Evaluation of New Coumarins for Anti-Cancer Activity in HL-60 Cell Line Supported by Molecular Docking, MD Simulation, and Binding Free Energy Calculations

Abstract

Introduction

Leukemia, a prevalent malignancy characterized by chromosomal aberrations, was targeted in this study to assess the anti-cancer potential of naturally isolated novel coumarins.

Methods

Novel coumarins (isolated from the natural source Sterculia colorata) were assessed against HL-60 leukemia cells using in vitro and in silico approaches.

Results

The MTT assay demonstrated significant cytotoxicity of the compounds against HL-60 cells. Molecular docking studies revealed strong binding interactions with CDK-2 and BCL-2 proteins, supporting the in vitro results. Compound 2 exhibited the highest binding energy with CDK-2. Molecular dynamics simulations, binding free energy calculations, principal component analysis (PCA), and free energy landscape analyses were performed to assess the stability of the Compound 2-CDK-2 complex. The results indicated that the complex remained stable during the MD simulation with favorable binding free energies. ADMET predictions confirmed favorable pharmacokinetic profiles and safety from hERG blockade.

Conclusion

These findings highlight the compounds’ promising pharmacological activity and potential for further drug development, though additional studies are needed to explore their clinical applications.

Graphical Abstract

Introduction

In developed nations, cancer is the leading cause of illness and mortality, while in underdeveloped countries, it ranks second in terms of cause of death.^1,2 The burden of cancer is increasing in developing countries due to factors such as population aging, urbanization, and the adoption of cancer-risk behaviors, including smoking, inactivity, and “Westernized” diets.³ Chemotherapy is the main therapeutic approach for leukemia, similar to all other cancers, and it involves using cytotoxic chemicals to eliminate aberrant leukemia cells.⁴ Studies conducted across multiple regions have consistently demonstrated the efficacy of chemotherapy in treating leukemia. For instance, a comprehensive analysis from high-income regions revealed that rigorous chemotherapy combined with supportive care can elevate the survival rates of leukemia patients to over 70%.^5,6 Various types of chemotherapy medications are used to treat cancers,^7–11 including both acute and chronic forms of leukemia. Docking is a core Computer-Aided Drug Design and Discovery (CADDD) component. The static or semi-flexible treatment of ligands and targets is a shortcoming of traditional docking approaches. Over the past 10 years, advances in proteomics, genomics, and computation have also led to the development of various docking techniques that account for the flexibility of ligands and proteins and their numerous binding conformations. Protein binding interactions with the ligand are shown more logically and precisely when binding poses are predicted due to receptor flexibility. Protein flexibility has been added through the use of dynamic docking techniques or the creation of protein ensembles. Dynamic docking analyses drug-receptor binding and interaction from a mechanistic and energetic perspective, considering solvation and entropic effects. Even though dynamic docking is computationally expensive in fast-moving drug development programs, it is being used increasingly to screen large chemical libraries for potential drugs.¹² The broad class of phenolic compounds known as coumarins comprises fused pyrone and benzene rings and is found in plants. Secondary metabolites from fungi, bacteria, and plants have been known to contain more than 1500 coumarins.¹³ Due to their varied pharmacological actions, coumarins have recently received much interest. Many coumarins and their derivatives have anticoagulant, anti-tumor, anti-viral, anti-inflammatory, antioxidant, and anti-microbial potential, and enzyme inhibitory activity.¹⁴ The many substituents in the coumarin nucleus significantly impact biological activity.¹⁵ The development of coumarin scaffolds for drug discovery as new anti-cancer medicines is progressing. Several researchers have thoroughly explored the anti-cancer activity of various natural and synthetic coumarin derivatives due to their stereochemistry, non-toxicity, and biological features.¹⁶ Coumarin-based anti-cancer compounds have been categorized into several classes, including alkylating agents, apoptosis inducers, angiogenesis inhibitors, topoisomerase blockers, hormone inhibitors, telomerase inhibitors, human carbonic anhydrase blockers, anti-mitotic drugs, and miscellaneous agents.¹⁷ Research indicates that the coumarin dicoumarol and its analogues, 3,3′-methylene bis (4-hydroxycoumarin) and others, may prevent cell division by interfering with spindle microtubule dynamics. Many of these agents have demonstrated notable antineoplastic efficacy and selectivity, with reduced toxicity and adverse effects. Numerous coumarin medication candidates displaying strong effectiveness and an appropriate pharmacological profile are currently undergoing clinical testing.¹⁸ The two main challenges coumarins are facing include translating existing knowledge into new applications potentially lead compounds, and the repurposing of established compounds for cancer treatment. Therefore, considering the multifaceted nature of leukemia and the critical need for novel therapeutic strategies, this study was designed to evaluate the anti-cancer potential of selected coumarin derivatives comprehensively. Our approach integrates both in vitro and in silico methodologies to provide a holistic assessment of these compounds’ efficacy and mechanism of action (Scheme 1). Specifically, we utilized the human promyelocytic leukemia (HL-60) cell line to conduct cytotoxicity assays to elucidate the compounds’ effects on cell viability and proliferation. Complementing these experimental findings, we employed advanced computational techniques, including molecular docking, molecular dynamics simulations, and binding free energy calculations, to explore the interactions of these coumarin derivatives with key oncogenic proteins, CDK-2 and BCL-2. Furthermore, we assessed the compounds’ pharmacokinetic profiles and ADMET properties to ensure their therapeutic potential and safety. This integrated approach aims to provide a robust foundation for developing these coumarin derivatives as novel anti-cancer agents, bridging the gap between in vitro efficacy and in silico predictive modeling.

Scheme 1.

Workflow for the anti-cancer evaluation of coumarin derivatives: In vitro to in silico analysis.

