GC SGC7901 cells in the logarithmic phase were harvested and counted using a hemocytometer to maintain a cell density of 5 × 10³ per well. The 96-well plates were incubated at 37°C with 5% CO₂ for 24 h before further use. For the methyl thiazolyl tetrazolium (MTT) assay, different drug concentrations were added to SGC7901 cells in the logarithmic growth phase. Two hundred microliters of drug-containing medium or blank medium was added to each well. For each drug group, six duplicate wells were used. The absorbance (optical density, OD) of each well was measured at 450 nm with a microplate reader. The OD value was used to calculate the survival rate of SGC7901 cells with each drug, and the IC₅₀ value was calculated using GraphPad Prism 9.0.

To reveal the relationship between electronic structure and anti-GC activity, density functional theory (DFT) studies of quantum chemistry were carried out on the ten EVO derivatives. First, the molecular structures of the EVO derivatives were fully optimized to obtain equilibrium geometries at the B3LYP/6-311++G^** level of theory. The frequency of the optimized geometry was then calculated to identify the nature of the stationary points. The results showed that there were no imaginary frequencies in the optimized structures of the EVO derivatives. This indicated that the optimized geometries corresponded to stationary points and minima. Finally, natural bond orbital charges and frontier molecular orbitals (FMOs) were calculated at the same level. All DFT calculations were performed using Gaussian 09 software (Frisch et al., 2009).

The initial structure of the receptor–ligand complex was obtained from molecular docking using the AUTODOCK 4.2 program (Morris et al., 2009). Ligand models were obtained from the optimized geometries of the DFT calculations. The initial receptor model was extracted from the X-ray crystal structure of TOP1 in complex with the inhibitor, topotecan, obtained from the Protein Data Bank (PDB:1K4T) (Staker et al., 2002). The crystal structure had 765 amino acid residues and a resolution of 2.10 Å. The ligand and receptor files were examined and prepared using AutoDock Tools 1.5.6.

The original binding inhibitor (topotecan) and crystal water molecules were deleted and polar hydrogen atoms and Gasteiger charges were added to the initial receptor structure. Nonpolar hydrogens were merged. For each docking ligand, rotatable bonds were defined. The receptor is considered a rigid molecule, whereas the ligand is flexible. A grid of 44 × 44 × 44 with 0.375 Å grid spacing was defined as the docking box for each complex. Docking studies were carried out using the Lamarckian genetic algorithm (Fuhrmann et al., 2010) in AutoDock 4.2.

For each system, one hundred conformations were generated. Finally, the optimal conformation with the best rational orientation in the DNA–TOP1 pocket was selected from the docking experiment.

The optimal docking conformation for the complex was used as the initial structure for the MD simulations. General AMBER force field (GAFF, version 2) (Wang et al., 2004) was employed to describe the force field for the ligand molecules. The parameters for the ligand were generated by the antechamber module. The topology parameters for TOP1 and the DNA were created using the Amber ff19SB force field (Tian et al., 2020).

For each system, a total of 100 ns MD simulations were performed to collect data. The system was solvated in a water box using an explicit TIP3P water model. The marginal distance of the water box was set at 15 Å to guarantee that the system was completely immersed in the solution. The total charges of the system were calculated and Na⁺ was added to neutralize them. First, to avoid unfavorable interactions caused by solvents and ions, the system was subjected to minimization using steepest descent and conjugate gradient methods. Subsequently, the complex system was heated from 0 to 300 K in an NVT ensemble using Langevin dynamics (Feller et al., 1995). The NPT ensemble was subjected to equilibrium for 50 ps at 300 K at a constant pressure of 1 atm. Finally, 10,000 frames were extracted from a total of 100 ns of simulations as the average conformation of the MD equilibrium for statistical analysis. All dynamic simulations were performed using the PMEMD CUDA version of AMBER 20 (Case et al., 2020). The MD trajectory data were analyzed by the CPPTRAJ module (Roe and Cheatham, 2018; Cheatham et al., 2019).

The molecular mechanics generalized Born surface area (MM/GBSA) method (Srinivasan et al., 1998; Lee et al., 2004) was used to calculate the binding free energy between the two most active EVO derivatives and TOP1. This is an effective method to quantitively evaluate the binding affinity between a ligand and receptor. The energy term for each complex was calculated by averaging over 2,000 frames of the last 20 ns of an MD trajectory. To evaluate the contribution of each residue, the total binding free energy between DNA–TOP1 and the ligand was decomposed using MM/GBSA binding free energy decomposition (Gaillard and Simonson, 2014). The contribution of entropy was not included in this approach. The MMPBA.py (Case et al., 2020) program was used to calculate free energy.

