Activation of EZH2 and SUZ12 Regulated by E2F1 Predicts the Disease Progression and Aggressive Characteristics of Bladder Cancer

Bladder cancer is the sixth most common cancer in men and women populations. In 2013, 72,570 new cases of bladder cancer were diagnosed and 15,210 deaths were due to bladder cancer in the United States (1). This cancer is characterized by 2 histologically distinct subtypes: non–muscle-invasive bladder cancer (NMIBC) and muscle-invasive bladder cancer (MIBC) at initial diagnosis (2). NMIBC is a heterogeneous disease (3), and patients frequently experience disease recurrence and 10% to 30% of them progress to MIBC, which is responsible for most bladder cancer–specific deaths (2). Because MIBC frequently leads to distant metastases (4), a major focus of research has been to understand the mechanisms that promote cancer progression. Although there have been many efforts to construct a robust model to predict progression of NMIBC using clinical information and pathologic classification (3, 5–7), precisely predicting the behavior of heterogeneous NMIBC remains challenging.

Previously, our genome-wide gene expression profile study using microarray technologies successfully identified a gene expression signature that could predict the likelihood of progression of NMIBC (8). Expression of E2F1 was significantly upregulated in the MIBC subtype, strongly indicating that activation of E2F1 might be a critical genetic event in the development of or progression to MIBC (8). Because E2F1 was not uniformly absent in all NMIBCs, we re-examined expression of E2F1 in NMIBCs and subdivided the patients into 2 groups according to the expression level of E2F1. The progression rate in the E2F1-high groups was profoundly higher than in the E2F1-low group, showing that E2F1 is strongly associated with NMIBC-to-MIBC progression (8).

In this study, on the basis of gene-to-gene network analysis, we found that the expression levels of EZH2 and SUZ12, binding partners of polycomb complex PRC2 (polycomb repressive complex 2) and direct targets of E2F1, were significantly higher in the E2F1-high subgroup than in the E2F1-low subgroup. Overexpression of several PcG proteins has been associated with many tumors and has also been identified as prognostic indicator in several tumors (9–13). The PRC2 core components are known such as EZH2, SUZ12, EED, and RBBP4 or RBBP7, which catalyze trimethylation of histone H3 lysine 27 (H3K27me3; ref. 14). Several studies reported that PRC2 was overexpressed in numerous cancer types and played a critical role in the aberrant silencing of tumor suppressor genes (15, 16).

Higher expression of the EZH2 and SUZ12 genes is clearly associated with tumor progression and overall survival (OS) in bladder cancer, but other genes including EED did not show significant level of prediction in this study. Therefore, we investigated the biologic activities of EZH2 and SUZ12, whose expression was significantly associated with poor prognosis and reflected the aggressive characteristics of bladder cancer. Moreover, we examined these genes' downstream pathways to mediate NMIBC progression, illustrating that E2F1 and EZH2 activated cancer stem cell (CSC) signaling pathways in anticancer drug-treated environments. Thus, we suggest that the transcriptional changes of E2F1, EZH2, and SUZ12 clearly predict bladder cancer aggressiveness, as well as anticancer drug resistance.

Human bladder cancer cell lines (EJ and 5637) were obtained from the ATCC. Other cell lines (UC5 and UC9) were provided by H. Barton Grossman (Department of Urology, University of Texas MD Anderson Cancer Center, Houston, TX: deposited into Public Health England, United Kingdom). The cells in this study were used within 6 months in our laboratory and were obtained from a cell bank that performed cell line characterizations. Cells from ATCC were certificated by the results of the short tandem repeat (STR) DNA profiling assay, cytochrome c oxidase I assay, and mycoplasma contamination assay. Eleven of UC series cells were characterized by the STR-PCR method and for mycoplasma contamination.

