The Epigenetic Modifier PRDM5 Functions as a Tumor Suppressor through Modulating WNT/β-Catenin Signaling and Is Frequently Silenced in Multiple Tumors

Background PRDM (PRDI-BF1 and RIZ domain containing) proteins are zinc finger proteins involved in multiple cellular regulations by acting as epigenetic modifiers. We studied a recently identified PRDM member PRDM5 for its epigenetic abnormality and tumor suppressive functions in multiple tumorigeneses. Methodology/Principal Findings Semi-quantitative RT-PCR showed that PRDM5 was broadly expressed in human normal tissues, but frequently silenced or downregulated in multiple carcinoma cell lines due to promoter CpG methylation, including 80% (4/5) nasopharyngeal, 44% (8/18) esophageal, 76% (13/17) gastric, 50% (2/4) cervical, and 25% (3/12) hepatocellular carcinoma cell lines, but not in any immortalized normal epithelial cell lines. PRDM5 expression could be restored by 5-aza-2′-deoxycytidine demethylation treatment in silenced cell lines. PRDM5 methylation was frequently detected by methylation-specific PCR (MSP) in multiple primary tumors, including 93% (43/46) nasopharyngeal, 58% (25/43) esophageal, 88% (37/42) gastric and 63% (29/46) hepatocellular tumors. PRDM5 was further found a stress-responsive gene, but its response was impaired when the promoter was methylated. Ectopic PRDM5 expression significantly inhibited tumor cell clonogenicity, accompanied by the inhibition of TCF/β-catenin-dependent transcription and downregulation of CDK4, TWIST1 and MDM2 oncogenes, while knocking down of PRDM5 expression lead to increased cell proliferation. ChIP assay showed that PRDM5 bound to its target gene promoters and suppressed their transcription. An inverse correlation between the expression of PRDM5 and activated β-catenin was also observed in cell lines. Conclusions/Significance PRDM5 functions as a tumor suppressor at least partially through antagonizing aberrant WNT/β-catenin signaling and oncogene expression. Frequent epigenetic silencing of PRDM5 is involved in multiple tumorigeneses, which could serve as a tumor biomarker.


Introduction
Tumor-specific epigenetic silencing of tumor suppressor genes (TSGs) through promoter CpG methylation and histone modification is frequently involved in multiple carcinogenesis [1]. CpG methylation can also be used as an epigenetic biomarker for novel TSG identification and tumor diagnosis. A series of TSGs such as p16, RASSF1A, PCDH10 and RASAL, have been identified with epigenetic inactivation in multiple cancers, through promoter CpG methylation [2][3][4].
PRDM (PRDI-BF1 and RIZ domain containing) proteins belong to the zinc finger protein family, harboring an evolutionarily conserved N-terminal PR domain followed by 16 zinc finger repeats [5]. The PR domain, a characteristic C2-H2 zinc finger motif, shares high homology with SET (Suvar3-9, Enhancer-ofzeste, Trithorax) domain that is involved in chromatin-mediated transcriptional regulation [6]. Thus far, seventeen human PRDM members have been identified, with several possessing growth inhibitory functions in tumor cells. PRDM1 (PRDI-BF1 or BLIMP1) functions as a transcriptional repressor of c-MYC and induces cell differentiation and apoptosis [7,8]. PRDM2 (RIZ1) is the first identified member of methyltransferase family with tumor suppressive function mediated by its PR domain [9,10]. The long isoform of PRDM3 (MDS1-EVI1) is commonly mutated in myeloid leukemia [11]. PRDM4 (PFM1) is located at a tumor suppressor locus 12q23-q24.1 commonly deleted in ovarian, gastric and pancreatic cancers [12].
PRDM5 (also known as PFM2) is a recently identified PRDM member, mapped to a commonly deleted region 4q25-26 [13]. PRDM5 was first identified from an EST database based on its conserved PR domain at the N-terminus followed by 16 zinc finger motifs [14]. It acts as a candidate TSG in some tumors [14,15]. Recent reports showed PRDM5 silencing by promoter methylation in multiple cancers including breast, ovarian, colorectal, gastric, and hepatocellular tumors. Ectopic expression of PRDM5 leads to G2/M arrest and apoptosis of tumor cells [14], although its molecular mechanism is still unclear. PRDM5 is a nuclear protein involved in the epigenetic regulation of various genes, through interaction with histone methyltransferase G9A and histone deacetylase 1 (HDAC1) [16].
Here, we studied the epigenetic abnormality and tumor suppressive functions of PRDM5 in multiple common tumors including nasopharyngeal, esophageal, gastric, hepatocellular and cervical cancers.

