Discovery of novel and potent dual-targeting AXL/HDAC2 inhibitors for colorectal cancer treatment via structure-based pharmacophore modelling, virtual screening, and molecular docking, molecular dynamics simulation studies, and biological evaluation

Abstract

Colorectal cancer (CRC) is one of the most common cancers worldwide. Nowadays, owing to the complex mechanism of tumorigenesis, simultaneous inhibition of multiple targets is an important anticancer strategy. Recent studies have demonstrated receptor tyrosine kinase AXL (AXL) and histone deacetylase 2 (HDAC2) are closely associated with colorectal cancer. Herein, we identified five hit compounds concurrently targeting AXL and HDAC2 using virtual screening. Inhibitory experiments revealed these hit compounds potently inhibited AXL and HDAC2 in the nanomolar range. Among them, Hit-3 showed the strongest inhibitory effects which were better than that of the positive control groups. Additionally, MD assays showed that Hit-3 could bind stably to the AXL and HDAC2 active pockets. Further MTT assays demonstrated that Hit-3 showed potent anti-proliferative activity. Most importantly, Hit-3 exhibited significant in vivo antitumor efficacy in xenograft models. Collectively, this study is the first discovery of dual-targeting AXL/HDAC2 inhibitors for colorectal cancer treatment.

Keywords:

• Colorectal cancer
• receptor tyrosine kinase AXL
• histone deacetylase 2
• dual-targeting AXL/HDAC2 inhibitor
• structure-based pharmacophore modelling

Introduction

Cancer is a multifaceted and complex disease that is the main cause of mortality worldwide and one of the primary public health challenges harming the health of people globally^1,2. Colorectal cancer (CRC) is the third leading cause of cancer-related death, with an estimated 153,020 persons diagnosed and 52,550 dying from the disease by 2023^2,3. The incidence of CRC has grown in recent years, and around 20% of CRC patients have metastasis at diagnosis, with up to 50% of individuals with initial localised disease developing metastasis⁴. Radiotherapy and chemotherapy are still the main strategies for cancer treatment^5,6. Many marketed drugs now only act on a single target, which may result in drug resistance throughout the treatment process due to mutation and plasticity-driven drug targets that do not achieve good therapeutic results^7–9. It has been reported that more than 90% of mortality in cancer patients is attributed to resistance to chemotherapy drugs, making drug resistance a major obstacle to cancer treatment^10,11. Combining various medications may improve efficacy while also reducing the emergence of drug resistance^12,13. However, side effects and limits of combination therapy are common, including drug-drug interactions (DDI), unpredictable pharmacokinetics, and enhanced restrictive toxicity^8,14. As an alternative strategy to combination therapy, dual-targeting drugs can effectively reduce drug interactions and toxicities while also having better pharmacokinetic profiles, which have been effective in an increasing number of clinical studies^15–17. Therefore, dual-targeting drugs have been widely used in more cancer treatments in recent years, providing more therapeutic alternatives for cancer treatment.

In 1991, the receptor tyrosine kinase AXL (AXL) transformed gene was identified from primary myeloid leukaemia cells from two patients with chronic myeloid leukaemia (CML)¹⁸. AXL proteins belong to the TAM receptor family (Tyro 3, AXL, and Mer), which are members of the receptor tyrosine kinase (RTK) family¹⁹. AXL protein overexpression has been associated with a variety of human cancers, such as liver cancer, non-small cell cancer, breast cancer, cervical cancer and so on, which are associated with poor prognosis and treatment resistance^20–23. Meanwhile, aberrant activation of AXL signalling induces an epithelial-mesenchymal transition (EMT) in tumour cells, leading to increased migration and invasiveness^24,25. Furthermore, the vitamin K-dependent protein growth arrest-specific 6 (Gas 6), a soluble glycoprotein, is a common ligand for the TAM receptor, with the highest affinity for AXL^26,27. Gas 6 promotes AXL activation, thereby facilitating cell survival and proliferation²⁸. In addition, the Gas 6/AXL signalling pathway is essential for cell invasion, cell migration and angiogenesis, highlighting AXL^28,29. Aside from the Gas 6/AXL signalling pathway, research has revealed that genetic and pharmacological AXL inhibition can affect downstream signalling pathways such as PI3K-AKT, AK-STAT, etc³⁰, which are crucial for the survival, migration, invasion, and angiogenesis of tumour cells²⁸. Combining AXL inhibitors and PI3K/AKT pathway inhibitors can promote the inhibition of PI3K/AKT pathway inhibitors, which increases the rate of apoptosis and slows down the growth and proliferation of tumour cells^31–33. Through downstream signal transduction of the JAK/STAT pathway, AXL also drives cancer development and metastasis cascades³⁴. Therefore, AXL may be a potential target for cancer treatment. The reported small molecule inhibitors of AXL (Figure 1) can be classified according to the inhibitor-kinase binding mode into Type I inhibitors such as TP0903³⁵ and BGB324³⁶, which preferentially occupy the ATP-binding pocket of the kinase in the active “aspartate-phenylalanine-glycine (DFG) – in” conformation, and Type II inhibitors, like BMS-777607³⁷ and NPS-1034³⁸, adopt linear structures to form inactive DFG-out conformations. However, these AXL small molecule inhibitors face challenges of innate or acquired resistance to single-targeting therapies or even radiotherapy in tumour treatment, which leads to severe limitations in the clinical application of AXL small molecule inhibitors^39–41. Therefore, we considered combining AXL with other potential targets to develop novel, potent, dual-targeting inhibitors to overcome these problems.

