Purpose:

The Mediator complex is a multiprotein assembly, which serves as a hub for diverse signaling pathways to regulate gene expression. Because gene expression is frequently altered in cancer, a systematic understanding of the Mediator complex in malignancies could foster the development of novel targeted therapeutic approaches.

Experimental Design:

We performed a systematic deconvolution of the Mediator subunit expression profiles across 23 cancer entities (n = 8,568) using data from The Cancer Genome Atlas (TCGA). Prostate cancer–specific findings were validated in two publicly available gene expression cohorts and a large cohort of primary and advanced prostate cancer (n = 622) stained by immunohistochemistry. The role of CDK19 and CDK8 was evaluated by siRNA-mediated gene knockdown and inhibitor treatment in prostate cancer cell lines with functional assays and gene expression analysis by RNAseq.

Results:

Cluster analysis of TCGA expression data segregated tumor entities, indicating tumor-type–specific Mediator complex compositions. Only prostate cancer was marked by high expression of CDK19. In primary prostate cancer, CDK19 was associated with increased aggressiveness and shorter disease-free survival. During cancer progression, highest levels of CDK19 and of its paralog CDK8 were present in metastases. In vitro, inhibition of CDK19 and CDK8 by knockdown or treatment with a selective CDK8/CDK19 inhibitor significantly decreased migration and invasion.

Conclusions:

Our analysis revealed distinct transcriptional expression profiles of the Mediator complex across cancer entities indicating differential modes of transcriptional regulation. Moreover, it identified CDK19 and CDK8 to be specifically overexpressed during prostate cancer progression, highlighting their potential as novel therapeutic targets in advanced prostate cancer. Clin Cancer Res; 23(7); 1829–40. ©2016 AACR.

Translational Relevance

Tight control of gene expression is needed for physiological cell homeostasis, but it is frequently deregulated in cancer. The multiprotein Mediator complex plays a pivotal role in transcriptional regulation being a central binding element between transcription factors and RNA polymerase II. A pan-cancer analysis of the Mediator complex transcriptome revealed distinct expression profiles distinguishing entities by hierarchical clustering. CDK19 was identified to be specifically overexpressed in the progression of prostate cancer and was associated with clinicopathologic parameters of aggressiveness and decreased survival. Knockdown of CDK19 and its paralog CDK8 or inhibition with senexin A, one of several small-molecule CDK19/CDK8 inhibitors that were recently developed, led to significantly decreased migration and invasion of prostate cancer cells in vitro. The prognostic and functional roles of CDK19 in prostate cancer and the ongoing development of selective inhibitors thereby highlight the potential of CDK19 and its paralog CDK8 as therapeutic targets in prostate cancer.

One of the crucial factors determining differentiation, growth, and development of cells is the time- and tissue-specific expression of protein-coding genes by RNA polymerase II (Pol II), which is therefore tightly controlled (1).The multiprotein Mediator complex (MED) has emerged as one of the main Pol II coactivators governing transcriptional regulation (2). By functioning as a hub for input from numerous signaling pathways and transcription factors, the Mediator complex induces or represses essential aspects of the transcription cycle such as initiation, elongation, and chromatin architecture (2, 3). In humans, the Mediator complex consists of 33 dynamically assembled protein subunits each of which is designated to 1 of 4 distinct MED modules “head,” “middle,” “tail,” and “kinase” (Fig. 1A). Generally, the head and the middle modules directly interact with Pol II and together with the tail module form a relatively stable “core” structure but are devoid of enzymatic activity. The kinase module including subunits cyclin-dependent kinase 8 (CDK8), MED12, MED13, and cyclin C (CCNC) can reversibly associate with the core structure, which leads to considerable changes of MED confirmation and functional activity (3). In addition, in vertebrates, the subunits CDK19, MED12L, and MED13L have been shown to take part in formation of the kinase module in a mutually exclusive manner with their paralogous counterparts CDK8, MED12, and MED13, respectively, indicating a high flexibility in MED composition and its quaternary structure (3–5).

Figure 1.

Expression profiling of the Mediator complex across tumor types. A, Schematic representation of the Mediator complex, its subunits, and interaction with Pol II. B, Heatmap representation for expression levels of Mediator complex subunits (columns) in primary tumor samples of patients within TCGA (rows). Patients and subunits are arranged on the basis of similarities in Mediator complex expression profile as determined by unsupervised hierarchical clustering. Color keys next to the rows and columns indicate cancer type and Mediator complex module per individual patient or subunit, respectively. Colored branches in the row dendrogram correspond to relevant subgroups (A–P) identified by the dynamic tree partitioning algorithm. Bold black boxes indicate high CDK19 expression in prostate cancer (PCa). C, Pairwise Spearman correlation coefficient of the Mediator complex subunits in PCa.

Figure 1.

Expression profiling of the Mediator complex across tumor types. A, Schematic representation of the Mediator complex, its subunits, and interaction with Pol II. B, Heatmap representation for expression levels of Mediator complex subunits (columns) in primary tumor samples of patients within TCGA (rows). Patients and subunits are arranged on the basis of similarities in Mediator complex expression profile as determined by unsupervised hierarchical clustering. Color keys next to the rows and columns indicate cancer type and Mediator complex module per individual patient or subunit, respectively. Colored branches in the row dendrogram correspond to relevant subgroups (A–P) identified by the dynamic tree partitioning algorithm. Bold black boxes indicate high CDK19 expression in prostate cancer (PCa). C, Pairwise Spearman correlation coefficient of the Mediator complex subunits in PCa.

Close modal

Recently, several subunits have been implicated in a growing number of human diseases including developmental disorders (3, 6) and cancer (4, 7). Most evidence has been gathered for the kinase module subunit CDK8, which has been linked to colon and pancreatic cancer (8, 9) and was also investigated as a potential target in cancer therapy (10, 11). Other subunits have also been reported in neoplastic conditions such as MED12 in uterine leiomyomas, fibroepithelial breast tumors, and prostate cancer (12–14), MED15 in prostate cancer and head and neck cancer (15, 16), and cyclin C in acute lymphoblastic leukemia (17). In addition, some subunits have been linked to altered hormone receptor signaling such as MED1 in breast cancer and prostate cancer (18, 19).

To systematically characterize Mediator complex expression profiles across human tumors, we performed a pan-cancer analysis of the Mediator complex transcriptome in 23 cancer entities for more than 8,000 patients from The Cancer Genome Atlas (TCGA), followed by an in-depth investigation of the role of the kinase subunits CDK19 (formerly known as Cdc2L6 or CDK11-like) and its paralog CDK8 in prostate cancer.

FirebrowseR

We developed the R-package FirebrowseR to facilitate access to data generated by the Broad Institute's Firehose pipeline and made it available for public use under MIT license (github.com/mariodeng/FirebrowseR). With FirebrowseR, TCGA datasets and results of Firehose analyses runs (ref. 20; including mutations, copy number, gene expression, and clinical information) can directly be imported into the R (21) programming environment. All functions, data types, and output formats provided by the Firehose web application programmable interface (API) are fully implemented.

Transcriptome data assembly

TCGA transcriptome sequencing data (RNA-Seq v2) was imported into R using FirebrowseR from the Broad Institute Firehose Pipeline standard data analysis run 2015-04-02 (20). Log2-transformed RSEM count values per gene were used for expression analysis (22) and unsupervised hierarchical clustering (Euclidean distance, Ward clustering; ref. 23). Subclasses generated by clustering patients according to MED expression were derived by partitioning the dendrogram with a dynamic tree cutting algorithm (24).

