Design, synthesis, structure, in vitro cytotoxic activity evaluation and docking studies on target enzyme GSK-3β of new indirubin-3ʹ-oxime derivatives

The addition of chalcone and amine components into indirubin-3′-oxime resulted in 15 new derivatives with high yields. Structures of new derivatives were also elucidated through 1D, 2D-NMR and HR-MS(ESI) spectra and X-ray crystallography. All designed compounds were screened for cytotoxic activity against four human cancer cell lines (HepG2, LU-1, SW480 and HL-60) and one human normal kidney cell line (HEK-293). Compound 6f exhibited the most marked cytotoxicity meanwhile cytotoxicity of compounds 6e, 6h and 6l was more profound toward cancer cell lines than toward normal cell. These new derivatives were further analyzed via molecular docking studies on GSK-3β enzyme. Docking analysis shows that most of the derivatives exhibited potential inhibition activity against GSK-3β with characteristic interacting residues in the binding site. The fast pulling of ligand scheme was then employed to refine the binding affinity and mechanism between ligands and GSK-3β enzyme. The computational results are expected to contribute to predicting enzyme target of the trial inhibitors and their possible interaction, from which the design of new cytotoxic agents could be created in the future.

www.nature.com/scientificreports/ of various diseases including cancers, diabetes, Alzheimer's disease 4,5 . GSK-3 was found to have two isoforms, GSK-3α (51 kDa) and GSK-3β (47 kDa). Both of which were observed in mammalian tissues, and together present an 84% overall similarity. Their kinase catalytic domains share around 98% similarity and are only distinguished by an extra Gly-rich stretch in the N-terminal region of GSK-3α 6 . GSK-3β was recognized to play a significant role in the Wingless (Wnt) signaling pathway, suggesting that the inhibition of GSK-3β could lead to decreased cancer cell proliferation, triggering the p53-dependent apoptosis and stimulate the TRAIL-induced cell death 7 . Indirubin and indirubin-3′-oxime are well known as potential anticancer agent 8,9 and thus, a myriad of indirubin derivatives have been devised exhibiting valuable biological activities, high selectivity, and drug-likeness 10 .
In addition, recent studies on interactions between indirubin-3′-oxime with the active site of GSK-3β (PDB ID: 1Q41) offered crucial insights into intermolecular interactions and the mechanism of specificity towards this kinase 11 .
In addition, to discover compounds with high biological activity for drug development, virtual screening has been considered as a convenient and economically efficient method. Previous investigations involving GSK-3β inhibitors have produced promising results in pharmacophore design 12,13 , docking 12,14,15 , and elaboration of structure-activity relationships 15,16 . However, the targeting behavior of GSK-3β toward indirubin, indirubin-3′oxime and its derivatives has not been clearly substantiated. In addition, investigations regarding the relationship between structural differences and the affinity of GSK-3β have been lacking in the literature. The present work focuses on indirubin derivatives due to their promising inhibition activity toward GSK-3β. First, the synthesis of new derivatives by adding chalcone and amine components into indirubin-3′-oxime was attempted with the aim of improving bioactivity and solubility. The structures of obtained derivatives were elucidated through 1D, 2D-NMR and HR-MS(ESI) spectra and X-ray crystallography. Second, due to the essential role of GSK-3β in the development of various cancer types 3,7 , cytotoxic and selectivity of the derivatives were also evaluated in vitro against four human tumor cell lines (HepG2, LU-1, SW480 and HL-60) and one human normal embryonic kidney cell line (HEK-293). Third, as indirubin and its derivatives are potential GSK-3β inhibitors 14,15 , molecular docking of studied compounds on GSK-3β enzyme were performed to explore the possible interaction within its active site. Fourth, the fast pulling of ligand (FPL) simulation was conducted to refine the binding affinity and mechanism of the ligand (molecule) to GSK-3β. Our results demonstrate the potential application of designed compounds in cancer treatment.