Figure 1. The reported AXL inhibitors and HDAC inhibitors.

Histone deacetylases (HDACs) are a crucial class of epigenetic enzymes that regulate gene expression by removing lysine residues from histones^42,43. HDACs are key regulators of several vital biological processes, including cell growth, cell differentiation, and cell apoptosis^44,45. In addition, HDACs are involved in the regulation of protein expression, including tumour suppressors (p53, p21, etc.) and transcription factors (TFIIE, TCF, SF1, etc.)⁴⁶. Abnormalities in HDACs are closely associated with cancer development, which is mainly manifested by high levels of HDAC expression and HDAC mutations⁴⁷. Among the HDACs, HDAC2 is a significant target in tumour development, as its abnormal expression promotes cancer progression primarily by silencing pro-apoptotic proteins such as NOXA and APAF1 (caspase 9-activating protein) and inactivating the tumour suppressor p53⁴⁸. In recent years, several HDAC inhibitors have been approved by the U.S. Food and Drug Administration (FDA) including Vorinostat (SAHA), Belinostat and so on (Figure 1)^49,50. Unfortunately, most of the HDAC inhibitors that have been reported with very limited efficacy against human solid tumours, which limits their partial application in oncological indications⁵. Furthermore, these small molecule inhibitors are associated with adverse gastrointestinal, hematological, and cardiac side effects^17,51.

Despite the limitations of AXL and HDAC single-targeting inhibitors in different aspects, it has been shown that combining AXL inhibitors with HDAC inhibitors is effective in the treatment of tumours^52,53. In addition, HDAC inhibitors with a high affinity for HDAC2 have been demonstrated to induce apoptosis and cell cycle arrest while suppressing mitomycin C (MMC) induced expression of AXL and the expression of related molecules⁵⁴. The multi-targeting HDAC, EGFR, and HER2 inhibitor CUDC-101 has been reported to promote the reduction of AXL levels to overcome treatment resistance caused by AXL overexpression⁵⁵. Although these studies have demonstrated the relevance of AXL and HDAC in tumour therapy, no dual-targeting AXL/HDAC2 inhibitors for tumour treatment have been reported so far.

Based on the correlation between AXL and HDAC2, we used structure-based virtual screening technology to screen and obtain the dual-targeting AXL/HDAC2 inhibitor^56,57, relying on its low cost, high efficiency, and ability to replace traditional screening methods, which avoid the drawbacks that exist in mono-targeting drugs and drug combination therapy. Here, we combined structure-based pharmacophore modelling, virtual screening, and molecular docking to obtain the hit compounds. Molecular dynamics simulations were performed to determine whether the best-scored compounds have good stability, and further in vitro and in vivo evaluations were performed to determine whether they have potential antitumor activity. The results showed that we successfully discovered a novel and potent dual-targeting AXL/HDAC2 inhibitor, providing a new candidate for tumour therapy.

Materials and methods

Materials

Cells and reagents. All cell lines were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA) and were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Gibco, New York, NY, USA) (A549, MCF7, HepG2, HCT116 (KRAS(mut)/p53(wt)) and Hela) or Roswell Park Memorial Institute (RPMI) −1640 medium (Gibco, Grand Island, NY, USA) (H1975 and Anoikis-resistant H460) with 10% FBS, 1% penicillin/streptomycin comprising 5% CO₂ controlling the temperature at 37 °C. All compounds were purchased from WuXi AppTec (Shanghai, China). Proteins were obtained from Abcam (Cambridge, UK). AXL and HDACs [HDAC1-11] proteins were obtained from Abcam (Cambridge, MA, USA). The 6-week-old BALB/c nude mice were bought from Changzhou Cavens Experimental Animal Limited Company (Changzhou, China).