An MTT assay was used to investigate the proliferative inhibitory effects of different concentrations of the target compounds on the GC SGC7901 tumor cell line, with EVO as the parent compound. The growth inhibition rate of the compound on tumor cells was determined by calculating the percentage of dead cells (Supplementary Figure S1). As shown in Supplementary Figure S1, the inhibition rate of these 10 derivatives gradually increased with increasing drug concentration. They displayed a concentration-dependent relationship. At 200 μmol·L⁻¹, the inhibition rate of each compound reached its maximum. The inhibition rates of the EVO derivatives in GC SGC7901 cells were higher than that of EVO, except EVO-5 and EVO-8. The order of the anti-GC activity of the ten tested derivatives was as follows: EVO-8 < EVO-5 < EVO-10 < EVO-3 < EVO-7 < EVO-2 < EVO-9 < EVO-4 < EVO-6 < EVO-1. The inhibition rates of EVO-1 and EVO-6 were 81.17% ± 5.08% and 80.92% ± 2.75%, respectively, indicating that among these 10 derivatives, EVO-1 and EVO-6 had the strongest anti-GC activity. Therefore, EVO-1 and EVO-6 were selected for subsequent analysis of the TOP1 inhibition mechanisms.

The chemical structure of a drug determines its physicochemical properties and directly affects the absorption, distribution, metabolism, and excretion of drug molecules in the body (Nehra et al., 2022). The relationship between their structure and activity is of great significance for the design and development of drugs. To investigate the relationship between structure and activity, we performed DFT calculations on the 10 EVO derivatives. The molecular structures of these 10 EVO derivatives were fully optimized using the DFT method at the level of B3LYP/6-311++G^**. The equilibrium geometries and atom numbers of these optimized derivatives are shown in Supplementary Figure S2. For compound EVO-4, the optimized structure was consistent with the X-ray structure (Nie et al., 2020). The optimized geometries of other derivatives were almost nonplanar, similar to the optimized geometry of EVO (Liu et al., 2022).

Compounds EVO-1 to EVO-5 underwent alkyl substitution on the thiophene ring. The anti-GC activity of these compounds decreased with the increasing steric hindrance of the alkyl substituent on the thiophene ring. Based on the structures of compounds EVO-1 to EVO-5, compounds EVO-6 to EVO-10 introduced a methoxy group at position 10 of the indole ring. When methoxy substituents were introduced at position 10 of the A ring, anti-GC activity decreased to a certain extent. For example, the activity of EVO-6 was lower than that of EVO-1. This indicates that the methoxy substituent at position 10 of the A ring decreased the anti-GC activity to some extent. One previous study also showed that 10-methoxy substitution reduces the antitumor activity of EVO derivatives (Fan et al., 2021). However, a more recent study (Liang et al., 2022a) showed that antitumor activity was improved to some extent after the introduction of methoxy, methyl, and chlorine substituents at position 10 of the A ring. In particular, 10-chlorine-substituted derivatives showed significant antitumor activity against different GC cell lines. However, by comprehensively considering both antitumor activity and toxicity, 10-methyl was found to be more suitable. Among these 10 derivatives, EVO-1 and EVO-6 showed the greatest inhibitory activity on the proliferation of SGC7901 cells. This may indicate that a less bulky substitute group results in better anti-GC activity. In addition, di-substitution between the A ring and thiophene ring may have a synergistic effect. EVO-1 and EVO-6 may serve as promising compounds for the development of potential anti-GC agents.

Molecular docking is an important method for structure-based drug design. It can predict the binding mode and affinity between small ligand molecules and receptor biomolecules (Meng et al., 2011; Luca and Giulio, 2019). To gain a deep understanding of the anti-GC activity of compounds EVO-1 and EVO-6, we utilized molecular docking techniques to investigate the ligand–receptor interactions of EVO-1 and EVO-6 with the known anticancer target TOP1. As shown in Figure 3, we plotted the binding models of EVO-1 and EVO-6 with TOP1. The binding conformations of EVO-1 and EVO-6 with TOP1 were very similar, which is consistent with the previous predicted binding mode of EVO with TOP1 (Liu et al., 2022). Despite the nonplanar structure of EVO and its derivatives, they can still bind to the active site of TOP1. Their orientation in the active site cavity is almost perpendicular to the main axis of the DNA and parallel to the base pairs. In contrast with the binding pattern of topotecan in the crystal structure (Staker et al., 2002), the EVO derivatives bind at the entrance of the active cavity. This suggests that pharmaceutical chemists can introduce other functional groups to occupy the active cavity.

