Cytotoxic lanostane-type triterpenoids from the fruiting bodies of Ganoderma lucidum and their structure–activity relationships

We conducted a study of Ganoderma lucidum metabolites and isolated 35 lanostane-type triterpenoids, including 5 new ganoderols (1-5). By spectroscopy, we compared the structures of these compounds with known related compounds in this group. All of the isolated compounds were assayed for their effect against the human breast carcinoma cell line MDA-MB-231 and hepatocellular carcinoma cell line HepG2. Corresponding three-dimensional quantitative structure–activity relationship (3D-QSAR) models were built and analyzed using Discovery Studio. These results provide further evidence for anti-cancer constituents within Ganoderma lucidum, and may provide a theoretical foundation for designing novel therapeutic compounds.

In particular, G. lucidum in cancer research has become more prominent over the recent decades. Triterpenoids and polysaccharides are believed to form the pharmacodyamic material basis of the demonstrated anticancer effects. Lanostane-type triterpenoids are typical constituents of G. lucidum. Since the first triterpenoids (ganoderic acid A) were reported by Kubota in 1982, over 150 compounds have been isolated and reported in G. lucidum [17], with the number continually increasing. In order to search for bioactive anti-tumor metabolites, we launched a systematic study of the chemical constituents extracted from G. lucidum and other members of medicinal mushrooms [18][19][20][21][22]. In the present study, 35 triterpenoids were isolated, including 5 novel compounds. We then performed structural elucidation and cytotoxic assays using these compounds, and built 3D-QSAR models to predict anti-cancer activity.
Compound 2 was isolated as a white amorphous powder with optical rotation of +22.6 (c 0.20; CHCl 3 ). The molecular formula of compound 2 was found to be C 31 (Table 2), combined with the DEPT-135 data, showed that compound 2 had 31 carbon signals consisting of eight methyls, ten methylenes, six methines and seven quaternary carbons. Comparison of the NMR data of compound 2 with those of ganoderiol D (6) [23] indicated that they were closely related to their structures, except of two keto carbonyl group at C-3 and C-7 in ganoderiol D being replaced by a hydroxyl and a methoxyl groups in compound 2, respectively. The difference was confirmed by the significant change of the chemical shift value for C-3 and C-7 from δ C 214.6 and 198.1 in ganoderiol D to δ C 78.5 and 76.1 in compound 2, which was consistent with its molecular formula. The linkage position of the hydroxyl at C-3 was further supported by significant HMBC correlations from δ H 3.25 (H-3) to δ C 35.0 (C-1) and 48.2 (C-5), from δ H 0.83 (H-28) and 1.06 (H-29) to δ C 78.5 (C-3) ( Figure 3). In addition, the HMBC correlations between δ H 4.12 (H-7) to δ C 139.5 (C-9) and δ C 49.7 (C-14), δ H 1.65 (H-5) to δ C 76.1 (C-7), and between δ H 3.38 (H-31) to δ C 76.1 (C-7) indicated the presence of a methoxyl groups at C-7. The configuration of compound 2 was determined by analyzing the NOESY spectrum ( Figure 3). Key ROESY correlations were observed between H-3 and H-5, indicating that H-3 was on the α-orientation same as H-5. H-7 resonated as a broad singlet, indicating the presence of β-orientation. On the basis of the above evidence, the structure of compound 2 was identified as 3β, 24S, 25R, 26-tetradroxy-7α-methoxy-8-ene-lanost-ol.
Compound 3 was isolated as a white amorphous powder, [α] 25 D = 24.8 (c 0.20, CHCl 3 ). The molecular formula of compound 3 was found to be C 31 H 50 O 4 on the basis of a molecular ion peak at an m/z value of 485.3642 [M -H] in the HR-ESI-MS. The 1 H and 13 C NMR spectra displayed were similar to those of compounds 1 and 16. The significant difference was the presence of a methoxyl group connecting to C-12 in compound 3 and a hydroxymethyl at C-26 in compound 16 was deoxidated to a methyl in compound 3. Location of this methoxyl (C-12) was assigned on the basis of the HMBC correlations from δ H 0.62 (H-18) to δ C 79.2 (C-12) and 52.5 (C-17) and the correlations from δ H 4.37 (H-12) to δ C 149.9 (C-9), 52.9 (C-14) and 52.5 (C-17). The relative configurations of H-7 and H-12 were assigned as α-and β-orientation, respectively, on the basis of ROESY correlations of H-12 with H3-18 and H3-19 (δ H 1.32, s) ( Figure 3). Therefore, the structure of compound 3 was elucidated as 12αmethoxy-ganodermanondiol.
The molecular formula of compound 4 was established as C 30 (Tables 1 and 2) between compound 5 and compound 16 indicated that compound 5 was a hydroxylated derivative of compound 16. The hydroxyl moiety was deduced from signals due to one    (Figure 3). Thus, compound 5 was identified as 15α-hydroxy-ganodermanontriol. It has been reported that triterpenoids possess cytotoxic activity on human cancer cells. We analyzed the effect of the isolated triterpenoids on human breast cancer cells MDA-MB-231 and hepatocellular cancer cells HepG2, and the results were summarized in Table 3. Among the compounds examined, compounds 1-3 were highly cytotoxic in both types of cancers.  (Table 4).