Results and discussion
Synthesis studies and structure determination. The click chemistry reaction between compound 4 and azide chalcones through a 3-step process resulted in 11 new indirubin-3′-oxime derivatives (compounds 6a-l). At first, compound 3 was obtained (yield 59%) by N-alkylation reaction between indirubin 1 and propagyl bromide at room temperature in anhydrous DMF solvent using K 2 CO 3 , KI, 1-(butyl)triethylammonium bromide as base catalyst. Then, compound 4 was formed with high yield (79%) by condensation reaction between 3 and hydroxylamine chlorohydric in reflux pyridine solvent. The structure of compound 4 was determined by NMR, HR-MS(ESI) method in combination with single crystal X-ray diffraction measurement which for the first time confirmed 4 with absolute configuration (2′Z indirubin , 3′E oxime ) (Figs. 1, 2). This result also confirms that indirubin 1 has absolute configuration 2′Z 17 . Finally, the azide-alkyne cyclisation reaction between compound 4 and various azide chalcones (5a-l) in DMSO at room temperature with CuI as catalyst afforded 11 new triazoles (6a-l) in 57-70% isolated yield (Scheme 1). Four new indirubin-3′-oxime derivatives 6m-p were synthesized with yield 56 ÷ 68% by reaction between compound 4 and solution A using CuI as catalyst at room temperature (Scheme 1).  www.nature.com/scientificreports/ The structures of obtained compounds were determined by 1D, 2D-NMR and HR-MS(ESI) spectroscopic methods in which compounds 6a was chosen as the representative molecule to determine the exact structure with the combination of COSY, HSQC and HMBC spectral evidence. In the 1 H-NMR spectrum of 6a, proton signals of the group OH oxime, OH chalcone and H-1′ indirubin appeared as singlet at 13.57, 13.35 and 11.78 ppm, respectively. A doublet signal at 8.7 ppm (J = 8 Hz) could be assigned to proton H-4. One triplet signal at 8.24 ppm is the interlacing proton signal of H-6‴ and H-4′. Singlet signal at 8.12 ppm corresponds to proton H-5″, characteristic of 1,2,3-triazole substituted at 1 and 4 position. Two doublet signals with the "roof effect" that   In addition, there are proton signal interaction between CH 2 -C-4″ and C-2 (168.6 ppm); C-4″ (142.8 ppm); C-7a (138.3 ppm) and C-5″ (123.9 ppm). The correlation signal between proton of -CH 2 -CH 2 -O-and C-5″ (123.9 ppm) in HMBC spectrum combined with HR-MS(ESI) method confirmed the exact structure of compound 6a. The structure of the remaining derivatives are also completely consistent with NMR and HR-MS(ESI) data. Combining NMR, HR-MS(ESI) spectra of compounds 6a-p with the absolute configuration of compound 4, it is determined that compounds 6a-l have absolute configuration (2′Z indirubin , 3′E oxime , E chalcone ). Compounds 6m-p have the absolute configuration (2′Z indirubin , 3′E oxime ) similar to the absolute configuration of compound 4. Detail spectroscopic datas of synthesized compounds are provided in Supplementary information S3 and S6. cell viability assay. To gain better insight into the potential anticancer activities of the synthesized compounds, we performed in vitro cytotoxicity assay on four human cancer cell lines and one human normal kidney cell line. In the present study, starting material compounds including indirubin (1), compound 3, compound 4 and indirubin-3′-oxime (2) along with other derivatives (6a-p) were tested for cytotoxicity against HepG2, LU-1, SW480, HL-60 and HEK-293 cell lines using Ellipticine and indirubin-3′-oxime as standard (Table 1). In the past, indole derivatives have been identified as GSK-3 inhibitors 18 . Indirubins are likewise indole derivatives and their biological activities have given rise to the widespread use, such as in remediation of leukemias, in traditional Chinese medicine 19 . According to obtained results, indirubin (1) was identified to have no activity due to their IC 50 value being higher than 20 μM. This was also observed in some of its derivatives including compounds 3 and 4 which exhibited weak signals of cytotoxicity against four cancer cell lines.
The chalcone component (6a-l) derivatives offer clear advantages in anti-proliferation activities in comparison to its parent compounds. Compound 6f exhibited the strongest activity with IC 50 values of 2.01 ± 0.43, 1.30 ± 0.14, 2.54 ± 0.25, 0.98 ± 0.12 and 1.03 ± 0.11 μM on HepG2, LU-1, SW480, HL-60 and HEK-293, respectively, which are considerably more toxic than indirubin-3′-oxime. In addition, structural data from compound 4 and series 6a-l in combination with cytotoxic and anti-proliferation activities prompt the speculation that the presence of oxime group in the structure of studied compounds could play an important role in inhibition of cancer cell activity. www.nature.com/scientificreports/ High IC 50 values of chalcone derivatives of compound 4 also suggest the role of OC 3 H 7 group in triggering cytotoxic activities through conformational changes. The exception was observed with compound 6b (IC 50 > 20 μM against all tested cell lines), which is in line with poor interaction with GSK-3β, calculated from the following molecular docking studies. This result suggests the simultaneous intervention of methoxy groups at R 2 , R 3 , R 4 position resulted in the decreased cytotoxicity. It is noteworthy that different substitution groups at R 3 position resulted in toxicity selectivity between cancer and normal cell line of compounds 6e, 6h and 6l. The selectivity is evidenced by the slightly lower IC 50 values of compounds 6e, 6h and 6l on HEK-293, at 6.98 ± 0.25, 9.74 ± 0.53 and 7.12 ± 0.18 μM, respectively, in comparison to IC 50 values obtained with the other four cancer cell lines.
