Design, synthesis, in vitro anti-α-glucosidase evaluations, and computational studies of new phthalimide-phenoxy-1,2,3-triazole-N-phenyl (or benzyl) acetamides as potential anti-diabetic agents

An important target in the treatment of type 2 diabetes is α-glucosidase. Inhibition of this enzyme led to delay in glucose absorption and decrease in postprandial hyperglycemia. A new series of phthalimide-phenoxy-1,2,3-triazole-N-phenyl (or benzyl) acetamides 11a–n were designed based on the reported potent α-glucosidase inhibitors. These compounds were synthesized and screened for their in vitro inhibitory activity against the latter enzyme. The majority of the evaluated compounds displayed high inhibition effects (IC50 values in the range of 45.26 ± 0.03–491.68 ± 0.11 µM) as compared to the positive control acarbose (IC50 value = 750.1 ± 0.23 µM). Among this series, compounds 11j and 11i represented the most potent α-glucosidase inhibitory activities with IC50 values of 45.26 ± 0.03 and 46.25 ± 0.89 µM. Kinetic analysis revealed that the compound 11j is a competitive inhibitor with a Ki of 50.4 µM. Furthermore, the binding interactions of the most potent compounds in α-glucosidase active site were studied through molecular docking and molecular dynamics. The latter studies confirmed the obtained results through in vitro experiments. Furthermore, in silico pharmacokinetic study of the most potent compounds was also performed.


Structure-activity relationship (SAR) for α-glucosidase inhibitory activity. As evidenced by
obtained results, all of the N-phenylacetamide derivatives, with the exception of 2,6-dimethyl derivative 11c and 4-nitro derivative 11j, were more potent than positive control acarbose. In contrast, N-benzylacetamide derivatives 11m-n did not show significant inhibitory activity against this enzyme. In the N-phenylacetamide derivatives 9a-l, the most potent compound was 4-bromo derivative 11j. Substitution of bromine with nitro group, in case of compound 11k, leads to the complete loss of inhibitory activity and substitution of bromine with ethyl group, in case of compound 11d, leads to a significant decrease in the inhibitory activity.
The second most potent compound was 2,6-dichloro derivative 11i. 2,3-Dichloro derivative 11g and 2,4-dichloro derivative 11j as regioisomers of compound 11i showed moderate anti-α-glucisidase activity in comparison to this compound. The third potent compound among the newly synthesized compounds was 3-chloro derivative 11f. Replacing of chloro substituent with fluorine atom, as in case of compound 11e (the fourth potent compound), caused to a moderate decrease in the anti-α-glucosidase activity. The fifth potent compound was un-substituted compound 11a. Placing two methyl substituents on the pendant phenyl group of compound 11a depending on their positions, has interesting effects on the inhibitory effect of this compound: 2,3-dimethyl derivative 11b was 1.8-fold less potent than compound 11a while 2,6-dimethyl derivative 11c was inactive. Another interesting point that can be observed about the effect of methyl substitution in the obtained inhibitory activities is that the 4-nitro derivative is an inactive compound, but by adding a methyl on 2-position of the 4-nitrophenyl group, potent compound 11l is obtained.
As can be seen in Table 1, N-benzylacetamide derivatives 11m-n did not show a significant inhibitory against α-glucosidase.
Comparison of the new compounds 11 with template compounds E and F. The comparison of IC 50 values of the new phthalimide derivatives 11 with their corresponding analogs of benzimidazole derivatives E revealed that with the exception of 2,6-dichloro and 4-bromo derivatives, benzimidazole analogs were more potent than their corresponding analogs of phthalimide series (Scheme 2).The mentioned trend is also observed in the comparison between phthalimide derivatives 11 and quinazolinone derivatives F (Scheme 2) 17 . To gain further insight into the mechanism of α-glucosidase inhibition of the title class of compounds, a kinetic study was performed on compound 11j as the most potent α-glucosidase inhibitor. For this purpose, the reaction rates in the presence of different concentrations of compound 11j were measured in various concentrations of substrate (p-nitrophenyl-a-D-glucopyranoside). Graphs of different concentrations of inhibitor were drawn by the Lineweaver-Burk plot (Fig. 2a). As the concentrations of inhibitors increased, V max values were not affected, but K m values gradually decreased, thereby indicating that compound 11j was a competitive inhibitor against α-glucosidase (Fig. 2a). The K i value was calculated directly by plotting the slope of each line in the Lineweaver-Burk plots into the different concentrations of compound 11j (Fig. 2b). The results proved that K i value of compound 11j was 50.4 µM.