UC5, UC9 (NMIBC cells), and EJ (MIBC cells) were cultured in DMEM (Hyclone) supplemented with 10% FBS (Hyclone) and 1% penicillin/streptomycin (Hyclone). 5637 (MIBC) cells were cultured in RPMI-1640 medium (Hyclone).

We used a gene expression dataset (GSE13507, n = 256) containing 165 primary patients with bladder cancer in a previous study (8). Among the 165 cancers, 102 were histopathologically proven to be primary NMIBC and remained 63 were primary MIBC [GSE13507; the Korean cohort, n = 165 (102 NMIBCs and 63 MIBCs)]. Clinical data including progression-free survival (PFS), updated in January 2010, were obtained from the Chungbuk National University Hospital. To validate a prognostic value of the signature, 3 other gene expression datasets of patients with bladder cancer from hospitals of the Swedish southern healthcare region [GSE32894, n = 308 (215 NMIBCs and 93 MIBCs); ref. 17], Skane University Hospital [GSE32548, n = 131 (93 NMIBCs and 38 MIBCs); ref. 18], and University Hospital of Lund [GSE19915, n = 146 (97 NMIBCs and 49 MIBCs); ref. 19] were collected. Among them, MIBC data from GSE32894 and GSE32548 were combined, and a total of 58 MIBCs, whose survival time data were available, were used to assess survival rate of MIBC. Additional gene expression dataset including 19 bladder cancer cell lines was also examined in this study (GSE48277, n = 349). All gene expression datasets (GSE13507, GSE32894, GSE32548, GSE19915, and GSE48277) were freely available at NCBI GEO database.

To estimate prognostic values (PFS of NMIBC and OS of MIBC) of a signature combined with E2F1, EZH2, and SUZ12 genes, we adopted a previously developed strategy using the Cox regression coefficient for the genes in the signature [prognostic index (PI); refs. 20, 21]. Additional analysis was carried out as described in Supplementary Methods S1. Gene network–based activation regulator analyses were performed using the Ingenuity Pathway Analysis (IPA) tool.

The plasmid construct was the pcDNA6-V5-His–tagged expression vector with fusion coding sequence (CDS) of E2F1, EZH2, and SUZ12 (Invitrogen) genes. Transfection of plasmids was carried out using the jetPRIME reagent (Polyplus Transfection Inc.) at a ratio of DNA to jetPRIME of 1:3 according to the manufacturer's protocol. The measurement of gene expression at 24 hours posttransfection was normalized with the corresponding empty vectors.

The M-MLV Reverse Transcription kit (Beams Biotechnology) along with 3 μg of total RNA and poly(dT) primers were used for synthesis of cDNA. RT-PCR was carried out using an Emerald Amp GT PCR Master Mix (Takara Bio Inc.) to detect the mRNA level of E2F1, EZH2, and SUZ12 with primer sets (in Supplementary Methods S2). PCR cycling conditions were 94°C for 2 minutes to activate DNA polymerase, followed by 25 to 28 cycles of 94°C for 30 seconds, 58°C for 20 seconds, and 72°C for 40 seconds, and 72°C for 7 minutes for postelongation. Real-time PCR was carried out TOPreal premix SYBR Green (Enzynomics) and β-actin was used as control.

Proteins from the 5637, EJ, and UC5 cells were homogenized in RIPA buffer containing protease inhibitor (Roche), and the protein concentration was determined by using the BCA Assay (Thermo Scientific; ref. 22). The antibodies used in immunoblotting were against E2F1 (A300-766A, Bethyl Laboratories), EZH2 (4905, Cell Signaling Technology Inc.) SUZ12 (A302-407A, Bethyl Laboratories), and β-actin (4967, Cell Signaling Technology Inc.). Immunoreactivity was detected using the ECL Detection System (GE Healthcare BioSciences Corp.). Films were exposed at multiple time points to ensure that images were not saturated.

siRNAs targeting E2F1, EZH2, and SUZ12 were used: the SMARTpool ON-TARGET plus siE2F1 (L-003259-00; Dharmacon, GE Healthcare Bio-Sciences Corp.), siEZH2 (L-004218-00; Dharmacon), and siSUZ12 (L006957-00; Dharmacon). The SMARTpool ON-TARGET plus siControl nontargeting pool (D-001810-10) was purchased from Dharmacon. Cell were grown on 60-mm dishes and transfected either with control siRNA, siE2F1, siEZH2, or siSUZ12 (siRNA; 100 nmol/mL). The cells were analyzed 24 hours posttransfection.

