Discovering novel P38α inhibitors for the treatment of prostate cancer through virtual screening methods

The incidence of prostate cancer, the second most diagnosed cancer and the fifth leading cause of death due to malignancy in men, continues to increase worldwide [1,2]. Pharmacotherapy including endocrinotherapy and chemotherapy has been the most frequent treatment for advanced prostate cancer [3]. However, a few months or years after pharmacotherapy, the tumor can obtain drug-resistance abilities. Therefore, it is crucial to discover novel targeting drugs for the treatment of prostate cancer.

P38 mitogen-activated protein kinases (MAPK), belonging to the MAPK signaling pathway family, are a conserved serine/threonine kinase signaling pathway which plays a crucial role in tumorigenesis and cancer progression [4]. P38α, one of the subunits of P38 MAPK, is expressed in numerous types of tissues including prostate tissue [5]. P38α could induce the proliferation, migration, invasion and differentiation of cancer while being activated by stress or certain cytokines [6,7]. A number of x-ray structures of P38α MAPK have been published in different functional and pharmacological journals and P38α can bind to many different inhibitors due to its conformation [8]. P38 MAPK inhibitors were used alone or in combination with chemotherapy to suppress tumor progression [9]. However, with the off-target effects of these inhibitors, the use of such inhibitors has been restricted [7,10].

In consequence, discovering novel inhibitors of P38α for the treatment of prostate cancer is necessary. Virtual screening has been a comprehensive and effective method of acquiring some compounds for the discovery of functional drugs for almost 20 years. In recent years, more and more computational approaches and screening methods have successfully identified novel agents to perturb and combat kinase function, besides providing a number of small molecules for cancer therapy [11]. Herein, we selected two databases of small molecules which contain more than 3 million compounds in total and screened them in multiple models to construct a P38α inhibitor-like library. As contrasted to the monotonous approach used in other studies, we employed an integrated virtual screening strategy using a combination of multiple classification models, molecular docking approaches and drug-screening principles to identify novel P38α inhibitors. After virtual screening and auto-docking, the top three compounds (compound 1, 5, 9), which demonstrated more than a 50% inhibition rate of P38α MAPK kinase activity at the 10 μM concentration, were finally selected. These hit compounds indicated that new inhibitor scaffolds could be discovered by utilizing a virtual screening method with the combination of multiple classifications. The top three compounds have also been selected for further in vitro cellular experiments.

Exhilaratingly, all three compounds can inhibit the progression of prostate cancer, especially compound 9. The novel P38α inhibitor can suppress the proliferation of prostate cancer by arresting the cell cycle at the G0/G1 phase and inducing cell apoptosis, besides being able to inhibit the migration ability of prostate cancer.

Materials & methods

Diverse small molecule database & receptor preparation

Two diverse small molecule databases, Chemdiv (Topscience, Shanghai, China) and Enamine (1,425,044 and 1,969,861 compounds, respectively), were filtered to construct the P38α inhibitor-like library. First, all compounds from these two databases were processed in the following manner (including addition of charge, transformation into valid 3D conformers and structure optimization) by using Openbabel software (Openbabel/2.4.0-gcc-4.8.5). Then, the refined database was filtered using drug-like analysis including Lipinski’s rule of five [12]. Lipinski’s rule states that, in general, an orally active drug has no more than one violation of the following criteria: log p ≤ 5, molecular weight between 150 and 500, five or less hydrogen bond donors and ten or less hydrogen bond acceptors and rotatable bonds. In the third step, pan assay interference compounds (PAINS) website was used to screen for false positives (www.cbligand.org/PAINS/) [13].

For the docking simulations, we searched the PDB database (www.rcsb.org/pdb/results/results.do?tabtoshow=Current&qrid=21E708F) – the text search applied for: P38α and TAXONOMY was ‘Homo sapiens’ (human) and the resolution was set between 1.5 and 1.999. In total, 21 outcomes were listed and the last crystallized structure of human P38α MAPK receptor protein (PDB code: 3ZS5) with a resolution factor of 1.60 Å, which contained the structure of known inhibitors, was retrieved and prepared in three steps [14]. First, the native SB2 ligand, ions and crystalline water were removed and BOG and EDO ligands were retained from the crystallized structure of P38α MAPK. Second, the missing hydrogen atoms were added. At last, the protein file was automatically prepared in the PDBQT format. Molecular Graphics Laboratory (MGL Tool) software was applied in the preparation of all the structure parameters of P38α MAPK protein.