Anoikis-resistant H460 cells. Anoikis-resistant H460 cells were established according to previously described methods⁵⁸. First, attached H460 cells were trypsinized with 0.05% trypsin/0.02% EDTA. The cell suspensions were then mixed with serum-free RPMI-1640 medium at a density of 2 × 10⁵ and seeded in 6-well plates (Corning, New York, NY, USA) for 48 h. After collection of the suspended cells, the cells were treated with 1 mM EDTA. After finally washing the cells with RPMI medium, cell viability was tested and viable cells were used for further experiments.

Establishment of AXL pharmacophore model

The X-ray 3D structures of AXL (PDB ID: 5U6B) and HDAC2 (PDB ID: 4LXZ) were obtained from the Protein Data Bank. All the structures were prepared in the Molecular Operating Environment (MOE) Program, which included partial charge calculation, protonation and energy minimisation of the two proteins. Subsequently, the structures of proteins were analysed, and the interactions between ligands and proteins were examined by using the ligand interaction tool. Based on the key interactions between AXL and small molecule inhibitors, the pharmacophore model of AXL was established with the pharmacophore query editor of MOE. All the established pharmacophores correspond to key amino acid residues of AXL.

Docking-based virtual screening

First, the In-house database of 43,000 compounds was established for virtual screening. These compounds were made up of 35,000 compounds reported in previous studies and 8000 compounds from the commercially available SPECS database (http://www.specs.net)⁵⁹. The MOE energy optimisation algorithm changed all of the compounds from two-dimensional structures to three-dimensional structures. The AXL pharmacophore model was selected as a constraint condition and the selected hit compounds interacted with AXL and HDAC2 active sites respectively. The native ligand and docking ligand were used for further docking verification, and the overlaps were assessed using the RMSD value calculation. MOE docking scores were used to evaluate the binding affinity and the docking scores of positive controls were set as the screening threshold value. Finally, five hits that met both scoring thresholds were chosen for further analysis.

Enzymatic assays for AXL and HDAC2

The in vitro HDACs enzymatic inhibitory experiments were performed as previously reported⁶⁰. 10 μL of HDAC2 enzyme solution and 50 μL different concentrations of compounds were added to the 96-well plate. Following 5 min incubation at 37 °C, 40 μL of the fluorescent substrate Boc-Lys (acetyl) – AMC was added to the plates. The 100 L developer containing trypsin and TSA were then added to the mixture after the plates had been incubated at 37 °C for 30 min. Finally, after 20 min, the fluorescence intensity at 390 nm and 460 nm were measured by a microplate reader Perkin Elemer.

The in vitro AXL enzymatic assay was performed using the method previously reported⁶¹. The reaction was performed in the buffer containing 100 mM Hepes, pH 7.5, 10 mM MgCl₂, 1 × Halt™ Protease and Phosphatase Inhibitor Cocktail, EDTA-Free, 1 mM DTT (Dithiothreitol), 0.003% of Brij35 and 0.004% of Tween-20. The total volume of the reaction was 26 μL including 1 nM AXL, 63 µM ATP (KmATP) and 1.5 μM compound. The plate was incubated at RT for 1.5 h and then stopped by adding 45 μL termination buffer containing 10 mM EDTA.

All the experiments were performed in triplicate. GraphPad Prism 6.0 software (GraphPad Software Inc, San Diego, CA) was used to analyse the competition binding data, and nonlinear curve fitting was used to get the inhibition constants (IC₅₀).

Molecular dynamics simulation

The GROMACS software (version 2021.5) was used to simulate the stability of Hit-3 over time in the AMBER99SB-ILDN field. The molecular dynamics simulation was performed under periodic boundary conditions^62,63. First, the Hit-3 was solvated with a single point charge (SPC) water molecule in a cube box that was 1.0 nm distant from the complex, and Na⁺ and Cl⁻ were used to neutralise the energy of the system instead of water molecules. The steepest descent algorithm was then used to set the energy of the system minimised with 5000 steps. A 100 ps NVT simulation with a V-rescale thermostat kept the system temperature at 300 K. This was followed by 100 ps NPT simulations using the Parinello-Rahman constant pressure device, which maintained the system pressure at 1 bar. Finally, the 50 ns MD simulation of the system was performed, and the RMSD and RMSF analysis of the system were recorded. Track data was saved every 10 ps.