We obtained shRNAs for E2F1, EZH2, and SUZ12 from Sigma-Aldrich (MISSION shRNA). Each shRNA for E2F1, EZH2, or SUZ12 was cloned into the pLKO.1-puro vector, using the Polymerase III U6-RNA promoter. A set of 5 shRNAs to each of E2F1, EZH2, and SUZ12 was tested for knockdown, and the shRNA containing the sequences for E2F1, EZH2, and SUZ12 was chosen for these experiments, because both mRNA and protein of E2f1, Ezh2, and Suz12 were effectively decreased. We used nontarget shRNA vector (Cat. No SHC016) as a control and selected stably expressing cells using puromycin (2 μg/mL).

Cell viability was detected using MTT assay. Cells were seeded in 96-well plates at a density of 1,000 cells per well, and then cells were incubated for 24 and 48 hours. Ten microliters of MTT (5 mg/mL; Sigma-Aldrich) was added to each well and incubated for 3 hours. At the end of the incubation, the supernatants were removed and 100 μL of dimethyl sulfoxide (Sigma-Aldrich) was added to each well, and absorbance at 490 nm was determined for each well using a Wallac Vector 1420 Multilabel Counter (EG&G Wallac). For each experimental condition, 3 wells were used.

For the colony formation assay, UC9 and EJ cells were transfected with expression vector or siRNAs. Trypsin-treated cells were suspended in medium containing DMEM or RPMI-1640 medium with 10% FBS, antibiotics, and 3 mL of 0.35% noble agar (Difco). Cells (1 × 10⁵ cells/well) were plated onto a solidified medium containing 3 mL of 0.7% noble agar in a 60-mm dish. The dishes were incubated at 37°C with 5% CO₂, and fresh medium was added every 4 to 5 days. UC9 were grown for 35 days, and EJ were incubated for 21 days before staining with 0.05% crystal violet. We counted forming colonies (>100 μm in diameter) using microscopy.

For cell invasion assays, we used a Boyden chamber (NeuroProbe) and membrane (8-μm pore size) precoated with growth factor–reduced Matrigel (BD Biosciences). After 24 hours of transfection, bladder cancer cells in 56 μL of medium without FBS were seeded (5 × 10⁴ cell/well) in the upper chamber. In the lower chamber, 27 μL of medium with 0.1% FBS medium (5637, EJ) and 10% FBS medium (UC5, UC9) was added as a chemoattractant. Then, cells were incubated for 12 hours (5637, EJ) and 24 hours (UC5, UC9).

For cell migration assays, the procedure was similar to the cell invasion assay, except Transwell membranes precoated with collagen (Sigma-Aldrich) were used, and cells were incubated for 12 hours (5637, EJ) and 24 hours (UC5, UC9). After staining the membrane using Diff-Quik reagents (Sysmex Co.), cells adhering to the lower surface were counted using a light microscope at 50× and 200× magnification and at least 4 wells were selected for each experimental group.

For the tumor xenograft assay, 4-week-old male BALB/c nude mice were obtained from SLC (Japan SLC, Inc.) and maintained under pathogen-free conditions. Knockdown- or overexpressed cells (KD-EJ, 1 × 10⁶ cells; UC9, 2 × 10⁶ cells) were suspended in 100 μL PBS. Cells were injected subcutaneously into both flanks on the top and bottom of mice. Tumor diameters were measured every 3 days for 3 weeks postinjection using digital calipers. Tumor volume in cubic millimeters was calculated using the formula: (L × W²) × 0.52, where L is the maximum length and W is the maximum width.