3D-QSAR was then used to investigate the structure-activity relationship for inhibiting human breast cancer cells MDA-MB-231. Illustrated in Table 5, the training and test set of the 17 compounds (1-9, 17, 19-24 and 28) with accurate IC 50 ranging from 21.2 to 163.5 μM was randomly selected for correlation analysis in due proportion that ratio of training set was 0.765, and ratio of test set was 0.235 by the Diverse Molecules method of Discovery Studio 3.1. The calculated pIC 50 values ranged from 3.78 to 4.71. The correlation coefficient (r 2 ) between the observed and predicted activity of the training set was found to be 0.968, whereas that of the test set was found to be 0.317, which proved that this QSAR model was acceptable. The predicted pIC 50 values and residual errors of the 17 compounds analyzed using this QSAR model were listed in Table 5. A plot of the observed pIC 50 versus the predicted data is provided in Figure 4, in which the plot of the actual IC 50 versus the predicted values indicated that this model was reliable in forecasting activity for G. lucidum triterpenoids. Moreover, the molecules aligned with the iso-surfaces of the 3D-QSAR model coefficients on van der Waals grids (Figure 5a) and electrostatic potential grids (Figure 5b). It was widely accepted that a better inhibitor based on the 3D-QSAR model should have strong van der Waals attraction in the green areas and a polar group in the blue electrostatic potential areas (which were dominant close to the skeleton).
According to the modeling result provided in Figure 5a-5b, introducing slight bulk and lowly negative charged groups at C-25 or C-26 may elevate the activity of compound 1. Meanwhile, introducing slightly bulk and low positive charged substitutes at C-7 and C-15 may increase the activity of compound 1. Inversely, introducing bulk substitutes at C-1, C-2, C-3, C-22 and C-24 may decrease the activity of compound 1. However, replacing C-1, C-2 and C-3 with negative charged moieties may increase the activity of compound 1, while replacing C-19, C-23 and C-24 with low positive moieties may raise the activity of compound 1. For compounds 9, 21 and 23, the introduction of methyl group of steric hindrance at C-2 and negative O atom such as carbonyl group at C-7 to replace neutral H atom may decline their activity from the IC 50 of 21.2 μM for compound 3 to about 160 μM for compounds 9, 21 and 23 (approximately 8 folds). www.impactjournals.com/oncotarget  As depicted in Table 3, compound 3 showed the highest cytotoxicity, which suggests that compound 3 may exhibit the most potent affinity for its target. Since experimentally identifying and validating a target for a biological agent is time-consuming and costly, we used Pharmaceutical Target Seeker (PTS) [40] to predict potential targets of compound 3 and found that its most possible target is tumor necrosis factor α (TNF-α; PDB code: 2AZ5). To gain better understanding on the potency of the compound, we analyzed the interaction of compound 3 with TNF-α. The molecular docking was performed by inserting the compound into the binding site of TNF-α. All docking runs were applied by Discovery Studio. The binding interaction energy (-142.896 ± 50.365 kJ/mol) was predicted. Figure 5c-5e showed the binding mode of compound 3 interacting with 2AZ5 protein and the docking results revealed that the amino acid Tyr119 located in the binding pocket of the protein played vital role in the conformation with compound 3, which were stabilized by one hydrogen bond. The hydrogen bond was formed relating to Tyr119, which connected to hydrogen atom of hydroxyl of compound 3 part with 2.8 °A. This molecular docking model suggests that compounds 3 may target TNF-α. TNF-α is an extraordinarily pleiotropic cytokine with a central role in immune homeostasis, inflammation, and host defense. TNF-α is a double-dealer. On one hand, TNF-α functions as an endogenous tumor promoter, because TNF-α stimulates cancer cell proliferation, invasion and metastasis. It also induces tumor angiogenesis. On the other hand, TNF-α functions as a cancer killer. Modulation of the activity of TNF-α will offer possibilities for cancer therapy. Since Ganoderma lucidum plays a central role in immune homeostasis, compound 3 may be developed as an agent for cancer immunotherapy.
a The activity was shown as IC 50 value, which was the concentration (μM) of tested compound that resulted in 50% inhibition of cell growth. Results were expressed as the mean value of triplicate data points.    The chemical structures of the new compounds were elucidated by spectroscopy. All of the compounds were assayed for their cytotoxic activity against the human breast carcinoma cell line MDA-MB-231 and hepatocellular carcinoma cell line HepG2, and the structure-activity relationships were revealed by 3D-QASR. Compound 3 showed the highest cytotoxic activity and it may target TNF-α. Our work may provide a guideline to design and optimize more effective inhibitors for human breast carcinoma based on the triterpenoids from Ganoderma lucidum. Our next task is to chemically synthesize this compound, confirm its activity, and explore its anticancer mechanism, and develop it as an agent for cancer immunotherapy.