MTT cell viability assay

According to the previously described method⁶⁴, cell suspensions were prepared at the concentration of 5 × 10⁴/ml and then cell lines were inoculated overnight in 96-well plates. Cells were exposed to different concentrations of inhibitors and incubated at 37 °C for 72 h. The incubation was maintained for an additional 4 h after adding 0.5 mg/ml MTT solution to each well. Insoluble crystals were dissolved with dimethyl sulfoxide and absorbance was measured at 570 nm according to the microplate reader. The data were analysed with the GraphPad Prism 6.0 software and half-maximal inhibitory concentration (IC₅₀) values and cell viability was calculated from dose-response curves displayed by nonlinear regression (curve fitting).

In vivo anti-tumour activity

We injected tumour cells (200 µL, 1 × 10⁷ cells) into the right subcutaneous space of 6-week-old BALB/c nude mice according to the previous in vivo antitumor activity assay method⁵⁶. The Animal Ethics Committee of China Pharmaceutical University gave its approval to all the experimental protocols.[药代动方法]

The BALB/c nude mice were divided into four groups at random, with five mice in each group, while the tumours of BALB/c nude mice reached 80–120 mm³. The BALB/c nude mice were then injected intraperitoneally with vehicle (PBS buffer pH 7.4), Hit-3 (1 mg/kg), Hit-3 (10 mg/kg), and Hit-3 (20 mg/kg), respectively. The injection volume capacity of each group was 0.2 ml and the Hit-3 was all in PBS vehicle. Each group was injected every 3 days for a total of five times. Tumour volume was measured and computed as (c × c × d)/2 (c, minimum diameter; d, maximum diameter).

After that, the BALB/c nude mice were injected intraperitoneally with vehicle (PBS buffer pH 7.4), TP0903 (10 mg/kg), SAHA (10 mg/kg), and Hit-3 (10 mg/kg), respectively. The injection volume capacity of each group was 0.2 ml and each group was injected every 3 d for a total of five times. The body weights of the BALB/c nude mice were recorded, and the tumour volumes were measured and calculated according to the above method.

Results

Structure-based virtual screening of dual-targeting AXL/HDAC2 inhibitors

In this study, we identified novel and potent dual-targeting AXL/HDAC2 inhibitors from the In-house database (Figure 2), and the identification steps were shown in Figure 3. Firstly, we used a structure-based technique to construct the AXL pharmacophore model depending on the AXL structure of the crystal (PDB ID: 5U6B). As shown in Figure 3, we generated four typical pharmacophore features (called F1-F4), two hydrophobic features (F1 and F2), one aromatic feature (F3), and one hydrogen bond donor feature (F4). Hydrophobic characteristics F1 and F2 corresponded to amino acid residues around the ligand, Leu620, Phe622, and Met623. The Arg676 residue was related to the aromatic character F3. The hydrogen bond interaction established by the amino acid residue Asp627 with the ligand corresponded to characteristic F4. According to the four pharmacophore characteristics of F1–F4, the pharmacophore model was established to prepare for the docking virtual screening.

Figure 2. The chemical structures of Hits 1–5.

Figure 3. Identification of dual-targeting AXL/HDAC2 inhibitors. (A) Pharmacophore model based on AXL (F1 and F2 are hydrophobic characteristics, F3 is aromatic characteristics, and F4 is hydrogen bond donor characteristics). Pharmacophore characteristics are represented by spheres. Hydrogen bonds are represented by the black dotted line. (B) The identification process of dual–targeting AXL/HDAC2 inhibitors.

