Significance and mechanism of androgen receptor overexpression and androgen receptor/mechanistic target of rapamycin cross-talk in hepatocellular carcinoma

HUMAN HCC PATIENTS AND TISSUE SAMPLES

One hundred forty-two clinical tissue samples of HCC were collected from HCC patients who underwent hepatic resection from January 2004 to August 2012 at the Sun Yat-Sen University Cancer Center. All the cases were diagnosed by pathological examination. The median follow-up is 48 months (range, 4-111). None of the patients received chemotherapy or radiotherapy treatment before surgery. Demographic and clinicopathological parameters of HCC patients are listed in Supporting Table S1. This study was reviewed and approved by the ethics committees of Sun Yat-Sen University Cancer Center. Informed consents signed by the patients were collected. An HCC transcriptome data set available from the Oncomine cancer microarray database (https://www.oncomine.org/) was used to analyze AR mRNA and AR target gene expression. This data set contains mRNA expression information of 10,802 genes from 104 human primary HCC tumors and 76 adjacent non-tumor liver tissues, of which 93 tumors are HBV⁺.28 We analyzed the mRNA expression of AR and AR target genes using the Student's t test. HCC proteomic data derived from 184 primary human HCC tumors (113 white Americans, 52 Asian Americans, and 12 black Americans) were downloaded from The Cancer Proteome Atlas (http://tcpaportal.org/tcpa/).29 Normalized phosphoprotein levels of AKT (S473) and phospho-p70S6K (T389) were used to represent AKT-mTOR signaling. The correlation between the level of AR protein and p-AKT(473) or p-p70S6K(389) was analyzed using Pearson's correlation test as described.30

ANIMAL STUDIES

Mouse procedures and protocols were approved by the Institutional Animal Care and Use Committee of Sun Yat-Sen University. For xenograft tumor formation assay, male BALB/c athymic nude mice aged 4 weeks were injected with 2 × 10⁶ MHCC-97L of tumor cells in 0.1 mL of phosphate-buffered saline subcutaneously into the flank of each mouse. When average tumor volume reached 300 mm³, tumor-bearing animals were randomized and treatment was initiated. Enzalutamide (10 mg/kg) was solubilized in 100 μL of critical micellar concentration Na⁺ 1%/Tween-80 1% and administered intragastrically daily. Rapamycin (10 mg/kg) was dissolved in 100 μL of 30% polyethylene glycol 300 and 5% Tween-80 and given intraperitoneally daily for 7 consecutive days, followed by 2 days without drug. Enzalutamide and rapamycin were freshly prepared daily just before administration. The control group was given the vehicle used for administration. Tumors were measured every 3 days and tumor volumes were calculated from the formula: Tumor volume = Length × Width² × (Л /6). After mice were euthanized, xenograft tumors were collected and fixed in formalin and embedded in paraffin block. The synergism of enzalutamide and rapamycin was calculated using the Chou-Talalay method with Calcusyn software (Biosoft, Cambridge, UK). A combination index (CI) less than 1 indicates the drug combination has synergism. The results show that the CI value for the combination of the two drugs is 0.28, indicating that they have synergistic effect.

STATISTICAL ANALYSIS

Statistical data analysis was performed by SPSS (version 11.0; SPSS, Inc., Chicago, IL) and GraphPad Prism software (version 5.0 for windows; GraphPad Prism Inc., San Diego, CA). Experimental data are expressed as mean ± SD. Receiver operating characteristic curve analysis was used to estimate the cut-off value for high and low AR expression in HCC patients. Kaplan-Meier plots and log-rank test were performed to compare cancer-specific survival rates between patients with high and low AR expression. Univariate and multivariate survival analyses were performed using a Cox proportional hazards regression model. Repeated-measures analysis of variance (ANOVA) was performed to analyze xenograft tumor size and proliferation of cultured cells. One-way ANOVA test and Student's t test were conducted to evaluate the variables of different groups. A paired t test was used to analyze xenograft tumor size and immunostaining scores of AR in paired clinical samples. A nonparametric Spearman correlation test was used to compare the correlation between AR expression and enzalutamide half maximal inhibitory concentration (IC₅₀) values. P < 0.05 was considered as statistical significant (indicated in corresponding figures with one asterisk for P < 0.05, two asterisks for P < 0.01 and three asterisks for P < 0.001).

For additional materials and methods, see the Supporting Information.

