Novel dual-targeting inhibitors of NSD2 and HDAC2 for the treatment of liver cancer: structure-based virtual screening, molecular dynamics simulation, and in vitro and in vivo biological activity evaluations

Abstract

Liver cancer exhibits a high degree of heterogeneity and involves intricate mechanisms. Recent research has revealed the significant role of histone lysine methylation and acetylation in the epigenetic regulation of liver cancer development. In this study, five inhibitors capable of targeting both histone lysine methyltransferase nuclear receptor-binding SET domain 2 (NSD2) and histone deacetylase 2 (HDAC2) were identified using a structure-based virtual screening approach. Notably, DT-NH-1 displayed a potent inhibition of NSD2 (IC₅₀ = 0.08 ± 0.03 μM) and HDAC2 (IC₅₀ = 5.24 ± 0.87 nM). DT-NH-1 also demonstrated a strong anti-proliferative activity against various liver cancer cell lines, particularly HepG2 cells, and exhibited a high level of biological safety. In an experimental xenograft model involving HepG2 cells, DT-NH-1 showed a significant reduction in tumour growth. Consequently, these findings indicate that DT-NH-1 will be a promising lead compound for the treatment of liver cancer with epigenetic dual-target inhibitors.

Keywords:

• Liver cancer
• NSD2
• HDAC2
• dual-targeting inhibitors
• virtual screening

Introduction

Liver cancer is the second most common cause of cancer-related deaths worldwide due to its invasiveness and lack of effective treatment.^1–3 Targeted therapy has become a new approach to cancer treatment because of its high specificity and low side effects.⁴ However, the majority of the current drugs developed for single-target cannot fully consider the essential characteristics of cancer with multiple pathogeneses, pathological links, and genetic correlations.^5,6 Although drug combination therapy and fixed-dose combinations (FDCs) therapy are utilised as alternative approaches to achieve clinical efficacy, they are associated with a greater risk of drug-drug interactions, unpredictable pharmacokinetics, and poor safety.^7,8 The ideal multi-target drugs can avoid drug resistance caused by single-target mutations and better balance the relationship among multiple pathological factors of the disease. In addition, multi-target drugs have exhibited a great potential in expanding the scope of treatment and improving patient compliance.^9,10 Therefore, it is imperative to research and develop multitarget drugs, especially dual-targeting drugs, for the treatment of liver cancer.¹¹

Histone lysine deacetylases (HDACs), as the most important histone post-translational modifiers, promote a closed chromatin conformation by removing acetyl groups from ε-amino group of histone lysine residues.¹² To date, four classes of human HDAC isoenzymes have been identified: class I (HDAC1, 2, 3, and 8), class II (HDAC4, 5, 6, 7, 9, and 10), and class IV (HDAC11) including Zn²⁺-dependent enzymes that adopt an arginase deacetylase fold, whereas class III (Sirtuins) are NAD⁺-dependent enzymes that adopt an unrelated fold.¹³ Among them, compared with other categories, HDAC1, 2, and 3 are not only expressed in the nucleus, but also widely expressed, and they may play the most important role in tumorigenesis.^14–16 In particular, clinical studies have demonstrated that serum HDAC2 level is higher in patients with liver failure than that in healthy individual people.¹⁷ In addition, Kim et al. found that the mechanism of HDAC2 upregulation in a large number of patients with hepatocellular carcinoma (HCC) may be that epidermal growth factor (EGF) induces the transcriptional activation of HDAC2 in hepatoma cells activation of through CK2α/Akt signalling pathway.¹⁸ HDAC2 knockdown increases the function of tumour suppressor p53, thereby inhibiting proliferation and inducing cellular senescence.^19,20 Moreover, recent studies have shown that inhibition of HDAC2 can regulate the acetylation of PD-L1 to improve human immunity.²¹ HDAC2 siRNA encapsulated in lipid nanoparticles was also found to inhibit the growth of HCC cells.^22,23 A growing body of evidence has highlighted that HDAC2 plays a key role in the tumorigenesis of liver cancer by regulating hepatocyte death and cell cycle components.^24,25 Therefore, the development of compounds that selectively inhibit HDAC2 is a very promising strategy for liver cancer treatment.