First, the small molecule database of 43,000 compounds was established for virtual screening and the 2D structure of each compound was converted into the 3D structure by energy minimisation. Following that, 379 hit compounds meeting all pharmacophore features were obtained by using the AXL pharmacophore model as a screening limitation. Before the docking-based virtual screening, the validations of native ligand and docking ligand with co-crystal ligand and protein complex were performed respectively. The validation results were shown in Figure 4, showing the high overlaps between the native and docking ligands. The RMSD values were calculated with 0.009 nm of AXL and 0.012 nm of HDAC2. These 379 compounds were then screened and docked to AXL active sites and HDAC2 active sites. The binding free energy was utilised to forecast the binding affinity of hit compounds to the active sites, usually, the lower binding free energy indicated the stronger binding affinity. The active site of the compound connected to AXL was screened, and the docking score of each hit was calculated. The molecular docking score of TP0903 was −10.02 kcal/mol. Therefore, the reasonable docking score of −10.02 kcal/mol was set as the AXL threshold. A total of 49 compounds were identified. Then, 379 compounds were docked to the crystal structure of HDAC2 (PDB ID: 4LXZ), and 30 compounds were screened with docking scores of less than −11.35 kcal/mol (The molecular docking score of SAHA was −11.35 kcal/mol). Based on docking scores of less than −10.02 kcal/mol and less than −11.35 kcal/mol, we identified five potential dual-targeting AXL/HDAC2 candidate compounds (called Hits 1–5) that simultaneously met both scoring thresholds. The docking scores of Hits 1–5 were shown in Figure 5. The interactions between Hits 1–5 and AXL/HDAC2 were further analysed.

Figure 4. The docking validation using the co-crystal ligand and protein complex before setting screening threshold values and docking-based virtual screening. (A) The docking validation of AXL (PDB ID: 5U6B). (B) The docking validation of HDAC2 (PDB ID: 4LXZ). (yellow for native ligand, green for docking ligand).

Figure 5. The docking scores of Hits 1–5, TP0903 and SAHA. ^aBinding free energy between the compounds and the targets (lower binding free energies show stronger binding affinities).

Interaction analysis

Docking studies predicted the binding conditions of Hits 1–5 to AXL and HDAC2. Figure 6 showed the interaction of Hits 1–5 with AXL. Figure 6(A,C,E,G,I) showed that a total of 11 amino acid residues (Ala565, Lys567, Leu620, Phe622, Met623, Asp627, Arg676, Leu693, Lys710, Met711, Pro712) were crucial for the binding of each compound to AXL. Notably, all of the hits generated hydrogen bonds with residues Asp627 and Arg676, which improved the stability of the compound binding to the AXL active site, resulting in a good binding between the compound and the active binding pocket. In addition, each hit formed hydrophobic interactions with Leu620, Phe622, and Met623, making the binding more stable. Figure 6(B,D,F,H,J) showed the binding surface maps of Hits 1–5 to AXL, and the compounds obtained from the screening were well accommodated by the AXL active site pocket. The docking binding pattern of Hits 1–5 to HDAC2 was shown in Figure 7. As shown in Figure 7(A,C,E,G,I), Hits 1–5 formed interactions with 8 amino acid residues (His146, Tyr209, Phe210, Leu276, Cys278, Tyr308, Met355, Thr356). The hydroxamic acid of the compound produced ionic bonds with zinc ions and hydrogen bonds with residues Phe210 and Tyr308, which could increase the stability of the compound. Besides, the surface maps in Figure 7(B,D,F,H,J) showed that the hydroxamic acid groups of Hits 1–5 were inserted into the HDAC2 active pocket and Hits 1–5 could fit comfortably in the active pocket of HDAC2. Therefore, these docking poses suggest that Hits 1–5 can interact with key residues of both AXL and HDAC2, indicating that Hits 1–5 may have the potential effect of inhibiting target proteins.

Figure 6. The X-ray crystal structure and the predicted docking poses of Hits 1–5 at the AXL active site: (A,B) Hit-1; (C,D) Hit-2; (E,F) Hit-3; (G,H) Hit-4; (I,J) Hit-5. Compounds are indicated as sticks with different atom colours (red for Hit-1, green for Hit-2, cyan for Hit-3, orange for Hit-4, dark blue for Hit-5). AXL is colour-coded by gray. The surfaces of AXL are colour-coded by gray. Hydrogen bonds are indicated by dashed black lines.

Figure 7. The X-ray crystal structure and the predicted docking poses of Hits 1–5 at the HDAC2 active site: (A,B) Hit-1; (C,D) Hit-2; (E,F) Hit-3; (G,H) Hit-4; (I,J) Hit-5. Compounds are indicated as sticks with different atom colours (red for Hit-1, green for Hit-2, cyan for Hit-3, orange for Hit-4, dark blue for Hit-5). HDAC2 is colour-coded by gray. The surfaces of HDAC2 are colour-coded in gray. Hydrogen bonds are indicated by dashed black lines.