Materials and Methods

Materials

All chemicals and reagents used in this study were of high purity and analytical grade. The following reagents were obtained from Sigma Chemicals Co., St. Louis, Missouri, USA: 3–4,5−dimethylthiazol−2−yl −2,5-diphenyltetrazolium bromide (MTT, analytical grade), dimethyl sulfoxide (DMSO, analytical grade), RPMI-1640 medium (cell culture grade), fetal bovine serum (FBS, 10% (v/v), sterile), L-glutamine (2 mM, analytical grade), streptomycin (0.1 mg/mL, sterile), and penicillin (100 U/mL, sterile). The docking studies (in silico) were conducted at Abasyn University, Peshawar. However, biological studies (in vitro) were performed at HEJ, ICCBS, University of Karachi, Pakistan. Coumarin derivatives from the natural source Sterculia colorata were isolated, and their structures were established using spectroscopic techniques (NMR and ESI-MS) as shown in Table 1.¹⁹

Table 1.

Structure, IUPAC Name, and SMILES Notation of Coumarin Derivatives

Compound	Ligand
1	IUPAC	4-(3,5-dihydroxyphenyl)-3,5,7-trihydroxy-2H-chromen-2-one
	SMILES notation	O=c3oc1cc(O)cc(O)c1c(c2cc(O)cc(O)c2) c3O
2	IUPAC	4-(3,4-dihydroxyphenyl)-5,7-dihydroxy-3-(((2R,3S,4R,5R,6S)-3,4,5-trihydroxy-6-(hydroxymethyl) tetrahydro-2H-pyran-2-yl) oxy)-2H-chromen-2-one
	SMILES notation	O=c3oc1cc(O)cc(O)c1c(c2ccc(O)c(O)c2) c3OC4OC(CO)C(O)C(O)C4O

Methods

Cell Culture

Cell viability was evaluated using the human promyelocytic leukemia (HL-60) cell line. The ATCC provided the human promyelocytic leukemia cells (HL-60) (ATCC^® CCL240TM). For HL-60 cells, standard culture conditions include 10% (v/v) FBS, RPMI 1640 media containing 2 mM L-glutamine, 0.1 mg/mL streptomycin, and 100 U/mL penicillin. Cells were cultured in flasks with a surface area of 75 cm² at 37 °C in a moist atmosphere with 5% CO₂. Continuous incubation was used to achieve confluent growth, and each experiment was started with freshly sub-cultured cells to make sure they were growing exponentially.²⁰

MTT Cytotoxicity Assay

Cell viability was measured using a colorimetric technique. It assessed cellular proliferation and analyzed cytotoxic and cytostatic substances, including anti-cancer medications and other pharmacological substances. According to the MTT experiment, live cells’ mitochondrial NAD(P) H-dependent cellular oxidoreductase enzyme transforms water-soluble yellow MTT into insoluble purple formazan crystals. When a substance (Formazan) is dissolved in DMSO, it emits a purplish color, and the absorption was measured at 540 nm. The intensity of the purple color, which represents cell viability, was inversely correlated with the number of cells.²¹

Designing and Optimizing Compound Structures

The two chosen compounds were created as 2D structures in ChemDraw Professional version 15.0 to initiate molecular docking studies.²² The Discovery Studio visualizer was used to transform the ligands’ 2D structures into 3D structures and minimize their energy.

Selection and Preparation of Protein Targets

The target protein’s (PDB ID: 4LVt and 2A4L) human B-cell lymphoma-2 (BCL-2) and Cyclin-dependent kinase-2 (CDK-2) proteins structure were extracted as a PDB files from the RCSB protein data bank for a molecular docking study (https://www.rcsb.org/).²³ With the help of Discovery Studio Visualizer, the protein was created by eliminating water molecules, co-crystallizing the ligand, assigning charges, adding polar hydrogen atoms, and adding any residue that might have been missing. Each targeted protein’s active site was identified using co-crystal ligands on all residues within ten of them.²⁴

Molecular Docking Studies

Molecular docking of coumarin derivatives was conducted against potential targets expressed in the leukemic cell line. It was utilized to predict the binding affinity and interactions between the tested compounds and the targeted proteins (ie, CDK-2, BCL-2). Van der Waals forces, hydrophobic contacts, and hydrogen bonds represent some of these interactions. Hydrogen bonding is crucial for the formation of the ligand-protein complex. The previously reported protocol was adhered to for executing the molecular docking studies. After preparing the protein and ligand structures, the molecular docking study was executed using AutoDock Vina 4.2.6. The PDBQT format was employed to convert the ligand and protein files, which are now stored in the same folder. The dimensions of the grid box were determined as follows: x = 69.62, y = 77.17, and z = 59.52. The coordinates of the grid center were established as x = 2.2958, y = −1.087, and z = 28.49. The ligand had sufficient time to rotate and translate, as the grid boxes were positioned at the enzyme’s binding site. For each ligand, ten poses were generated using AutoDock’s default settings, and the pose with the lowest docking score was selected for further analysis. The binding interactions of the best docked poses were illustrated using the Protein Plus web server and the Discovery Studio client v16.1.0. Ten poses were generated for each ligand using AutoDock’s default settings, and the lowest docking score pose was chosen. The binding interactions of the optimal docked poses were displayed using the Protein Plus web server and the Discovery Studio client v24.1.0.ult settings, and the lowest docking score pose was chosen. The binding interactions of the optimal docked poses were displayed using the Protein Plus web tool server and the Discovery Studio client v24.1.0.

Furthermore, the redocking was performed for the selected target proteins. The native ligands, ABT-199 and roscovitine, for the BCL-2 and CDK-2, were extracted using UCSF-Chimera. The inverse selection method was used for removing protein, and the ligands were saved in the mol2 format. Openbabel was employed for the conversion of ligands from mol2 to PDBQT format. The exact coordinates for the dimension box were used as mentioned above, and redocking was performed. At least ten poses were generated for each complex, and the lowest score for each complex was selected for further analysis.

ADMET Predictions

Physicochemical properties were calculated using SWISS ADMET software. The selected phyto-compounds were subjected to toxicity and ADMET predictions. The features of ADMET (absorption, distribution, metabolism, excretion, and toxicity) were predicted using the mentioned software.²⁵

Physicochemical Properties, Medicinal Chemistry, and Drug-Likeness

Compounds 1 and 2 were evaluated using SMILES notation for their physicochemical characteristics, medicinal chemistry properties, and drug-likeness.²⁶ The physicochemical properties were calculated using the following software tools: SWISS ADME, ADMET Lab version 2.0, and pkCSM. Additionally, the selected phyto-compounds underwent toxicity and ADMET predictions.