Liver cancer is a heterogeneous disease with different molecular and phenotypic characteristics, and it typically involves mutations in two or more carcinogenic pathways.²⁶ As an important class of epigenetic modifiers, histone lysine methyltransferase (HKMTase) nuclear receptor binding SET domain 2 (NSD2) specifically dimethylates H3K36 to form and maintain chromatin structure, regulate gene transcription and participate in the epithelial-mesenchymal transition (EMT).^27,28 NSD2 is a multi-domain protein, containing two proline-tryptophin-tryptophin-proline (PWWP) domains and a SET domain that can catalyse methylation.²⁹ The NSD2 has been proven to be overexpressed in HCC cells, and it played an oncogene role.^30,31 NSD2 overexpression resulted in a global increase in H3K36me2 expression level and it acted as a potent coactivator to mediate NF-κB signalling pathway to promote tumour growth, while NSD2 knockdown inhibited cancer cell proliferation.^32–36 Importantly, Qiu et al. demonstrated that NSD2 promoted the proliferation, migration, and invasion of HCC cells by activating mTORC1 signalling pathway, and determined that NSD2 could be a promising therapeutic target for HCC.³⁷ The recent development of small-molecule inhibitors targeting NSD2 has emerged as a very promising strategy to treat cancer progression and metastasis.³⁸ In 2021, Allali-Hassan et al. discovered a chemical antagonist MR837 that could specifically target the aromatic cage in the NSD2-PWWP1 domain by virtual screening, and successfully proposed a new method to target NSD2.³⁹ Subsequently, based on the cocrystal structure of the NSD3-PWWP1 domain chemical probe BI-9321 and MR837, Li et al. optimised a new small-molecule compound 38 targeting NSD2-PWWP1, and biological evaluation indicated that the compound partially inhibited the proliferation of tumour cells.⁴⁰ Therefore, the number and types of NSD inhibitors are relatively limited, and the development of small molecule inhibitors is still in its infancy.⁴¹ In addition, the recent Food and Drug Administration (FDA) approval of Tazemetostat, a lysine methyltransferase (PKMT) inhibitor, for epithelioid sarcoma and follicular lymphoma further strengthens the therapeutic relevance and drug availability of the NSD2 target.⁴² Therefore, additional research is urgently required to develop NSD2 inhibitors with high cellular potency or in vivo efficacy.

To date, it has become a consensus that inhibitors developed against HDAC induce cell cycle arrest, differentiation, and apoptosis by preventing aberrant HDAC deacetylation in cancer patients.⁴³ Some HDAC inhibitors that inhibit HDAC2 have been approved by the FDA, such as Vorinostat and Panobinostat.^44,45 However, clinical results indicated that the therapeutic efficacy of these drugs is unsatisfactory, and patients still experience tumour invasion and metastasis.⁴⁶ Recent date demonstrated that NSD2 is involved in tumour cell invasion and metastasis through EMT.⁴⁷ Moreover, studies have found that NSD2 is closely associated with histone methyltransferase EZH2, and EZH2 mediates trimethylation of H3K27 (H3K27me3) to further upregulate NSD2 expression level.²⁸ The combined treatment of human acute myeloid leukaemia (AML) cells with EZH2 inhibitor 3-deazaneplanocin A and Panobinostat caused more depletion of EZH2 and inhibition of trimethylation of lysine 27 on histone H3 than treatment with each drug alone, and significantly improved the survival rate of immunodeficient mice with leukemia.⁴⁸ Hence, it is imperative to design NSD2/HDAC2 dual-targeting small-molecule compounds to prevent tumorigenesis and inhibit tumour metastasis to improve the efficacy of liver cancer treatment (Scheme 1).

Scheme 1. Mechanistic diagram of NSD2/HDAC2 inhibitors in the treatment of HCC.

With the recent advances in the understanding of tumour molecular biology, molecular targets for treatment have been identified, such as tumour antigens, growth factors, transcription factors, and epigenetic modifiers.^49,50 Small-molecule drugs targeting tumours have noticeable advantages of reaching intracellular targets and greater drug exposure in the tumour microenvironment. Therefore, there is an urgent need to develop small molecules to modulate the immune system and treat diverse type of cancer.⁵¹ From discovery of bioactive compounds to corresponding biological verification, computer-aided drug design (CADD) has remarkably accelerated the process of drug research and reduced the overall cost.⁵² The virtual screening method based on structure is robust and efficient, and it is one of the most promising in silico techniques.^53–55 Through virtual screening and functional verification, Ucar et al. and Thiyagarajan et al. identified the lead compound of dual-targeting inhibitors of focal adhesion kinase (FAK) and other factors (insulin-like growth factor 1 and ribosomal S6 kinase 1).^56,57 Moreover, the TAOK1 and MAP4K5 dual inhibitor compounds identified using a structure-based virtual screening approach arrested cancer cells at G0/G1 phase.⁵⁸