Enzymatic inhibition assays of AXL and HDAC2 by Hits 1–5

To further evaluate the inhibitory effects of the selected Hits 1–5 on the dual targets, in vitro inhibitory effect assays were performed on Hits 1–5. The IC₅₀ values for the inhibitory effects of Hits 1–5 against AXL were 19.52, 14.33, 9.41, 24.38 and 21.25 nM, respectively, all of which were lower than the positive control of 26.23 nM (Table 1). The IC₅₀ values for HDAC2 inhibitory effects were 7.49, 6.65, 4.06, 8.37 and 8.89 nM, respectively, all with lower IC₅₀ values than the positive control of 9.02 nM. Among the positive controls, TP0903 inhibited AXL, but not HDAC2. SAHA inhibited HDAC2 but did not affect AXL activity. Notably, among Hits 1–5, Hit-3 showed the best inhibitory activity against AXL and HDAC2, and the inhibitory effects were stronger than that of the positive controls, with IC₅₀ values about 2.8 times that of TP0903 and 2.2 times that of SAHA. In addition, we selected Hit-3 to evaluate the IC₅₀ values of other HDAC enzymes. The results show that the inhibitory effect of Hit-3 on other HDAC enzymes was not as good as that of HDAC2 (Table 2). Therefore, we further performed molecular dynamics simulations of Hit-3.

Table 1. The enzymatic inhibitory effects of Hits 1–5, TP0903 and SAHA in cell-free assays.

Compounds	AXL (IC50, nM)	HDAC2 (IC50, nM)
Hit-1	19.52 ± 0.63	7.49 ± 0.85
Hit-2	14.33 ± 0.78	6.65 ± 0.48
Hit-3	9.41 ± 0.76	4.06 ± 0.13
Hit-4	24.38 ± 1.27	8.37 ± 0.76
Hit-5	21.25 ± 1.81	8.89 ± 0.72
TP0903	26.23 ± 1.09	no inhibition
SAHA	no inhibition	9.02 ± 1.81

Table 2. The inhibitory effects of Hit-3 on other HDAC enzymes.

Target	IC50 (nM)	Target	IC50 (nM)
HDAC1	7.51 ± 0.64	HDAC7	9.54 ± 0.43
HDAC3	11.33 ± 0.82	HDAC8	8.46 ± 0.22
HDAC4	10.95 ± 0.91	HDAC9	16.73 ± 0.85
HDAC5	13.12 ± 0.76	HDAC10	14.56 ± 0.59
HDAC6	9.24 ± 0.53	HDAC11	12.24 ± 0.56

Molecular dynamics simulations of Hit-3

The binding stability of the system over time was analysed using 50 ns molecular dynamics simulations of the Hit-3-AXL complex and Hit-3-HDAC2 complex. The RMSD of Hit-3 in AXL and HDAC2 complexes were depicted in Figure 8(A,B). RMSD was one of the most essential factors in describing the stability. Better binding stability is indicated by lower RMSD values. The final average value of the Hit-3-AXL complex was less than 0.4 nm, and the final average value of the Hit-3-HDAC2 complex was less than 0.2 nm. These show that Hit-3 can tightly bind both AXL and HDAC2 and the Hit-3-AXL and Hit-3-HDAC2 complexes are stable in the molecular dynamics simulation. The RMSF of the C atom of the amino acid residues was then measured. The RMSF values were related to the interactions between AXL/HDAC2 and Hit-3. Generally, low RMSF indicates minimal residual motion and a stable system. As shown in Figure 8(C,D), there were large fluctuations in the RMSF of both complexes at the C-/N-terminus, which might be connected with less interaction. Moreover, the fluctuations of key amino acid residues related to Hit-3-AXL and Hit-3-HDAC2 complexes were mostly below 0.4 nm. Figure 8(E,F) showed the secondary structures of the complexes including A-Helix, Coil, Turn, B-Sheet, Bend, 3-Helix, B-Bridge, and 5-Helix in Hit-3-AXL and Hit-3-HDAC2. Within 50 ns, these secondary structures were fluctuating stably, indicating that the proteins were in a stable state. Therefore, the molecular dynamics simulations show that Hit-3 can stably bind to AXL and HDAC2 and Hit-3 may be a novel and potent dual-targeting inhibitor of AXL/HDAC2.