Microarray data from a prostate cancer progression cohort (25) were downloaded from Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/, GSE6919) for log-transformed, prenormalized array signal intensities. Multiple samples per patient were aggregated using the median normalized intensity per gene. A second dataset (GSE21032) included primary prostate carcinoma samples (pPCa; n = 131) and prostate cancer metastases (n = 19) with gene expression z-scores after normalization to normal samples. Clinical data were obtained from http://cbio.mskcc.org/cancergenomics/prostate/ (26). Biochemical disease recurrence based on prostate-specific antigen (PSA) elevation was defined according to Taylor and colleagues (26).

Immunohistochemical analysis

Tissue microarrays (TMA) from paraffin-embedded prostatic tissue provided from cohorts of the University Hospitals of Örebro, Basel, and Tübingen were assessed as described previously (Supplementary Table S3; refs. 14, 27, 28). Approval for the present study was obtained from the ethics committee of the University Hospital of Bonn. Immunohistochemical (IHC) staining was conducted using the Ventana Discovery automated staining system (Ventana Medical System). In brief, slides were incubated at room temperature with primary antibodies: anti-CDK19 rabbit polyclonal (1:100, HPA007053, Sigma-Aldrich) and anti-CDK8 rabbit polyclonal (1:200, A302-501A, Bethyl Laboratories). Primary antibodies were detected with the ultraView Universal DAB Detection Kit (Ventana Medical System). IHC staining quality was validated independently by 2 observers. Samples with a lack of tissue or absence of carcinoma in all cores present on the TMA were excluded.

For quantification, slides were scanned (Panoramic Desk, 3DHistech) and tumor tissue in each sample was marked manually blinded to patient and type of tissue. These user-specified regions of interest were analyzed with the image analysis software Definiens Tissue Studio 2.1 (Definiens Inc.). Average nuclear staining intensity (SI, corresponding to the mean brown chromogen intensity) was quantified as a continuous value (arbitrary units). Specificity of the CDK19 antibody for IHC staining was confirmed by co-incubation with the corresponding CDK19 antigen (APREST71438, Sigma-Aldrich).

Cell lines

Cell lines RWPE (benign prostate) and prostate cancer cell lines PC-3, DU-145, and LnCaP were obtained from the ATCC and maintained according to provider's instructions. Cell lines were authenticated by Multiplex Human Cell Line Authentication Test (June 2013; Multiplexion). C4-2 and 22Rv1 were a kind gift of A. Merseburger and M. Cronauer (UKSH Lübeck).

Quantitative reverse transcription PCR

RNA was isolated using RNeasy Mini Kit (Qiagen) and reverse transcribed using iScript cDNA synthesis kit (Bio-Rad). PCR reactions were performed as duplicates with Power SYBR Green kit (Thermo Fisher Scientific) using the Light-Cycler 480 II (Roche). Expression levels of CDK8 and CDK19 were normalized to β-actin using 2−ΔΔCT method. Primers for CDK8 (29), CDK19 (6), and β-actin (Applied Biosystems):

  • CDK8-Fwd: 5′-CGGGTCGAGGACCTGTTTG-3′, CDK8-Rev: 5′-TGCCGACATAGAGATCCCAGT-3′; CDK19-Fwd: 5′-CGGAACCTATTTTTCACTGTCG-3′, CDK19-Rev: 5′-TGTGGGATATTCTGGCATCTT-3′; β-Actin-Fwd: 5′-CCTGTGGCA TCCACGAAACT-3′, β-actin-Rev: 5′-CACTGTGTTGGCGTACAGGTCTT-3′.

Western blot

Protein extraction and Western blotting were performed as described previously (17); primary antibodies used were: anti-CDK19 rabbit polyclonal (HPA007053, Sigma-Aldrich), anti-CDK8 rabbit polyclonal (A302-501A, Bethyl Laboratories), anti-β-actin (Sigma-Aldrich), anti-vimentin rabbit monoclonal (D21H3, Cell Signaling), and anti-androgen receptor mouse (554225, BD Pharmingen). Development was carried out with horseradish peroxidase (HRP)-conjugated secondary antibodies using ECL Western Blotting Reagent (GE Healthcare).

CDK8 and CDK19 knockdown

Pools of 3 siRNAs targeting CDK8 or CDK19 (CDK8: sc-29267, CDK19/Cdc2L6: sc-72844, Santa Cruz) and a nontargeting scrambled siRNA as control (sc-37007, Santa Cruz) were transfected with Screenfect A (Genaxxon Bioscience GmbH). Optimal CDK8/19 siRNA concentrations (10–50 nmol/L) were determined by Western blotting. For proliferation assays, siRNA transfection was performed in 96-well plates and MTT was added 72 hours later. For Western blot analyses and migration/invasion assays, siRNA transfection was performed in 6-well plates and cells were seeded into migration/invasion chambers 48 hours after transfection. Proteins were extracted 72 hours after transfection.

CDK8/CDK19 inhibitor senexin A

Sensitivity of cell lines to senexin A (Cat. no. 4875, Tocris Bioscience) was assessed by dilution series in 96-well plates compared with DMSO control. Viability was measured by MTT assay 72 hours later. For migration/invasion, cells were pre-treated with senexin A for 24 hours before plating into migration/invasion chambers.

Cell viability assay

Viability following senexin A treatment or knockdown was assessed using thiazolyl blue tetrazolium bromide (MTT). Five thousand PC-3 or 7000 LnCaP cells per well were plated in 96-well plates (Corning), and 500 μg/mL MTT (Sigma-Aldrich) dissolved in PBS (Gibco Life Technologies) was added 72 hours after treatment. Four hours later, 100 μL solubilization buffer [40% vol/vol dimethylformamide (Alfa Aesar), 2% glacial acetic (Merck), 16% SDS (Applichem), pH 4.7)] was added and absorbance at 595 nm was measured the next day. For LnCaP cells, the experiments were performed in medium containing charcoal-stripped FCS (ThermoFisher Scientific) without (=androgen deprivation) and with supplementation of dihydrotestosterone (DHT).

Migration and invasion assays

Cell lines were pretreated with senexin A or subjected to siRNA knockdown as described above. A total of 5 × 105 cells were plated in the upper chamber of migration inserts (VWR) or Matrigel-coated Transwell inserts (VWR) containing 0% FCS medium. The lower chamber was filled with medium containing 10% FCS. After 24 hours, cells were fixed with 4% paraformaldehyde (Merck), stained with hematoxylin (Waldeck), and washed with water. Membranes were scanned and cells were counted using Definiens software (Definiens Inc.).

3′-RNA sequencing

Total RNA of DU-145 cells treated for 48 hours with 10 μmol/L Senexin A or DMSO was extracted and transcriptome sequencing with the 3′ QuantSeq FWD Kit (Lexogen) and data analysis was performed as described previously (30) and in Supplementary Methods. The experiment was performed in triplicates.

Statistical analyses

Group comparisons were done using nonparametric Mann–Whitney U or Kruskal–Wallis test unless indicated otherwise. Migration/Invasion assays were analyzed by t tests. To assess the classification of samples as prostate cancer or cancer of another entity by CDK19 mRNA expression in male patients, a logistic regression model was applied. A Hosmer–Lemshow Goodness-of-Fit Test, a receiver operator curve (ROC), and its area under the curve (AUC) were used to evaluate model accuracy. Association of Mediator subunits among each other and with androgen receptor (AR) expression was evaluated by Spearman rank correlation. All tests were done 2-sided and P values were adjusted for multiple testing using Bonferroni–Holm correction as indicated. Survival was evaluated by Kaplan–Meier estimator and log-rank test and high CDK19 expression of pPCa samples was defined as values greater than 1.5 SDs above the average expression level in normal samples. Adjustment for co-variables was done using Cox regression models (31). All statistical analyses were done with Microsoft Excel and R (21).