Simultaneous targeting of HDAC2/NSD2 by a small-molecule compound is a promising cancer treatment that has rarely been reported, making them an attractive strategy for the development of drugs to treat liver cancer. In the present study, the first NSD2/HDAC2 dual-targeted inhibitor (DT-NH-1) was successfully identified using structure-based virtual screening. Subsequently, DT-NH-1 exhibited encouraging in vitro enzymatic inhibitory activity, anti-cell proliferation activity, and biosafety. Finally, it was found to display significant antitumor efficacy in a HepG2 xenograft mouse model. In summary, DT-NH-1 is a promising dual-targeting inhibitor in the treatment of cancer.

Materials and methods

Materials

Four HCC cell lines (HepG2, HuH-1, SK-HEP-1, and Bel-7405) and human normal liver L-02 cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA). All cells were cultured in a Dulbecco’s modified Eagle’s medium (DMEM; Gibco, New Nork, USA), supplemented with 100 U/ml penicillin, 100 mg/ml streptomycin, and 10% foetal calf serum, and maintained in a humidified incubator with 5% CO₂ and 95% air at 37 °C. All compounds were purchased from WuXi AppTec (Shanghai, China). NSD2 and HDAC2 proteins were obtained from Abcam (Cambridge, UK).

Establishment of a structure-based pharmacophore model for NSD2

Pharmacophore models based on validated protein structures have been widely utilised in the drug design and development.^59,60 The X-ray crystal structure of NSD2 (PDB ID: 7VLN) was obtained from Protein Data Bank (PDB). The protein preparation tool in the Molecular Operating Environment program (MOE, Chemical Computing Group Inc, Montreal, Quebec, Canada) was used to pretreat the protein structure of NSD2 by removing water molecules and adding hydrogen atoms. Then, the energy was minimised through Amber14: EHT force field. The representative pharmacophore features in NSD2 active site, including hydrogen bond acceptor (Acc), hydrogen bond donor (Don), aromatic centre (Aro), and hydrophobic (Hyd) features, were selected by the pharmacophore query editor of MOE.

Validation of the pharmacophore model

To validate the reliability of the NSD2 model described above, the Güner–Henry (GH) scoring method was utilised as described previously.⁶¹ A test database consisted of 1981 inactive molecules from the Enhanced Database of Useful Decoys (DUDe) and 19 reported active molecules.^38,39,41 The proposed model was subsequently used as a three-dimensional (3D) query to filter the test database using the pharmacophore search protocol available in MOE. Finally, statistical parameters such as total hits (Ht), yield of actives (%), ratio of actives (%), the goodness of hit score (GH), and enrichment factor (E) were calculated. The E value is proportional to the ability of pharmacophore model to identify active compounds. A GH score greater than 0.7 represents a higher reliability of the model. The GH score was calculated based on the following formula:⁶² $$GH=\left(\frac{Ha\left(3A+Ht\right)}{4HtA}\right)\left(1-\frac{\left(Ht-Ha\right)}{(D-A)}\right)$$

Docking-based virtual screening

Molecular docking is a computational process that simulates the interaction between a ligand and a target protein, aiming to predict the binding affinity and the specific binding orientation (pose) of the two molecules. This is achieved through the calculation of various physicochemical parameters.⁶³ To reduce computational effort and obtain the optimal conformation, the semi-flexible docking method was employed, which was to set the target protein to be rigid and the ligand compound to be flexible. The energy minimisation principle of MOE was followed to convert 35,000 molecules from the obtained ligand database into 3D compounds. The crystal structures of NSD2 (PDB ID: 7VLN) were utilised as a screening template, and the docking score of MR837 was used as a threshold to screen compounds in the database for subsequent filtering of HDAC2 targeted compounds. The crystal structures of HDAC proteins co-crystallized by the FDA-approved drug Vorinostat (PDB ID: 4LXZ) were obtained from the PDB. The two proteins were imported into the MOE program and prepared by removing water, adding hydrogen atoms, calculating partial charges, and performing energy minimisation in Amber14: EHT force field.