Pharmacokinetic Profile

The pharmacokinetic profile details were computed using the SMILES notation. Using ADMETlab version 2.0, data on absorption, distribution, metabolism, excretion, and toxicity were obtained along with sub-descriptors.²⁶ The Boiled-Egg graph, which provides information on blood-brain barrier penetration, human intestinal absorption, and whether or not it is a substrate for p-glycoprotein, was obtained using the Swiss ADME database. The “bioavailability radar chart” was anticipated to link physicochemical characteristics to bioavailability.

The Human Ether-a-Go-Go Related Gene (hERG) and Other Predictions of Toxicities

High drug attrition rates can be attributed to drug toxicity, which is a limiting factor. To build a safety profile for both of the chosen drugs, several characteristics related to drug responses were identified. Arrhythmias of the heart result from agonism of the hERG K+ channels.²⁷ Therefore, it is critical to identify this side effect as early in the drug discovery process as possible.

Molecular Dynamics (MD) Simulation

After molecular docking analysis, the MD simulation analysis was performed to assess the stability of the ligand-protein complex using GROMACS software, version 24.1.²⁸ The ligand and protein topologies were initially generated using the SwissParam online server. The CHARMM27 force field was used during the MD simulation, and the solvation TIP3 box was used. As reported previously, the Steepest Descent Model minimized the system’s energy. The system was equilibrated in two phases using the NVT and NPT ensembles for 1 ns. After the equilibration, the final MD run was performed for 50 ns. The various parameters were evaluated using RMSD (Root Mean Square Deviation), RMSF (Root Mean Square Fluctuations), RoG (Radius of Gyration), SASA (Solvent Accessible Surface Area), and H-bonds, as reported previously.

MM-PBSA, MM-GBSA, and Per-Residue Decomposition

After the MD simulation analysis using GROMACS software, the binding free energy was calculated using Molecular Mechanics Poisson-Boltzmann Surface Area (MM-PBSA) and Molecular Mechanics Generalized Born Surface Area (MM-GBSA).²⁹ The MM-PBSA and MM-GBSA provide insight into the molecular mechanics and strength of the ligand-protein complex. The various parameters were investigated, including the electrostatic energy, polar solvation energy, non-polar solvation energy, Gas solvation energy, van der Waals energy, and total energy using the mm_pbsa.py software. After the MD simulation, 500 total trajectory frames were processed. Furthermore, the per-residue decomposition analysis was performed to assess the interaction of the ligand with the different residues and their binding free energies, as reported.

Principal Component Analysis (PCA) and Free Energy Landscape (FEL)

The PCA analysis and FEL were performed after the MD simulation to assess the dynamic changes of the ligand-protein complex using the covariance matrix from which eigenvectors and eigenvalues were derived.³⁰ The initial two components, ie, PC1 and PC2, were employed for the ligand-protein complex motions, effectively decreasing the MD trajectories’ dimensionality. The 3D graph was generated for the PCA analysis to assess the clustering conformational changes of the principal component. The 3D FEL was created to measure the free energy as a function of the initial two PCA components, providing insight into the complex conformational stability. This analysis reveals the stable and unstable states of the complex corresponding to the regions of low energy and high energy, respectively. After the MD analysis using GROAMCS 24.1 software, the FEL and PCA analysis was performed using the gmx anaeig and gmx covar tools by employing the in-house scripts in GROAMCS software. As reported previously, the FEL and PCA were visualized using the Python environment and Python tool, ie, Matplotlib.

Data Analysis

An assay was carried out in triplicate, and the results obtained will be reported as the mean of the IC₅₀ and its standard deviation. GraphPad Prism (version 9.0, San Diego, CA) will be used for this analysis. Autodock Vina was used to predict the binding affinity and the binding interaction between the tested compounds and targeted proteins (ie, CDK-2, BCL-2).

Results and Discussion

Results

Cell Viability Assay

The HL-60 cell line was treated to study the anti-cancer properties of Compound 1–2. The 3- [4, 5-dimethylthiazol-2-yl] 2,5 2,5-diphenyltetrazolium bromide (MTT) test was used to assess the cell viability of the treated cells. The cell line responded differently to each treatment. There was a notable and more pronounced cytotoxic effect. The influence of DMSO was ruled out, as the concentration of DMSO employed in the test samples had no significant impact on the outcomes (Table 2). HL-60 cells exhibited a significant cytopathic effect across different groups, with doxorubicin serving as the standard reference drug. The results indicated that both compounds markedly decreased the viability of HL-60 cells, as depicted in Figure 1.

Table 2.

Cell Viability Study of HL-60 Cells Using Coumarin Derivatives

S. No.	Treatment	IC50, μg/mL
1.	DMSO (Control)	100 ± 0.42
2.	Compound 1 (10 μM)	35.24 ± 0.82
3.	Compound 2 (10 μM)	34.32 ± 1.04
4.	Doxorubicin (Standard)	19.425 ± 0.90

Note: DMSO as vehicle (negative control).

Figure 1.

Dose-response curves showing the effects of coumarin derivatives on HL-60 cell viability, with doxorubicin used as the reference drug (n = 3, mean ± SD).

Molecular Docking Analysis

The software Auto Dock vina Version 4.2 was used to carry out the molecular docking study. The compounds reported from the Sterculia colrata plant docked against the two-protein target, playing a critical role in cancer, and are the legitimate target for anti-cancer drugs. The various compounds reported from Sterculia colorata are listed in Table 3. The compounds were computationally evaluated against the CDK-2 and BCL-2 proteins. The docking results of the ligand and protein target were expressed as binding energy, ie, kcal/mol. The ligand with maximum binding energy to the protein target was visualized in 3D and 2D views.

Table 3.