Figure 8. Molecular dynamics simulation of the Hit-3 with AXL and HDAC2. (A,B) RMSD of Hit-3 in complex with AXL and HDAC2, respectively. (C,D) RMSF of amino acid residues of Hit-3 in complex with AXL and HDAC2, respectively. (E,F) The secondary structures analysis of Hit-3 in complex with AXL and HDAC2, respectively. (black for Structure, red for Coil, green for B-Sheet, dark blue for B-Bridge, yellow for Bend, pink for Turn, grey for A-Helix, purple for 5-Helix and blue for 3-Helix).

Cell inhibition rate and cell viability analysis

To further analyse the anti-proliferative effect of the Hit-3, we also evaluated the cell inhibition rate of 5 hits on the HCT116 cells at 0.7 μM concentration. The experimental results are shown in Figure 9. The cell inhibition rate of Hits 1–5 on the HCT116 cells was more than 70%, with the best inhibitory effect of Hit-3 reaching more than 90%. In addition, the IC₅₀ value of Hit-3 was determined by the MTT method. In general, the lower the IC₅₀ values indicated the stronger ability of the compound to induce apoptosis. As shown in Figure 10, Hit-3 had growth inhibitory activity on HCT116 cells with an IC₅₀ value of 0.07 μM. At the concentration of 0.01 μM, the cell viability rate was about 80%, which decreased with increasing concentration. Therefore, these indicate that Hit-3 has good anti-tumour ability and maintains good cell viability at IC₅₀ concentration. We also evaluated the IC₅₀ values of A549, MCF-7, HepG2 and HeLa cells. The results show that the inhibitory effect of Hit-3 on HCT116 cells was stronger than that of other cells. In the evaluation of the inhibition effect of EGFRi-resistant H1975 and AXLi-resistant H460, Hit-3 shows an improvement in drug resistance of them (Table 3).

Figure 9. Inhibitory effects of Hits 1–5, TP0903 and SAHA on HCT116 cells.

Figure 10. Inhibitory effects of Hit-3 on HCT116 cells detected by using MTT assay.

Table 3. The inhibitory activity of Hit-3 on other cancer cells and drug-resistant cancer cells.

Name	IC50 (μM)^a
	A549	MCF-7	HepG2	HeLa	EGFRi-resistant H1975	AXLi-resistant H460
Hit-3	0.14	0.52	0.26	0.63	1.18	2.76

^aIC₅₀ (μM) is the concentration of compound needed to reduce cell growth by 50% following 48 h cell treatment with Hit-3.

In vivo anti-tumour activity

The metabolic stability and pharmacokinetics profiles of Hit-3 were evaluated before in vivo antitumor activity. The results in Table 4 show that the T_1/2 of Hit-3 is 8.41 h and the C_max of Hit-3 is 1307 ng ml⁻¹ at 10 mg/kg dose. To further analyse the in vivo anti-tumour activity of Hit-3, HCT116 cells were injected into BALB/c nude mice to establish the tumour models, and the mice were divided into four groups: vehicle, 1 mg/kg Hit-3, 10 mg/kg Hit-3, and 20 mg/kg Hit-3. As shown in Figure 11, compared with the vehicle, the tumour volumes of mice given injections of Hit-3 at doses of 1 mg/kg, 10 mg/kg, and 20 mg/kg were substantially reduced. Therefore, these indicate that Hit-3 has a strong in vivo anti-tumour activity. Furthermore, compared to mice given injections of 1 mg/kg Hit-3 and 10 mg/kg Hit-3, the mice treated with 20 mg/kg Hit-3 showed a better notable reduction in tumour volume. These suggest that Hit-3 inhibits tumour growth in a dose-dependent manner.

Figure 11. Changes of tumour volume with injecting HCT116 (KRAS(mut)/p53(wt)) cells in BALB/c nude mice over time. The data are shown as mean ± SD, n = 5. ***p < .001 means a significant difference versus the vehicle group.

Table 4. In vivo pharmacokinetic parameters of Hit-3 after intraperitoneal administration in micea.

Name	route	dose (mg/kg)	T1/2 (h)	Cmax (ng ml^−1)	Tmax (h)	AUC (h*ng/mL)
Hit-3	i.p.	10	8.41	1307	0.25	3962

^aT_1/2: elimination half-life; C_max: maximum plasma concentration; T_max: the peak time of plasma after administration; AUC: area under the drug − time curve.