Enzymatic assays for HDAC2 and NSD2-PWWP1

In vitro HDAC2 inhibition experiments were carried out as described previously.⁶⁴ In the 96-well plate, 10 μL of HDAC2 enzyme solution was first added, and 50 μL of inhibitor compounds with different concentrations were mixed. After incubating the mixture at 37 °C for 5 min, 40 μL of the fluorescent substrate Boc-Lys (acetyl) -AMC was added to the mixture. The plates were incubated at 37 °C for an additional 0.5 h. Finally, the developer (100 μL) containing trypsin and TSA was injected, and the fluorescence intensity at wavelengths of 390 and 460 nm was measured 20 min later by a microplate reader (BioTek Cytation 5; Agilent Technologies, Inc., Santa Clara, CA, USA).

The enzymatic activity of NSD2-PWWP1 was determined by a previously described homogeneous time-resolved fluorescence (HTRF) assay.⁴⁰ First, different concentrations of inhibitor compounds were prepared and dissolved in dimethyl sulfoxide (DMSO). Subsequently, the mixture of 5 μL of NSD2-PWWP1 (4 nM) dilution buffer and 5 μL of substrate polypeptide dilution buffer was added to 96-well plates according to the manufacturer’s protocol. Then, 5 μL of the final 1:1000 dilution of the inhibitor compound was sucked into the mixture, and 5 μL of the Tb/d2 mixture was added to each well. The 96-well plates were incubated at room temperature for 3 h, and the fluorescence intensity was recorded using a microplate reader (BioTek Cytation 5) at test wavelengths of 665 and 620 nm. All experiments were repeated three times under the same conditions.

Molecular dynamics (MD) simulation

GROMACS 2021.5 was employed to perform the MD simulations of NSD2-hit complex and HDAC2-hit complexes. The topology files for the NSD2 and HDAC2 proteins were created using GROMACS, utilising the AMBER99SB-ILDN force field. The topology file for the compound of interest (hit) was generated through the Acpype Server (www.bio2byte.be) and employed GAFF force field. NSD2-hit complex was placed in the centre of a cubic box with each side length of 71 Å, and HDAC2-hit complex was placed in the centre of a cubic box with each side length of 82 Å. TIP3P water model was added into the boxes to solvate the complex systems, and appropriate Na⁺ and Cl^- ions were also added to neutralise the complex systems. Thereafter, the two complex systems were subjected to energy minimisation using the steepest descent method for 5000 steps. Furthermore, the two systems were gradually heated from 0 to 300 K using the V-rescale thermostat in a 1-ns NVT simulation to equilibrate systems. Subsequently, 1-ns NPT simulation was performed to relax systems (pressure = 1 atm) using the Parrinello-Rahman barostat. Finally, 50 ns MD was carried out with the integration step of 1 fs, and hydrogen-bonds were constrained using the linear constraint solver (LINCS) algorithm. Trajectory data were recorded every 10 ps. The root-mean-square deviation (RMSD) calculation was conducted using the gmx_rms module, the root-mean-square fluctuation (RMSF) calculation was performed using the gmx_rmsf module, and the secondary structure was evaluated via DSSP 4.0 software.

MTT assay

The effects of candidate compounds on cell proliferation were determined as previously reported by MTT assay.⁶⁵ Firstly, the most potent compound was selected by the inhibition rate of all compounds on the HepG2 cell line under the same condition (the same concentration), and then the IC₅₀ value of the compound on more HCC cell lines was determined.^60,66 The cells were seeded into 96-well culture plates at a density of 5 × 10⁴ cells/well. After overnight incubation, cells were treated with various concentrations of the candidate compounds for 72 h. Each well was filled with 100 μL MTT (0.5 mg/mL) solution dissolved in phosphate-buffered saline (PBS), and the 96-well plates were placed at 37 °C for 4 h until a purple precipitate appeared. Then, 200 μL dimethyl sulfoxide (DMSO) was injected into the mixture and shaken gently for 10 min. The absorbance of the samples at 570 nm was measured using a Synergy 4 microplate reader (BioTek Instruments Inc., Winooski, VT, USA). Each set of experiments was performed three times in a parallel manner.