Pan-cancer transcriptome profiling of the Mediator complex

To comprehensively analyze the transcriptomic profile of the multiprotein Mediator complex across a large spectrum of the most common tumor entities from thousands of patients, TCGA presents an invaluable resource. To facilitate retrieval of large-scale data sets provided by TCGA and to avoid complicated and error-prone manual editing of data files, we developed an R-package called FirebrowseR. FirebrowseR enables users to query servers at the Broad Institute from within the statistical programming environment R and directly imports desired cohorts and data types for further processing. Using FirebrowseR, RNA-Seq–based expression levels of all 33 Mediator complex subunits across all 23 tumor entities publicly available at TCGA at the time of analysis for all 8,137 patients was obtained (see Supplementary Tables S1 and S2). A heatmap representation of mRNA levels shows that the majority of Mediator complex subunits are expressed at varying degrees across patient samples at medium to high levels (Fig. 1B). A notable exception is MED12L, which shows very low mRNA expression in almost all samples except for a subpopulation of patients with acute myeloid leukemia (LAML, cluster P, Fig. 1B). In an exploratory approach, unsupervised hierarchical clustering of patient samples on the basis of their similarity of Mediator complex subunit expression profiles was performed and patients (rows) of the heatmap were reordered according to these groupings. Relatedness of clusters is indicated by a dendrogram (Fig. 1B). A dynamic partitioning algorithm was used to derive relevant subclasses in these clusters (24). This resulted in identification of 16 subclasses (labeled A-P). Upon inspection, it was evident that whereas some of these subclasses contain a heterogeneous mixture of samples from different tumor entities (e.g., clusters A, D, F, and M), others were primarily composed of samples from a single tumor entity (Fig. 1B). Examples of tumor entities that separated well on the basis of their Mediator complex expression profile were low-grade gliomas (LGG) and glioblastomas (GBM, cluster N), thyroid cancers (THCA, cluster J), hepatocellular carcinoma (LIHC, cluster G), and LAML (cluster P). Cluster L contained the majority of pPCa and no samples of other tumor entities. This cluster was marked by the high expression of CDK19, which is part of the kinase module and has been implicated in prostate cancer before (32). High CDK19 expression was also present in the second group of pPCa samples (black boxes, Fig. 1B). Overall, prostate cancer samples showed highest CDK19 expression levels (P = 1 × 10−224, Supplementary Fig. S1A). It is known that either CDK19 or its paralog CDK8 but not both can be a member of the Mediator complex to form a functioning kinase (5). In contrast to CDK19, highest levels of CDK8 are observed in rectal cancer (READ) and colon cancer (COAD) samples (P = 1 × 10−152, Supplementary Fig. S1B). This is in line with findings indicating a tumor-promoting role of CDK8 in colorectal cancer (9) and corresponds with the group of COAD and READ samples in cluster C marked by high CDK8 expression levels (Fig. 1B). In contrast to CDK19, CDK8 levels were not increased in pPCa compared with other entities (Supplementary Fig. S1B). Within pPCa samples, CDK19 has the highest expression of all subunits, whereas CDK8 is significantly lower (P = 1 × 10−163, Supplementary Fig. S1C). To further unravel the Mediator complex profile in pPCa, all pairwise correlation coefficients of expression levels of Mediator complex subunits were calculated from the TCGA prostate cancer cohort. Hierarchical clustering on the basis of these correlation coefficients revealed 2 major clusters (Fig. 1C). All kinase modules exhibit strong correlation with each other and were grouped in cluster 1. Interestingly, a finer subdivision of the dendrogram revealed that almost all kinase modules fell into cluster 1.1 with the exception of CDK19 and CCNC (Fig. 1C), potentially indicating differences in expression regulation. Because AR signaling is one of the main determinants in prostate cell regulation, we additionally performed a correlation analysis of AR mRNA levels with the expression of Mediator complex subunits. In pPCa, tumors revealed a strong positive correlation of AR expression with several subunits pre-eminently from the kinase module such as CDK8 (r = 0.60, P < 1 × 10−16, Supplementary Fig. S1D). Interestingly, the highest correlation was observed with MED1 (r = 0.81, P < 1 × 10−16), which has been shown to be involved in AR signaling and in prostate cancer before (19). In contrast, other subunits showed a strong negative correlation, whereas CDK19 and cyclin C showed only a weak association with AR expression (r = 0.19, P = 1 × 10−6, Supplementary Fig. S1D). This may indicate a distinct regulation of the kinase module and its constituents in prostate cancer compared with other entities even though directionality and causality of interaction cannot be inferred from correlation.

To further assess the distinct role of CDK19 in pPCa, a logistic regression model was used to predict the probability of a tumor sample from a male patient (n = 3,617) being a prostate cancer or any other entity based on CDK19 mRNA expression alone. Interestingly, the odds for a sample being a prostate cancer were 9.1-fold increased per one-unit increase of (log-transformed) CDK19 expression [P = 2 × 10−11; OR, 9.1, 95% confidence interval (CI), 8.3–11.2]. In addition, the Goodness-of-Fit test was nonsignificant (P = 0.08), thereby not indicating a lack of fit and the ROC had an AUC of 0.93 (Supplementary Fig. S1E and S1F). Overall this provides evidence that CDK19 expression as a single marker can successfully discriminate tumor samples from males as prostate versus non–prostate cancer and emphasizes that CDK19 plays a specific role in prostate cancer.

Evaluation of CDK19 and CDK8 in prostate cancer progression

When comparing the normal prostate samples available from TCGA (n = 52) with TCGA primary tumors (n = 496), CDK19 was significantly increased in both an unpaired analysis (P = 2.1 × 10−6; Fig. 2A) and in an analysis of matched tumor–normal pairs (P = 9.9 × 10−5, Supplementary Fig. S2A). However, analysis of TCGA data regarding expression of CDK19 and its paralog CDK8 in prostate cancer is limited by the restriction to primary tumors. We therefore additionally analyzed publicly available microarray gene expression data (GSE 6919) including normal prostate tissue (n = 81), pPCa (n = 65), and also some metastatic samples (n = 4; ref. 25). CDK19 expression increased significantly during prostate cancer progression (P = 5.4 × 10−5; Supplementary Fig. S2B), whereas CDK8 remained at similar levels between normal and primary tissue and was even decreased in metastases (Supplementary Fig. S2C). To increase sample size, we extended our analysis with a second publicly available microarray data set (GSE 21032; ref. 26) with clinically annotated pPCa (n = 131) and metastases (n = 19). This validated a significantly higher CDK19 expression in metastatic samples than in pPCa (P = 0.002, Fig. 2B). In pPCa, increased CDK19 expression was significantly associated with higher pathologic T stage (T2 vs. ≥T3; P = 0.02, Supplementary Fig. S2D), extracapsular extension (present vs. absent; P = 0.03), presence of an ERG rearrangement (P = 4 × 10−4, Supplementary Fig. S2E), and also with a higher Gleason score (<7 vs. ≥7; P = 0.001, Fig. 2C). CDK19 overexpression in pPCa specimens, defined as 1.5 SDs above the average expression in normal tissues, was moreover associated with a significantly decreased disease-free survival (log-rank: P = 0.002, Fig. 2D) and a 3.4 times increased hazard for biochemical disease recurrence (HR, 3.4, 95% CI; 1.5–7.8; P = 0.004). This increased hazard remained significant even after adjustment for Gleason score and age (CDK19: HR, 2.7; 95% CI; 1.13–6.3; P = 0.02) resulting in 5-year disease-free survival rates of 51.3% and 82.6% for high and low CDK19-expressing patients, respectively.