Chromatin immunoprecipitation (ChIP) assay was carried out as previously described (22) with primer sets (in Supplementary Methods S2) for EZH2 and SUZ12 promoter used for the qPCR.

Cells were seeded at a density of 1 × 10⁵ cells per well, and transfection of plasmids was carried out using the jetPRIME reagent (Polyplus Transfection Inc.) according to the manufacturer's protocol for 12 hours. After transfection, the medium was changed with 0 or 5 μmol/L of mitomycin C (MMC, Sigma) and 10 μmol/L of cisplatin (Dong-A ST) for 12 hours.

Using gene expression data of 102 NMIBCs in the Korean cohort (GSE13507; ref. 8), we selected in trans genes correlated with the E2F1 (total 1441 genes by the Pearson correlation test, P < 0.001, r > ∣0.4∣). Gene-to-gene network and upstream regulator analyses were performed using IPA tool displayed several important regulators with their effectors (i.e., gene networks) associated with disease progression of NMIBC (Supplementary Table S1). The path-exploring function of IPA revealed that an interconnection of network hubs composed of E2F1, EZH2, and SUZ12 is involved in a signaling pathway strongly associated with NMIBC progression (Fig. 1A). The patients in the Korean cohort were separated into the EH (E2F1 high expression) subgroup and the EL (E2F1 low expression) subgroup (two-sample t test, P < 0.001; Fig. 1B), indicating that activation of E2F1, the most predominant regulator, might be a key event associated with the progression of NMIBC. E2F1 was interconnected with another gene network hub composed around EZH2 that was interconnected with SUZ12 network (Fig. 1A). EZH2 was more highly expressed in the EH subgroup than in the EL subgroup (two-sample t test, P < 0.001; Fig. 1B). The expression of SUZ12 was also significantly higher in the EH subgroup than in the EL subgroup (two-sample t test, P < 0.001; Fig. 1B).

The PFS analysis in NMIBCs (n = 102) using a signature combined with E2F1, EZH2, and SUZ12 (the 3-gene signature) showed a significant difference of progression rates in NMIBC between poor- and good-prognosis subgroups (P = 0.008; Fig. 1C, left). To validate a prognostic value of the signature of 3 genes in NMIBC progression, we tested the signature in independent patient cohorts (GSE32894, GSE32548, and GSE19915). Because all validation cohorts did not contain PFS time data, we alternatively validated the signature by receiver operating characteristic (ROC) analysis comparing PI scores and NMIBC progression events. High or moderate area under curve (AUC) values were observed in all 3 patient cohorts (AUCs = 0.71, 0.59, and 0.65 in GSE32894, GSE32548, and GSE19915, respectively; Supplementary Fig. S1), indicating that the 3-gene signature would be highly associated with NMIBC progression. Using PI scores, patients were divided into 2 groups (poor- or good-prognosis), in which proportions of disease progression were also assessed. Significant differences of progression between poor and good prognosis groups were obtained from the datasets except for GSE32548 (Fisher exact test: P < 0.001, P = 0.459, and P = 0.049 in GSE32894, GSE32548, and GSE19915, respectively). However, the ratios of disease progression in the poor-prognosis group (21.3%, 12.2%, and 22.2%) were still higher than in the good-prognosis group (5%, 5.8%, and 7%) in all 3 validation cohorts (GSE32894, GSE32548, and GSE19915), respectively (Supplementary Fig. S1). In addition to NMIBC progression, we also assessed OS of MIBC using the 3-gene signature. The OS of MIBC (n = 63) in the poor-prognosis subgroup was also significantly worse than that in the good-prognosis subgroup (P = 0.021; Fig. 1C, right). For validation in OS of MIBC, we tested the signature in an independent combined cohort with GSE32984 and GSE32548, in which the survival rate of MIBC in the poor-prognosis subgroup classified by the signature of 3 genes was significantly worse than that in the good-prognosis subgroup (P = 0.044; Supplementary Fig. S2A).