The ligand intercalated at the site of DNA cleavage, with the A ring directed toward the major groove and the E ring pointing toward the minor groove. Base-stacking interactions occurred between the ligand and DNA base pairs. The long axis of the ligand molecule was parallel to the axis of base pairing. This intercalation-binding mode may explain how EVO-1 and EVO-6 specifically block DNA relaxation. Therefore, they could have inhibitory effects on TOP1, exerting anti-GC activity. This implies that EVO-1 and EVO-6 have the potential to be developed as lead compounds against GC.

The docking model also predicted that only one direct hydrogen bond was formed between the enzyme and EVO-1. The distance between the carboxyl oxygen of Asp533 and the carbonyl oxygen of EVO-1 is 3.2 Å, and they can therefore form a hydrogen bond. In contrast, this hydrogen bond is replaced by the NH of Arg364 and the carbonyl of EVO-6. This hydrogen-bond interaction can be interpreted by MEPs. The MEP around the carbonyl oxygen atom is negative, which facilitates formation of the hydrogen bond. The importance of the carbonyl group is also supported by structure–activity studies of EVO (Dong et al., 2012). For example, when the carbonyl O is replaced by S, the antitumor activity of the molecule is weakened. This indicates that hydrogen bonding between the ligand and Asp533 is important for antitumor activity. This hydrogen-bonding interaction needs to be maintained or replaced by new hydrogen bonding during the modification of the EVO structure. Asp533 was also coordinated to the e-nitrogen of Arg364. The distance between the hydrogen atom of the e-nitrogen of Arg364 and the carbonyl group of Asp533 is 2.2 Å for EVO-1 and 1.9 Å for EVO-6. This demonstrates that Arg364 and Asp533 are critical for ligand binding.

Molecular docking revealed that TOP1, DNA, and EVOs formed ternary complexes. The ternary TOP1–DNA–ligand complex is stabilized by base-stacking interactions and several protein–DNA interactions. The binding energy of EVO-1 with TOP1 was −7.28 kcal·mol⁻¹, which was lower than that for EVO-6 (−7.16 kcal·mol⁻¹). This indicates that EVO-1 has a better anti-GC effect than EVO-6. Thus, it was further demonstrated that EVO-1 can be used as a potential anti-GC candidate.

To simulate the dynamic characteristics of the ternary TOP1–DNA–ligand complex, MD simulation studies are required. A total of 100 ns MD simulations were performed to estimate the stability of the TOP1/EVO-1 and TOP1/EVO-6 systems. The root-mean-square deviation values of the DNA and the heavy atoms of the proteins and inhibitors for the TOP1/EVO-1 and TOP1/EVO-6 systems are shown in Figure 4. The results indicate that fluctuations in the protein, DNA, and inhibitor structures were relatively small during the MD simulation process. This illustrates that a stable TOP1–DNA–ligand ternary complex was formed. In other words, the ligand can stably bind at the cleaved DNA-binding site. This stable ternary complex may lead to DNA damage, which can, in turn, induce cancer cell apoptosis. Thus, EVO-1 and EVO-6 can exert anti-GC effects by stabilizing the covalent TOP1–DNA complex. The two TOP1–DNA–ligand ternary complex structures were used for subsequent analyses.

The fluctuations of each atom relative to its average position were calculated by root-mean-square fluctuation (RMSF), which can characterize the average effect of structural changes over time and the flexibility of each region of the protein. During MD simulations the stability of specific residues can be quantified by RMSF. RMSF was used to examine whether the simulation results were consistent with the crystal structure. The RMSF values for the TOP1/EVO-1 and TOP1/EVO-6 systems are illustrated in Supplementary Figure S4. The residues at positions 235–248, 287–299, 283–338, 388–400, 510–524, 635–646, and 656–697 of TOP1 exhibited a certain degree of fluctuation (RMSFs >1.50 Å). This demonstrates that the core domain of the enzyme is flexible. The linker region containing amino acids 636–712 showed large fluctuations with RMSFs >2.50 Å. The calculated results support that the linker region has little influence on the affinity of ligand binding to TOP1 (Stewart et al., 1997).

During MD simulation, the flexibility and extensibility of a ligand binding inside the pocket can be described by the radius of gyration. In Supplementary Figures S5, S6, we plotted the radius of gyration of the protein, DNA, and ligand for the TOP1/EVO-1 and TOP1/EVO-6 systems, respectively. It is evident that these two systems reached a stable state in the 100 ns MD simulations. The surfaces of the protein, DNA, and ligands of the TOP1/EVO-1 and TOP1/EVO-6 systems are shown in Supplementary Figures S7, S8. The surface of the protein did not show any significant fluctuations in the binding pocket throughout MD simulation. The small ligand surface area indicated that it was completely enclosed in the cavity of the active site. A clamp around the DNA was formed between the C-terminal domain and the core domain (Redinbo et al., 1998; Chillemi et al., 2001). Thus, most of the solvent-accessible surfaces of the ligands were covered by TOP1.