Figure 2.

Association of CDK19 expression with clinical prostate cancer (PCa) disease stage. CDK19 mRNA levels are increased in primary PCa compared with normal controls (TCGA gene expression; A) and during progression from pPCa to metastatic tissue (GSE 21032; B). C, Association of CDK19 expression with Gleason score. D, Disease-free survival of patients with pPCa with low versus high CDK19 expression (≥1.5 SDs above the average expression in normal prostate tissue).

Figure 2.

Association of CDK19 expression with clinical prostate cancer (PCa) disease stage. CDK19 mRNA levels are increased in primary PCa compared with normal controls (TCGA gene expression; A) and during progression from pPCa to metastatic tissue (GSE 21032; B). C, Association of CDK19 expression with Gleason score. D, Disease-free survival of patients with pPCa with low versus high CDK19 expression (≥1.5 SDs above the average expression in normal prostate tissue).

Close modal

To also assess the expression CDK19 and CDK8 on the protein level and to validate their implication in prostate cancer progression, we subsequently analyzed a large TMA including 622 patients with a substantial number of metastatic and castration-resistant samples (Supplementary Table S3) by IHC. The Mediator complex is predominantly interacting with transcription factors and DNA in the nucleus (3, 19), and following co-incubation of the CDK19 antibody with the recombinant CDK19 antigen, only nuclear expression was considered specific (Supplementary Fig. S2G). Therefore, analysis of specimens was restricted to measurement of average nuclear CDK19 and CDK8 staining intensity and was considered for further analyses (Fig. 3A and B). Comparably to the mRNA data, CDK19 protein expression was low or absent in normal prostate glands but increased significantly from benign tissue to primary tumors (P = 9 × 10−4) and from primary tumors to metastases/castration-resistant prostate cancer (CRPC; P = 4 × 10−10, Fig. 3C). CDK8 expression was comparable in benign and primary tissues (P = 0.28) but was significantly increased in metastases/CRPC (P = 2 × 10−8, Fig. 3D). Subsets of the current cohort had been analyzed for the presence of a ERG rearrangement (deletion or translocation) by FISH and ERG/Ki-67 protein expression by IHC in previous studies (Supplementary Table S3; refs. 27, 28, 33). A combined analysis of the datasets showed that presence of an ERG rearrangement was significantly associated with higher CDK19 levels in primary tumors (adj. P = 3 × 10−7), but high CDK19 levels in more advanced stages were independent of ERG rearrangement status (Fig. 3E). No association of CDK8 expression level and ERG rearrangement was present (Supplementary Fig. S2F). Similarly, the detection of ERG protein expression correlated with significantly increased CDK19 levels in primary tumors (adj. P = 5 × 10−9) but not in increased CDK19 levels in metastases/CRPC (adj. P = 0.51) or in altered CDK8 levels (Supplementary Fig. S2H and S2I). Ki-67, an indicator of high cell division rate and associated with more aggressive phenotype, was also significantly associated with increased CDK19 levels in primary tumors (adj. P = 0.012, Fig. 3F). For lymph node metastases, a trend toward increased CDK19 expression with higher Ki-67 index was present but remained nonsignificant, potentially due to low sample numbers (adj. P = 0.14), whereas no effect was evident in distant metastatic CRPC (Fig. 3F). Overall, these results indicate that CDK19 is elevated already in early prostate cancer progression and that CDK8 protein levels are upregulated in metastatic tumors only. Moreover, CDK19 is significantly associated with ERG rearrangements, ERG protein expression, and with Ki-67 in primary tumors only, but CDK19 levels become independent of ERG in advanced stages.

Figure 3.

IHC staining analysis of CDK19 and CDK8 expression in a large cohort of clinical specimens. A and B, Representative images of the CDK19 and CDK8 IHC staining for benign prostate, pPCa, lymph node metastases (LNPC), and distant metastases of CRPC. −, samples with weak or absent nuclear staining in comparison to samples with strong nuclear expression (+). 20× (top) and 40× (bottom) objective magnification. C, Average nuclear CDK19 staining intensity (mean brown) of tissue samples increases significantly throughout disease stages. D, CDK8 nuclear staining intensity is in contrast only elevated in metastatic CRPC samples. E, In primary tumors, CDK19 protein expression is significantly associated with the presence of an ERG rearrangement (gene fusion or deletion, determined by FISH). This association is lost in more advanced stages. F, In primary cancers, CDK19 staining is significantly increased with higher Ki-67 index, a marker of cell proliferation, aggressive behavior, and worse outcome. Of note, benign prostate samples all had a Ki-67 index of 0%, whereas distant metastases/CRPCs had all more than 0% Ki-67–positive cells.

Figure 3.

IHC staining analysis of CDK19 and CDK8 expression in a large cohort of clinical specimens. A and B, Representative images of the CDK19 and CDK8 IHC staining for benign prostate, pPCa, lymph node metastases (LNPC), and distant metastases of CRPC. −, samples with weak or absent nuclear staining in comparison to samples with strong nuclear expression (+). 20× (top) and 40× (bottom) objective magnification. C, Average nuclear CDK19 staining intensity (mean brown) of tissue samples increases significantly throughout disease stages. D, CDK8 nuclear staining intensity is in contrast only elevated in metastatic CRPC samples. E, In primary tumors, CDK19 protein expression is significantly associated with the presence of an ERG rearrangement (gene fusion or deletion, determined by FISH). This association is lost in more advanced stages. F, In primary cancers, CDK19 staining is significantly increased with higher Ki-67 index, a marker of cell proliferation, aggressive behavior, and worse outcome. Of note, benign prostate samples all had a Ki-67 index of 0%, whereas distant metastases/CRPCs had all more than 0% Ki-67–positive cells.