In vivo antitumor assay

To investigate the antitumor activity of DN-NH-1 in vivo, a human tumour xenograft model was established as described previously.⁵⁵ The animal experiments were approved by the Ethics Committee of China Pharmaceutical University (Nanjing, China). The specific pathogen-free grade 6-week-old male BALB/c nude mice were purchased from Changzhou Cavens Experimental Animal Limited Company (Changzhou, China). HepG2 cells (1 × 10⁷ cells) were injected into the right subcutaneous space of mice. When the average tumour volume reached about 80–100 mm³, mice were randomly divided into 4 groups: vehicle-control group, low-dose treatment group (3 mg/kg), middle-dose treatment group (6 mg/kg), and high-dose treatment group (12 mg/kg). All mice were injected intraperitoneally. Subcutaneous tumour volume was measured every 3 days for a total of seven times during treatment and was calculated by the formula: (c × c × d)/2, where c and d represent the smallest and largest diameters of tumours, respectively (Figure 1).

Figure 1. Structures of the representative NSD2 and HDAC inhibitors.

Results

Generation of NSD2 pharmacophore model

The retrieved X-ray crystal structures of NSD2 (PDB ID: 7VLN) co-crystallized with imidazole-5-ylbenzonitrile compounds were utilised to characterise the NSD2 pharmacophore model. PWWP1 is the reader domain of NSD2, which performs the function of preferentially combining with nucleosome containing H3K36me2 and maintaining a proper nuclear localisation, which depends on the conservative aromatic cage composed of Trp236, Tyr233, and Phe266. As illustrated in Figure 2, the pockets of the aromatic cage can accommodate reasonably designed compounds (e.g., the cyclopropyl group of MR837). In this study, four important pharmacophore features were generated (F1-F4): feature F1 corresponds to hydrophobic amino acid residues Trp236 and Leu318, features F2 and F3 are both aromatic centres, corresponding to amino acids Tyr233 and Phe266 that are components of the aromatic pocket, and the F4 feature is a hydrogen bond receptor, forming hydrogen bond interactions with Gln321, Ala270, and Asp269. According to the above-mentioned four features, the NSD2 pharmacophore model was finally established for the subsequent molecular docking.

Figure 2. Details of the NSD2-based pharmacophore model features (F1: hydrophobic feature, F2 and F3: aromatic features, F4: hydrogen-bond acceptor feature).

Validation of the pharmacophore model

A database of 1981 inactive and 19 active molecules for the pharmacophore model validation was used for screening. The Güner–Henry (GH) scoring method assessed the specificity and accuracy of the structure-based model. The value of the enrichment factor (E) in Table 1 is 95, and 18 active hits were identified from 20 molecules, demonstrating that the pharmacophore model is more capable of recognising active molecules. Moreover, a GH score of 0.91 (greater than 0.7) was obtained, indicating an excellent ability to distinguish between active and inactive compounds.

Table 1. The validation of the pharmacophore model by the GH scoring method.

Parameter	Pharmacophore Model
Total molecules in database (D)	2000
Total number of actives in database (A)	19
Total hits (Ht)	20
Active hits (Ha)	18
% Yield of actives [(Ha/Ht) × 100]	90%
% Ratio of actives [(Ha/A) × 100]	95%
Enrichment factor (E) [(Ha × D) / (Ht × A)]	95
False negatives (A − Ha)	1
False positives (Ht − Ha)	2
Goodness of hit score (GH)	0.91

Virtual screening of HDAC2/NSD2 dual-targeting inhibitors

In the present study, the steps of multi-step virtual screening are depicted in Figure 3(A). Firstly, a two-dimensional (2D) small-molecule database containing 35,000 compounds was established, and each compound was replaced with a 3D structure based on the principle of energy minimisation. The above-validated NSD2 pharmacophore model features were used as the screening constraints to obtain 207 hit compounds matching with all pharmacophore features. The reliability of the docking method was verified in NSD2 and HDAC2 before molecular docking. The original ligand was utilised as a template, and the docking ligand was redocked flexibly to the active sites of NSD2 and HDAC2 by the Triangle Matcher placement method. The results revealed that the poses of the original and docking ligands were almost identical (Figure 3(B and C)). Subsequently, these 207 compounds were further screened by docking with the active sites of NSD2 and HDAC2, respectively. The docking-binding free energy can be used to predict the binding affinity between each hit compound and the active site of the target. A lower value indicates a stronger binding affinity. MR837 was selected as a positive control for NSD2 molecular docking with a docking score of −10.51 kcal/mol. The compounds were screened for the active site connected to NSD2, and the docking score of each hit was calculated. The reasonable docking score threshold of < −10.51 kcal/mol was set as the screening criterion (Table 2), and 37 compounds were obtained from 207 compounds. Subsequently, a high-resolution crystal structure of HDAC2 (PDB ID: 4LXZ) was obtained, and based on docking scores below −12 kcal/mol, 19 compounds were achieved from 207 compounds for HDAC2 docking screening. As shown in Figure 3(D), based on docking scores below < −10.51 and −12 kcal/mol, five potential dual-targeting NSD2/HDAC2 candidate compounds (referred to as DT-NH-1–5) were ultimately obtained that met both scoring thresholds. Five screened compounds were finally purchased for subsequent biological experiments (Figure 4).