The Ligand’s Binding Energy to the Target Protein

S. No.	Sample	CDK-2	BCL-2	RMSD
1.	4-(3,5-dihydroxyphenyl)-3,5,7-trihydroxy-2H-chromen-2-one	−8.4	−6.008	–
2.	4-(3,4-dihydroxyphenyl)-5,7-dihydroxy-3-(((2R,3S, 4R,5R,6S)-3,4,5-trihydroxy-6-(hydroxyllmethyl) tetrahydro-2H-pyran-2-yl) oxy)-2H-chromen-2-one	−5.83	−5.785	–
3.	Roscovitine	−5.375	–	160.515
4.	ABT-199	–	−7.822	18.315

Molecular Docking of Compounds 1 and 2 with the CDK-2

CDK-2 inhibits acute myeloid leukemia cell development by directly targeting and activating PRDX2. The molecular docking analysis showed significant binding affinity and interaction with both compounds, as evident from Table 3. Compounds showed multiple hydrophilic and hydrophobic interactions with the CDK-2, as shown in Figure 2A and B.

Figure 2.

Molecular docking (3D and 2D) of CDK-2 with (A) Compound 1 and (B) Compound 2.

Molecular Docking of Compounds 1 and 2 with the BCL-2

BCL-2 significantly influences intrinsic apoptosis. Resistance to chemotherapy and apoptosis can be prevented in acute myeloid leukemia by overexpressing BCL-2 proteins. The molecular docking analysis showed significant binding affinity and interaction with both Compounds, as evident in Table 3. Compounds 1 and 2 showed multiple hydrophilic and hydrophobic interactions with BCL-2, as shown in Figure 3A and B, respectively.

Figure 3.

Molecular docking (3D and 2D) of the BCL-2 with (A) Compound 1 and (B) Compound 2.

Redocking

However, in comparison with the redocking performed against the native ligands ABT-199 and roscovitine, our compounds showed better results in terms of docking score and binding affinity. RMSD values were also compared for understanding the behavior of the complexes, which exhibited improved conformation in the context of our compounds compared to the native ligands.

Physicochemical Properties

The potential of a molecule to develop into a drug is determined by its physicochemical profile. Compound 1 complies with the widely accepted Lipinski’s Rule of Five, while compound 2 is non-compliant. This is critical for a drug’s oral bioavailability. Compound 1–2 exhibits the parameters defined by Lipinski, ie, molecular weight, Topological Polar Surface Area, Hydrogen bond donors, and several hydrogen bond acceptors (nHA). Compound-1 possesses a single rotatable bond, while compound-2 possesses 4. The molar refractivity of compounds 1–2 is 78 and 110, respectively, as shown in Table 4. The radar chart, which represents a ligand’s drugability, comprises thirteen physicochemical parameters. Some characteristics are molecular weight, H-bond donors, acceptors, heteroatoms, maximal rings, Formal charge, number of rings, rotatable bonds, stiff bonds, and Topological Polar Surface Area. Table 4 provides a summary of the physicochemical characteristics of compounds 1 and 2. As seen in Figure 4A and B, the majority of its characteristics were within the permitted limits for drugs.

Table 4.

Physicochemical Properties of Compound 1 and 2

S. No.	Physicochemical Properties	Value
		Compound 1	Compound 2
1.	Molecular formula	C15H10O7	C21H20O12
2.	Molecular weight	302.24	464.38
3.	No. of hydrogen bond acceptors	7	12
4.	No. of hydrogen bond donors	5	8
5.	Topological polar surface area	131.36	210.51
6.	Molar refractivity	78.03	110.16
7.	Number of rotatable bonds	1	4
8.	Aqueous solubility	−2.9	−2.909
9.	Octanol-water partition coefficient	1.848	−0.534
10.	Distribution coefficient	2.072	0.156

Figure 4.

Physicochemical properties radar chart of: (A) Compound 1 and (B) Compound 2.

Medicinal Chemistry and Drug-Likeness

The quantitative estimate of drug similarity (QED score) calculates a drug-likeness score based on eight physicochemical factors. Maximum absorption and bioavailability, reduced dosage, fewer interactions, and fewer p-glycoprotein interactions correlate with higher QED scores. According to the QED score, compounds 1 and 2 are complex molecules that could be utilized in medication development. Compounds 1 and 2 have a proportion of sp3 hybridized carbon atoms (Fsp3) of 0.0 and 0.286, respectively. Compound 1 has a medicinal chemistry evolution (MCE) score of 19, less than the optimal limit of 45 based on various medicinal chemistry parameters. In contrast, compound 2 has an MCE-18 score of 91, which exceeds the ideal limit (≥ 45), as determined by several medicinal chemistry factors (Table 5). The characteristics of compounds 1 and 2 adhere to the Golden Triangle, Pfizer, GSK, Lipinski RO5, and other medication similarity standards. Drug permeability or absorption may be hindered by violating even one of Lipinski’s guidelines. However, only one of these attributes may change without causing harm. While compound 2 is non-compliant, compound 1 complies with the widely recognized Lipinski’s Rule of Five. Compounds 1 and 2 also adhere to the Pfizer rule. Pharmacokinetic profiles of therapeutic molecules that follow GSK guidelines for drug similarity are typically acceptable Compound 1 complies with GSK’s medication similarity guidelines, while compound 2 disregards them. The ADMET profiles of drug compounds following the “Golden Triangle” are favorable. Both molecules meet the Golden Triangle’s requirements: two hundred ≤ MW > 50 and two logD ≤ 5. Compound 1 has one alert for chelator rules, three ALARM NMR alerts, zero for PAINS, and one for BMS. In contrast, compound 2 has one chelator rule alert, four ALARM NMR alerts, zero BMS alerts, and one PAINS alert. Reactive chemicals, frequent hitters, and alpha-screen artifacts may all comprise a component of this molecule. Both compounds adhere to the chelator rule. Based on ALARM NMR alerts, adverse effects may be associated with a thiol-reactive moiety.

Table 5.