Molecular docking study.
In order to explain interactions of the most potent compounds 11j, 11i, and 11f. in α-glucosidase active site, molecular docking simulation was carried out 18 . The superposed structure of acarbose and the title new compounds in the active site of target enzyme is shown in Fig. 3. Interaction modes of the positive control acarbose and selected compounds 11j, 11i, and 11f. were showed in the Fig. 3 and details of their interactions in the active site of target enzyme were listed in Table 2. www.nature.com/scientificreports/ The 2D interaction mode of acarbose demonstrated that this positive control created eight hydrogen bonds with active site residues Thr307, Asn241, Glu304, Pro309, Ser308, Thr301, Arg312, and Gln322 (Fig. 4). Acarbose also formed a hydrophobic interaction with residue His279, two non-classical hydrogen bonds with residues Val305 and His239, and two unfavorable interactions with residues Thr307 and Arg312. Moreover, BE of acarbose was -4.04 kcal/mol.
The most potent compound 11j created three hydrogen bonds with residues Asn241, His239, and Arg312 via oxygen atom of phenoxy moiety, 1,2,3-triazole ring, and carbonyl unit of acetamide moiety, respectively (Fig. 4). Phthalimide moiety of this compound established three interaction with active site: a π-anion interaction with Glu304 and two hydrophobic interactions with Pro309. Two π-cation interactions were also observed between phenyl ring of phenoxy moiety with residue His279 and 1,2,3-triazole ring with His239. 4-Bromophenyl ring of compound 11j established two hydrophobic interactions with Phe158 and Arg312 via bromo substituent and phenyl ring, respectively. The second potent compound 11i established two hydrogen bonds with residues Asn241 and Phe157 via 1,2,3-triazole ring and NH unit of acetamide moiety, respectively (Fig. 4). The latter amino acid also interacted with carbonyl unit of acatamide moiety via a non-classical hydrogen bond. Compound 11i formed a π-cation interaction and a π-anion interaction with residues His279 and Asp408. This compound also established several hydrophobic interactions with residues Val305, Pro309, His239, Phe311, Tyr313, Arg312, and Phe157.  www.nature.com/scientificreports/ The third potent compound 11f. created three classical hydrogen bonds with residues Asn241 and Asp408 via acetamide moiety and two non-classical hydrogen bonds with residues Phe311 and Arg312 via 1,2,3-triazole ring (Fig. 4). Phenyl ring of phenoxy moiety established a π-anion interaction with Glu304. Compound 11f. also formed an unfavorable interaction with residue Asp408 via carbonyl unit of acetamide moiety.
Docking study on the inactive compounds 11c and 11m showed that this compounds formed only a hydrogen bond with the active site of target enzyme. As can be seen in Fig. 5, compound 11c formed a hydrogen bond with Asn241, a π-cation interaction with His279, two non-classical hydrogen bonds with Thr301 and Arg312, an unfavorable interaction with Thr307, and several hydrophobic interactions with Val305, His239, His279, Pro309, and Arg312. Inactive compound 11m established a hydrogen bond with Phe157, two π-anion interactions with Glu304, two non-classical hydrogen bonds with His239 and His279, and several hydrophobic interactions with Pro309 and Arg312.