Classifier model & P38α MAPK inhibitor-like library construction

Multiple classification models were employed to filter compounds in the Chemdiv and Enamine libraries in order to construct the P38α inhibitor-like library. The activity data of P38α MAPK inhibitors were downloaded from the ChEMBL (www.ebi.ac.uk/chembl/) and Binding DB (www.bindingdb.org/bind/index.jsp) databases. These inhibitors are filtered according to the following classification requirements: only human P38α MAPK inhibition assay data were selected; only P38α MAPK assay data based on enzyme or enzyme regulation were selected; a Naive Bayesian (NB) model based on molecular fingerprints was constructed by the best cut-off value; the models were validated with the Y-scrambling test. Duplicated compounds, compounds with Na⁺ or K⁺, or compounds without detailed assay values (IC₅₀ or Ki) were excluded.

By using these criteria, compounds from small molecule libraries were considered to be ‘inhibitors’ if their reported IC₅₀ or Ki was below 10 μM [15]. The cut-off value appeared to be a starting point for hit-to-lead activity. To estimate the influence of these cut-off values on the performance of the classification models, four threshold values (1, 5, 10 and 20 μM) were used to divide the data into inhibitor and noninhibitor classes.

In the present study, two types of fingerprints were used to construct the classifier model, namely SciTegic extended-connectivity fingerprints (ECFP, FCFP and LCFP) and Daylight-style path-based fingerprints (EPFP, FPFP and LPFP). The first letter of each fingerprint represent different rules of initial assignment of atom identifiers (E, F or L): F represents atom’s functional role code, E represents the Daylight atomic invariants rule, and L represents AlogP atom type code. C or P, in the name of fingerprint was used to represent different type of fingerprint: C represents extended-connectivity fingerprints and P represents path-based fingerprints. These fingerprints are widely used for the prediction of other absorption, distribution, metabolism and excretion (ADME), quantitative structure–activity relationship (QSAR) and quantitative structure–property relationships (QSPR) [16–18]. Using the Discovery Studio molecular simulation package (version 3.5, Accelrys, Inc., CA, USA), 12 fingerprints were successfully calculated at the end of the model construction. The LCFP-12 and EPFP-10 models were first applied for a total of 2,829,942 molecules, and 557,394 compounds were then retained, where all of these compounds screened by the model were capable of hopping new inhibitor scaffolds (Supplementary Table 1) [15,19].

Virtual screening & clustering analysis

AutoDock Vina was employed to screen the refined 557,394 compounds from the Chemdiv and Enamine libraries. During the docking process, semiflexible docking simulations were performed by employing the Lamarckian genetic algorithm, and the receptor was kept rigid, while the ligands were flexible to rotate and explore the most probable binding conformations. All steps were performed in the Tianhe-2 supercomputing platform using the self-built procedure. After the docking-based virtual screening, the compounds in the P38α inhibitor-like library were sorted by docking score, and the top 2000 compounds with docking scores were obtained [20].

Clustering and artificial visual analyses were carried out to exclude molecules that do not have medicinal properties. By analyzing the drug-like properties which included absorption, distribution, metabolism, excretion and toxicity of the molecule, 58 compounds were selected and purchased for further investigation.

In vitro P38α MAPK kinase inhibition assay

The 58 compounds were purchased and purified for the testing of their half maximal inhibitory concentration (IC₅₀) of P38α MAPK kinase (www.carnabio.com/). The inhibition assay was performed by ChemParner (Shanghai, China). P38α MAPK activity was evaluated using purified P38α MAPK (Cat. No. 04-152, Carna). In short, 1 × kinase base buffer (combination of 50 mM HEPES, pH 7.5 and 0.0015% Brij-35) and stop buffer (combination of 100 mM HEPES, pH 7.5, 0.015% Brij-35, 0.2% Coating Reagent and 50 mM EDTA) was prepared for the testing of kinases. Then, 10 μl of the compound and 90 μl of 1 × kinase buffer was transferred to 96-well plates after diluting the compound to 500 μM in DMSO. A total of 100 μl of this compound dilution was transferred into a well for the enzyme-free control and 100 μl of 100% DMSO was added to two empty wells for the compound-free control and enzyme-free control in the same 96-well plate. At last, 5 μl from each well of the 96-well intermediate plate was transferred to a 384-well plate in duplicates.