Figure 3. Identification of hit compounds targeting NSD2 and HDAC2 through a comprehensive strategy of structure-based virtual screening and biological validation. (A) Workflow of the combinatorial screening and biological validation process of NSD2/HDAC2 dual-targeting inhibitors. (B) The docking validation of NSD2 before molecular docking (yellow indicates docking pose, and cyan indicates original pose). (C) The docking validation of HDAC2 before molecular docking (yellow indicates docking pose, and cyan indicates original pose). (D) Identification of five hit compounds with lower binding energies than both NSD2 (-10.51 kcal/mol) and HDAC2 (-12 kcal/mol).

Figure 4. The chemical structures of five hit compounds.

Table 2. The docking scores and enzymatic inhibitory activities of 5 hit compounds.

Compounds	NSD2 Binding free Energy^a (kcal/mol)	HDAC2 Binding free Energy^a (kcal/mol)	NSD2 (IC50, μM)	HDAC2 (IC50, nM)
DT-NH-1	−11.25	−12.29	0.08 ± 0.03	5.24 ± 0.87
DT-NH-2	−10.74	−12.13	0.57 ± 0.08	7.35 ± 1.41
DT-NH-3	−10.69	−12.02	0.64 ± 0.12	11.28 ± 1.26
DT-NH-4	−10.98	−12.17	0.52 ± 0.11	6.77 ± 1.15
DT-NH-5	−11.07	−12.21	0.43 ± 0.06	6.21 ± 1.01
MR837	−10.51	–	17.13 ± 1.95	no inhibition
Vorinostat	–	−12.00	no inhibition	9.49 ± 1.48

Interaction analysis

The potential five compounds were sequentially docked into NSD2 and HDAC2 proteins to assist in studying their binding modes. Figure 5 shows the interaction between five hit compounds and NSD2 proteins. The nitrogen atoms of all the compounds not only formed stable hydrogen bonds with Gln321, Ala270, Asp269, and Glu272, but also provided additional hydrogen bonds with Lys317 by the oxygen atoms at the end of the fatty chain. In addition, each compound exhibited hydrophobic interactions with hydrophobic amino acids such as Tyr233, Trp236, and Phe266 (Figures 5(A, C, E, G, and L)). Furthermore, five potential compounds that matched well with the active pockets of NSD2 were identified (Figures 5(B, D, F, H, and M)). Next, the binding modes of the five compounds to the HDAC2 protein are illustrated in Figures 6(A, C, E, G, and L). Each hit compound contained an N-Hydroxyformamide group, which not only formed important ionic bonds with positively charged zinc ions, but also exhibited hydrogen bond interactions with His145, His146, and Tyr308. In addition, the fatty chains of each compound showed hydrophobic interactions with Phe155 and Phe210. These potent interactions are consistent with the fit results of each compound with the HDAC2 surface active pocket, as displayed in Figures 6(B, D, F, H, and M). Therefore, the docking results of the hit compounds with NSD2 and HDAC2 indicated that DT-NH-1–5 could simultaneously interact with the key residues of the active pockets of both NSD2 and HDAC2, which might lead to the occurrence of dual inhibitory effects on NSD2 and HDAC2.

Figure 5. Predicted docking poses of five hit compounds at the NSD2 active site. Compounds in (A), (C), (E), (G), and (L) are represented by different colours (red for DT-NH-1, purple for DT-NH-2, brown for DT-NH-3, light pink for DT-NH-4, and cyan for DT-NH-5). (B), (D), (F), (H), and (M) are their corresponding surface maps. The coded green bars represent the active amino acid residues of NSD2. Hydrogen bonds are indicated by dashed black lines.

Figure 6. Predicted docking poses of five hit compounds at the HDAC2 active site. Compounds in (A), (C), (E), (G), and (L) are represented by different colours (red for DT-NH-1, purple for DT-NH-2, brown for DT-NH-3, light pink for DT-NH-4, and cyan for DT-NH-5). (B), (D), (F), (H), and (M) are their corresponding surface maps. The coded green bars represent the active amino acid residues of HDAC2. Hydrogen and ionic bonds are indicated by dashed black lines.