Results
Frequent silencing of PRDM5 in multiple tumor cell lines due to promoter methylation We first examined PRDM5 expression in a panel of human normal adult and fetal tissues by semi-quantitative RT-PCR. Results showed that PRDM5 was broadly expressed in all normal tissues ( Figure 1A). We then checked PRDM5 expression in multiple carcinoma cell lines. Results revealed that PRDM5 was frequently downregulated or silenced in cell lines of nasopharyngeal, esophageal, gastric, hepatocellular and cervical carcinomas, but readily detected in immortalized normal epithelial cell lines ( Figure 2).
To confirm the MSP results, we further examined PRDM5 promoter methylation by high-resolution bisulfite genomic sequencing (BGS) analysis of 43 individual CpG sites within its CGI. The BGS results were consistent with those of MSP, with all promoter alleles intensively methylated in silenced cell lines but rare methylated CpG sites detected in expressing normal cell lines ( Figure 3A). Taken together, these results revealed a strong correlation between PRDM5 promoter CpG methylation and its transcriptional silencing in tumor cell lines.

Pharmacologic demethylation restores PRDM5 expression
To determine whether promoter methylation directly contributes to PRDM5 silencing, several silenced tumor cell lines were treated with DNA methyltransferase inhibitor 5-aza-29-deoxycytidine (Aza) alone or combined with histone deacetylase inhibitor trichostatin A (TSA). This pharmacologic demethylation restored PRDM5 expression in all silenced cell lines, accompanied by the increase of demethylated promoter alleles ( Figure 3B and 3C), indicating that PRDM5 silencing in tumor cells was directly mediated by promoter methylation.
PRDM5 is a stress responsive gene but its response is disrupted by promoter methylation Analysis of regulatory elements of the PRDM5 promoter by TFSearch (www.cbrc.jp/research/db/TFSEARCH) revealed two HSF (heat shock factor) and five Sp1 binding sites ( Figure 4A), suggesting that PRDM5 might be a stress-responsive gene. We then examined its response to environmental stress stimuli. We found that PRDM5 expression was dramatically induced in HK1 cells with unmethylated promoter alleles, after exposure to heat shock or UV irradiation. However, this stress response was significantly reduced or totally abolished in cell lines with methylated promoter (C666-1 and CNE2) ( Figure 4B). Moreover, ChIP assay showed significant enrichment of HSF1 binding to PRDM5 promoter after heat shock treatment of HK1 cells ( Figure 4C), suggesting that PRDM5 is indeed stress-responsive but this response is disrupted when the promoter is methylated.

PRDM5 inhibits tumor cell growth and proliferation
Frequent silencing of PRDM5 in multiple cancer cell lines and primary tumors indicated that PRDM5 likely functions as a tumor suppressor. Thus, we examined the effects of ectopic PRDM5 expression on tumor cell clonogenicity. A mammalian expression vector encoding full-length PRDM5 was transfected into nasopharyngeal, esophageal and gastric cancer cell lines with completely methylated and silenced PRDM5 (HONE1, KYSE140 and MKN28). Ectopic PRDM5 expression dramatically reduced the colony formation efficiencies of all cell lines in monolayer culture, as compared to vector controls (down to 18%, 32% and 56%, respectively, Figure 7A upper panel and 7B). Expression of PRDM5 in these cell lines was confirmed by RT-PCR and Western blot ( Figure 7A bottom panel). Consistently, knock-down of endogenous PRDM5 using siRNA in HEK293 cells significantly increased the cell proliferation rate ( Figure 7C), suggesting that PRDM5 does function as a tumor suppressor.