In the kinase reaction phase, 2.5x enzyme solution (kinase was added into 1 × kinase base buffer) and 2.5x peptide solution (FAM-labeled peptide and ATP was added into 1 × kinase base buffer) was first prepared. Then, the enzyme solution and compound were combined onto the assay plate. Incubation was performed at room temperature for 10 min and the peptide solution was the added to each well. At last, 25 μl of stop buffer was added to stop the reaction after incubation at room temperature for a specified period of time. Conversion data from the Caliper program was collected and copied, and then conversion values were converted to inhibition values by using the following formula: percent of inhibition (max-conversion)/(max-min)*100 (‘max’ represents DMSO control, ‘min’ represents low control). At the end of the assay, the data were added into XLFit excel (version 5.4.0.8) to obtain IC₅₀ values (the equation used: Y = Bottom + (Top-Bottom)/(1 + (IC₅₀/X)∧HillSlope)).

Cell viability assay

Cell proliferation of RWPE1, LNCap and DU145 was evaluated using Cell Counting Kit-8 (CCK-8) (Cat. No. C0039; Beyotime Institute of Biotechnology, Shanghai, China). A total of 5 × 10³ cells/well were seeded into 96-well plates and cultured with compound 1, 6 and 9 for 24 h at 37°C. The concentration of compound 1, 6 and 9 ranged from 0 to 1 × 10³ μM. Each cell line was seeded into three wells at each concentration of these compounds. Cells were then incubated with 20 μl CCK-8 for 2 h at 37°C. The 450 nm wavelength absorbance was measured with a spectrophotometer (MultiskanTM MK3, Thermo Fisher Scientific, Inc., MA, USA). The IC₅₀ values of selected compounds were presented as mean ± standard deviation of three independent experiments.

Apoptosis

Apoptosis was analyzed using an Annexin V-allophycocyanin (APC)/7-aminoactinomyocin (7-AAD) Apoptosis Detection kit (BD Pharmingen; BD Biosciences, NJ, USA). LNCap and DU145 cells were incubated with compound 9 for 24 h at 37°C. Adherent cells were digested and the supernatant was collected from the medium after these cells were cultured. The cells were centrifuged for 5 min at 1000 rpm. The cultured cells were incubated with Annexin V-APC/7AAD for 15 min at room temperature and then were collected and suspended in 1 × Binding Buffer (Annexin V-APC/7AAD kit, Cat. No. 4224750; Multisciences, Zhejiang, China). Apoptotic analyses were conducted on a BD FACSCalibur flow cytometer (BD Biosciences), and a minimum of 1 × 10⁶ cell counts were used for each experimental sample. All the experiments were repeated [21].

Cell cycle analysis

Adherent LNCap and DU145 cells were digested and the supernatant was removed from the medium after treatment with compound 9 for 24 h at 37°C. Cells were centrifuged for 5 min at 1000 rpm. LNCaP and DU145 cells were washed twice with precooled phosphate-buffered saline and treated with 70% precooling ethanol fixation overnight at 4°C, 100 μg/ml RNase A treatment and 5 μg/ml propidium iodide staining at 37°C for 20 min, and were then detected under flow cytometry by FACSCalibur (BD Biosciences and NovoCyte, NJ, USA). Cell populations at the G0/G1, S and G2/M phases were quantified by Flowjo software which excluded the calculation of cell debris and fixation artifacts [22].

Transwell migration & invasion assay

The CytoSelect Cell Migration and Invasion kit (Cell Biolabs, Inc., CA, USA) was used according to the manufacturer’s protocol. After being cultured with compound 9 for 24 h at 37°C, LNCap and DU145 cells were re-suspended in a serum-free medium (HyClone, Cat. No. sh30809.01B) to a density of 2 × 10⁵/ml for DU145 and a density of 3 × 10⁵/ml for LNCaP. The suspended cells (200 μl) were seeded in the upper chambers, whereas the lower chambers contained 10% fetal bovine serum as a chemoattractant. Following incubation at 37°C for 24 h (LNCaP cell line was incubated at 37°C for 48 h), the cells that had migrated through the membrane were fixed with 4% paraformaldehyde (YJ-ZC-003, YongJin, China) for 15 min, then stained with 0.1% crystal violet (BS234b, Biosharp, China) at room temperature for 15 min and quantified via the counting of nine independent symmetrical visual fields under an ordinary optical microscope (Nikon, Tokyo, Japan). Data were presented as mean ± standard deviation of three independent experiments.