Drug Likeness/Medicinal Chemistry of Compound 1 and 2

S. No.	Drug Likeness / Medicinal Chemistry	Value
		Compound 1	Compound 2
1.	QED	0.434	0.181
2.	Synthetic accessibility score	4.113	3.24
3.	Fraction of sp3-hybridized carbon	0.0	0.286
4.	Medicinal chemistry evolution	19.0	91.0
5.	Natural product likeness	1.189	1.845
6.	Lipinski Rule	Accepted	Rejected
7.	Golden Triangle	Accepted	Accepted
8.	GSK Rule	Accepted	Rejected
9.	Pfizer Rule	Accepted	Accepted
10.	PAINS	0 Alerts	1 Alert
11.	ALARM NMR	3 Alerts	4 Alerts
12.	BMS Rule	1 Alert	0 Alerts
13.	Chelator Rule	1 Alert	1 Alert

Pharmacokinetic Profile

Bioavailability

A radar map based on six criteria (size, flexibility, lipophilicity, instauration, polarity, and insolubility) can be used to predict the oral bioavailability of a drug candidate (Figure 5). Except for polarity in compound 2 (Figure 5B) and instauration in compound 1 (Figure 5A), all characteristics are within the ranges shown on the radar chart. The pink zone designates an acceptable range for the aforementioned physicochemical characteristics that will have oral bioavailability.

Figure 5.

Oral bioavailability radar of (A) compound 1 and (B) compound 2. The red line depicts the characteristics of compounds 1–2, and the pink zone defines the advantageous area for pharmaceuticals that are orally accessible.

Abbreviations: INSOLU, Insolubility; LIPO, Lipophilicity; INSTU, instauration; POLAR, polarity; FLEX, flexibility.

Boiled Egg Analysis

Compound 1 demonstrated intestine route absorption, BBB impairment, and the inability to function as a p-glycoprotein substrate (Figure 6A), whereas compound 2 did not demonstrate any of these properties (Figure 6B). The grey part of the graph indicates routes other than oral, the yellowish space indicates CNS penetration, and the white portion defines human intestinal absorption. Before formulation and clinical trials, it is crucial to understand how a drug candidate enters the human gut and how it permeates the central nervous system.

Figure 6.

The boiled egg graph of (A) compound 1, and (B) compound 2 (does not cross the BBB, showing human intestinal absorption and not being a P-gp substrate).

ADMET

Table 6 presents the parameters associated with the distribution. For compounds 1 and 2, the estimated PPB was 95.105% and 88.960%, respectively. A PPB range greater than 90% for a therapeutic medicinal molecule is preferable. Drugs with high PPB have low levels of free drugs in circulation, which reduces their therapeutic impact. Compounds 1 and 2’s unbound fractions (Fu) in plasma were determined to be 5.867% and 11.861%, respectively. For compounds 1 and 2, the Vd was determined to be 0.651 L/kg and 0.798 L/kg, respectively, falling within the optimum range. There is no tendency for either molecule to penetrate the BBB. As shown in Table 6, the metabolic fates of both compounds were assessed with different CYP 450 isoforms, acting as either substrates or inhibitors. The drug excretion profile was also expected to be established by the half-life and clearance of both molecules. Both compounds’ half-lives seemed longer than three hours, indicating a variable dosage schedule. The clearance rate was found to be moderate (5–15 mL/min/kg), with values of 8.815 mL/min/kg and 6.417 mL/min/kg, respectively (Table 6).

Table 6.

Pharmacokinetic Parameters of Compound 1 and 2

Sample	Absorption and Bioavailability										Distribution
	Caco-2		Pgp-Inhibitor		Pgp-Substrate			HIA	F20%	F30%	PPB%	Vd		BBB			Fu %
1	−5.467		0.1		0.1			0.05	0.995	0.999	95.105	0.651		−1.521			5.867%
2	−6.319		0.1		0.7			0.8	0.9	0.9	88.960	0.798		−2.165			11.861
Sample	Metabolism															Excretion
	Inhibition of Cytochrome P450 Isoforms								Substrate for Cytochrome P450 Isoforms
	1A2	2C19		2C9		2D6	3A4		1A2	2C19	2C9		2D6		3A4	t1/2		CL
1	0.9	0.1		0.6		0.7	0.7		0.3	0.1	0.9		0.5		0.1	0.911		8.815
2	0.1	0.1		0.1		0.1	0.1		0.1	0.1	0.6		0.3		0.1	0.798		6.417

hERG and Other Toxicities

Compound 1–2 toxicity was assessed using many criteria, including skin sensitization, hERG blocking, ocular corrosion/irritability, hepatotoxicity, rat oral acute toxicity, drug-induced liver damage, respiratory toxicity, and mutagenicity. Table 7 presents an outline of all these parameters. ADMETlab and pred-hERG were used to calculate the hERG toxicity. They confirmed its safety since, according to calculations with 69.15% and 88.12% confidence, no substances had demonstrated hERG cardiotoxicity. The hERG toxicity diagram is shown in Figure 7, with the adversely contributing molecules to hERG Blockade defined by the red /pinkish zones. Figure 7 shows that functional groups that positively contribute to hERG Blockade are indicated by prominent greenish lines. Compounds 1 (Figure 7A) and 2 (Figure 7B) do not appear to be the cause of human hepatotoxicity, ocular corrosion, AMES mutagenicity, carcinogenicity, rat oral acute toxicity, or respiratory toxicity. But it can also cause skin hypersensitivity, eye irritation, and liver damage. hERG: human ether-a-go-go related gene; FDAMDD: Federal Drug Administration maximum drug dose; DILI: drug-induced liver injury; H-HT: human hepatotoxicity.

Table 7.

Toxicity Profile of Compound 1 and 2

S. No.	Toxicity	Value
		Compound 1	Compound 2
1.	hERG Blockers	0.3	0.1
2.	Rat Oral Acute Toxicity	0.1	0.1
3.	AMES Toxicity	0.1	0.7
4.	DILI	0.9	0.9
5.	H-HT	0.3	0.3
6.	FDAMDD	0.7	0.1
7.	Eye Corrosion	0.1	0.1
8.	Carcinogenicity	0.1	0.1
9.	Eye Irritation	0.9	0.1
10.	Skin Sensitization	0.9	0.3
11.	Respiratory Toxicity	0.1	0.1

Figure 7.

The hERG blockage probability map for compound 1 (A) and compound 2 (B).