Close modal

Functional evaluation of CDK19 and CDK8 in prostate cancer cell lines

To evaluate the functional role of the paralogs CDK19 and CDK8 in prostate cancer, first the benign cell line RWPE and the 2 malignant cell lines LnCaP (androgen-sensitive) and PC-3 (androgen-independent) were investigated. Similar to the IHC and mRNA results in patient samples, mRNA and protein expression of CDK19 were higher in the malignant cell lines (Fig. 4A and B). Interestingly, CDK19 and CDK8 increased in LnCaP cells upon androgen deprivation (Fig. 4C). siRNA-mediated knockdown of CDK19 and CDK8 did not impact proliferation of either LnCaP or PC-3 cells (Fig. 4D and E). To investigate involvement of CDK8 and CDK19 in metastatic spread, Boyden chamber assays were used to assess invasion and migration. This was done with PC-3 cells, as RWPE and LnCaP cells showed no migratory potential. Following a transient CDK19 knockdown, both migration and invasion of PC-3 cells were significantly reduced to 44% (P = 6 × 10−4) and 58% (P = 0.01) compared with control cells, respectively (Fig. 4F). The effect of CDK8 knockdown led to similar but less pronounced effects (migration: 61%, P = 0.02; invasion: 68%, P = 0.06). To further assess CDK8 and CDK19 as potential therapeutic targets in prostate cancer, we used the selective CDK8/CDK19 inhibitor senexin A (11) and extended analyses in a larger panel of prostate cancer cell lines (LnCaP, C4-2, 22RV1, PC-3, DU-145; refs. 14, 34). All cell lines expressed CDK8/CDK19 (Supplementary Fig. S3A). In accordance with the knockdown results, senexin A treatment showed only modest impact on proliferation (Supplementary Fig. S3B) with slightly higher effects in LnCaPs after androgen deprivation (Supplementary Fig. S3C). However, pretreatment of cells with senexin A significantly reduced migration and invasion in all assessable cell lines except for C4-2 (Fig. 5A) with dose-dependent effects in PC-3 cells (P = 0.01, Supplementary Fig. S3D). Effects were most pronounced in DU-145 (migration: 21% of control, P < 0.001; invasion: 18%, P < 0.001; Fig. 5A). Accordingly, treatment of DU-145 with senexin A, knockdown of CDK19 and to a lesser extent CDK8 led to reduced vimentin expression (Fig. 5B), which has been described as a relevant determinant of prostate cancer migratory capacity (35). To systematically investigate the mechanism of action, senexin A–treated DU-145 cells were subjected to RNA sequencing. Among the genes most strongly upregulated, several have previously been implicated in reduced migration of cancer cells or as being repressed during prostate cancer progression such as such as TAGLN (36, 37), CDH1 (38, 39), CDH3 (40), TPM1 (41), or MYL-9 (refs. 39, 42; Fig. 5C). A systematic evaluation of transcriptional changes using an enrichment analysis of Gene Ontology (GO) terms and KEGG pathways revealed that upregulated genes were enriched for processes governing cytoskeletal structure, modulation of cell motility and focal adhesion, whereas downregulated processes were primarily involved in transcription and translation (Fig. 5D). This is in line with the observed phenotypic changes and the role of the Mediator complex in transcriptional control and provides an explanation how interference with the kinase module of the Mediator complex may impact prostate cancer cell migration.

Figure 4.

In vitro analysis of CDK19 and CDK8 in prostate cancer (PCa) cell lines. A and B, CDK19 mRNA (mean + SD) and protein levels as measured by reverse transcriptase qPCR normalized to β-actin using the 2−ΔCT method and by Western blotting were increased in malignant cell lines LnCaP (androgen-sensitive) and PC-3 (androgen-independent) compared with the benign cell line RWPE (n = 3). CDK8 levels were similar in all cell lines. C, CDK19, CDK8, and AR expression in LnCaP cells after androgen deprivation (AD) or after treatment with DHT at indicated concentrations. D, Transient knockdown of CDK19 or CDK8 by siRNA was established at the respective optimal concentrations and verified by Western blotting for LnCaP and PC-3 cells. E, No effect of CDK19 or CDK8 knockdown on cell viability after 72 hours as measured by MTT assay in comparison to scrambled (scr) RNA control (n = 3; mean + SD). F, Migration and invasion were significantly decreased in PC-3 cells after CDK19 and CDK8 knockdown compared with control (n = 4, mean + SD).

Figure 4.

In vitro analysis of CDK19 and CDK8 in prostate cancer (PCa) cell lines. A and B, CDK19 mRNA (mean + SD) and protein levels as measured by reverse transcriptase qPCR normalized to β-actin using the 2−ΔCT method and by Western blotting were increased in malignant cell lines LnCaP (androgen-sensitive) and PC-3 (androgen-independent) compared with the benign cell line RWPE (n = 3). CDK8 levels were similar in all cell lines. C, CDK19, CDK8, and AR expression in LnCaP cells after androgen deprivation (AD) or after treatment with DHT at indicated concentrations. D, Transient knockdown of CDK19 or CDK8 by siRNA was established at the respective optimal concentrations and verified by Western blotting for LnCaP and PC-3 cells. E, No effect of CDK19 or CDK8 knockdown on cell viability after 72 hours as measured by MTT assay in comparison to scrambled (scr) RNA control (n = 3; mean + SD). F, Migration and invasion were significantly decreased in PC-3 cells after CDK19 and CDK8 knockdown compared with control (n = 4, mean + SD).

Close modal
Figure 5.

Inhibition of CDK19 and CDK8 in prostate cancer (PCa) cell lines with senexin A. A, Migration and invasion of PCa cell lines in Transwell Boyden chamber assays following 24-hour senexin A (10 μmol/L) pretreatment compared with DMSO control (n = 3, mean + SD). 22Rv1 cells did not invade under control or treatment conditions. n.s., nonsignificant; *, P < 0.05; **, P < 0.01; ***, P < 0.001. B, Treatment of DU-145 cells with senexin A (10 μmol/L) or siRNA knockdown of CDK19/CDK8 led to reduced expression of vimentin. C, Volcano plot of RNAseq results for senexin A versus DMSO-treated DU-145 cells (n = 3) indicating log2 fold change (x-axis) and false discovery rate (FDR)-adjusted significance (y-axis). The 15 most significantly up- or downregulated genes are indicated by name. D, Top five most significantly enriched GO processes and KEGG pathways inferred from genes significantly (adj. P < 0.01) up- (Up) or downregulated (Down). Dashed line indicates adjusted GO enrichment: P = 0.05; ***, adjusted P < 0.001.

Figure 5.

Inhibition of CDK19 and CDK8 in prostate cancer (PCa) cell lines with senexin A. A, Migration and invasion of PCa cell lines in Transwell Boyden chamber assays following 24-hour senexin A (10 μmol/L) pretreatment compared with DMSO control (n = 3, mean + SD). 22Rv1 cells did not invade under control or treatment conditions. n.s., nonsignificant; *, P < 0.05; **, P < 0.01; ***, P < 0.001. B, Treatment of DU-145 cells with senexin A (10 μmol/L) or siRNA knockdown of CDK19/CDK8 led to reduced expression of vimentin. C, Volcano plot of RNAseq results for senexin A versus DMSO-treated DU-145 cells (n = 3) indicating log2 fold change (x-axis) and false discovery rate (FDR)-adjusted significance (y-axis). The 15 most significantly up- or downregulated genes are indicated by name. D, Top five most significantly enriched GO processes and KEGG pathways inferred from genes significantly (adj. P < 0.01) up- (Up) or downregulated (Down). Dashed line indicates adjusted GO enrichment: P = 0.05; ***, adjusted P < 0.001.

Close modal

Recently, an increasing number of Mediator complex subunits have been identified as relevant components in a range of human diseases including different cancer entities (3, 4, 7). To foster the knowledge of Mediator complex biology in human cancers and to evaluate its complex and potentially entity-specific subunit composition, we performed a pan-cancer analysis of Mediator complex transcriptome profiles. To facilitate entity-wide analyses using TCGA data, we developed the R-package FirebrowseR, which we made available to the public. In an exploratory approach, RNA-Seq expression data of all 33 MED subunits for more than 8,000 patients across 23 cancer entities available from TCGA were subjected to an unsupervised hierarchical clustering and closely related subgroups were determined by a dynamic partitioning algorithm (24). Using mRNA expression data, other modes of changes in the Mediator complex such as mutations and copy number alterations cannot be evaluated and the correspondence of mRNA levels to protein levels and to MED composition can be inferred only indirectly. Nevertheless, this analysis resulted in several subclusters that were almost entirely composed of only one tumor entity, whereas others showed a more heterogeneous mixture of samples from different entities. This segregation of tumor entities solely by MED expression indicates that the Mediator complex plays distinct roles in different cancer types and allows generation of several novel hypotheses regarding MED biology and potential therapeutic targets. The LAML samples (cluster P) are for example further subdivided into 2 groups on the basis of differences in MED12L expression, which presents a valuable topic for further research. Very high CDK8 expression was observed in the group of colorectal cancers (cluster C, Fig. 1B, Supplementary Fig. S1B), which is in accordance with prior data (9, 43) and presents validation for our approach. One of the entities resulting in a very homogenous group was pPCa (cluster L, Fig. 1B). Compared with all other entities, pPCa samples were marked by exceptionally high CDK19 expression and an increase from normal prostate tissue to pPCa, which indicates a distinct role of this kinase subunit in prostate cancer and a potential candidate for targeted cancer-specific therapy. High CDK19 expression in prostate cancer has also been observed by studies not primarily investigating the Mediator complex (32, 44) and was linked to migratory potential in prostate cancer cells (45). However, an in-depth evaluation of CDK19 had not been performed yet.