Enzymatic inhibitory assay of NSD2 and HDAC2

Five potential dual-targeting NSD2/HDAC2 candidate compounds were obtained through virtual screening and interaction analysis. Their enzymatic inhibitory activities against NSD2 and HDAC2 were measured using homogeneous time-resolved fluorescence (HTRF) technology and fluorescence signal-based methods, respectively. The reported NSD2 inhibitor MR837 and the approved HDAC inhibitor Vorinostat were utilised as positive controls. As shown in Table 2, the half maximal inhibitory concentration (IC₅₀) of the five hit compounds for NSD2 in the range of 0.08–0.64 μM was significantly smaller than that of MR837 (17.13 ± 1.95 μM). The reliability of feature screening based on the pharmacophore model was also confirmed. In addition, apart from DT-NH-3, the other four dual-targeting compounds exhibited superior inhibitory effects on HDAC2 versus Vorinostat. Notably, the enzymatic inhibitory activity of DT-NH-1 was optimal, with 214.1- and 1.8-fold inhibition of NSD2 and HDAC2 proteins compared with the controls MR837 and Vorinostat, respectively.

MD Simulation

To further evaluate the time-dependent binding stability of NSD2-DT-NH-1 and HDAC2-DT-NH-1 complexes, 50 ns MD simulations were carried out. The RMSD generally represents the system stability during MD simulation. The lower RMSD indicated a greater stability. The RMSDs of the NSD2-DT-NH-1 and HDAC2-DT-NH-1 complexes are respectively illustrated in Figures 7(A and B). The RMSDs of the hit in the NSD2-DT-NH-1 and HDAC2-DT-NH-1 complexes are respectively depicted in Figures 7(C and D). It was found that the RMSDs of the two complexes were relatively stable at about 0.38 and 0.26 nm, respectively. The RMSDs of the hit in the two complex systems were also stabilised at about 0.42 and 0.28 nm, respectively. These results indicated the appropriate structural stability of the hit and the two complexes during the MD simulation, and the hit could simultaneously bind stably with NSD2 and HDAC2. Furthermore, RMSF is generally utilised to descript the structural adaptability of each protein residue. The lower RMSF presents less flexibility, which is mainly related to the interactions between protein and ligand. As illustrated in Figures 7(E and F), RMSF values of all residues within NSD2 and HDAC2 were lower than 0.4 nm, suggesting the appropriate stability of protein residues and the promising binding affinity between the two proteins and the DT-NH-1 during the MD simulation. Finally, to determine whether the secondary structure of the NSD2 and HDAC2 was stable during the MD simulation, DSSP analysis was conducted. As depicted in Figures 7(G and H), no noticeable alterations were found in secondary structural elements of NSD2 and HDAC2 proteins, suggesting the proper structural stability of the two proteins. Collectively, these MD simulation results suggested that the DT-NH-1 could simultaneously bind stably to NSD2 and HDAC2 with promising binding stability. Thus, the DT-NH-1 could potentially serve as a potent dual-target inhibitors for NSD2 and HDAC2.

Figure 7. MD simulation of DT-NH-1 in complex with NSD2 and HDAC2. (A), (B) RMSD of NSD2-DT-NH-1 and HDAC2-DT-NH-1 complexes. (C), (D) RMSD of DT-NH-1 in NSD2-DT-NH-1 and HDAC2-DT-NH-1 complexes. (E), (F) RMSF of Cα atoms of NSD2 residues and HDAC2 residues respectively in NSD2-DT-NH-1 and HDAC2-DT-NH-1 complexes. (G), (H) The secondary structural analysis of DT-NH-1 in complex with NSD2 and HDAC2.