As shown in Figure 12, BALB/c nude mice were divided into four groups, one of which was used as the vehicle group, and the other three groups were injected with TP0903, SAHA and Hit-3 at a dose of 10 mg/kg respectively to evaluate the in vivo anti-tumour effect. Figure 12(A) showed that the tumour volume of mice injected with positive control drugs and the Hit-3 group was smaller than that of the vehicle group, and the tumour volume of Hit-3 was the smallest. In Figure 12(B), the body weight changes of the four groups showed a stable trend. Therefore, in vivo experimental results suggest that Hit-3 is effective in inhibiting tumour growth with good safety.

Figure 12. In vivo anticancer activity. (A) Tumour volume evaluations at the dose of 10 mg/kg for TP0903, SAHA, Hit-3 and vehicle. (B) Body weight evaluations after injecting vehicle, TP0903, SAHA and Hit-3 at the dose of 10 mg/kg. The data are shown as mean ± SD, n = 5. ***p < .001 means a significant difference versus the vehicle group.

Conclusion

Single-targeting medications and combination therapy have shown limitations in various parts of cancer therapy, thus the development of novel dual-targeting drugs is essential. AXL and HDAC2 are highly correlated with targeting cancer therapy, and the dual-targeting AXL/HDAC2 inhibitor may provide new options for cancer treatment. However, no dual-targeting AXL/HDAC2 inhibitors have been reported so far. In this study, five hit compounds targeting both AXL and HDAC2 were rationally identified through pharmacophore modelling, virtual screening, and molecular dynamics simulations. Interaction analysis showed that Hits 1–5 formed key interactions with residues Ala565, Lys567, Leu620, Phe622, Met623, Asp627, Arg676, Leu693, Lys710, Met711, and Pro712 of AXL. All hit compounds formed hydrogen bonds with residues Asp627 and Arg676 and hydrophobic interactions with Leu620, Phe622, and Met623, making the compounds occupy the AXL pocket completely. Meanwhile, Hits 1–5 formed interactions with HDAC2 residues His146, Tyr209, Phe210, Leu276, Cys278, Tyr308, Met355, and Thr356, which ensured that the compounds were stably bound to the active site. Further in vitro inhibitory experiments showed that the affinity of Hits 1–5 for both AXL and HDAC2 were within the micromolar range and the cellular inhibition rate was greater than 70%. The most effective inhibitor Hit-3, was selected for molecular dynamics simulations to analyse the binding stability of Hit-3 to AXL and HDAC2. The results showed that Hit-3 was able to bind stably to AXL and HDAC2. In the MTT assay, Hit-3 had a significant inhibitory effect on HCT116 cells and ensured cell viability well. According to in vivo data, Hit-3 suppressed tumour growth in a dose-dependent manner in the HCT116 cells xenograft tumour mouse model.

However, it is worth noting that although the in vivo activity evaluation of Hit-3 indicates that it has good anti-tumour activity, the anti-tumour mechanism of Hit-3 remains to be further explained. Previous research has revealed that increased HDAC2 expression promotes the positive feedback mechanism, which causes phosphoinositol molecules to activate AKT phosphorylation⁶⁵. Previously reported HDAC2 inhibitors have been shown to exert anti-tumour activities by significantly inhibiting the PI3K/AKT/mTOR pathway^66,67. Furthermore, studies have shown that AXL-HER2 dimerisation can also affect the PI3K/AKT pathway, hence affecting treatment efficacy⁶⁸. As mentioned above, we conclude that Hit-3 may have activation or inhibitory effects on other pathways, implying that we can further research the anti-tumour mechanism of Hit-3 in the future. Based on existing experimental studies, we reported and identified the dual-targeting AXL/HDAC2 inhibitor for the first time, and evaluated both the HDACs enzyme activity and the anti-tumour activity of Hit-3, which may provide a new therapeutic strategy for future cancer treatment.

Authors contributions

YZ, HH and XQ contributed to the conception of the study. XW and MMN performed experiments and acquired data. XQ, SC, XW and MMN performed statistical analysis. YZ, HH, XQ, XW and MMN contributed significantly to manuscript preparation. All authors had contributed and approved the manuscript.

Ethics statement

All experimental protocols were reviewed and approved by the Animal Ethics Committee of China Pharmaceutical University.

Acknowledgements

We gratefully acknowledge the support from the Taizhou People’s Hospital.

Disclosure statement

The authors report no conflicts of interest.

Additional information

Funding

This work was supported by Huai’an Science and Technology Bureau (HAB202210), and Jiangsu Pharmaceutical Association-CHIATAI TIANQING Hospital Pharmacy Research Fund Program (Q202239).