To provide comparative results with other signatures, we additionally illustrated Supplementary Fig. S2 and described a comparative analysis between the signatures in “Comparison of other signatures with the three-gene signature” subsection in Supplementary Text S1. Because a previously published signature for predicting progression of NMIBC (23) consisted of small number of genes (the 11-gene signature) like our signature, we tried to compare them (Supplementary text S1). As shown in Supplementary Text and Supplementary Fig. S2, the 3-gene signature is validated a significant prognostic value.

We tested the characteristics of various bladder cancer cell lines that were derived from diverse stages of bladder cancer tissues. Unsupervised hierarchical clustering analysis of gene expression data from 19 bladder cancer cell lines yielded 3 major clusters, one representing the more aggressive (MIBC-like) and the other less aggressive (NMIBC-like) cancer cells (Supplementary Fig. S3A). These gene expression patterns may reflect the molecular configurations that are readily distinguishable between more aggressive (MIBC-like) and less aggressive (NMIBC-like) cancer cells.

To determine the microarray data that divided cases into NMIBC and MIBC subgroups, the biologic characteristics of total 12 bladder cancer cell lines were assessed by their invasiveness (Supplementary Fig. S3B). UC5, UC9, UC1, and UC6 cell lines showed characteristics of NMIBC, whereas others (UC3, UC10, UC14, 5637, EJ, T24, KU7, J82) were similar to MIBC (Supplementary Fig. S3B). Thus, we selected 4 cell lines as representative members in 2 characterized subgroups (NMIBC and MIBC) for further studies; UC5 and UC9 were NMIBC and EJ and 5637 were MIBC cells. The expression levels of E2F1, EZH2, and SUZ12 in the 2 groups of bladder cancer cells were evaluated by quantitative RT-PCR and Western blot assay. The mRNA expression of E2F1, EZH2, and SUZ12 in NMIBC was significantly lower than in MIBC cells (P < 0.01; Supplementary Fig. S3C). Protein levels of these genes were also increased in MIBC cells compared with NMIBC cells (Supplementary Fig. S3D). Thus, higher expression of these genes is clearly preserved in invasive cancer cells (Supplementary Fig. S3C and S3D), indicating possible mechanisms that contribute to progression to invasive or metastatic bladder tumors.

To determine whether the E2F1–EZH2–SUZ12 signature mediates progression from NMIBC to MIBC, we performed a number of in vitro and in vivo assays. As previously described, we examined the endogenous expression levels of E2F1, EZH2, and SUZ12 and compared them between NMIBC and MIBC (Supplementary Fig. S3C and S3D). Expression changes in the overexpressed cells with E2F1, EZH2, and SUZ12 may reflect cancer cell invasion and/or migration properties if these genes are strongly related to tumor progression. Comparisons of the expression levels of these genes between UC9 cells transfected with the pcDNA6 control vector and with pE2F1, pEZH2, or pSUZ12 are shown in Fig. 2A. Cells overexpressing E2F1, EZH2, and SUZ12 showed significant increases in both invasion and migration (Fig. 2B). In addition, in NMIBC UC5 cells, the effects of increasing the expression of E2F1, EZH2, and SUZ12 were also examined (Supplementary Fig. S4).

We also investigated the effects of silencing E2F1, EZH2, and SUZ12 expression using siRNA in EJ (Fig. 2C). EJ cells with decreased E2F1, EZH2, and SUZ12 expression displayed a significant decrease in invasiveness and migration (Fig. 2D). These results demonstrate that the increased expression of E2F1, EZH2, and SUZ12 is related to the invasiveness and migratory characters of bladder cancer cells. In addition, MIBC-5637 was also examined to confirm the effect of silencing of E2F1, EZH2, and SUZ12 (Supplementary Fig. S5).