IC 50 values of the series 6m-p were slightly lower than those of 6a-l and were approximately equivalent to each other. However, the addition of amine components into indirubin-3′-oxime, generating the series 6m-p, seemed to result in better cytotoxicity and improved solubility properties and better cytotoxicity.
Docking studies. Autodock 4.2.6 was utilized to predict the interaction of all molecules with GSK-3β enzyme model. The X-ray crystallographic structure complex with co-crystalized ligand (indirubin-3′-oxime) was collected from Protein Data Bank (RCSB). Based on literature studies, two known inhibitors of GSK-3β (CHIR-98014 and BIO-acetoxime) were chosen as standard ligands for docking validation 20,21 . Obtained dock score for these two inhibitors were − 11.82 kcal/mol and − 10.69 kcal/mol, respectively, thus, the threshold value of the docking energy also determined as − 10.69 kcal/mol and any molecules whose docking energies are close to this threshold would be viewed as potential inhibitors of GSK-3β in the virtual screening stage. In general, results from the docking procedures showed that the compound 6f has the highest docking score, at − 14.09 kcal/ mol, far exceeding than that of the standard ligand (Table 2). It is noted that the initial compounds including 1, 3 and 4 had lower absolute dock score than indirubin-3′-oxime (2). However, higher absolute values of docking scores of chalcone-added components (6a-l) and amine-added components (6m-p) through triazole bridge into the compound 4 structure suggested better binding affinities toward GSK-3β. The dock score values were further converted to the prediction inhibition constants (K i , pred ). Through comparing the K i values in Table 2, high correlation between order of cytotoxic activity (IC 50 ), dock score and K i of tested compounds was observed. For example, compound 6f was recorded to have the most negative value (46.6E−12M) and its IC 50 and dock score was the most potential in comparision to those of others.
The relationship between dock score and experimental binding free energies on four cell lines was shown in Fig. 3. The experimental binding free energy was calculated from formula ΔG exp = RT lnK i where gas constant R = 1.987 × 10 −3 kcal K −1 mol −1 and absolute temperature T = 300 K. Here, in this equation the IC 50 was assumed as equal to the inhibition constant K i and measured in moles. The obtained correlation coefficient implies the high accuracy of the computations for all four models HepG2, LU-1, SW480 and HL-60 with R = 0.90; 0.93; 0.93 and 0.90, respectively.
The role of polar residue groups such as Asp133, Tyr134 and Val135 situating on the binding cavity in the ligand-ATP recognition and affinity has been confirmed and emphasized, due to the interaction of Asp200 with the phosphate hydroxyl group of ATP 22 . In addition, previous crystallographic evidence on GSK-3β interaction with various inhibitors has highlighted the role of Asp133 and Val135 in enhancing affinity to GSK-3β. Other residues such as Ile62, Gln185 and Cys199 were also shown to contribute to the interaction of the inhibitor to the backbone (Table 3). Docking validation analysis reveals that CHIR-98014 creates five H-bonds toward GSK-3β through four residues Ile62, Thr138, Val135 and Gln185. Bio-acetoxime, another known inhibitor, formed 3 hydrogen bonds with the enzyme target including amino acids Asp133 and Val135 (Fig. 4). In this study, the following compounds 6a, 6c, 6f, 6i with substituents changes in R 1 position and 6b were taken as the representatives to analyze the mechanism of actions. The potential interactions between the ligands and GSK-3β were further elaborated by evaluating residues within 5 Å of the ligands.
The carbonyl and two amine groups of compound 1 formed three hydrogen bonds toward carbonyl oxygen of Asp133 and Val135 in the hinge region however its docking score (− 9.08 kcal/mol) is lower than Indirubin-3′-oxime (2) (− 9.51 kcal/mol) thus excluding this compound as an appropriate inhibitor (Table 3 and Fig. 4, S3). www.nature.com/scientificreports/ Compound 3 had slightly lower dock score (− 8.49 kcal/mol) and was recognized to form only one hydrogen bond with Val135 on the side that is opposite to the Val135-binding side in compound 1. This is possibly due to a conformational change exhibited by alkyl group attached to amidine position ( Figures S1, S3).
