Oridonin Targets Multiple Drug-Resistant Tumor Cells as Determined by in Silico and in Vitro Analyses

Drug resistance is one of the main reasons of chemotherapy failure. Therefore, overcoming drug resistance is an invaluable approach to identify novel anticancer drugs that have the potential to bypass or overcome resistance to established drugs and to substantially increase life span of cancer patients for effective chemotherapy. Oridonin is a cytotoxic diterpenoid isolated from Rabdosia rubescens with in vivo anticancer activity. In the present study, we evaluated the cytotoxicity of oridonin toward a panel of drug-resistant cancer cells overexpressing ABCB1, ABCG2, or ΔEGFR or with a knockout deletion of TP53. Interestingly, oridonin revealed lower degree of resistance than the control drug, doxorubicin. Molecular docking analyses pointed out that oridonin can interact with Akt/EGFR pathway proteins with comparable binding energies and similar docking poses as the known inhibitors. Molecular dynamics results validated the stable conformation of oridonin docking pose on Akt kinase domain. Western blot experiments clearly revealed dose-dependent downregulation of Akt and STAT3. Pharmacogenomics analyses pointed to a mRNA signature that predicted sensitivity and resistance to oridonin. In conclusion, oridonin bypasses major drug resistance mechanisms and targets Akt pathway and might be effective toward drug refractory tumors. The identification of oridonin-specific gene expressions may be useful for the development of personalized treatment approaches.


INTRODUCTION
Chemotherapy is a mainstay of cancer treatment in addition to surgery, radiotherapy, and antibody-based immunotherapy. Conventional chemotherapy fails for many cancer patients due to various factors, drug resistance being one of the main reason together with severe side effects. Therefore, drug research constantly attempts to improve treatment results by the preclinical development of new drugs and the optimization of therapy regimens in the clinic. Natural products always played an important role in cancer pharmacology (Newman and Cragg, 2007), they are not only well-established cytotoxic anticancer drugs (e.g., anthracyclines, Vinca alkaloids, taxanes, camptothecins, etc.), but also valuable lead compounds for the development of novel targeted chemotherapy approaches (Walkinshaw and Yang, 2008;Gallorini et al., 2012;Garcia-Carbonero et al., 2013). Natural products can exert synergistic interaction with other natural or synthetic drugs (Efferth, 2017;Wagner and Efferth, 2017;Zacchino et al., 2017a,b), they can overcome drug resistance (Guo et al., 2016;Reis et al., 2016;Teng et al., 2016;Zuo et al., 2016), reduce side effects of chemotherapy and stimulate the immune system (Lacaille-Dubois and Schad et al., 2017).
Abnormal activation of signal transduction pathways may lead to carcinogenesis, invasion, and metastasis of tumors (Leber and Efferth, 2009;Spano et al., 2012). Signaling pathways related to the epidermal growth factor receptor (EGFR) such as EGFR/PI3K/AKT-mTOR pathway command a unique position in cancer biology (Efferth, 2012). Targeting those proteins led to the development of cancer therapeutics such as erlotinib and gefitinib (EGFR inhibitors), LY294002 (PI3K inhibitor), peritosine (Akt inhibitor), rapamycin and sirolimus (mTOR inhibitors), and many others.
Amplification of the EGFR gene (with a frequency of ∼50% in glioblastoma multiforme-GBM) (Furnari et al., 2007) is often associated with a tumor-specific mutation encoding a truncated form of the receptor, which lacks the extracellular binding domain, known as EGFR (also named de2-7EGFR or EGFRvIII) leading to ligand-independent, constitutive tyrosine kinase activity. Expression of EGFR is connected with glioma cell migration, tumor growth, invasion, survival, and resistance to treatment, and correlates with decreased overall survival in GBM patients (Heimberger et al., 2005;Liu et al., 2010). Drug resistance mediated by EGFR is not restricted to established anticancer drugs but also occurs toward other cytotoxic compounds of natural origin. Hence, EGFR-mediated resistance may represent a general type of cellular defense mechanisms toward a broad range of toxic xenobiotics (Kadioglu et al., 2015).
A well-known tumor suppressor gene, TP53 is one of the main guardian of normal cell proliferation by preventing cells with DNA damage to proliferate. Mutations or deletions in the TP53 gene are observed in approximately 50% of human cancers, leading to impaired tumor suppressor function (Wang et al., 2017). Proliferation of cells with DNA damage rises the risk of transferring mutations to the next generation upon loss of p53 functionality; therefore, deregulation of p53 often leads to tumor formation (Khoury and Domling, 2012). Abnormal p53 status is also linked with drug resistance and chemotherapy failure (Muller and Vousden, 2013).
ATP-binding cassette (ABC) transporters play crucial role to regulate absorption, distribution, metabolism, and excretion in normal tissues (Natarajan et al., 2012). Overexpression of certain ABC transporters such as ABCG2/BCRP and ABCB1/Pgp in tumor cells is linked with resistance to chemotherapy. P-glycoprotein (P-gp) encoded by the ABCB1/MDR1 gene is an important mechanism of MDR and is upregulated in many clinically resistant and refractory tumors (Kuete et al., 2015a). Overexpression of P-gp is causatively linked to accelerated efflux of chemotherapeutic agents (Kadioglu et al., 2016b) such as doxorubicin, daunorubicin, vincristine, etoposide, colchicine, camptothecins, and methotrexate (Dean, 2009). For instance, P-gp-overexpressing leukemia cells involve doxorubicin resistance compared to the sensitive subline (Kadioglu et al., 2016a). BCRP is involved in the efflux of mitoxantrone, topotecan, doxorubicin, daunorubicin, irinotecan, imatinib, and methotrexate (Dean, 2009).
Oridonin is a diterpenoid isolated from Rabdosia rubescens and reveals anticancer activity in vitro and in vivo (Xiao et al., 2016;Lu et al., 2017;Yao et al., 2017), but its mode of action and effect on drug resistance have not been well studied. R. rubescens inhibited breast cancer growth and angiogenesis (Sartippour et al., 2005) and overcame drug resistance in ADR/MCF-7 breast cancer cells by increasing doxorubicin accumulation (Li et al., 2013). Therefore, it is reasonable to investigate oridonin's mode of action on MDR in more detail.
In this study, we analyzed molecular factors determining the response of tumor cells to oridonin. Various drug resistance mechanisms were investigated. We addressed three main questions: (1) Is oridonin able to bypass resistance caused by different mechanisms such as P-gp, EGFR, p53, and BCRP? Moreover, can oridonin selectively target tumor cells rather than normal cells? To address these questions, we performed cytotoxicity assays.
(2) Are there other determinants predicting sensitivity or resistance of cancer cells to oridonin? For this reason, we performed COMPARE-and hierarchical cluster analyses of transcriptome-wide mRNA expression profiles of cancer cells.
(3) Can oridonin interact with EGFR pathway proteins? To answer this question, we applied molecular docking, MD, and Western blot.