To investigate whether the E2F1, EZH2, and SUZ12 affect cell proliferation and viability of NMIBC cells, cell proliferation was evaluated by counting the number of cells every day. As shown in Fig. 3A, top (left), E2F1-, EZH2-, or SUZ12-overexpressing UC9 significantly promoted cell proliferation compared with pcDNA. Otherwise, the depletion of these genes by siRNAs suppressed proliferation of EJ cancer cells (Fig. 3A, top right). We also found that the viability of cells overexpressing these genes was significantly higher than in the control group at 96 hours (Fig. 3A, bottom). In contrast, depletion of these genes by siRNA suppressed the viability of EJ cancer cells, compared with siControl-transfected cells at 72 and 96 hours (Fig. 3A, bottom right). These results suggest that elevated these genes may be related to bladder cancer cell proliferation and viability.

To obtain in vivo insight for these observations, nude mice were inoculated with E2F1-KD, EZH2-KD, or SUZ12-KD cell lines. The decreased mRNA and protein levels of shE2F1, shEZH2, or shSUZ12 cells are shown in Supplementary Fig. S6. All mice inoculated with shE2F1, shEZH2, or shSUZ12 cell lines showed a significant decrease in tumor volume compared with control-treated groups (Fig. 3B). Decreased expression levels of E2F1, EZH2, or SUZ12 from the resected tumors were also detected (Fig. 3B, right).

To examine the redundant role of elevated E2F1–EZH2–SUZ12 in proliferation and tumorigenesis in NMIBC cells, we also performed the reverse (overexpression) model in a tumor xenograft assay (Fig. 3C). All mice inoculated with overexpressed elevated E2F1–EZH2–SUZ12 cells showed a significant increase in tumor volume compared with control (pcDNA6) groups (Fig. 3C). Increased mRNA and protein levels from the resected tumors were also represented (Fig. 3C, right).

To identify whether E2F1 directly regulates transcription of the EZH2 and SUZ12 genes, we used the EZH2 and SUZ12 promoter vectors to drive a luciferase reporter gene in transient cotransfections with an E2F1 expression plasmid in 5637. Ectopic E2F1 strongly upregulated transcription from the both promoters in the 5637 (Fig. 4A). To determine the in vivo interaction of E2F1 with 3 potential E2F1-binding sites in the EZH2 promoter region or 1 potential E2F1-binding site in the SUZ12 promoter region (Fig. 4B), ChIP assays were performed using an E2F1 antibody. The appropriate EZH2 and SUZ12 promoter regions were immunoprecipitated with the E2F1 antibody (Fig. 4C and D). ChIP-qPCR analyses revealed that PCR fragments containing the potential E2F1-binding site at 3 regions in the EZH2 promoter and 1 region in the SUZ12 promoter were markedly increased in DNA samples from E2F1-transfected cells compared with DNA from pcDNA-transfected cells (Fig. 4C and D). No detectable band was observed in the control IgG precipitations.

To further define the mechanistic link between E2F1 and EZH2 or SUZ12, we tested whether expression levels of EZH2 or SUZ12 showed a transient change in E2F1-overexpressing UC9 (pE2F1, Fig. 5A, left) and in siE2F1-treated EJ (siE2F1, Fig. 5A, right). These experiments demonstrated that the overexpression of E2F1 activity induced EZH2 and SUZ12 expression, and the loss of E2F1 activity reduced EZH2 and SUZ12 expression (Fig. 5A). Consistent with the ChIP assays, these results suggest that E2F1 directly regulates transcription of the EZH2 and SUZ12.