For confirmation of the TCGA results, additional cohorts including normal tissue, primary tumors, and metastatic cancers of publicly available mRNA expression data and an IHC cohort of more than 600 clinical prostate specimens were used. Because of the predominant localization and function of the Mediator complex in the nucleus (3, 19), nuclear protein expression was evaluated in the IHC. We could thereby validate the findings regarding CDK19 and CDK8 in pPCa and also established a further increase of CDK19 in metastatic and CRPC (Figs. 2 and 3). CDK8 was in contrast strongly elevated in distant metastases only on a protein level but showed slightly decreased mRNA expression. Very interestingly, in the 2 cohorts with information on ERG rearrangement, CDK19 was significantly associated with ERG rearrangements and ERG expression in primary tumors but not in more advanced stages. This is in agreement with the observation that rearrangements and expression of ERG, a member of the ETS transcription factor family, are early events in prostate cancer carcinogenesis but most likely lose relevance in more advanced stages (46). Similar to our results, a recent study reported CDK19 as 1 of only 2 candidate genes associated with invasion and aggressiveness from in vivo and in vitro gene expression studies of prostate tumors with rearrangement of another gene from the ETS transcription factor family called ETV-1 (47). In addition, in primary tumors, a correlation of CDK19 levels with Ki-67 protein expression was observed, which has been linked to aggressive prostate cancer and was shown to be an independent prognostic biomarker (48). Given the low impact of CDK19 and CDK8 knockdown and inhibition on prostate cancer cell viability in vitro suggests that CDK19 and CDK8 do not causally influence proliferation but are instead concordantly upregulated in more aggressive tumors.

Because of the limited availability of patient data and follow-up information for most of the datasets, evaluation of clinical associations and outcome could unfortunately only be done for one of the cohorts. In this cohort, CDK19 expression was significantly associated with clinical parameters of malignancy such as advanced stage and higher pathologic Gleason score. Moreover, high CDK19 expression was significantly correlated to worse outcome for patients with pPCa with a 30% absolute difference of freedom from disease at 5 years. It has to be noted, however, that due to the low incidence of fatal events, biochemical recurrence was used as a surrogate endpoint and that the cohort consisted exclusively of patients who had undergone surgical prostatectomy (26). Extrapolation of conclusions drawn from these findings to other populations and different primary treatment strategies such as “watchful waiting” is therefore limited. Instead, these association and potential causal links with CDK19 should be investigated in subsequent molecular biological and epidemiologic studies, ideally in a prospective manner.

In prostate cell lines, increased expression of CDK19 mRNA and protein could be detected in the malignant cell lines compared with nonmalignant RWPE cells. Interestingly, proliferation of prostate cancer cell lines was not reduced following knockdown of CDK8 or CDK19 and only modestly impacted by treatment with the selective CDK8/CDK19 small-molecule inhibitor senexin A (11) with viability reduction of 50% at concentrations well above 20 μmol/L. This is in contrast to colon cancer and melanoma cell lines with high endogenous CDK8 levels in which it has been shown that CDK8 knockdown leads to reduced cell proliferation (9, 29). Other reports paradoxically suggest that CDK8 may possess tumor-suppressive properties (17, 49). For CDK19, no such knowledge has been available. In contrast, strong reductions of migration and invasion, which are essential traits in metastatic spread and constitute an aggressive phenotype, could be achieved in PC-3 cells with CDK19 and CDK8 knockdown (Fig. 4E and F) and by pretreatment with senexin A in 22Rv1, PC-3, and especially in DU-145 cells (Fig. 5A). This inhibition of migration and invasion corresponds well to the high expression of CDK19 and CDK8 in metastatic samples and indicates their functional role in prostate cancer. Interestingly, the increasing recognition of the Mediator complex and its kinase module as a potential therapeutic cancer target has recently led to the development of additional selective CDK8/CDK19 inhibitors (10), which could further enhance research into novel targeted treatment strategies for patients with prostate cancer. A focused assessment of vimentin, a major determinant of prostate cancer motility (35), showed decreased protein expression in DU-145 cells after senexin A treatment and knockdown of CDK19 or CDK8. Moreover, a systematic investigation via RNA sequencing following senexin A treatment showed differential regulation of several genes and processes involved in cancer cell migration, cytoskeleton, and focal adhesion (Fig. 5C and D; refs. 36–42). These observations therefore provide a mechanism for decreased motility, which corresponds well to the observed phenotype and the expected role of the Mediator complex. Interestingly, some genes differentially expressed in response to senexin A have been linked to Wnt/β-catenin pathway before (50). Nevertheless, it should be kept in mind that RNA interference and small-molecular inhibitors may possess off-target effects and that interference with a fundamental cellular component such as the Mediator complex may provoke differential responses depending on the cell type. Further, evaluation of the MED kinase module in prostate cancer and other entities will therefore be needed to gain an improved understanding of the Mediator complex as a therapeutic target in cancer.

Taken together, our data for the first time provide evidence that tumor entities can be differentiated by their Mediator complex expression profile and that CDK19—and to a lesser extent its paralog CDK8—plays an important and distinct role in prostate cancer, which gains impact throughout disease progression. Thereby, our study yields new insights into the differential regulation of the Mediator complex across tumor entities and its role in human disease, fosters the understanding of prostate cancer biology, and moreover presents CDK19 and CDK8 as novel potential therapeutic targets especially in the metastatic setting.

No potential conflicts of interest were disclosed.

Conception and design: J. Brägelmann, N. Klümper, I. Syring, P. Brossart, S. Perner

Development of methodology: J. Brägelmann, N. Klümper, S. Perner

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Brägelmann, A. von Mässenhausen, A. Queisser, I. Syring O. Andrén, P. Brossart, M.A. Svensson, J. Kirfel, S. Perner

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Brägelmann, N. Klümper, A. Offermann, M. Deng, I. Syring, A.S. Merseburger, J. Carlsson, S. Duensing, J. Kirfel, S. Perner

Writing, review, and/or revision of the manuscript: J. Brägelmann, N. Klümper, A. Offermann, I. Syring, A.S. Merseburger, I. Vlasic, J. Carlsson, O. Andrén, P. Brossart, S. Duensing, M. Svensson, S. Perner

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Offermann, D. Böhm, M. Deng, C. Sanders, I. Syring, W. Vogel, E. Sievers O. Andrén, P. Brossart, M.A. Svensson, S. Perner

Study supervision: S. Perner

Other (review/revision of article): D. Adler

The results shown here are in part based upon data generated by the TCGA Research Network: http://cancergenome.nih.gov/. We thank Anke Röper, Kerstin Ludwig and Sugirthan Sivalingam for help with experiments and Marcus Cronauer for C4-2 and 22Rv1 cells.