PRDM5 modulates WNT/b-catenin signaling and represses oncogene expression
As a zinc finger transcription factor and epigenetic modifier, PRDM5 should exert its tumor suppressive functions through modulating cell signaling and gene transcription. We thus analyzed the effect of PRDM5 expression on cell signaling using the TCF/bcatenin-(TOPFlash) and Ras/MAPK-ERK-signaling (SRE) luciferase reporter constructs. Although PRDM5 expression did not affect SRE luciferase reporter activity (data not shown), it did significantly inhibit TOPFlash reporter activity, an indicator of TCF/LEF-dependent transcription, with FOPFlash reporter construct containing mutant TCF/LEF binding sites as a control ( Figure 8A). Moreover, the promoter reporter activity of CCND1, a WNT/b-catenin downstream target gene, was markedly decreased  when PRDM5 was expressed ( Figure 8B). We also examined whether PRDM5 methylation/silencing is correlated with enhanced WNT/b-catenin signaling in tumor cell lines, and found that the active form of b-catenin was accumulated in cell lines with methylated and silenced PRDM5, with significant lower level of active b-catenin detected in PRDM5-expressing cell lines ( Figure 8C). We further examined the expression changes of multiple oncogenes and tumor suppressor genes after ectopic PRDM5 expression in both normal and tumor cell lines (HEK293 and HONE1). Real-time PCR showed that several oncogenes were dramatically repressed by PRDM5, such as CDK4, TWIST1, and MDM2 ( Figure 9A, Table S1). We further performed ChIP assay, and found direct binding of PRDM5 to the promoters of CDK4 and TWIST1 ( Figure 9B). Moreover, PRDM5 expression resulted in significant decreased levels of H3K4me3 and acetyl-histone H4 in CDK4 and TWIST1 promoters ( Figure 9C), both as active  transcription marks. Thus, PRDM5 does exert its tumor suppressive functions through modulating WNT/ b-catenin signaling and the expression of multiple oncogenes as an epigenetic modifier ( Figure 10).