To assess whether the lack of E2F1 is complemented by EZH2 or SUZ12 in cancer progression, we independently overexpressed the EZH2 or SUZ12 in E2F1-KD cells. In both E2F1-KD cells (shE2F1#1 and shE2F1#2), mRNA expression of EZH2 or SUZ12 was reduced compared with controls (shCon), whereas it was effectively restored by EZH2 or SUZ12 overexpression (Fig. 5B). Then, to verify whether the invasion ability of E2F1 was also rescued by EZH2 or SUZ12, we determined the level of invasion of E2F1-KD cells and cells overexpressing EZH2 or SUZ12 in E2F1-KD cells. E2F1-KD decreased invasion ability, which was significantly restored by EZH2 or SUZ12 overexpression (Fig. 5C and D).

Recent investigations demonstrated that tumorigenicity and tumor progression are driven by CSC characteristics, and the expression of EZH2 was consistently upregulated in CSCs (24–26). It also has been reported that SUZ12 is important for the function of CSCs, and ectopic expression of SUZ12 in transformed cells is sufficient to generate CSCs (27, 28). To elucidate the relationship between the E2F1–EZH2–SUZ12 signature and the CSC characteristics, we analyzed the ability of sphere formation and chemoresistance in bladder cancer cells.

Colony-forming assay was performed to verify whether the elevated expression of E2F1, EZH2, or SUZ12 is critical for sphere formation. Overexpression of E2F1, EZH2, or SUZ12 in UC9 significantly increased the number of large colony (>100 μm of diameter) formation compared with control (pcDNA6-empty; Fig. 6A, top). Otherwise, the depletion of E2F1, EZH2, or SUZ12 in the invasive EJ cells by each siRNAs for E2F1, EZH2, or SUZ12 decreased the number of large colony formation compared with control treated with scRNA (Fig. 6B).

To determine the chemoresistance which is known as one of CSC characteristics, the cell viability of E2F1-, EZH2-, or SUZ12-overexpressing cells was determined after the treatment of 5 μmol/L MMC (Fig. 6C, top) and 10 μmol/L cisplatin (Fig. 6C, bottom). Overexpression of E2F1, EZH2, or SUZ12 in UC9 and UC5 significantly increased the viability compared with control, whereas the depletion of E2F1, EZH2, or SUZ12 in the invasive EJ cells by each siRNAs decreased viability compared with control (Fig. 6C and Supplementary Fig. S8). Furthermore, mRNA levels of CD44, KLF4, OCT4, and ABCG2 known as CSC markers (29–32) significantly affected in under MMC- or cisplatin-treated conditions in bladder cancer cells with altered E2F1, EZH2, and SUZ12 expression than in matched control group (Fig. 6D and Supplementary Fig. S7).

According to the success of recent genome-wide gene expression profile studies (33–35), we previously reported a prognostic signature for predicting the progression of superficial tumors to invasive ones (8). The higher biologic activity of E2F1 suggests that it may be the major driving force during the progression of bladder cancer. Gene network analyses of the signature revealed that E2F1 and its downstream effectors EZH2 and SUZ12 could be important mediators for the invasive and metastatic progression of superficial tumors (Fig. 1). Consistent with other cancers, the relationship between cancer progression and the overexpression of these genes was observed (36–41). Recently, Santos and colleagues (42) also reported that the increased tumor recurrence and progression in patients with NMIBC is associated with increased E2F and EZH2 expression.

Overexpression of EZH2 and SUZ12 is directly controlled by E2F1 (Fig. 4), and their expression is associated with poor prognosis and indicative of invasion and metastasis in many cancers (Figs. 2 and 3). Our results show that these prognostic molecules can predict the likelihood of progression of NMIBC. Furthermore, unequal distribution of expression patterns reflecting activation of E2F1 in subgroups (EL, EH) with different progression rates supports the notion that distinct molecular features of the tumor govern the clinical phenotypes of NMIBC. As a result, we speculate that the overexpression of E2F1–EZH2–SUZ12 may play a role in proliferation, migration, and invasion of cancer cells.