MD Simulation

The MD simulation was performed for 50 ns using GROMACS Software 24.1 to investigate the stability of the ligand-protein complex. Based on the MD simulation analysis, the RMSD value shows that the complex remains stable for 50 ns, and the value of the RMSD remains within the acceptable range, as shown in Figure 8A. Furthermore, the RMSF analysis was performed to determine the fluctuation of the residue for 50 ns. Figure 8B shows that no significant fluctuations have been observed, and the complex remains stable during the simulation. Additionally, the RoG was performed after the MD simulation to assess the compactness of the protein after interaction with the ligand. Figure 8C results revealed that the ligand-protein complex system remains stable and compact for 50 ns. The SASA analysis was performed to determine the exposure of the ligand-protein complex to the solvent system. Figure 8D shows that SASA remains stable during the simulation, and no drastic fluctuations have been observed. The number of H-bonds was calculated for 50 ns, and the results showed the complex formed a maximum of 8 H-bonds during the entire simulation, as shown in Figure 8E.

Figure 8.

Molecular dynamic simulation (A) RMSD (B) RMSF (C) RoG (D) SASA (E) Number of hydrogen bonds.

Binding Free Energy Calculations

The binding free energy was calculated using various parameters from the MM-PBSA and MM-GBSA analyses. The MM-PBSA analysis showed that the total energy of the complex remains negative, which indicates a favorable binding energy, and the complex remains thermodynamically stable during the simulation time, as shown in Figure 9A–H. Furthermore, the MM-PBSA analysis confirmed the MM-GBSA results, and the system’s total energy remains negative, which indicates favorable binding free energy and stability of the ligand-protein complex, as evident from Figure 10A–H. The per-residue decomposition analysis showed the contribution of each amino acid with the ligands and the corresponding binding energy of each residue using MM-PBSA and MM-GBSA analysis, as shown in Figures 11A–C and 12A–C.

Figure 9.

MM-PBSA Analysis of the Ligand–Protein Complex. Energetic components System −1 / Normal / PB / Delta, including (A) EEL, (B) ENPOLAR. (C) EPB, (D) GGAS, (E) GSOLV, (F) VDWAALS, (G) TOTAL, (H) System −1 / Normal / PB / Delta.

Figure 10.

MM-PBSA Analysis of the Ligand–Protein Complex. Energetic components System-1 / Normal / GB / Delta, including (A) EEL, (B) ENPOLAR. (C) EPB, (D) GGAS, (E) GSOLV, (F) VDWAALS, (G) TOTAL, (H) System −1 / Normal / GB / Delta.

Figure 11.

Per-residue decomposition analysis using MM-PBSA. Energetic components [per-residue] System −1 / Normal / PB / Delta including (A–C) TDC.

Figure 12.

Per-residue decomposition analysis using MM-GBSA. Energetic components [per-residue] System −1 / Normal / GB / Delta including (A–C) TDC.

PCA Analysis and Gibbs Free Energy Calculations

After the MD simulation analysis, the PCA and Gibbs free energy calculations were commenced to investigate the behavior of the ligand-protein system and its interactions. The 2D graph was generated, representing the complex system’s FEL using the initial two PCA components. The red and blue color gradients represented the Gibbs free energy. The red colors represent the higher energy regions, while the areas of low energy are represented by the blue color, as shown in Figure 13A. Based on the color representation, the ligand-protein complex with a lower energy basin is shown in blue, representing the complex’s transition into thermodynamically stable regions and favorable interactions. Similarly, a 3D scatter plot was generated for the initial three PCA components, while the points within the scatter plot show the conformational state diversity during MS simulation. The clusters with the 3D graph show the stability of the conformations, while the dispersed point within the graph shows the transition between the different states, as shown in Figure 13B. Furthermore, Figure 13C shows the FEL utilizing the initial two components of the PCA. The regions of high and low energy are demonstrated by the valleys and peaks, which further provide the stability of the system. The smooth surface represents the region of a low-energy basin and stable conformation, while the peak represents the region of a less favorable and unstable conformation (Figure 13C).

Figure 13.

PCA analysis and Gibbs free energy calculations, (A) Free energy Landscape, (B) 3-D PCA Plot, (C) 3-D Gibbs free Energy Landscape.

Discussion

MTT assay has been tested for its potential to serve as a prediction tool for drug selection in leukemia cases since this assay approach is more straightforward to implement in cell cultures grown in suspension. The MTT test is most frequently used to investigate treatment with cytotoxic drugs. The MTT assay demonstrates well both in vitro and in vivo testing, and it can be performed rapidly and reliably.³¹ However, it’s essential to consider a few disadvantages of the MTT test. It is plausible that a decrease in the amounts of D-glucose, NADPH, or NADH in the culture medium might coincide with a reduction in the production of MTT-formazan. Early apoptotic stages are characterized by the preservation of mitochondria, which may lead to an MTT decrease in these cells. This investigation utilized the MTT test to analyze compounds (1–2). Findings demonstrated that compounds 1 and 2 were more effective against the HL-60 cell line. The cytotoxicity of these two active compounds against normal cells was confirmed by analyzing them at 10µM against HL-60 cell lines.