AR IS OVEREXPRESSED IN HCC, WHICH IS ASSOCIATED WITH DISEASE PROGRESSION AND IS AN INDEPENDENT PREDICTOR OF OS

Although there were several studies on AR mRNA or protein expression in human HCCs, the results and clinical significance were contradictory.31, 32 Limiting factors include small sample sizes and nonpaired tumor/normal tissues and the use of radiolabeled ligands to detect AR in tissues. We analyzed an HCC gene expression data set available in Oncomine from a transcriptome study of 104 primary HCC and 76 normal liver tissues,28 which shows that AR mRNA expression is highly variable in tumors, but not in normal liver tissues (Fig. 1A). Overall, AR mRNA level is higher in tumors. To verify this finding, we examined AR protein expression by immunohistochemistry (IHC) in a cohort of 142 paired human HCCs and adjacent noncancerous liver tissues. AR is expressed in both normal hepatocytes and tumor cells with predominant nuclear localization (Fig. 1B), indicating that AR is largely in the activated form. Nuclear AR protein is significantly overexpressed in tumors compared to the adjacent noncancerous liver tissues (P < 0.001; Fig. 1C), with 37% (53 of 142) cases with 2-fold or higher nuclear AR in tumors (Fig. 1C, lower panel). In contrast, there is no significant difference in cytoplasmic AR expression between tumor and adjacent liver tissues (Fig. 1D).

We next examined the relationship between nuclear AR protein expression and clinicopathological parameters. Advanced tumor, node, and metastasis (TNM) stage, multiple tumor nodules, and high HBV viral copies were positively correlated with high nuclear AR expression (Supporting Table S2), indicating that AR is associated with HCC disease progression. There was no statistically significant correlation between AR protein expression and age, sex, tumor size and grade, hepatitis B surface antigen, and alpha-fetoprotein (Supporting Table S2). Kaplan-Meier analysis demonstrates that high nuclear AR staining was significantly correlated with poorer OS of HCC patients (P < 0.001; Fig. 1E). Furthermore, multivariate analysis shows that nuclear AR overexpression in tumors is an independent predictor for poor OS (hazard ratio = 2.423; 95% confidence interval = 1.053-5.577; P = 0.037; Supporting Table S2). These results indicate that nuclear AR protein is highly expressed in a subset of HCC tumors characteristic of advanced disease stage and poor prognosis.

AR OVEREXPRESSION ALTERS AR-DEPENDENT TRANSCRIPTOME IN HCC

We also examined AR protein expression in a panel of immortalized hepatocyte and HCC cell lines. Compared with the immortalized LO2 hepatocyte cell line, AR is significantly higher in SNU423 and SNU387 cells and moderately higher in BEL7402 and MHCC-97L cells (Fig. 2A). Immunofluorescence (IF) staining indicates that AR is primarily localized in the nucleus of these HCC cells (Fig. 2B). These results indicate that HCC cell lines share the same heterogeneity in AR expression as primary liver tumors, and that they are useful in vitro models for studying AR functions. AR is a transcription factor that regulates expression of genes with diverse functions, especially those related to cell growth in prostate cancer.33, 34 Therefore, we investigated the transcriptional function of AR in HCC using a luciferase reporter driven by an ARE. Treatment of SNU423 and MHCC-97L cells with testosterone stimulates AR reporter activity in these AR-positive HCC cells (Fig. 2C), demonstrating that AR regulates target gene expression in HCC cells.

To assess the transcriptional function of AR in HCC, we overexpressed AR in SNU449 cells (Supporting Fig. S1), a cell line with relatively low AR level. The global transcriptional effect of AR overexpression was examined using the Human Androgen Receptor Signaling Targets PCR Array that consists of 84 key AR target genes (Qiagen, Valencia, CA). AR overexpression leads to up-regulation of 38 AR target genes and down-regulation of 18 AR target genes by >2-fold (Fig. 2D,E), representing 67% key AR target genes. Analysis of eight randomly selected AR target genes by qRT-PCR verified the AR target gene PCR array results (Fig. 2F). Significantly altered AR target genes were enriched in Androgen, PI3K, insulin-like growth factor 1 (IGF1), and Wnt and β-catenin signaling (Fig. 2D), pathways well known for their crucial role in liver cancer cell proliferation and survival, which is consistent with a tumor-promoting role of AR overexpression in HCC. In the same HCC transcriptome data set,28 expression of AR target genes was significantly altered. For example, APPBP2, Rab4A, SMS, ZNF189, SLC26A2, and TPD52, which were up-regulated by AR overexpression (Supporting Table S3), were significantly overexpressed in primary HCC samples compared to adjacent non-tumor liver tissues (Supporting Fig. S3). Similarly, VIPR1, MME, and TMPRSS2, which were down-regulated by AR (Supporting Table S3), were underexpressed in primary HCC tumor samples compared to adjacent non-tumor liver tissues (Supporting Fig. S3). Together, these results support the notion that AR pathway is activated in human primary HCCs.