It has been known that CSC characteristics may lead to cancer aggressiveness, chemoresistance to anticancer drugs, and a high risk of recurrence in patients with cancer (29, 31, 32, 43). Recent reports suggest that tumorigenicity and tumor progression is driven by CSC characteristics, and CSCs consistently showed elevated EZH2 and SUZ12 expression (24–28). It also has been reported that EZH2 and SUZ12 is important for the function of CSCs and its ectopic expression in transformed cells is sufficient to generate CSCs (24–28). Thus, we investigated the association of the E2F1–EZH2–SUZ12 downstream targets with the characteristics of CSCs. To identify the characteristics of CSCs, the abilities of sphere formation and chemoresistance were determined. The capacity of sphere formation is determined by both the proliferation rate and cell adhesion ability and has been used to identify the characteristic of CSCs in the previous studies (44–46). In this study, overexpression of E2F1–EZH2–SUZ12 increased the formation of large sphere, and the depletion of these genes decreased the formation of large sphere, which suggest that these genes might play important roles in sphere formation. Also, an increase in viability of the E2F1–EZH2–SUZ12–overexpressing cells under the treatment of MMC or cisplatin reflects that E2F1, EZH2, or SUZ12 might be related with resistance of the cells to anticancer drug. In addition, the activation of stem cell–like molecules (CD44, OCT4, KLF4, and ABCG2) was detected in bladder cancer cells overexpressing E2F1, EZH2, and SUZ12. Moreover, the activation of ABCG2 and CD44 could be related to the chemoresistance of the cells and, consequently, enriching CSCs by drug treatment might induce cancer aggressiveness (30, 47, 48).

We would suggest that the activation of EZH2 and SUZ12 expression under the control of E2F1 might play a role of switch in the development of bladder cancer according to tumor microenvironment. If the tumor microenvironment is favorable for cancer growth, the overexpression of EZH2–SUZ12 signature by E2F1 regulation contributes to the proliferation and invasiveness of bladder cancer cells. However, under the condition of anticancer drug treatment, the EZH2–SUZ12 signature by E2F1 control might activate CSCs signatures. As results, overexpressed E2F1–EZH2–SUZ12 cells have significantly higher capacities for sphere formation and activate the stem cell–like molecules when cells were put on an anticancer drug-treated condition.

Taken together, the elevated E2F1–EZH2–SUZ12 expression in bladder cancer cells might play important roles in proliferation, migration, and invasiveness. Moreover, the enrichment of cells with the characteristics of CSCs by overexpression of these genes might play critical roles in chemoresistance and tumorigenicity, which might be associated with poor prognosis of bladder cancer cells. Therefore, our findings show that a prognostic molecular signature, E2F1–EZH2–SUZ12, can predict the likelihood of progression of NMIBC. Furthermore, our study could provide useful information to predict an individual's risk of progression and to establish a suitable chemotreatment for disease.

No potential conflicts of interest were disclosed.

Conception and design: J. Heo, I.-S. Chu, S.-H. Leem

Development of methodology: H.-H. Lee, I.-S. Chu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.-R. Lee, Y.-G. Roh, S.-K. Kim, J.-S. Lee, S.-Y. Seol, H.-H. Lee, W.-T. Kim, I.-S. Chu

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.-K. Kim, J.-S. Lee, J. Heo, T.-H. Kang, I.-S. Chu, S.-H. Leem

Writing, review, and/or revision of the manuscript: S.-R. Lee, Y.-G. Roh, S.-K. Kim, J. Heo, T.-H. Kang, I.-S. Chu, S.-H. Leem

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W.-J. Kim, H.-J. Cha, J.W. Chung, I.-S. Chu, S.-H. Leem

Study supervision: J. Heo, I.-S. Chu, S.-H. Leem

This research was supported by the Mid-Career Researcher Program through the National Research Foundation of Korea (NRF) grant (NRF-2013R1A2A2A04008115), an NRF grant (2008-0062611 and 2011-0019745) funded by the Korea government (MEST), and a grant from the KRIBB Research Initiative Program.

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.