Cell Lines
CCRF-CEM leukemia cells were cultured as previously described (Efferth et al., 2003b). Drug resistance of P-gp/MDR1/ABCB1overexpressing CEM/ADR5000 cells was maintained in 5000 ng/mL doxorubicin (Kimmig et al., 1990). Breast cancer cells transduced with a control vector (MDA-MB-231-pcDNA3) or with cDNA for the breast cancer resistance protein BCRP/ABCG2 (MDA-MB-231-BCRP clone 23) were generated and maintained as reported (Doyle et al., 1998). The mRNA expression of MDR1 and BCRP in the resistant cell lines has been reported (Efferth et al., 2003a;Gillet et al., 2004). Human wild-type HCT116 colon cancer cells (p53 +/+ ) as well as knockout clones (p53 −/− ) derived by homologous recombination (Bunz et al., 1998) were a generous gift from Dr. B. Vogelstein and H. Hermeking (Howard Hughes Medical Institute, Baltimore, MD, United States) and cultured as described (Bunz et al., 1998). Human GBM U87MG cells transduced with an expression vector harboring an EGFR gene with a deletion of exons 2-7 (U87MG. EGFR) has been previously reported (Huang et al., 1997). Transduced and non-transduced cell lines were kindly provided by Dr. W. K. Cavenee (Ludwig Institute for Cancer Research, San Diego, CA, United States). Human HepG2 hepatocellular carcinoma cells and AML12 normal hepatocytes were obtained from the American Type Cell Culture Collection (ATCC, United States).