Molecular dynamics.
Interaction of a ligand with a protein is a dynamic phenomenon which takes place on a very small time scale. Molecular dynamic simulation helps to grasp the interaction between ligand and protein and evaluate the stability and flexibility of the resulting complex. To this end, dynamics of the protein-ligand complex is simulated in an environment very similar to the natural environment that includes water and ions. Based on in vitro studies compound 11j has the most potential to inhibit α-glucosidase. Therefore the complex of α-glucosidase-11j was simulated in an explicit hydration environment by molecular dynamics simulation to evaluate the stability, flexibility and intermolecular interactions between α-glucosidase and this compound 19 . Moreover molecular dynamics of acarbose as a standard inhibitor in complex with α-glucosidase was simulated in an explicit hydration environment to have a decent reference for comparison. In this study, molecular dynamic simulation was performed in two steps. A 10 ns simulation at the first step to investigate if the ligands i.e. acarbose and 11j were stable at their binding site on α-glucosidase. After confirming the stability of ligands in Table 2. Interaction mode details of the compounds 11j, 11i, and 11f. www.nature.com/scientificreports/ their binding site, simulation time was extended for another 10 ns to gain a better comprehension of the behavior of these compounds in the active site of α-glucosidase. Stability of 11j and α-glucosidase was confirmed in this step too. For further evaluation the trajectory file was analyzed by several tools. To evaluate the stability of the complexes, root-mean-square deviation (RMSD) and radius of gyration (Rg) of all structures of the trajectory were calculated and the related graphs were drawn. To assess the residual flexibility and the flexibility of ligand atoms, the root mean square fluctuation (RMSF) of backbone atoms of α-glucosidase and heavy atoms of ligands were calculated. Figures 6 and 7 show the result of RMSD calculations. According to Fig. 6 that shows the RMSD of backbone atoms of α-glucosidase in complex with 11j and acarbose, RMSD of α-glucosidase does not change very much and is less than 0.25 nm in all the simulation trajectory. RMSD less than 0.3 nm indicates a stable structure. The average RMSD values of α-glucosidase in the complex with acarbose and/or 11j were 0.17 and 0.16 nm, respectively. RMSD less than 0.25 nm was observed for acarbose and/or 11j in the complex with α-glucosidase too that shows they had the least conformational changes and were completely stable during the simulation time (Fig. 7). The average RMSD values of acarbose and/or 11j in complexes with α-glucosidase were 0.14 and 0.14 nm, respectively.
All atoms and consequently residues in a protein have some fluctuations. The fluctuation of α-glucosidase residues in complexes with acarbose and 11j are depicted in Fig. 8. According to this figure RMSF of α-glucosidase residues in the complex of this enzyme with acarbose and/or 11j are very similar and almost match. α-Glucosidase is a big protein with 579 residues and several domains with different structure and functions. As it could be seen  Fig. 9. As can be seen in this figure, the RMSF of all heavy atoms in these ligands is less than 0.2 nm. This low RMSF indicates that these compounds have a stable structure in complex with α-glucosidase and intermolecular interactions limit their fluctuations. Among the heavy atoms of acarbose and 11j those that were part of a ring had the lowest RMSF. Rings usually have less fluctuations as atoms of the ring limit their movements and make stable non-bond interactions such as π-Anion, π-Cation, π-Alkyl, π-π T-shaped, π-sigma, and hydrogen bonds with binding site residues of the protein. The radius of gyration (Rg) of α-glucosidase was calculated for evaluation of protein compactness during simulation (Fig. 10). The average Rg of α-glucosidase was 2.530 and 2.54 nm in the complex of α-glucosidase with acarbose and 11j, respectively. The Rg value of α-glucosidase in complexes with both acarbose and 11j was only in the narrow range of 2.45 to 2.52 nm and did not show a significant upward or downward trend during the simulation time that shows stable protein structures.   (Table 3). As can be seen in Table 3, acarbos did not follow of Lipinski 'Rule of five' while compounds 11j and 11i followed of this rule. Acarbose and compounds 11j and 11i had poor permeability to Caco-2. Permeability to blood brain barrier (BBB) and skin for the title compounds is in the acceptable range. Compounds 11j and 11i had high human intestinal absorption (HIA) while acarbose did not have HIA. Prediction of mutagenicity of acarbose and compounds 11j and 11i demonstrated that these compounds are mutagen. Moreover, this study predicted that acarbose had carcinogenic effect on mouse and did not have this effect on rat while compound 11j has carcinogenic effect on rat and mouse. Unlike the compound 11j, compound 11i has not carcinogenicity on the latter animals. Cardiotoxicity (hERG inhibition) of the positive acarbose is ambiguous while compounds 11jand 11i in term of this type of toxicity have medium risk.