Results

Binding domain analysis

The binding site at 21.723, 35.246, 15.333 Å exhibited the potential to inhibit kinase activities of P38α MAPK [23,24]. The representative structure of the electrostatic potential of Inhibitor P38α MAPK complex is shown in Figure 1A. The green molecule is a hypothetical inhibitor surrounded by the basic lobes of the N and C domains at the edge of a slot at the top of the molecule. Therefore, we chose this binding site as the active docking site and the grid box was set as 40 × 40 × 40 Å in the x, y, and z directions. The binding of this site to the active site of P38α is shown in Figure 1B.

The three classes of small molecule protein kinase inhibitors have been labeled as types I, II and III. Type I inhibitor is ‘a small molecule that binds to the active conformation of a kinase in the ATP pocket’, type II is ‘a small molecule that binds to an inactive (usually DFG-OUT) conformation of a kinase’ and type III is ‘a non-ATP competitive inhibitor’ or an allosteric inhibitor [25]. The conformation of the activation segment (and in particular, the DFG motif) is important in the structural studies of these kinases. Two distinct conformations can be differentiated: the DFG-in conformation and the DFG-out conformation. In the DFG-in conformation, phenylalanine is located in the back pocket and the first aspartate is pointed toward the ATP pocket. The DFG-out conformation is flipped by 180° with respect to the active conformation, phenylalanine is in the front pocket and the aspartate is pointed toward the back pocket [26,27]. The conformation of inhibitor-P38α MAPK complexes was markedly different as compared with the DFG-in conformation, and it has shown an uncommon DFG-out conformation. Phe169 is positioned outside the DFG pocket but points away from the hinge. This conformation is quite rare; it has been observed only twice and only in P38 [14]. Therefore, in this structure, the DFG-motif has a ‘DFG-out’ conformation facilitating aromatic stacking interactions between the imidazole core of the inhibitor and the phenyl side chain of Phe169 of the DFG sequence [14].

Virtual screening for P38α MAPK inhibitor

Virtual screening is gaining an increasingly important influence in modern drug discovery [11]. The flowchart of the present virtual screening is shown in Figure 1C. First, all compounds from the Chemdiv and Enamine libraries (1,425,044 and 1,969,861, respectively) were preliminary screened by drug-like analysis and PAINS-removal. After the primary screening, the drug-like library which included 2,829,942 small molecules from Chemdiv and Enamine library (995,581 and 1,834,361, respectively) were further analyzed. Then, a series of multiple classification models were developed to construct the P38α MAPK inhibitor-like library. For example, we used the two best NB models (LCFP-10+ EPFP-12) for the Y-scrambling test to validate that the models were not constructed by accident. In the Y-scrambling testing, we randomly scrambled the molecules’ labels for the training and test sets and then we used the scrambled data to reconstruct the NB models which was applied for a total of 2,829,942 molecules, which resulted in a reserve of 557,394 compounds. These 557,394 compounds which made up the P38α MAPK inhibitor-like library were first docked and scored by AutoDock Vina on the Tianhe-2 supercomputing platform.

After this VS docking procedure, the top 2000 compounds with docking scores were obtained for clustering and visual analyses. We then analyzed the drug-like properties (absorption, distribution, metabolism, excretion and toxicity) and checked whether they had interactions with the ATP binding pocket of P38α MAPK kinase, including hydrogen bond interactions with LYS-53, and π–π stacking interaction with TYR-35. Here, we obtained 58 final hit compounds (Supplementary Table 2) and selected them for further bioassay. This step has ensured that the selected compounds not only had a higher docking score but also a rational binding mode.