Resazurin Cell Growth Inhibition Assay
The resazurin (Promega, Mannheim, Germany) reduction assay (O'Brien et al., 2000) was used to assess the cytotoxicity as previously described (Kuete et al., 2015b(Kuete et al., , 2016. Each assay was conducted at least three times, with two replicates each. Cell viability was evaluated based on a comparison with untreated cells. IC 50 values were determined as concentrations required to inhibit 50% of cell proliferation and were calculated from a calibration curve by linear regression using Microsoft Excel.

Molecular Docking
The protocol for molecular docking was previously reported by us (Kadioglu et al., 2016b). An X-ray crystallography-based structure of wild-type Akt2 kinase domain (PDB ID: 3E87), EGFR (PDB ID: 1M17), mTOR (PDB ID: 4JSP), STAT3 DNA-binding domain, and VEGFR1 (PDB ID: 3HNG) were obtained from Protein Data Bank 1 . Homology model of STAT3 DNA-binding domain was created by us using MODELLER 9.11 (Fiser and Sali, 2003;Venkatachalam et al., 2003) and a Swiss-MODEL structure assessment tool 2 based on the wild-type structure (PDB ID: 1BG1) as template. In order to assess the effect of an Akt2 mutation and EGFR mutation on oridonin binding, one point mutation-R274H on Akt2 was selected which has been shown to be critical for phosphatase resistance and keeping the phosphorylated status on Akt2 (Chan et al., 2011) and one point mutation-T790M on EGFR which has been shown to cause resistance to EGFR tyrosine kinase inhibitors (Zhou et al., 2018). Homology model of R274H mutant Akt2 kinase domain was created in the same manner by using wild-type Akt2 kinase domain as template. T790M-mutant EGFR structure is available in PDB database (PDB ID: 5XDK). A grid box was then constructed to define docking spaces in each protein according to their pharmacophores. Docking parameters were set to 250 runs and 2,500,000 energy evaluations for each cycle. Docking was performed three times independently by Autodock4 and with AutodockTools-1.5.7rc1 (Morris et al., 2009) using the Lamarckian Algorithm. The corresponding lowest binding energies and pKi were obtained from the docking log files (dlg). Mean ± SD of binding energies were calculated from three independent docking. Visual Molecular Dynamics (VMD) was used to depict the docking poses of oridonin and the inhibitors for each target protein.

Molecular Dynamics
Lowest binding energy conformation of oridonin on Akt2 kinase domain prior to molecular docking analyses was picked to create ligand-protein complex structure for the MD simulations. QwikMD tool (Ribeiro et al., 2016) was used to perform 15 ns MD simulations after the equilibration of the protein-ligand complex. Stability of the docking pose was evaluated by root mean square deviation (RMSD) distance of the conformation throughout the MD simulation with the starting conformation. Total energy of the ligand-protein complex was calculated as well.

Western Blot
In order to evaluate the effect of oridonin on EGFR pathway proteins and validate the in silico results, varying concentrations of oridonin (IC 50 /4, IC 50 /2, IC 50 , 2xIC 50 , and 4xIC 50 ), determined after the cytotoxicity test on U87MG. EGFR cell line, were applied in a similar way as described previously . Briefly, 1 million cells per well were seeded in

COMPARE and Hierarchical Cluster Analyses
The microarray-based mRNA expression values of genes of interest and log 10 IC 50 values for oridonin of 49 tumor cell lines were selected from the NCI database 3 . The COMPARE analyses were performed to produce rank-ordered lists of genes expressed in the NCI cell lines. The methodology has been previously described in detail (Wosikowski et al., 1997). Briefly, every gene of the NCI microarray database was ranked for similarity of its mRNA expression to the log 10 IC 50 values for oridonin based on Pearson's rank correlation test. To derive COMPARE rankings, a scale index of correlations coefficients (R-values) was created. CIM miner software was used to perform the hierarchical clustering and heat map analysis 4 . FIGURE 3 | Western blot analysis of oridonin on EGFR pathway proteins. The effects of oridonin on phosphorylation of EGFR, STAT3, and Akt were evaluated. Bands were normalized to β-actin in order to obtain numerical values (mean ± SEM of three independent experiments). Total EGFR, STAT3, and Akt protein levels are also shown. A representative blot is shown and statistical analysis was done by paired Student's t-test. * * p < 0.01, * p < 0.05.

Statistical Analyses
Results were represented as mean ± SD. Student's t-test was performed in order to evaluate the statistical significance with two tails and unequal variance. Experiments with p-values lower than 0.05 were accepted as statistically significant.