In compound 4, the carbonyl group has been replaced with an oxime group, this modification could be the crux to ameliorate docking energy of compound (− 9.30 kcal/mol). Ile62, Asp133, Tyr134, Pro136, Gln185 and Cys199 were the key residues involved in hydrophobic interactions ( Figure S2). The key hydrogen bond with Val135 appears to be unique to this oxime group which further stabilizes the interaction of compound and provides an interesting model to explain the biological activities observed for the designed compounds (Table 3 and Figures S1, S3). Therefore, this group should be kept in lead optimization process.
Removal of alkynyl group in compound 4 resulted in indirubin-3′-oxime (2). The absence of this group might have made the space, which possibly caused the two amine groups and carbonyl group to interact with residues Asp133, Val135 through 3 hydrogen bonds with distance of 2.85, 2.56 and 2.85 Å, respectively (Table 3 and Figures S1, S3). The interaction was also indicated by a significant docking score (− 9.51 kcal/mol). The oxime group in this case do not interact with Val135 but contribute one hydrogen bond with Ile62 to stabilize   www.nature.com/scientificreports/ the ligand. In addition, an array of hydrophobic interactions was observed for Gly63, Val70, Ala83, Tyr134, Pro136, Leu188 and Cys199 ( Figure S2). Compounds 6a, 6c, 6f and 6i are chalcone-added components and are differentiated by the substituents at R 1 position (H, OCH 3 , OC 2 H 5 , OC 3 H 7 ). Binding energy values to GSK-3β of those ligands were reported to be − 13.64, − 13.17, − 14.09 and − 12.85 kcal/mol, respectively ( Table 2). It is noted that the oxime group in compounds 6c and 6f were observed to initiate hydrogen bond with Val135. Compound 6a formed H-bonding with two residues Ile62 and Val135 similar to those formed by indirubin-3′-oxime. The third H-bond interaction between hydroxyl group of 6a and the carboxyl oxygen of Gln185 could be the key for its high bioactivity. Figure S1 showed the binding of 6c, which was characterized by H-bonding with Ile62, Asn186 and Cys199. Hydrophobic bondings were also found to stabilize interaction of 6c with GSK-3β through residues such as Val70, Lys85, Arg141, Gln185 and Leu188 ( Figure S2). The substituent OC 2 H 5 located at the R 1 position of compound 6f was found to interact with Thr138 active site region, which suggests the importance of this residue in the formation of ligand binding affinities. Besides, the hydrophobic pockets formed with 6f including Ile62, Asn64, Val70, Ala83, Lys85, Asp133, Arg141, Cys199 also strengthen the interaction between ligand and the protein target ( Figure S2). Being different from compounds 6c and 6f, compound 6i showed the formation of the hydrogen bond between its oxime group and carbonyl oxygen of Gln185. Presumably, this phenomenon is due to conformational change caused by the bulky substituent OC 3 H 7 . The conformational change also causes the carbonyl and hydroxyl group of this ligand anchored to Asn64 and Tyr134 via two hydrogen bonds (Figures S1, S3).
The bulky substituents situated at R 2 , R 3 and R 4 position in compound 6b could be the reason leading to drastically decreased docking energy (− 9.34 kcal/mol), and the formation of only one hydrogen bond with Ile62 at the oxime group (Table 3 and Figure S1). This result suggests that without simultaneous interaction with Ile62 and Val135, ligand could not trigger inhibition activity against GSK-3β.