AR OVEREXPRESSION RENDERS AR-DEPENDENT AND ENZALUTAMIDE-SENSITIVE GROWTH OF HCC CELLS

To investigate the role of AR in HCC growth, we treated a panel of immortalized liver and HCC cell lines with enzalutamide (MDV3100). The IC₅₀ for enzalutamide was highly variable, ranging from 27.8 to 418.3 μM (Fig. 3A). There was a striking inverse relationship between AR protein expression and enzalutamide IC₅₀ (R = 0.79; P < 0.043; Fig. 3B), indicating that the growth of HCC cells with high AR expression is more AR dependent and hence more sensitive to AR blockage. Consistently, enzalutamide blocks testosterone stimulated growth of SNU423 and MHCC97L cells (Fig. 3C). We further tested the effect of AR knockdown by small interfering RNA (siRNA; siAR). Four siARs were tested and found to produce significant AR knockdown (Supporting Fig. S1B). Two siARs were selected for further studies, resulting in more growth inhibition in high AR-expressing SNU423 and MHCC97L cells (Fig. 3D) than low AR-expressing LO2 and Huh7 cells (Fig. 3E). These results suggest that AR overexpression renders AR-dependent and enzalutamide-sensitive growth in HCC cells.

INHIBITION OF AR LEADS TO FEEDBACK ACTIVATION OF AKT-mTOR SIGNALING THROUGH FKBP5

The fact that AR blockage only produces a modest growth inhibition raises an interesting possibility of a feedback mechanism that limits the impact of AR loss. Indeed, AR knockdown in SNU423 and MHCC-97L cells leads to elevated phosphorylation of AKT, mTOR, and S6 kinase 1 (S6K1), indicating that AKT-mTOR pathway is activated (Fig. 4A and Supporting Fig. S3A). Similarly, pharmacological inhibition of AR by enzalutamide treatment also results in activation of AKT-mTOR signaling in SNU423 and MHCC97L cells (Fig. 4B and Supporting Fig. S3B). Moreover, AR inhibition by siRNA-mediated knockdown or enzalutamide results in increased phosphorylation of glycogen synthase kinase 3, a known AKT substrate.35 Thus, blockage of AR activity triggers feedback activation of AKT-mTOR signaling. In contrast, AR inhibition has little effect on c-MYC expression or extracellular signal-regulated kinase (ERK) phosphorylation (Supporting Fig. S3), suggesting that blockage of the AR pathway selectively activates AKT-mTOR signaling in HCC.

However, it takes 6-12 hours for AKT-mTOR signaling to become activated by enzalutamide (Fig. 4B), suggesting that the feedback mechanism involves a transcriptional mechanism. FK506 binding protein 5 (FKBP5) is an AR target gene in prostate cancer that is known to inhibit AKT activity by promoting AKT dephosphorylation through scaffolding AKT with PH domain leucine-rich repeat protein phosphatase (PHLPP), an AKT phosphatase.36-38 Interestingly, AR also promotes FKBP5 expression in HCC cells (Fig. 2F). Enzalutamide causes down-regulation of FKBP5 mRNA and protein expression in MHCC-97L and SNU423 cells (Fig. 4C,D). Curiously, the level of PHLPP1 is also decreased by enzalutamide (Fig. 4D), though the mechanism is currently not understood. In contrast, enzalutamide does not affect the expression of FKBP5 and PHLPP1 in the low AR-expressing SNU449 cell line (Fig. 4D). Moreover, ectopic AR expression in SNU449 and PLC5 cell lines stimulates FKBP5 and PHLPP1 expression, with concurrent suppression of AKT/mTOR signaling (Fig. 4E). These observations demonstrate that blockage of AR leads to activation of AKT-mTOR signaling, which is, at least in part, attributed to down-regulation of the AR target gene, FKBP5, and AKT phosphatase, PHLPP1.