Response of Drug-Resistant Tumor Cell Lines Toward Oridonin
Cytotoxicity of oridonin and doxorubicin toward sensitive and drug-resistant cancer cell lines and normal cells were determined by the resazurin reduction assay ( Table 1) (Table 1). Oridonin was tested against multidrug-resistant P-gp (MDR1/ABCB1)-overexpressing CEM/ADR5000 cells and drug-sensitive parental CCRF-CEM cells using a resazurin assay. Although a weak cross-resistance of the CEM/ADR5000 cells was obtained (5.17-fold), this was much lower than that obtained with doxorubicin (975.60-fold). In another cell model for MDR, we compared the cytotoxicity of oridonin toward MDA-MB-231 cells transfected with BCRP/ABCG2 and cells transfected with pcDNA control vector. The BCRP transfectants were 1.61-fold more resistant to oridonin than their sensitive counterparts. The activities of oridonin in knockout HCT116 (p53 −/− ) cells and their sensitive wild-type HCT116 (p53 +/+ ) cells were also compared. TP53-knockout cells were crossresistant to this compound than the TP53 wild-type cells (degree of resistance: 1.92). However, the degree of resistance was slightly less resistant than that obtained with doxorubicin (2.84-fold). Interestingly, U87MG cells transfected with a deletion-activated EGFR cDNA were considerably more sensitive to oridonin than their wild-type counterpart (degree of resistance: 0.88). Normal AML10 hepatocytes were more resistant to oridonin than HepG2
The docking pose of oridonin on Akt2 kinase domain was used as starting conformation for the MD simulation. As can be seen in Figure 2, the LBE conformation of oridonin was stable, since the RMSD value was below 1 Å (0.567 ± 0.117) throughout 15 ns simulation.

Western Blot
In order to validate the in silico analyses, the phosphorylation status of EGFR signaling proteins as parameter of their activation during signal transduction was investigated. Oridonin revealed a dose-dependent inhibition of Akt and STAT3 phosphorylation supporting the in silico analyses, but no change in EGFR phosphorylation was observed (Figure 3). There was no change at the total Akt, EGFR, and STAT3 protein levels.

Pharmacogenomics
We investigated the transcriptome-wide RNA expression using COMPARE analysis and mined the database of the NCI by correlating the mRNA expression data with the log 10 IC 50 values for oridonin. This is a hypothesis-generating bioinformatical approach allowing to find novel putative molecular determinants of cellular response to oridonin. The scale rankings of genes obtained by COMPARE computation were subjected to Pearson's rank correlation tests. The thresholds for correlation coefficients were R > 0.50 for direct correlations and R < −0.50 for inverse correlations. As shown in Table 3, the identified genes can be assigned to different functional groups such as apoptosis regulation (CARD8, ANXA5), transcriptional and protein synthesis (FUBP1, RFXAP, CELF2, TWF1, COQ9), DNA repair and maintenance (RAD51C, BLM), signal transduction (PRKXP1, SRPK1, GNG12, IL6ST, PTPRK, MET), cell cycle regulation (MND1, PDS5B, PHF8), transport functions (ABCC5), cellular energy regulation (PKM2), cell adhesion (LAMB1, ADAM9), and ubiquitination (USP44).
The mRNA expression values of all NCI cell lines for the genes listed in Table 3 were subsequently subjected to agglomerative hierarchical cluster analysis, in order to find out, whether clusters of cell lines could be identified with similar behavior after exposure to oridonin. The dendrogram of the cluster analysis showed three clusters (Figure 4). As a next step, the log 10 IC 50 values for oridonin, which were not included in the cluster analysis, were assigned to the corresponding position of the cell lines in the cluster tree. The distribution among the clusters was significantly different from each other as determined by Chisquare test (p-value = 0.0014). Cluster 1 contained in its majority of cell lines resistant to oridonin, whereas Cluster 3 contained in its majority sensitive ones (Cluster 1: 17 resistant and 7 sensitive; Cluster 2: 8 resistant and 8 sensitive; Cluster 3: 0 resistant and 9 sensitive).