In vitro P38α MAPK kinase assay

The P38α MAPK kinase inhibitory activities of the 58 final virtual hits were determined using in vitro activity assays. P38α protein was used as a specific substrate. All the inhibitive activities of candidates were measured at a concentration of 10 μM. If the inhibition of P38α MAPK kinase activity of the compounds was greater than 20%, it would be retested to exclude false positives. The compounds with more than 20% inhibition activity are shown in Table 1 and the chemical structures of these molecules are shown in Figure 2. Among the 58 tested compounds, 6 out of 13 compounds (compound 1, 5, 6, 9, 42 and 47) demonstrated more than 20% inhibition of P38α MAPK kinase activity, and were further investigated to determine their IC₅₀ values (Table 1). Dose-response curves for P38α MAPK kinase inhibition by active compounds are shown in Supplementary Figure 2. In addition, original docking ranks and AutoDock docking scores are listed in Supplementary Tables 3 & 4 .

As shown in Table 1, three small molecules (compound 1, 5 and 9) exhibited IC₅₀ values of less than 10 μM. Compound 9 with an IC₅₀ value of 2.78 μM prevailed as the most potent inhibitor against P38α MAPK. Other promising compounds were compound 1 and 5, which showed IC₅₀ values of 4.21 and 6.34, respectively.

Structural novelty & binding mode analysis for potential candidates

The chemical structures of 13 active molecules are shown in Figure 2. To evaluate the novelty of these hits with respect to known P38α MAPK kinase inhibitors, we checked top hits from the ChEMBL website (www.ebi.ac.uk/chembl/index.php/target/results/ids/true). Our screened top hits were not present in the reported activity data of P38α MAPK inhibitors. All these results suggested that these P38α MAPK inhibitors discovered in this study are structurally novel.

The top three compounds (1, 5, 9) were chosen to have their binding modes analyzed. As shown in Figure 3A, compound 1 is surrounded by amino acids including VAL-30, THR-35, VAL-38, ALA-51, GLU-71, THR-106, MET-109, PHE-169 and GLY-170. Compound 1 could form hydrogen bonds with LYS-53 and π–π interaction with TYR-35. The other two small molecules were analyzed in the same process. Compound 5 (Figure 3B) is surrounded by amino acids including VAL-30, THR-35, VAL-38, ALA-51, GLU-71, LEU-75, LEU-86, LEU-104, VAL-105, THR-106, MET-109, PHE-169, GLY-170, and LEU-171. Some amino acids including VAL-30, THR-35, VAL-38, ALA-51, LEU-55, SER-56, THR-68, GLU-71, LEU-75, LEU-104, THR-106, PHE-169, and GLY-170 surrounded compound 9 (Figure 3C). In addition, compound 5 and 9 could also form hydrogen bonds with LYS-53 and π–π interaction with TYR-35.

Novel inhibitors of P38α MAPK inhibit the cell function of prostate cancer

The three compounds with the most potential (1, 5 and 9) were selected for in vitro antiproliferative assays using two prostate cancer cell lines (LNCaP & DU145) and one normal prostate cell line (RWPE-1). All of the selected compounds could suppress the proliferation of prostate cancer. However, compound 1 and 5 exhibited higher IC₅₀ values for prostate cancer cell lines compared with normal prostate cell line (Supplementary Figure 1). Only compound 9 inhibited the proliferation of prostate cancer without suppressing normal prostate cells in low concentrations. Compound 9 showed inhibitory effects on the proliferation of LNCaP, DU145 and RWPE-1 in a dose-dependent manner, with IC₅₀ values of 125.8, 280.4 and 507.5 μM, respectively (Figure 4). According to the results of the antiproliferative assay, compound 9 is shown to be the most promising inhibitor.

To evaluate how compound 9 suppressed the proliferation of prostate cancer, we analyzed the apoptotic cells of LNCaP and DU145 after treatment with compound 9. Apoptosis was analyzed using double-staining with Annexin V-allophycocyanin (APC)/7-aminoactinomyocin (7-AAD) and detected by flow cytometry. The percentage of apoptotic LNCaP and DU145 cells were 7.6 and 7.98% at the concentrations of IC₅₀ value, respectively (Figure 5A & B). These results demonstrated that compound 9 can induce the apoptosis of prostate cancer cells to inhibit the proliferation of prostate cancer.

To further elucidate the mechanism of cell growth suppression of compound 9, we examined the effects of compound 9 on cell cycle distribution in LNCaP and DU145 cells using flow cytometry. As shown in Figure 5C & D, the G1/G0-phase was increased compared with controls. These results showed that compound 9 inhibited cell growth by inducing arrest of the cell cycle at the G1/G0-phase.