mTOR PROMOTES AR TRANSCRIPTIONAL ACTIVITY BY ENHANCING AR STABILITY AND NUCLEAR LOCALIZATION

Inhibition of mTOR by rapamycin is known to stimulate AR transcriptional activity in prostate cancer.39 Surprisingly, however, we found that rapamycin treatment in MHCC-97L and SNU423 cells leads to marked inhibition of AR transcriptional activity under both basal and testosterone-stimulated conditions (Fig. 5A). Rapamycin treatment significantly decreases AR protein level (Fig. 5B). mTOR is regulated by both growth factors and amino acids. However, starvation from fetal bovine serum (FBS), not amino acids, causes AR down-regulation (Fig. 5C), indicating that regulation of AR by mTOR responds specifically to growth factor, not nutrient signaling. Under the condition when new protein synthesis is blocked by cycloheximide (CHX), AR is more rapidly degraded in rapamycin-treated cells than control cells (Fig. 5D), demonstrating that rapamycin accelerates AR degradation. Moreover, treatment with the proteasome inhibitor, MG132, abrogates rapamycin-induced AR degradation (Fig. 5E). Thus, mTOR negatively regulates the proteasome-dependent AR degradation in HCC. We have also examined AR localization by IF staining. As shown earlier, AR is predominantly localized in the nucleus in SNU423 and MHCC-97L cells. Consistent with proteasome-dependent degradation of AR, rapamycin treatment leads to a marked decrease in AR staining (Fig. 5F). Interestingly, AR becomes predominantly cytoplasmic in rapamycin-treated HCC cells (Fig. 5F). Rapamycin also inhibits AR nuclear localization in the presence of MG132. These results demonstrate that mTOR positively regulates both AR stability and nuclear localization.

COTARGETING AR AND mTOR SHOWS SYNERGISTIC ACTION AGAINST HCC AND REDUCES TUMOR BURDEN BY INDUCING APOPTOSIS IN VIVO

AR inhibition by enzalutamide or siRNA displays modest anticancer activity in HCC cells (Fig. 6A-D), which is, at least in part, attributed to the feedback activation of AKT-mTOR signaling (Fig. 4). We therefore explored the utility of combinating AR and mTOR inhibitors. Enzalutamide and rapamycin together indeed produce more potent inhibition of HCC cell growth and proliferation than each drug individually (Fig. 6A-C and Supporting Fig. S4). Moreover, the drug combination leads to a marked increase in apoptotic cell death (Supporting Fig. S5). A similar effect was observed with AR knockdown in the presence of rapamycin (Fig. 6D). mTOR kinase inhibitors target mTOR kinase domain and inhibit both mTORC1 and mTORC2.40 They exhibit excellent anticancer activity in preclinical models and are currently under clinical trials for various malignancies.40 We hence tested two mTOR kinase inhibitors, AZD8055 and PP242. Each agent significantly improves the anticancer activity of enzalutamide (Fig. 6E,F). These results indicate that combination of enzalutamide with rapamycin or mTOR kinase inhibitor yields more potent anti-HCC activity in vitro than each drug alone.

We next assessed the in vivo antitumor activity of enzalutamide alone or in combination with rapamycin in xenograft tumors derived from MHCC-97L cells. Enzalutamide alone has statistically significant, albeit modest, antitumor activity, whereas rapamycin alone is more potent, as measured by slower increase in tumor burden (Fig. 7A). Remarkably, combination of enzalutamide and rapamycin not only blunts tumor growth, but also significantly reduces tumor burden (Fig. 7A). No significant weight loss occurs in different treatment groups (Fig. 7B), indicating that each drug and their combination are well tolerated. As observed in vitro, rapamycin reduces AR protein expression, whereas enzalutamide causes mTOR activation as indicated by increased p-S6 staining (Fig. 7C). The drug combination much more potently inhibits tumor cell proliferation, as judged by Ki-67 staining (Fig. 7C and Supporting Fig. S6A). Consistent with the tumor burden reduction, the drug combination, not each drug individually, triggers elevated apoptosis, as judged by positive terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining (Fig. 7D and Supporting Fig. S6B). These results indicate that the combination of enzalutamide and rapamycin generates cytotoxic antitumor activity against HCC, which is much more robust in vivo than in vitro. A CI of enzalutamide and rapamycin, as calculated by the Chou-Talalay method using Calcusyn software (Biosoft), yielded a value of 0.28. A CI value of less than, equal to, and greater than 1 indicates that two drugs are synergistic, additive, and antagonistic, respectively. Our result demonstrates that enzalutamide and rapamycin have strong synergism against HCC.