Discussion
Identification of TSGs silenced by CpG methylation elucidates the molecular mechanisms of carcinogenesis and develops epigenetic biomarkers for cancer detection and prognosis prediction. In the present study, we showed that PRDM5, an epigenetic modifier gene, was frequently inactivated by promoter methylation in multiple common cancers. Ectopic expression of PRDM5 inhibited tumor cell clonogenicity, at least partially through antagonizing WNT/b-catenin signaling and oncogene expression such as CDK4, TWIST1, and MDM2.
Epigenetic silencing of TSGs is critically involved in the onset and progression of multiple cancers, even preceding genetic changes during tumorigenesis [17]. CpG methylation of TSG promoters is a fundamental epigenetic mechanism leading to their inactivation in tumors [1]. Previous studies showed that PRDM5 silencing was mediated by either DNA methylation or trimethylation of H3K27 [15], but with limited cell lines and tumors studied. Here, we systematically studied PRDM5 silencing and methylation in a large collection of cell lines and primary tumors. Our results showed that aberrant methylation of PRDM5 was associated with its inactivation in multiple tumor cell lines and primary tumors, with the silencing/methylation of PRDM5 in cell lines of gastric cancer (MKN45) and HCC (huH1, Hep3B and HepG2) consistent with previous studies [14,15]. Our results demonstrate that tumorspecific promoter methylation represents the predominant mechanism of PRDM5 inactivation, although other mechanisms like gene micro-deletion or mutations might also exist.
We also found that PRDM5 is a stress-responsive gene, but its response is impaired when the promoter is methylated. When inspecting the PRDM5 promoter, we found two predicted HSF binding sites, with the one at 2390 to 2380 (CGGAAATTTCC) containing a CpG site. It is likely that methylation of this CpG site would affect the binding of HSF, leading to the disruption of its stress response. Thus, epigenetic silencing of PRDM5 could abolish  PRDM5 is a zinc finger protein. As a sequence-specific DNAbinding transcription factor, PRDM5 functions as a repressor targeting multiple protein-coding and miRNA genes involved in cell cycle, signal transduction and protein modification [16]. Although with a PR/SET domain in the N-terminus, PRDM5 itself does not possess histone methyltransferase (HMTase) activity. Instead, PRDM5 acts as an epigenetic modifier by recruiting histone-modifying enzymes like HMTase G9A and HDAC1 to target genes [16]. Our study indicates that if the epigenetic modifier PRDM5 is inactivated in tumor cells, its repression to target oncogenes would be released and thereby contributing to tumor initiation and progression.
Previous report showed that ectopic PRDM5 expression caused cell cycle G2/M arrest and induced apoptosis in tumor cells [14], while the molecular mechanisms remain unknown. Recently, the zebrafish PRDM5 was found to regulate Wnt/b-catenin signaling at the early stages of zebrafish development [18]. Our results are in agreement with this study. Aberrant activation of WNT/b-catenin signaling is frequently involved in cancers, accompanied with elevated levels of active b-catenin [19]. In addition to genetic defects, epigenetic silencing of WNT/b-catenin antagonists also leads to aberrant WNT/b-catenin signaling in tumors [19]. Frequent methylation-mediated silencing of extracellular (SFRPs and DKKs) [20][21][22] and cytosolic (APC, AXIN2 and DACT3) [23][24][25] WNT antagonists, nuclear proteins (SOX7 and SOX17) [26,27], and non-transforming WNT families (WNT5A, WNT7A and WNT9A) occurs in cancers [28][29][30], indicating that the epigenetic inactivation of WNT-signaling negative regulators plays a critical role in tumor pathogenesis. Our finding that ectopic expression of PRDM5 leads to the inhibition of WNT/b-catenin signaling in tumor cells, probably through upregulating DKK1, DKK2, and WNT5A [18], suggests a novel mechanistic link between PRDM5 and WNT signaling in human tumorigenesis. PRDM5 also significantly inhibited the promoter activity of CCND1, a direct downstream target oncogene of WNT/b-catenin signaling which acts as a mitogenic signal sensor [31,32]. CDK4 is one of the binding partners of CCND1 and also transcriptionally downregulated by PRDM5. PRDM5 silencing is also linked to the accumulation of active b-catenin in tumor cells, which leads to constitutive activation of WNT/b-catenin signaling.
In summary, we found that PRDM5 was frequently silenced by promoter methylation in multiple tumors. PRDM5 suppresses tumor cell growth through antagonizing WNT/b-catenin signaling and suppressing multiple oncogenes expression. The high incidence of PRDM5 methylation in some carcinomas indicates that it is a potential epigenetic biomarker for these tumors.

Cell lines and tumor samples
A series of tumor cell lines were used, including nasopharyngeal, esophageal, gastric, hepatocellular, lung, colorectal, and cervical cell lines [33,34]. Immortalized epithelial cell lines NP69, NE083, Het-1A, NE1 and NE3 were used as normal controls. Carcinoma cell lines were maintained in RPMI 1640 (Invitrogen) supplemented with 10% fetal bovine serum. Cell lines were purchased from the American Type Culture Collection (ATCC, Manassas, VA) or obtained from collaborators. Human normal adult and fetal tissue RNA samples were purchased commercially (Stratagene, La Jolla, CA or Millipore Chemicon, Billerica, MA). Samples of normal nasopharyngeal and esophageal epithelial tissues from healthy individuals were described previously [33]. Nude mice-passaged undifferentiated NPC tumors from North Africans, C15, C17 and C18 were also used [34]. DNA samples of NPC, gastric tumors, esophageal squamous cell carcinomas, hepatocellular carcinomas (T) and their corresponding surgical marginal normal tissues (N), were described previously [34].

Pharmacologic demethylation and stress treatments
Cell lines were treated with DNA methyltransferase inhibitor 5aza-29-deoxycytidine (Aza) (Sigma, St. Louis, MO) and histone deacetylase inhibitor Trichostatin A (TSA) as reported previously [3]. Cells were subjected to heat shock treatment with incubation at 42 o C for 1 hour followed by recovery at 37uC for 2 hours. For UV treatment, medium was removed and the flask was turned upside down to face the light source in a UV cross-linker (Amersham Biosciences, Piscataway, NJ). Cells were irradiated for a dose of 70 J/m 2 . After irradiation, fresh medium was added, and cells were recovered at 37uC for 1 hour and then harvested [33].