This study utilized BCL-2 and CDK-2 as target proteins, which were based on their established and well-understood roles in cancer pathogenesis. BCL-2 is a crucial anti-apoptotic modulator that plays a key role in eluding programmed cell death. Different studies reported the overexpression of BCL-2 in certain malignancies, including leukemia, breast cancer, hematological cancer, and hepatocellular carcinoma.³² This aberrant expression contributes to chemoresistance and tumor progression. Targeting BCL-2 can sensitize cancer cells to apoptotic signals, thereby enhancing the effectiveness of anti-cancer therapies. At the same time, CDK-2 is a key regulator of the cell cycle, particularly in the G1-S phase transition. Mutation or overexpression of CDK-2 leads to the unrestrained proliferation, a hallmark of cancer. Inhibiting CDK-2 can effectively arrest the cell cycle, preventing further proliferation of cancer cells. Given the critical roles of BCL-2 and CDK-2 in leukemia pathogenesis, targeting these proteins represents a rational strategy for developing new anti-cancer agents. Our molecular docking studies were therefore designed to examine the interactions between the coumarin derivatives and these targets, providing insights into their potential mechanisms of action. The strong binding interactions observed with both BCL-2 and CDK-2, along with the favorable binding free energies obtained from the MD simulations, support the hypothesis that these coumarins may function as dual inhibitors. Molecular docking is a valuable tool in computer-aided drug design, as it predicts the binding affinity and interaction modes between ligands and target proteins. Consistent with this, compounds 1–2 demonstrated meaningful molecular interactions with CDK-2 and BCL-2 in our docking investigation. Because of its therapeutic relevance in modern structure-based drug design, protein–ligand docking remains an essential approach for identifying and optimizing potential drug candidates.³³ In this work, we summarize the latest advances in the three key domains of protein-ligand docking: scoring functions, ligand selection, and protein adaptability. Results of other anti-cancer research studies provide substantial support for the conclusions of our investigation, as per the published literature. There have been numerous reports of substances with anti-cancer properties that come from both natural and synthetic sources. Future breakthroughs could bring about a plethora of intriguing discoveries, eg, tailored compounds with potentially very precise actions against cancer cells could be beneficial for therapeutic success. In vitro synthesis and screening of novel compounds with excellent safety and efficacy, and computational investigations can establish a foundation for identifying anti-cancer drugs. The approach of administering a drug and its formulation as a pharmaceutical drug product is determined by its drug-like attributes. Since orally active agents are the focus of the original rule of five. Approximately ninety percent of orally active medications advancing to Phase II clinical trials comply with RO5.³⁴ Compliance with RO5 is not a guarantee of becoming a successful therapeutic molecule. However, the likelihood of an oral medication development project failing if a molecule does not conform to RO5 is significant. Following RO5, Compound 1 possesses an appropriate quantity of hydrogen bond donors and acceptors. According to Bitew et al, orally active medications typically have fewer rotatable bonds and H-bond acceptors/donors.³⁵ A drug’s octanol/water partition log is called the partition coefficient (LogP). Increased log P values are associated with lipophilicity. LogP is influenced by polarity, molecule size, and hydrogen bonding. This value agrees with the previously mentioned logS. Another comparable metric is logD, which is LogP computed at a physiological pH of 7.4.

Compound 2 (210.51 Å2) has a TPSA outside the optimum range, while Compound 1 (131.36 Å2) is within. Molecules with a TPSA greater than 140 Å2 absorb relatively little of the drug portion. According to Clark (1999), molecules with a TPSA of 60 Å2 absorb more than 90% of the GIT.³⁶ Aqueous solubility, oral bioavailability, lipophilicity, plasma albumin binding, hERG inhibition, and CYP450 inhibition are associated with aromatic and hetero-aromatic moiety components. The likelihood of becoming a medicine in both compounds increases with fewer aromatic rings. The possibility of pharmacological knockout is elevated when a molecule has more than three aromatic rings. More aromatic rings reduce a molecule’s water solubility even if it is within the permissible lipophilicity range.³⁷

The QED is influenced by several factors, including molecular weight, number of aromatic rings, rotatable bonds, H-bond donors and acceptors, structural alarms (alarms), and octanol-water partition coefficient (LogP). Unwanted chemical characteristics are known as ALERTS and can include perceived toxicity, chemical reactivity, etc. High QED scores indicate that a molecule is more bioavailable and well-absorbed, has a low dose of administration, little p-glycoprotein-mediated drug interactions, and a lower risk of food-drug interactions. Nonetheless, an astounding resemblance exists between medications with low and high QED scores in terms of their equivalency in terms of clearance, frequency of dose, elimination half-life, degree of gut-wall metabolism, first-pass impact, plasma unbound percentage, and distribution volume.³⁸ Both molecules stayed on the border of being attractive or unattractive. Since it complies with the vast majority of other drug-like attributes, it might not significantly affect the biopharmaceutical properties of compounds 1 and 2.

Compounds 1 and 2 have Fsp3 values of 0.286 and zero, respectively. The percentage of C-atoms in a molecule is hybridized to a sp3 state out of all C-atoms. This refers to a molecule’s carbon saturation and describes how intricate its spatial structure is.³⁹ The value of Fsp3 is higher in natural products than in synthetic ones. They are, therefore, a rich source of drugs.⁴⁰ Nevertheless, compound 1 has no sp3 content, which is a different discovery. This parameter was utilized by Yan and Gesteiger (2003) to estimate solubility.⁴¹ According to Kombo et al, the fsp3 value for medicine is > 0.42, nearly equivalent to 84% of marketed drugs. Higher Fsp3 values, however, may not support drug-likeness and may instead make synthesis more challenging.^42,43 MCE-18 is a more useful metric to help in the selection of compounds, profiling HTS-focused libraries, and the design of possible drugs.⁴⁴ Compound 1 complies with the golden triangle, Pfizer, GSK, and Lipinski rules, among other important drug similarity descriptors. Compound 1 adheres to the universal drug discovery guidelines (MW ≤ 500, logP < 5, Hdon ≤ 5, and Hacc ≤ 10) that were previously covered.^45–47 This parameter provides an appropriate pharmacokinetic profile and can be used to produce metabolically stable and permeable drug candidates. According to Johnson, Dress, and Edwards (2009), Compounds 1 and 2 subscribe to the golden triangle rule.⁴⁸ PAIN alerts for compounds 1 and 2 are zero and one, respectively. PAINS ALERTS are linked with non-covalent interactions and protein reactivity, though their exact mechanism is unknown.⁴⁹ ALARM NMR is another screening factor that helps to select compounds for lead optimization.^50–52 Studies on drug absorption usually use Caco-2 cell permeability.^53–56 The predicted Caco-2 cell permeability is lower for compounds 1 and 2. However, another permeability screening model called MDCK permeability indicated moderate permeability.⁵⁷ The boiled-egg plot revealed that both compounds prefer the intestinal route of absorption, do not cross the blood-brain barrier, and are not p-glycoprotein substrates. Furthermore, safety from p-glycoprotein-related drug interactions has been demonstrated.⁵⁸ Drug interactions typically involve a substrate or p-glycoprotein inhibitor.^59,60 Compound 1 is a non-P-gp substrate and is more likely to be free of drug interaction and resistance mechanisms than compound 2. Compared to compound 2, which has a lower unbound fraction in the plasma, compound 1 has a higher plasma protein binding. The drug fraction in unbound plasma is cleared quickly and may have a shorter half-life.