Conclusion
In conclusion, we designed and synthesized a novel series of phthalimide-phenoxy-1,2,3-triazole-N-phenyl (or benzyl) acetamide 11a-n. The synthesized compounds 11a-n were evaluated against α-glucosidase because were designed based on active pharmacophores of potent reported α-glucosidase inhibitors. The majority of the title compounds displayed high α-glucosidase inhibitory activity. Among them, compounds 11j and 11i represented the most potent α-glucosidase inhibitory activities with IC 50 values of 45.26 ± 0.03 and 46.25 ± 0.89 µM

Experimental
General. Melting points of compounds 11a-n were measured with a Kofler hot stage apparatus. IR spectra of these compounds were recorded with a Nicolet Magna FTIR 550 spectrophotometer (KBr disks). 1 H and 13 C NMR spectra of the title phthalimide derivatives were obtained with a Bruker FT-400 (TMS was used as an internal standard). Mass spectrometry results were obtained with an Agilent Technology (HP) mass spectrometer (Ionization potential: 70 eV). Elemental analysis was determined with an Elementar Analysen system GmbH VarioEL CHNS mode.

N-(2,3-dichlorophenyl)-2-(4-((4-(1,3-dioxoisoindolin-2-yl)phenoxy)methyl)-1H-1,2,3-triazol-1-yl)acetamide (11 g).
In the third step docking study of acarbose and the selected compounds was performed in the active site of modeled α-glucosidase. The 3D structures of the acarbose and selected inhibitors were built by MarvineSketch 5.8.3, 2012, ChemAxon (http:// www. chema xon. com) and converted to pdbqt coordinate using Auto dock Tools. The pdbqt coordinate of enzyme was produced using the same software. Prepared pdbqt files were used as input files for the AUTOGRID program. In this program for each atom type in the selected ligand, maps were calculated with 0.375 Å spacing between grid points and the center of the grid box was placed at x = 12.5825, y = 7.8955, and z = 12.519. The dimensions for the active site box were set at 40 × 40 × 40 Å. Flexible ligand dockings were accomplished for the selected ligands. Each docked system was carried out by 50 runs of the AUTODOCK program search by the Lamarckian genetic algorithm (LGA). The best poses of the selected ligands were selected for analyzing the interactions between target enzyme and the selected ligand. The obtained results were visualized using BIOVIA Discovery Studio v.3.5. Molecular dynamics. MD simulations were performed using Groningen machine for chemical simulations (GROMACS) 5.1.2 22 . Topology files and other force field parameters of the selected compounds were made by SwissParam server 23 . Protein topology file was constructed by using the pdb2gmx command and CHARMM27 all-atom force field (CHARM22 plus CMAP for proteins). The protein-ligand complex (in.gro format) was created in Notepad + + and the topology file of the protein was edited to include topology parameters of ligand as well. The resulting complex was centered in a cubic box with a side length of 2.0 nm and the SPC216 water model was used to fill the system. The net negative charge of the protein was neutralized by fifteen Na + ions which replaced the same number of water molecules. The Steepest descent minimization algorithm was used for the minimization of the system in a maximum number of 50,000 steps until the maximum force became less than 10.0 kJ/mol. For NVT equilibration the v-rescale algorithm was used in 300 K with a coupling constant of 0.1 ps and time duration of 500 ps. The last phase in preparation of the system was NPT equilibration. In this step, Berenson pressure coupling algorithm with a coupling constant of 5.0 ps was applied for 1000 ps of NPT simulation. Particle Mesh Ewald (PME) algorithm was used for long-range electrostatics and cut-off method for van der Waals interactions. Cut off distances were set at 1.0 nm for the calculation of the electrostatic and 1.2 nm for van der Waals interactions. Finally, 20 ns MD simulation was performed for the protein-ligand complex.
In silico pharmacokinetic and toxicity predictions. In silico prediction of pharmacokinetic properties and toxicity profile of the positive control acarbose and the most potent compounds 11j and 11i was performed using by the preADMET online server 24 . Ethical approval and consent to participate. The ethics code for this work is IR.NIMAD. REC.1400.155 (https:// ethics. resea rch. ac. ir/ IR. NIMAD. REC. 1400. 155).

Data availability
The datasets used or analyzed during the current study are available from the corresponding author on reasonable request.