Calculated pharmacokinetic parameters including miLogP (octanol-water partition coefficient), TPSA (total molecular polar surface area), MW (molecular weight) and toxicity prediction of indirubin-3′-oxime derivatives are shown in Table 4. www.nature.com/scientificreports/ In general, although the molecular weights of designed compounds were high, their TPSA were significantly higher than that of Ellipticine, indicating that designed molecules might pass through membrane more easily, making them particularly suitable for oral use. In addition, the calculated miLogP of derivatives 6m-p were lower than those of 6a-l, indicating better solubility of these compounds than those of the others. Interestingly, predicted toxicities of all the designed derivatives were lower than that of the commercial drug Ellipticine. The LD 50 value of Ellipticine was 178 mg/kg, classifying it as a toxic compound. Meanwhile, predicted LD 50 values of all designed compounds was 1,000 mg/kg, indicating that they are ranked as moderately toxic compounds. All these results suggest the potential of indirubin-3′-oxime derivatives in the treatment of cancer. fpL simulation. Computer-aided drug design is often used to explore the probable inhibitor, resulting in a reduction of cost and time for therapeutic growth 23 . During the issue, the ligand-binding affinity is normally required to be calculated 24 since a more efficient ligand often adopts a stronger binding affinity 25 . Numerous approaches were developed in order to resolve the problem [26][27][28][29] . Among these techniques, the fast pulling of ligand (FPL) is a very efficient method with low-required CPU time consumption 30 . In particular, a ligand is forced to travel from bound to unbound states via a harmonic-external force. The physical details during unbinding process reveal the binding affinity and mechanism of a ligand to GSK-3β enzyme.
The FPL technique was applied to 10 complexes as shown in Table S3 of the Supplementary (SI file). The relative binding affinity of a ligand to GSK-3β enzyme was estimated using 8 independent FPL calculations. The pulling force was recorded every 0.1 ps and other metrics were monitored every 10 ps. All of the computed values were averaged over 8 independent trajectories. The recorded pulling force and work were shown in Figures S4  and S6. The displacement of a ligand during FPL simulations was mentioned in Figure S5. The ligand-binding affinity is able to estimate via the difference of interaction energy between a ligand and GSK-3β enzyme over the FPL simulations referring to the previous work 30 .
The interaction energy term was computed via formula , where E full cou and E 0 cou are electrostatic interaction energy between a ligand and GSK-3β enzyme at bound and unbound states, respectively; E full vdW and E 0 vdW are vdW interaction energy between a ligand and GSK-3β enzyme at bound and unbound states, respectively. The interaction energy curve was shown in Figure S7. The obtained results were described in Table S3. The total interaction energy difference adopts a high correlation with respected experiments ( R = 0.88 ) shown in Fig. 5. The obtained correlation coefficient implies the high accuracy of the computations. Moreover, the precision of the calculated results is also high because the computed error is small, which was estimated using root-mean-square error (RMSE) with linear regression RMSE = 0.8 kcal/mol. In addition, the binding mechanism of a ligand to GSK-3β enzyme was also characterized. The vdW interaction energy controls the binding process of a ligand to GSK-3β enzyme. In particular, the vdW term accounts for ca. 79% of total interaction energy that is significantly stronger than the electrostatic term, which accounts for ca. 21% only. When designing new inhibitor for GSK-3β enzyme, the issue should be considered carefully. Table 4. Pharmacokinetic parameters and toxicity prediction of research compounds. *Toxicity prediction class: 1 → 6 (High toxicity to non-toxic).

conclusion
In conclusion, we have described an efficient pathway for the synthesis of novel indirubin-3′-oxime derivatives with advantages of simple operation conditions and high yields. Structures of compounds were elucidated through spectroscopic methods and crystal structure of compound 4 has been clearly determined using X-ray crystallography method. Testing of synthesized derivatives for in vitro cytotoxic activity towards four cancer cell lines (HepG2, LU-1, SW480, HL-60) and one human normal kidney cell line (HEK-293) demonstrated considerable cytotoxic and antiproliferative potential, especially for compound 6f. Compounds 6e, 6h and 6l with substitution groups at R 3 position exhibited toxicity selectivity between cancer and normal cell lines. Molecular docking studies have shed light on the interaction of these compounds towards GSK-3β. The binding affinity of tested compounds are in high correlation with experimental binding free energy, besides, the probable binding poses proposed by the docking simulation could offer a plausible explanation for the impact of different substituent components to the bioactivity of derivatives. The FPL approach evaluates the ligand-binding affinity of GSK-3β enzyme with high accuracy (R = 0.88) and precision (RMSE = 0.8 kcal/mol). The vdW term dominates over the electrostatic ones during the binding process of a ligand to GSK-3β. When designing a new inhibitor for GSK-3β enzyme, these issues should be carefully considered. In general, our results suggest that the proposed routine is an important orientation for the design and synthesis of new anticancer agents from indirubin-3′-oxime.