Due to the extensive role of the P38α MAPK signaling pathway, in addition to researching the effects of compound 9 on the proliferation of prostate cancer, the impact on the migration ability was also analyzed. As shown in Figure 5E & F, the number of migrated cells of DU145 and LNCaP decreased significantly after the treatment of compound 9, indicating that compound 9 could inhibit the migration of prostate cancer by inhibiting the P38α MAPK signaling pathway.

Discussion

The P38 MAPK pathway is an important member of the MAPK family, along with c-Jun N-terminal kinases and the extracellular signal-regulated kinases [4]. P38α is one of four crucial members of the P38 MAPK family [5]. In addition to its effects on proliferation and differentiation, P38α can also affect metastasis of cancer [7]. In previous studies, it was found that the P38α signaling pathway can both suppress and promote the progression of cancer. Different functions of P38α exists depending on the different stages of the tumor. P38α MAPK can activate P53 and enhance apoptosis induced by P53 to suppress the genesis of malignancy. However, P38α MAPK was activated for facilitating the growth and metastasis of cancer [9,28]. Currently, some inhibitors of P38α MAPK have been tested in clinical trials, but most of them failed because of their side effects [10,28]. The main intent of this present study is to screen for small molecules with a variety of virtual screening methods, in order to find some specific inhibitors of P38α MAPK.

In the beginning, a series of novel compounds were selected to create a P38α inhibitor library. After virtual screening and multiple classification model research, we found that the IC₅₀ value of compound 1, 5 and 9 was less than 10 μM and their inhibitory effects against prostate cancer were detected by cell experiments. Compound 9 showed the best inhibitory effect and specificity while all the top three compounds could suppress the proliferation of prostate cancer. At the same time, we also found that compound 9 had significantly better inhibitory effects against cancer than normal tissue, which may be due to the different proteins and pathways activated by P38α MAPK in tumor and normal tissue [7,28]. The results indicated that compound 9 could specifically suppress the proliferation of prostate cancer.

Previous studies have shown that P38α MAPK influenced the proliferation of mammalian cells by mainly regulating cell apoptosis [9,29]. P38α MAPK can inhibit the apoptosis of cancer, especially apoptosis induced by TGF-β. Moreover, some inhibitors could promote the apoptosis of tumor cells by affecting Bax and FOXA proteins, thereby inhibiting the progress of tumors. In our study, compound 9 could also induce apoptosis of prostate cancer cells. We also found that compound 9 exhibited the tendency to arrest the cell cycle at phase G1/G0, however its effect on cell cycle arrest was not as strong as that on apoptosis induction. Therefore, we concluded that compound 9 mainly inhibited proliferation by inducing apoptosis of prostate cancer cells. This also established the specificity of the inhibitors for P38α MAPK signaling pathway.

The survival rate drops dramatically when prostate cancer progresses from localized to metastatic cancer. Therefore, inhibiting or slowing down the progression of metastatic prostate cancer is a problem to be solved by new therapeutic drugs. Several studies have shown that P38α MAPK could promote the metastasis of cancer by inducing epithelial–mesenchymal transition [30,31]. P38α MAPK can also activate hypoxia-inducible factor-1α and remodel the actin cytoskeleton to promote cancer metastasis [28]. Hence, we further investigated the inhibitory ability of compound 9. The result showed that after the treatment of prostate cancer with compound 9, the migration ability of prostate cancer was significantly reduced.

From the research results, the novel inhibitors we obtained from screening multiple classifications could be utilized for further investigation of anticancer drugs targeting P38α MAPK. The top three hits (compound 1, 5 and 9) showed some inhibitory activity at the cellular assay. Further studies were considered on the structure–activity relationship of compound 1, 5 and 9. In the future, we intend to improve the specificity and targeting abilities of the novel inhibitors by optimizing their structure.

Conclusion

The potent compounds discovered in this study through virtual screening methods possess anticancer function by targeting the P38α MAPK signaling pathway. They can inhibit the proliferation of prostate cancer by inducing apoptosis and cell cycle arrest at G0/G1. Moreover, they can suppress the metastasis of cancer cells. All of them are worthy of further investigation.

Future perspective

Via virtual screening, three compounds (compound 1, 5, 9) with kinase IC₅₀ of below 10 μM were discovered. As we research these drugs, they may serve as effective drugs for treating prostate cancer or other tumors in the future.