosen for further biological evaluation. The enzyme kinetics assessments indicated that compound 9m inhibited α-glucosidase in a competitive inhibition manner (K i = 18 µM). According to the molecular dynamics simulations, the α-glucosidase-9m got stabilized after 2 ns at (1.8 Å) and then leaped to 2.75 Å at 20 ns and remained at the same interval with a slightly decreasing trend for the rest of the simulation  www.nature.com/scientificreports/ time. Also, 9m recorded several H-bond interactions and multiple hydrophobic interactions with the binding site of the enzyme. Based on these results, thio-quinoline derivatives could be considered an attractive candidate for further investigations.

Experimental
Chemistry. All the reagents were purchased from commercial sources. 1 H and 13 C NMR spectra were determined by a Bruker FT-400 MHz spectrometer in DMSO-d 6 . All the chemical shifts were reported as (δ) values. The MS spectra were recorded using an Agilent 7890A spectrometer at 70 eV. CHNOS analysis was performed using ECS4010 Costech Company. IR spectra were obtained with a Nicolet, FR-IR Magna 550. Melting-point were also recorded using Kofler hot-stage apparatus.

2-Mercaptoquinoline-3-carbaldehyde (3).
Then, to a solution of 2-chloroquinoline-3-carbaldehyde 2 (0.01 mol) in dry DMF (50 mL), sodiumsulphide (0.015 mol) was added and stirred for 2 h at room temperature. Then, the reaction mixture was poured into crushed ice and made acidic with acetic acid. The product was filtered off, washed with water, and dried to give desired 2-mercaptoquinoline-3-carbaldehyde 3 that was further purified by recrystallization in ethanol 26,27 .  (7) was added. The reaction mixture was stirred at room temperature for 12 h followed by addition of cold water. The resulting solid was washed with water three times and then with petroleum ether giving solid compounds 8a-p 28 .

2-((3-((2-Benzoylhydrazineylidene)methyl)quinolin-2-yl)thio)-N-benzylacetamide(9o
IC 50 values were obtained from the nonlinear regression curve using the Logit method 15,29 . Enzyme kinetic studies. The  Homology modeling. The α-glucosidase sequence of Saccharomyces cerevisiae was downloaded from uniprot.org by the UniProt code of P38158 31 . The isomaltase enzyme (PDB ID: 3A47) of Saccharomyces cerevisiae was chosen as the template in the previous reports 32 . The homology modeling was conducted using the maestro prime 33 .

Molecular docking.
The modeled protein in the previous stage was prepared using the protein preparation wizard 34 . And the missing sidechains and loops were filled using the prime tool and H-bonds assigned by PROPKA at pH = 7.4. The 2D structure of the ligand was drawn in ChemDraw (ver. 16) and exported as SDF files to use by the ligprep module in the next step. Ligand prepared by OPLS_2005 forcefield using EPIK at a target pH of 7.0 ± 2 35 . Site map tool used to find the potential binding sites of the Enzyme-substrate complex 36 . The site map report included 5 potential binding sites with at least 15 site points per each reported site and more restrictive definition of hydrophobicity. Grid box generated for each binding site using sites as entries with the box size of 25 A, afterward compound rf-16 docked on binding sites using glide 37 with standard precision and flexible ligand sampling reporting 20 poses per ligand. Molecular dynamic simulation. MD simulation was performed using desmond from Schrodinger Maestro interface 38 . Results of the MD simulation conducted on the complex from the previous docking stage. An orthorhombic cell filled with TI3P model water molecules have been defined and adequate Na ions have been added to the system to neutralize the overall charge of the complex. The simulation time was 100 ns. The NPT ensemble (constant number of atoms; constant pressure, i.e., 1.01325 bar; and constant temperature, i.e., 300 K) were applied with the 1.0-ps interval Nose-Hoover chain method as the default thermostat with and 2.0-ps interval Martyna-Tobias-Klein as the default barostat. The results of the molecular dynamic simulation were analyzed using the maestro graphical interface 39 .

Data availability
All data generated or analyzed during this study are included in this published article and its supplementary information files.