Materials and methods
General synthesis procedures. All chemicals were purchased from Sigma-Aldrich (USA), VWR (EU) except for 2, 5a-l were prepared according to previous procedures 31,32 . The isolation of Indirubin (1) was carried out using Strobilanthes cusia plant materials growing in Vietnam in (2′Z) configuration according to the procedures of Cuong et al. 33  cell viability assay. Cell viability assay was conducted using SRB method for HepG2, LU-1, SW480 and HEK-293 cell lines, based on the measurement of protein content 35 . Cells were plated with 180 μl growth medium into 96-microwell plates (4 × 10 4 cells per well) and allowed to grow overnight. Test compounds at various concentrations including 20 μM; 4 μM; 0.8 μM; 0.16 μM with incubation medium were supplemented and cultivated for additional 48 h in the same conditions. Afterwards, the removal of the medium was performed and  For HL-60 cell line, 5 × 10 3 cells/well were seeded in 96-multiwell plate and cultured overnight. The test compounds were serially diluted to the wells with concentrations ranging at 20 μM; 4 μM; 0.8 μM and 0.16 μM then incubated for 72 h. Thereafter, each well was added with 10 μl of MTT (5 mg/ml in 1X PBS), followed by incubation for 3 h and centrifugation at 1,500 rpm for 10 min. The media were discarded then added 150 μl DMSO to each well. Optical density was measured using a microplate reader at 540 nm. IC 50 values were calculated as the concentrations that show 50% inhibition of proliferation on the HL-60 cell line.
Ligands and receptor preparation. Crystal structure of GSK-3β (PDB ID: 1Q41) was employed for the docking studies 11,36 . The 3D structure of protein was obtained from the Protein Data Bank (PDB) and prepared using Autodock Tools (ADT). The designed molecules (ligands) were prepared using MarvinSketch version 19.27.0 and PyMOL version 2.2.2 37 . The energy minimization was carried out using Gabedit version 2.5.0. The Molinspiration and ProTox-II cheminformatic server were utilized to predict bioactivity and assess the toxicity of all research compounds. Docking using AutoDock4. The docking simulation procedure was performed by AutoDock 4.2.6 utilizing Lamarckian genetic algorithm and an empirical binding free energy function. Docking modes produced by this procedure are expected to agree with X-ray crystal structures 38,39 . In addition, the method also allows for simulation of interactions between candidate compounds and their three dimensional structures receptors, thereby improving ligand flexibility. The detail of molecular docking simulation is given in Supplementary information S2. fpL simulations. Structure of complexes was obtained via molecular docking method. Caver 2.1 40 was employed to evaluate the disassociate direction of a ligand referring previous works 30 . All complexes were then aligned to be the disassociate pathway oriented to the Z-axis. The complex was inserted into a periodic boundary condition rectangular box (8.15 × 8.21 × 12.75 nm), which consists of an enzyme GSK-3β, a ligand, 26,000 water molecules and 9 Cl-ions as shown in Fig. 6. In particular, the protein was parameterized via the AMBER99SB-ILDN force field 41 , and the ligand was represented using general Amber force field 42 . The water molecule was topologized using TIP3P water model 43 .
GROMACS version 5.1.5 44 was employed to carry out the molecular dynamics (MD) simulations. The simulation was performed according to the following four steps: energic minimization, NVT, NPT, and steered-MD (SMD) simulations. In particular, the non-covalent pair was affected within a range of 0.9 nm and the pair list was updated every 5 fs. The Particle mesh Ewald method 45 was employed to mimic the electrostatic interaction with cut-off of 0.9 nm. The van der Waals interaction was affected in a range of 0.9 nm. Both of NVT and NPT simulations were carried out with the length of 100 ps at 300 K. During simulations, the GSK-3β C_α atoms were restrained by using a weak harmonic force of 1,000 kJ/mol nm 2 . The last snapshot of NPT simulations was used as initial structure of SMD simulations. In this scheme, a harmonic-external force with cantilever spring constant of k = 600 kJ/mol/nm 2 and pulling speed v = 0.005 nm/ps was put on the ligand center of mass along the Z-direction. The ligand was disassociated out of the binding site of GSK-3β enzyme over SMD simulations. www.nature.com/scientificreports/ The data was recorded every 0.1 ps during SMD simulations. The calculations were independently repeated 8 times with the same initial conformation to guarantee the sampling of calculation.
Statistical analysis. All data are presented as mean ± SD. Experiments were carried out in triplicate for the accuracy of data. Statistical significant differences were realized at p < 0.05 via Student's t test.

Data availability
Supplementary data in this article are provided in Supporting information. 1 H NMR and 13 C NMR data and spectrums of compounds 3, 4, 6a-p were available free of charge via the Supporting Information.