Plasmid construction
The full length open reading frame (ORF) of PRDM5 [14] was subcloned into the NotI and KpnI site of the pcDNA3.1 mammalian expression vector to generate pcDNA3.1-PRDM5 with sequence and orientation confirmed.

Semi-quantitative RT-PCR and real-time PCR
Total RNA was extracted from cell lines using TRI reagent. Reverse transcription (RT) using random hexamer, and RT-PCR using Go-Taq (Promega, Madison, WI) were performed as previously, with GAPDH as a control [34]. Primers used were PRDM5F: 59-CAGGTTCTCCCTGAAGTCCT and PRDM5R:  59-TGAGATGGTGCCTCATGAAC. RT-PCR was performed for 32 cycles for PRDM5, while 23 cycles for GAPDH.
Real-time PCR was performed using SYBR Green master mixture and an HT7900 system according to the manufactures' instruction (Applied Biosystems, Foster City, CA). To screen for PRDM5 modulated target genes, total RNA extracted from PRDM5 or vector-transfected cells was treated with DNase I, and reverse transcripted as described previously [34]. Real-time PCR was performed for 34 selected candidate target genes. For each gene, the expression level in PRDM5-transfected cells was normalized to control. Primer sequences were listed in Table S2.

Colony formation assay
Cells were seeded at 1610 5 cells/ml and cultured overnight in 12-well plates. Cells were then transfected with pcDNA3.1-PRDM5 or empty vector using FuGene6 (Roche). Forty-eight hours post-transfection, cells were collected and seeded in 6-well plates at an appropriate density with G418 selection (400 mg/ml). After 2-3 weeks, cells were stained with Gentian Violet with survived colonies ($50 cells/colony) counted.
siRNA transfection and cell proliferation assay PRDM5 siRNA and siRNA negative control were obtained from Invitrogen. siRNAs were transfected using Lipofectamine 2000 (Invitrogen) according to manufacturer's protocol. Twenty-four hours after transfection, cells were collected and seeded in 96-well plates (10 3 cells per well). Cell proliferation rates were determined at indicated time points using Cell Counting Kit-8 (Dojindo, Rockville, MD). Triplicates for each sample were performed.

Dual-luciferase reporter assay
Luciferase reporter construct of CCND1 promoter (+ ,2 kb upstream of transcription start site), or TOPFlash reporter containing 4xTCF-binding sites (kind gift from Dr. Christof Niehrs, German Cancer Research Center (DKFZ)), or FOPFlash reporter containing mutant TCF/LEF binding sites (kind gift from Dr. Jin Dong-Yan, University of Hong Kong) was cotransfected with either pcDNA3.1-PRDM5 or empty vector, together with an internal control Renilla luciferase reporter pRL-CMV vector using FuGene6 [35]. Forty-eight hours after transfection, cells were harvested and analyzed by the dualluciferase assay kit (Promega, Madison, WI). Each experiment was repeated for three times. Statistical analysis was carried out by Student's t-test, p,0.05 was considered as statistically significant difference.

Western blot
Cell pellets were incubated in lysis buffer (50 mmol/L Tris-HCl, pH 8.0; 150 mmol/L NaCl; 0.5% NP40) for 30 minutes on ice, followed by centrifugation at 14,000 rpm for 15 minutes at 4uC. Cell lysates were resolved using SDS-PAGE gels and transferred onto nitrocellulose membranes. Membranes were then incubated with primary antibodies for 1 hour at room temperature or overnight at 4uC, followed by incubation with a secondary antibody at room temperature for 1 hour. Immunoreactive bands were detected using Western blot Luminol reagent (GE Healthcare Bio-Sciences) according to the manufacturer's protocol. Antibody used were FLAG M2 (F3165, Sigma), active b-catenin (05-665, Upstate) and a-Tubulin (DM1A, Lab Vision). Table S1 Real-time PCR of PRDM5 target genes. Modulation of multiple genes by PRDM5 expression in HEK293 and HONE1 cells was evaluated by real-time PCR. Data were obtained from three independent experiments and presented as relative average ratio. The expression of target genes in PRDM5-transfected cells were normalized to those in vector controls, GAPDH was used as internal control. p,0.05 was considered as statistical significance by Student's t-test. (DOC)