In vitro anti-proliferative assay

To evaluate the inhibitory effects of five hit compounds on liver cancer cells, MTT assay was carried out. As shown in Figure 8, all hit compounds exhibited stronger inhibitory effects on HepG2 cells than MR837 and Vorinostat at a concentration of 0.6 μM. Among them, DT-NH-1 was evaluated as the strongest inhibitor with > 90% inhibition of HepG2 cells. Considering the need to evaluate the biosafety of DT-NH-1 and its inhibitory effects on multiple liver cancer cells, the effects of DT-NH-1 on normal human liver cell lines and four liver cancer cell lines (HepG2, HuH-1, SK-HEP-1, and Bel-7405) were further assessed. As illustrated in Figure 9, DT-NH-1 exhibited an excellent biosafety for normal human liver cells (IC₅₀ > 10 μM), while DT-NH-1 showed strong inhibitory effects on four HCC cell lines (IC₅₀ = 0.05–0.15 μM), especially on HepG2 cells (IC₅₀ = 0.05 μM). Importantly, the selectivity index (SI) of DT-NH-1, which is the IC₅₀ value of a compound on normal cells divided by the IC₅₀ value of cancer cells.⁶⁷, ranged from 66.7 to 200, confirming the high selectivity and excellent biosafety of DT-NH-1. The above-mentioned results of in vitro anti-proliferation experiments indicated that DT-NH-1 in this study has promising anti-tumour potential, which lays the foundation for the future dual-targeted anti-tumour drug design. Therefore, we will further investigate the anti-tumour effects of DT-NH-1 on HepG2 xenograft in nude mice.

Figure 8. Inhibitory effects of five hits, MR837 and Vorinostat on HepG2 cells at a concentration of 0.6 μM. The results are representative of three independent experiments and are expressed as mean ± SD.

Figure 9. Inhibitory effects of DT-NH-1 on normal liver cell lines and four liver cancer cell lines. The results are representative of three independent experiments and are expressed as mean ± SD.

In vivo antitumor efficacy of DT-NH-1

The in vivo antitumor efficacy of DT-NH-1 was evaluated based on the inhibitory effects of the target protein and the anti-cell proliferation potency in vitro. A xenograft model of BALB/c nude mice was first successfully established, carrying the HepG2 cell line, and vehicle and different doses of DT-NH-1 were then intraperitoneally injected, respectively. Compared with the vehicle-control group, the tumour growth rate of the DT-NH-1 treated group was significantly reduced in a dose-dependent manner, and DT-NH-1 had the optimal anti-tumour effect at the dose of 12 mg/kg (Figure 10). The above-mentioned findings demonstrated that DT-NH-1 exerted a strong anticancer effect in vivo.

Figure 10. The change in tumour volume with varying concentrations of DT-NH-1. The results are presented as mean ± SD, n = 5. ***p < 0.001 mean significant difference versus the vehicle group.

Conclusions

Genomic changes and epigenetic modifications jointly promote the progression and metastasis of cancer, and immune checkpoint inhibitors have been widely studied in cancer treatment. However, the severe reality of drug resistance in patients and the urgent need to develop new targets have attracted scholars’ attention in the research of epigenetic modifications-related cancer treatment targets. Moreover, due to the strong efficacy and satisfactory safety, designing drugs with dual-targeting combined effects is a promising strategy for treating cancer. The NSD2 and HDAC2 play a crucial role in the cancer cell proliferation and cancer metastasis. In the present study, five novel NSD2/HDAC2 dual-targeting small-molecule compounds were identified using a comprehensive virtual screening scheme. Particularly, enzymatic inhibitory experiments confirmed that compound DT-NH-1 has a strong inhibitory effect on NSD2 (IC₅₀ = 0.08 ± 0.03 μM) and HDAC2 (IC₅₀ = 5.24 ± 0.87 μM). Moreover, DT-NH-1 exhibited a potent antiproliferative activity against various liver cancer cell lines, particularly HepG2 cells, and exhibited a high level of biological safety according to the results of in vitro cell proliferation experiments. Afterwards, the HepG2 cell line, which was the most sensitive to DT-NH-1, was selected to established a xenograft model for in vivo evaluation experiments. The results revealed that DT-NH-1 significantly inhibited tumour growth in mice. In summary, designing small-molecule compounds that simultaneously target NSD2 and HDAC2 has become a new therapeutic strategy for hindering tumour progression. In addition, our laboratory is currently conducting further optimisation research on DT-NH-1.

Ethical statement

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

Author contributions

GB, QZ and XJ contributed to the conception of the study. YW, XJ and MMN performed experiments and acquired data. YW, JC, MMN, and YY performed statistical analysis. XJ and YW contributed significantly to manuscript preparation. All authors had contributed and approved the manuscript.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was financially supported by Guangzhou Science and Technology Plan Projects [2023A03J0755]; “Young Talents Program” of Guangdong Academy of Traditional Chinese Medicine [SZ2022QN09].