The study was supported by a grant of the Rudolf Becker-Foundation, a grant of the Wilhelm Sander Foundation (2011.077.2), and of the German Research Foundation (DFG PE1179/9-1 and PE1179/11-1) to S. Perner. I. Syring was supported by the Ferdinand Eisenberger-Fellowship of the German Society of Urology (DGU, SYI1/FE-13). J. Brägelmann was supported by a Gerok-Fellowship grant (2014-11-06) of the Medical Faculty of the University of Bonn. A. Offermann was supported by a medical doctoral fellowship grant (BONFOR) of the Medical Faculty of the University of Bonn and C. Sanders by a “Mildred-Scheel medical doctoral programme” grant of the German Cancer Aid.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Poss
ZC
,
Ebmeier
CC
,
Taatjes
DJ
.
The Mediator complex and transcription regulation
.
Crit Rev Biochem Mol Biol
2013
;
48
:
575
608
.
2.
Malik
S
,
Roeder
RG
.
The metazoan Mediator co-activator complex as an integrative hub for transcriptional regulation
.
Nat Rev Genet
2010
;
11
:
761
72
.
3.
Allen
BL
,
Taatjes
DJ
.
The Mediator complex: a central integrator of transcription
.
Nat Rev Mol Cell Biol
2015
;
16
:
155
66
.
4.
Schiano
C
,
Casamassimi
A
,
Rienzo
M
,
de Nigris
F
,
Sommese
L
,
Napoli
C
.
Involvement of Mediator complex in malignancy
.
Biochim Biophys Acta
2014
;
1845
:
66
83
.
5.
Daniels
DL
,
Ford
M
,
Schwinn
MK
,
Benink
H
.
Mutual exclusivity of MED12/MED12L, MED13/13L, and CDK8/19 paralogs revealed within the CDK-Mediator kinase module
.
J Proteomics Bioinform
2013;S2:004
.
6.
Mukhopadhyay
A
,
Kramer
JM
,
Merkx
G
,
Lugtenberg
D
,
Smeets
DF
,
Oortveld
MAW
, et al
.
CDK19 is disrupted in a female patient with bilateral congenital retinal folds, microcephaly and mild mental retardation
.
Hum Genet
2010
;
128
:
281
91
.
7.
Syring
I
,
Klümper
N
,
Offermann
A
,
Braun
M
,
Deng
M
,
Boehm
D
, et al
.
Comprehensive analysis of the transcriptional profile of the Mediator complex across human cancer types
.
Oncotarget
2016
;
7
:
23043
55
.
8.
Xu
W
,
Wang
Z
,
Zhang
W
,
Qian
K
,
Li
H
,
Kong
D
, et al
.
Mutated K-ras activates CDK8 to stimulate the epithelial-to-mesenchymal transition in pancreatic cancer in part via the Wnt/β-catenin signaling pathway
.
Cancer Lett
2015
;
356
:
613
27
.
9.
Firestein
R
,
Bass
AJ
,
Kim
SY
,
Dunn
IF
,
Silver
SJ
,
Guney
I
, et al
.
CDK8 is a colorectal cancer oncogene that regulates β-catenin activity
.
Nature
2008
;
455
:
547
51
.
10.
Dale
T
,
Clarke
PA
,
Esdar
C
,
Waalboer
D
,
Adeniji-Popoola
O
,
Ortiz-Ruiz
M-J
, et al
.
A selective chemical probe for exploring the role of CDK8 and CDK19 in human disease
.
Nat Chem Biol
2015
;
11
:
973
80
.
11.
Porter
DC
,
Farmaki
E
,
Altilia
S
,
Schools
GP
,
West
DK
,
Chen
M
, et al
.
Cyclin-dependent kinase 8 mediates chemotherapy-induced tumor-promoting paracrine activities
.
Proc Natl Acad Sci U S A
2012
;
109
:
13799
804
.
12.
Tan
J
,
Ong
CK
,
Lim
WK
,
Ng
CCY
,
Thike
AA
,
Ng
LM
, et al
.
Genomic landscapes of breast fibroepithelial tumors
.
Nat Genet
2015
;
47
:
1341
5
.
13.
Turunen
M
,
Spaeth
JM
,
Keskitalo
S
,
Park
MJ
,
Kivioja
T
,
Clark
AD
, et al
.
Uterine leiomyoma-linked MED12 mutations disrupt mediator-associated CDK activity
.
Cell Rep
2014
;
7
:
654
60
.
14.
Shaikhibrahim
Z
,
Offermann
A
,
Braun
M
,
Menon
R
,
Syring
I
,
Nowak
M
, et al
.
MED12 overexpression is a frequent event in castration-resistant prostate cancer
.
Endocr Relat Cancer
2014
;
21
:
663
75
.
15.
Shaikhibrahim
Z
,
Offermann
A
,
Halbach
R
,
Vogel
W
,
Braun
M
,
Kristiansen
G
, et al
.
Clinical and molecular implications of MED15 in head and neck squamous cell carcinoma
.
Am J Pathol
2015
;
185
:
1114
22
.
16.
Shaikhibrahim
Z
,
Menon
R
,
Braun
M
,
Offermann
A
,
Queisser
A
,
Boehm
D
, et al
.
MED15, encoding a subunit of the mediator complex, is overexpressed at high frequency in castration-resistant prostate cancer
.
Int J Cancer
2013
;
135
:
19
26
.
17.
Li
N
,
Fassl
A
,
Chick
J
,
Inuzuka
H
,
Li
X
,
Mansour
MR
, et al
.
Cyclin C is a haploinsufficient tumour suppressor
.
Nat Cell Biol
2014
;
16
:
1080
91
.
18.
Cui
J
,
Germer
K
,
Wu
T
,
Wang
J
,
Luo
J
,
Wang
SC
, et al
.
Cross-talk between HER2 and MED1 regulates tamoxifen resistance of human breast cancer cells
.
Cancer Res
2012
;
72
:
5625
34
.
19.
Vijayvargia
R
,
May
MS
,
Fondell
JD
.
A coregulatory role for the Mediator complex in prostate cancer cell proliferation and gene expression
.
Cancer Res
2007
;
67
:
4034
41
.
20.
Broad Institute TCGA Genome Data Analysis Center
.
Analysis-ready standardized TCGA data from Broad GDAC Firehose stddata__2015_04_02 run
.
Broad Inst MIT Harvard Dataset
2015
. http://doi.org/10.7908/C15X282W.
21.
R Development Core Team
.
R: A language and environment for statistical computing [Internet]
.
Vienna, Austria
:
R Foundation for Statistical Computing
;
2011
.
Available from
: http://www.R-project.org/.
22.
Li
B
,
Dewey
CN
.
RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome
.
BMC Bioinformatics
2011
;
12
:
323
.
23.
Murtagh
F
,
Legendre
P
.
Ward's hierarchical agglomerative clustering method: which algorithms implement ward's criterion?
J Classif
2014
;
31
:
274
95
.
24.
Langfelder
P
,
Zhang
B
,
Horvath
S
.
Defining clusters from a hierarchical cluster tree: the Dynamic Tree Cut package for R
.
Bioinformatics
2008
;
24
:
719
20
.
25.
Chandran
UR
,
Ma
C
,
Dhir
R
,
Bisceglia
M
,
Lyons-Weiler
M
,
Liang
W
, et al
.