For each TOP1–DNA–ligand system, snapshots were taken from MD trajectories at intervals of 10 ns. They were aligned as shown in Supplementary Figures S9, S10. The conformation of the ligand underwent minor changes, while the conformation of the protein showed little change except in the linker region. This phenomenon can be explained by the significant RMSF fluctuations of the linker region. A stable ternary complex was formed between the ligand and the covalent complex of TOP1–DNA. A close examination of the ternary complex revealed that there was sufficient space within the binding pocket to allow the introduction of other groups.

The binding free energy can be used to quantitatively predict the binding affinity between the ligand and receptor. The binding free energies of the two complex systems were calculated by MM/GBSA and are summarized in Supplementary Tables S1, S2. The calculated binding free energy was −38.33 kcal·mol⁻¹ for the TOP1/EVO-1 system and −33.25 kcal·mol⁻¹ for the TOP1/EVO-6 system. This can explain the activity results of the MTT assay at the molecular level, which showed that EVO-1 had a stronger anti-GC effect than EVO-6. The results of free energy calculations agree with those of molecular docking. The binding free energy results also indicated that EVO-1 is a promising candidate as a potential chemotherapeutic agent against GC. This conclusion is also consistent with the molecular docking results.

Meanwhile, the van der Waals (vdW) interactions for the TOP1/EVO-1 and TOP1/EVO-6 systems were −48.07 and −53.01 kcal·mol⁻¹, respectively. This suggests that the vdW interactions are favorable for the formation of the ternary TOP1–DNA–ligand complex. The contributions of the electrostatic interactions to the binding free energies of EVO-1 and EVO-6 were −9.96 and 2.62 kcal·mol⁻¹, respectively. The results indicate that, for the two binding models, the vdW interactions contributed more to the binding free energy than the electrostatic interactions. This suggests that vdW interactions play a crucial role in ligand binding with the covalent complex TOP1–DNA.

It is known that the polar (E_ele + E_GB) and nonpolar (E_vdw + E_surf) terms affect the binding of inhibitors to their targets. The polar contributions for TOP1/EVO-1 and TOP1/EVO-6 were 11.88 and 22.78 kcal·mol⁻¹, respectively. The positive polar contribution values indicated that polar interactions were not favorable for binding between the ligand and receptor. Therefore, for the TOP1/EVO-1 and TOP1/EVO-6 systems, polar interactions were not favorable for ligand–receptor binding. This means that the lower the energy of the polar term, the more favorable the binding between the ligand and receptor. This can be used to explain why EVO-1 has stronger binding ability than EVO-6. Therefore, EVO-1 exhibited stronger anti-GC activity than EVO-6.

For TOP1/EVO-1 and TOP1/EVO-6, the total nonpolar values were −51.14 and −56.94 kcal·mol⁻¹, respectively. This indicates that the nonpolar terms contribute significantly to the formation of ternary TOP1–DNA–ligand complexes. Thus, it can be inferred that the nonpolar term plays a dominant role in ligand binding. Hydrophobic interactions are the main determinant factors involved in the binding of EVO-1 and EVO-6 to TOP1–DNA.

The main idea of free energy decomposition is to decompose the free energy of the receptor–ligand complex to each amino acid residue, thereby obtaining the energy contribution of each amino acid to the binding. The MM/GBSA decomposition protocol was employed to calculate the energy contribution of each amino acid in the binding complex (Supplementary Tables S3, S4). 5′-thio-2′-deoxyguanosine phosphonic acid (TGP) made a significant contribution to the complex, with energy values of −8.59 kcal·mol⁻¹ for the EVO-1/TOP1 system and −4.35 kcal·mol⁻¹ for the EVO-6/TOP1 system. This demonstrates that the Π–Π stacking interaction between TGP and the ligand is important for ligand binding.

For EVO-1/TOP1, residue Arg364 contributed −3.43 kcal·mol⁻¹ because of the hydrogen bonding interaction; for EVO-6/TOP1, the hydrogen bonding interaction between Arg364 and EVO-6 contributed −0.84 kcal·mol⁻¹. This indicates that the hydrogen bonding interaction between the ligand and Arg364 plays an important role in the binding models. Additionally, other hydrophobic residues such as DT10 and DA113 had substantial binding free energies. These residues are located at cleaved DNA-binding sites. Thus, DT10 and DA113 are key DNA residues for ligand binding with TOP1–DNA. These results explain the stronger anti-GC effect of EVO-1 than that of EVO-6.