The lack of ability to cross the BBB, which may cause CNS adverse effects, is a positive point. Compound 1 was shown to be a substrate of CYP2C9, although it also showed inhibition of CYP1A2, CYP2C9, CYP2D6, and CYP3A4. Likewise, compound 2 was found to be a CYP2C9-specific substrate. From the viewpoint of metabolism, substances that promote or inhibit CYP 450 isoforms can also impact the plasma concentration of a concurrently administered drug that functions as an enzyme’s substrate. The drug’s pharmacokinetic profile disturbance results in a drug interaction.^61,62 Since both drugs appear to have a moderate clearance rate and half-life, a flexible dosage regimen might be possible. Both compounds were determined to be safe regarding hERG blocking, acute toxicity in rats, hepatotoxicity, carcinogenicity, and AMES mutagenicity during the toxicity evaluation process. During drug discovery, one of the side effects that can occur is cardiac arrhythmia. This action results from suppressing the cardiac K+ channel gene linked to the human ether-à-go-go. Regulatory bodies such as the FDA and the European Medicines Agency advise evaluating hERG toxicity at the preclinical stage.²⁷ It looked like both compounds were non-hERG blockers. Hepatotoxicity is another obstacle that can lead to clinical trial failure or the withdrawal of an approved medication from distribution. Since over 600 medications have hepatotoxic side effects, liver toxicity accounts for ≥ 50% of reports of abrupt hepatic failure.⁶³ There are no corrosive adverse effects for compounds 1 or 2. Clinical trial failure during the early stages can also be attributed to drug-induced liver impairment. Therefore, to minimize drug knockout, it must also be evaluated in the early stage.^64,65 The potential of mutagenic substances must be assessed, which requires considering the AMES toxicity. Compound 1 seems to be safer from these kinds of toxicities; however, compound 2 displayed AMES toxicity.⁶⁶ Regarding the two selected molecules, no respiratory toxicity concern has been identified.

The MD simulation analysis was performed for the top hit, ie, Compound 2, with the CDK-2 protein. The MD simulation analysis showed that the Compound 2 CDK-2 complex remains stable during the MD simulation analysis. Furthermore, the binding free energy calculations were performed using the MM-PBSA and MM-GBSA analyses.²⁹ The MM-PBSA and MM-GBSA analyses showed negative and favorable binding energies. The Per-residue analysis was performed to assess the binding energy of the individual with Compound 2. Additionally, the PCA and Gibbs free energy landscape analyses were performed to determine the complex stability during the MD simulation analysis.

MM-PBSA per-residue decomposition showed key residues involved in the stabilization of the BCL-2 complex (Figure 12). Stable binding was noted throughout the simulation in terms of total decomposition (Figure 12A). Average energy analysis unveiled the involvement of GLU81, ASP86, LYS89, GLN131, and ASP145 in achieving the most favorable ΔG, signifying the strong hydrogen bond and electrostatic interactions inside the BH3 binding groove. The hydrophobic residues like ILE10, VAL18, and LEU134 moderate interaction via van der Waals forces. Figure 12C further confirms the interaction of the mentioned residues across the frames. These results are consistent with already reported BCL-2 inhibitors.⁶⁷ However, the CDK-2 complex showed consistent negative total decomposition energy throughout the trajectory (Figure 13A). Among the average residues that contributed are ASP86, LYS89, GLN131, and ASP145 (Figure 13B), which have a significant role in the stabilization and exhibit strong hydrogen and electrostatic interactions. These residues resemble the hinge and catalytic region of CDK2. Persistent low-energy interactions were observed with ASP86 and LYS89, indicating the most reliable energetic favorability.

While the current study provides valuable insights into the cytotoxic potential of the coumarin derivatives against HL-60 leukemia cells using the MTT assay, further research is warranted to fully elucidate the mechanisms underlying the cytotoxic effects of the coumarin derivatives. Future studies may include employing apoptosis assays, such as annexin/PI staining by flow cytometry, to investigate the induction of apoptosis. Additionally, cytotoxicity assays on normal cell lines could determine the selectivity index, providing insights into the therapeutic potential and safety profile of these compounds. Exploring the binding interactions of the coumarin derivatives with other relevant oncogenic proteins may also identify additional targets. These studies could contribute to the development of novel therapeutic agents for leukemia.

Conclusions

Herbal natural products exhibit distinct biological activities compared to synthetic compounds. Harnessing these unique properties, this study comprehensively evaluated the anti-cancer potential of two coumarin derivatives against HL-60 leukemia cells using both in vitro and in silico approaches. Both compounds demonstrated significant cytotoxicity, with compound 2 showing a marginally lower IC₅₀ value (34.32 ± 1.04), suggesting enhanced potency. The molecular docking studies confirmed strong interactions with CDK-2 and BCL-2 proteins, which were further validated by molecular dynamics simulations showing stable compound-protein complexes. Binding free energy calculations using MM-PBSA and MM-GBSA methods indicated favorable thermodynamic stability. ADMET predictions revealed favorable pharmacokinetic profiles and a low risk of hERG-related cardiotoxicity, supporting the compounds’ safety and drug-like properties. In conclusion, the coumarin derivatives exhibited promising anti-cancer activity, a stable interaction with target proteins, and a favorable safety profile. These findings suggest their potential as novel therapeutic agents for leukemia, warranting further investigation.

Funding Statement

The authors gratefully acknowledge financial support from the Postdoctoral Research Start-Up Fund of Lishui People’s Hospital, Lishui, Zhejiang, China (Grant No. 2024bsh002), and the Joint Fund of Zhejiang Provincial Natural Science Foundation of China (Grant No. LKLY25H180011).

Data Sharing Statement

The supporting data can be obtained from the corresponding authors upon reasonable request.

Author Contributions

All authors made a significant contribution to the work reported, whether in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising, or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.