DISCUSSION
Oridonin is a diterpenoid isolated from R. rubescens with previously reported anticancer activity. It targets PI3K/Akt pathway causing G2/M arrest in prostate cancer cells (Lu et al., 2017). Oridonin also targets Notch signaling leading to inhibition of breast cancer progression (Xia et al., 2017). According to the literature, 4 µg/mL or 10 µM is the upper IC 50 limit considered for a promising cytotoxic compound after incubation for 48 and 72 h (Boik, 2001;Brahemi et al., 2010;Kuete and Efferth, 2015). In the present work, oridonin displayed IC 50 values (<10 µM) within this threshold value toward four tested cancer cell lines as determined by resazurin assay. These data show the antiproliferative potential of oridonin against drugsensitive and -resistant cancer cell lines since identification of compounds able to overcome MDR is an attractive strategy in drug research (Efferth, 2001;Gottesman and Ling, 2006;Gillet et al., 2007). Interestingly, the resistant U87MG. EGFR cells were even more sensitive to oridonin than their corresponding sensitive counterparts (U87MG cells). If cross-resistance was obtained, the degree or resistance was lower in all cases than that of the reference compound, doxorubicin. This suggests that oridonin could be explored further to develop a cytototoxic drug to combat MDR phenotypes. Overexpression of P-gp, a broad spectrum drug transporter, leads to the efficient extrusion of a large number of established anticancer drugs and cytotoxic natural products out of cancer cells. This is the main reason, why tumors with P-gp overexpression exert a MDR phenotype limiting the success of established drugs. Therefore, it was a pleasing result that the expression of P-gp/MDR1 in the NCI cell line panel did not correlate with cellular response to oridonin, which implies that P-gp does not confer resistance to oridonin. In addition, multidrug-resistant CEM/ADR5000 cells with overexpression of various ABC transporters including P-gp/MDR1 (400-fold) (Kadioglu et al., 2016a) revealing high degrees of resistance to well-known anticancer drugs such as doxorubicin (1036fold), vincristine (613-fold), docetaxel (435-fold), and many others (Efferth et al., 2008) were even slightly more sensitive to oridonin than the parental, wild-type, drug-sensitive CCRF-CEM tumor cells. It can be speculated that oridonin successfully kills otherwise unresponsive, multidrug-resistant tumors.
Oridonin revealed comparable binding energies to EGFR pathway proteins as the known inhibitors. It shares the same docking pose with GSK690693 on Akt2 and NSC74859 on STAT3, implying the inhibitory potential of oridonin toward Akt2 and STAT3. MD study revealed that oridonin docking pose on Akt2 is stable throughout the simulation with a relatively small RMSD deviation (<1 Angström). In order to validate the in silico findings and further evaluate the mode of action of oridonin, western blot experiments for the EGFR pathway proteins regarding their phosphorylation status upon oridonin treatment were performed. Results implied that cytotoxicity of oridonin is dependent on the EGFR pathway influence.
Various other resistance factors in addition to EGFR and P-gp determine the success rate of chemotherapy. In order to achieve a deeper understanding of drug response determinant mechanisms, microarray technology is widely used. This methodology is especially helpful to identify potential mechanisms of novel, still incompletely understood cytotoxic compounds. For this purpose, we performed COMPARE and hierarchical cluster analyses of transcriptome-wide, microarraybased mRNA expression of the NCI cell line panel. The expression of the genes identified via COMPARE analyses determined cellular response to oridonin in this panel of cell lines.
Despite P-gp expression was not correlated to oridonin resistance, our COMPARE analysis revealed that another member of ABC superfamily, i.e., ABCC5, was a molecular determinant to mediate resistance to oridonin in the NCI cancer cell line panel. In this context, it is worth to mention that a member of the ABC sub-family C, ABCC1 (multidrug protein 1, MRP1), which is known to confer MDR phenotype differs from that one caused by P-gp (ABCB1/MDR1) and BCRP/ABCG2 (Efferth, 2001).
In conclusion, oridonin targeted various resistance mechanisms and inhibited Akt2 and STAT3 phosphorylation. The resistance and sensitivity genes identified may be helpful for the development of personalized therapy approaches, as their expression in an individual patient may suggest potentially successful oridonin treatment in the future. However, further preclinical and clinical studies are required to assess the therapeutic potential of oridonin for cancer therapy.

AUTHOR CONTRIBUTIONS
TE conceived the study. OK performed the in silico experiments. MS contributed to the in silico experiments. OK, MS, HG, and TE wrote the manuscript. OK, MS, and VK performed the in vitro experiments. All the authors read the manuscript.