Gene expression profiles of prostate cancer reveal involvement of multiple molecular pathways in the metastatic process
.
BMC Cancer
2007
;
7
:
64
.
26.
Taylor
BS
,
Schultz
N
,
Hieronymus
H
,
Gopalan
A
,
Xiao
Y
,
Carver
BS
, et al
.
Integrative genomic profiling of human prostate cancer
.
Cancer Cell
2010
;
18
:
11
22
.
27.
Nowak
M
,
Svensson
MA
,
Carlsson
J
,
Vogel
W
,
Kebschull
M
,
Wernert
N
, et al
.
Prognostic significance of phospho-histone H3 in prostate carcinoma
.
World J Urol
2013
;
32
:
703
7
.
28.
Braun
M
,
Scheble
VJ
,
Menon
R
,
Scharf
G
,
Wilbertz
T
,
Petersen
K
, et al
.
Relevance of cohort design for studying the frequency of the ERG rearrangement in prostate cancer
.
Histopathology
2011
;
58
:
1028
36
.
29.
Kapoor
A
,
Goldberg
MS
,
Cumberland
LK
,
Ratnakumar
K
,
Segura
MF
,
Emanuel
PO
, et al
.
The histone variant macroH2A suppresses melanoma progression through regulation of CDK8
.
Nature
2010
;
468
:
1105
9
.
30.
Moll
P
,
Ante
M
,
Seitz
A
,
Reda
T
.
.
QuantSeq 3 [prime] mRNA sequencing for RNA quantification
.
Nat Methods
2014
;
11.12
.
31.
Cox
DR
.
Regression models and life-tables
.
J Royal Stat Soc B (Methodological)
1972
;
34
:
187
220
.
32.
Saeed
K
,
Östling
P
,
Björkman
M
,
Mirtti
T
,
Alanen
K
,
Vesterinen
T
, et al
.
Androgen receptor-interacting protein HSPBAP1 facilitates growth of prostate cancer cells in androgen-deficient conditions
.
Int J Cancer
2014
;
136
:
2535
45
.
33.
Scheble
VJ
,
Scharf
G
,
Braun
M
,
Ruiz
C
,
Stürm
S
,
Petersen
K
, et al
.
ERG rearrangement in local recurrences compared to distant metastases of castration-resistant prostate cancer
.
Virchows Arch
2012
;
461
:
157
62
.
34.
Zarif
JC
,
Lamb
LE
,
Schulz
VV
,
Nollet
EA
,
Miranti
CK
.
Androgen receptor non-nuclear regulation of prostate cancer cell invasion mediated by Src and matriptase
.
Oncotarget
2015
;
6
:
6862
76
.
35.
Wei
J
,
Xu
G
,
Wu
M
,
Zhang
Y
,
Li
Q
,
Liu
P
, et al
.
Overexpression of vimentin contributes to prostate cancer invasion and metastasis via src regulation
.
Anticancer Res
2008
;
28
:
327
34
.
36.
Prasad
PD
,
Stanton
J-AL
,
Assinder
SJ
.
Expression of the actin-associated protein transgelin (SM22) is decreased in prostate cancer
.
Cell Tissue Res
2010
;
339
:
337
47
.
37.
Thompson
O
,
Moghraby
JS
,
Ayscough
KR
,
Winder
SJ
.
Depletion of the actin bundling protein SM22/transgelin increases actin dynamics and enhances the tumourigenic phenotypes of cells
.
BMC Cell Biol
2012
;
13
:
1
.
38.
Umbas
R
,
Schalken
JA
,
Aalders
TW
,
Carter
BS
.
Expression of the cellular adhesion molecule E-cadherin is reduced or absent in high-grade prostate cancer
.
Cancer Res
1992
;
52
:
5104
9
.
39.
Gorlov
IP
,
Byun
J
,
Gorlova
OY
,
Aparicio
AM
,
Efstathiou
E
,
Logothetis
CJ
.
Candidate pathways and genes for prostate cancer: a meta-analysis of gene expression data
.
BMC Med Genomics
2009
;
2
:
48
.
40.
Jarrard
DF
,
Paul
R
,
van Bokhoven
A
,
Nguyen
SH
,
Bova
GS
,
Wheelock
MJ
, et al
.
P-Cadherin is a basal cell-specific epithelial marker that is not expressed in prostate cancer
.
Clin Cancer Res
1997
;
3
:
2121
8
.
41.
Raval
GN
,
Bharadwaj
S
,
Levine
EA
,
Willingham
MC
,
Geary
RL
,
Kute
T
, et al
.
Loss of expression of tropomyosin-1, a novel class II tumor suppressor that induces anoikis, in primary breast tumors
.
Oncogene
2003
;
22
:
6194
203
.
42.
Stephenson
AJ
,
Smith
A
,
Kattan
MW
,
Satagopan
J
,
Reuter
VE
,
Scardino
PT
, et al
.
Integration of gene expression profiling and clinical variables to predict prostate carcinoma recurrence after radical prostatectomy
.
Cancer
2005
;
104
:
290
8
.
43.
Firestein
R
,
Shima
K
,
Nosho
K
,
Irahara
N
,
Baba
Y
,
Bojarski
E
, et al
.
CDK8 expression in 470 colorectal cancers in relation to β-catenin activation, other molecular alterations and patient survival
.
Int J Cancer
2010
;
126
:
2863
73
.
44.
Sowalsky
AG
,
Xia
Z
,
Wang
L
,
Zhao
H
,
Chen
S
,
Bubley
GJ
, et al
.
Whole transcriptome sequencing reveals extensive unspliced mRNA in metastatic castration-resistant prostate cancer
.
Mol Cancer Res
2015
;
13
:
98
106
.
45.
Nishikawa
R
,
Goto
Y
,
Kurozumi
A
,
Matsushita
R
,
Enokida
H
,
Kojima
S
, et al
.
MicroRNA-205inhibits cancer cell migration and invasion via modulation of centromere protein Fregulating pathways in prostate cancer
.
Int J Urol
2015
;
22
:
867
77
.
46.
Perner
S
,
Mosquera
JM
,
Demichelis
F
,
Hofer
MD
,
Paris
PL
,
Simko
J
, et al
.
TMPRSS2-ERG fusion prostate cancer: an early molecular event associated with invasion
.
Am J Surg Pathol
2007
;
31
:
882
8
.
47.
Mesquita
D
,
Barros-Silva
JD
,
Santos
J
,
Skotheim
RI
,
Lothe
RA
,
Paulo
P
, et al
.
Specific and redundant activities of ETV1 and ETV4 in prostate cancer aggressiveness revealed by co-overexpression cellular contexts
.
Oncotarget
2015
;
6
:
5217
36
.
48.
Zhao
L
,
Yu
N
,
Guo
T
,
Hou
Y
,
Zeng
Z
,
Yang
X
, et al
.
Tissue biomarkers for prognosis of prostate cancer: a systematic review and meta-analysis
.
Cancer Epidemiol Biomarkers Prev
2014
;
23
:
1047
54
.
49.
McCleland
ML
,
Soukup
TM
,
Liu
SD
,
Esensten
JH
,
de Sousa e Melo
F
,
Yaylaoglu
M
, et al
.
Cdk8 deletion in the Apc(Min) murine tumour model represses EZH2 activity and accelerates tumourigenesis
.
J Pathol
2015
;
237
:
508
19
.
50.
Herbst
A
,
Jurinovic
V
,
Krebs
S
,
Thieme
SE
,
Blum
H
,
Göke
B
, et al
.
Comprehensive analysis of β-catenin target genes in colorectal carcinoma cell lines with deregulated Wnt/β-catenin signaling
.
BMC Genomics
2014
;
15
:
74
.

Supplementary data