Fungi material
The quality of Ganoderma lucidum fruit body, collected in Dabie Moutain, Anhui, China, was inspected and analyzed by Yuewei Edible Fungi Technology Co. Ltd., Guangzhou, China. The voucher specimen (No. GL20151010) was deposited in State Key Laboratory of Applied Microbiology Southern China, Guangdong Institute of Microbiology.

Extraction and isolation
The air-dried and powdered fruit bodies of Ganoderma lucidum (6 kg) were extracted with ethyl acetate (48 L × 2, v/v). The crude extract (130 g) was then incubated in chloroform (3 L × 3) to generate the total triterpenoids fraction (78 g). An aliquot of the chloroform extract (70 g) was applied to a silica gel column (200-300 mesh) eluted successively with CHCl 3

Viability and cell death assay
The cell culturing conditions, viability, and cell death assay was performed as described [41][42][43]. In brief, human breast carcinoma cell line MDA-MB-231 and hepatocellular carcinoma cell line HepG2 were used in the study. The cells were cultured in DMEM supplemented with 10%FBS, 100 U/mL penicillin/ streptomycin at 37°C, 5%CO 2 in an incubator (SANYO, MCO-18AIC). Cells (1 x 10 5 cells/mL) were seeded in 24-well plates (500 μL/well). Four hours after inoculation, Compounds 1-35 were added individually into the cultured cells at different concentrations. After 48 h incubation, the cells were detached by trypsin, collected and analyzed by trypan blue staining for cell viability. Each experiment was repeated three times. Each treatment was performed with three replicates. Cell inhibition rate (IR) was expressed as follows: IR (%) = (total cell number -living cell number)/ total cell number×100% IR and the corresponding concentrations of the compounds were inputed into SPSS and the Probit analysis was uesd for IC 50 calculation. Data were expressed as mean ± SD (standard deviation).

Cell survival assay
The survival assay was performed as described [44,45]. MDA-MB-231 cells (5 x 10 4 cells/well) were seeded in 24-well plates with 500 μL DMEM containing 10% FBS. 24-hour after cell inoculation, cells were washed gently with PBS twice and cultured in serum-free DMEM. The compounds 1-4, 8, 17, 19, 20, 22 and 24 were added to the wells at different concentrations. The medium used to dissolve compounds served as a control. After 24 h treatment, the cells were harvested, counted using trypan blue staining. The experiments of each compound were repeated three times. Each treatment contained three replicates. Cell inhibition rate (IR) was expressed as follows: IR (%) = (total cell number -living cell number)/ total cell number×100% IR and the corresponding concentrations of the compounds were inputed into SPSS and the Probit analysis was uesd for IC 50 calculation. Data were expressed as mean ± SD.

QSAR model
A subset of 13 compounds was utilized as a training set for QSAR modeling using the procedure as described [46,47]. Because it is essential to assess the predictive power of the resulting QSAR models on an external set of inhibitors, the remaining 4 molecules (ca. 25 % of the dataset) were employed as an external test subset for validating the QSAR models by the Diverse Molecules protocol in the Discovery Studio 3.1. The selected test compounds were compounds 6, 9, 19 and 24.
The inhibitory effect of the compounds observed (IC 50 ; μM) was changed to a negative logarithmic scale (pIC 50 ; μM), and then used for subsequent QSAR analyses as a response variable.
In the Discovery Studio 3.1, the CHARMm force field was used and the electrostatic potential and the van der Waals potential were treated as separate terms. A +1e point charge was used as the electrostatic potential probe and the distance dependent dielectric constant was used to mimic the solvation effect. For the van der Waals potential, a carbon atom with a radius of 1.73 °A was used as a probe. The truncation for both steric and the electrostatic energies was set to 30 kcal/mol. Standard parameters were implemented in the Discovery Studio 3.1. A partial least-squares (PLS) model was built using energy grids as descriptors. QSAR models were built using the 3D-QSAR protocol of Discovery Studio 3.1.

Target seeking and docking
For the molecular docking model, the threedimensional X-ray structure of searched target acquired from the RCSB protein data bank (http://www.pdb.org) was selected as the template. All bound water and ligands were eliminated from the protein and the polar hydrogen was added to the proteins. The docking procedure was carried out using CDOCKER protocol for receptor-ligand interaction section of Discovery Studio [48]. Initially, the three-dimensional structures of the compound in this paper were built and energetically minimized by using MMFF94 with 5000 iterations and minimum RMS gradient of 0.10. Molecular docking of all compounds was then performed using the Discovery Studio as implemented through the graphical user interface CDOCKER protocol. CDOCKER is an implementation of a CHARMm based molecular docking tool using a rigid receptor.