|Home | About | Journals | Submit | Contact Us | Français|
While tumor suppressor genes frequently undergo epigenetic silencing in cancer, how the instructions directing this transcriptional repression are transmitted in cancer cells remain largely unclear. Expression of cyclin-dependent kinase inhibitor 1C (CDKN1C), an imprinted gene on chromosomal band 11 p15.5, is reduced or lost in the majority of breast cancers. Here, we report that CDKN1C is suppressed by estrogen through epigenetic mechanisms involving the chromatin-interacting noncoding RNA KCNQ1OT1 and CCCTC-binding factor (CTCF). Activation of estrogen signaling reduced CDKN1C expression 3-fold (P < 0.001) and established repressive histone modifications at the 5′ regulatory region of the locus. These events were concomitant with induction of KCNQ1OT1 expression as well as increased recruitment of CTCF to both the distal KCNQ1OT1 promoter-associated imprinting control region (ICR) and the CDKN1C locus. Transient depletion of CTCF by small interfering RNA increased CDKN1C expression and significantly reduced the estrogen-mediated repression of CDKN1C. Further studies in breast cancer cell lines indicated that the epigenetic silencing of CDKN1C occurs in part as the result of genetic loss of the inactive methylated 11p15.5 ICR allele (R2 = 0.612, P < 0.001). We also found a novel cis-encoded antisense transcript, CDKN1C-AS, which is induced by estrogen signaling following pharmacologic inhibition of DNA methyltransferase and histone deacetylase activity. Forced expression of CDKN1C-AS was capable of repressing endogenous CDKN1C in vivo. Our findings suggest that in addition to promoter hypermethylation, epigenetic repression of tumor suppressor genes by CTCF and noncoding RNA transcripts could be more common and important than previously understood.
Through aberrant epigenetic modification of the genome, affected cells may acquire new phenotypes that promote tumorigenesis. Repressive histone modification and CpG island hypermethylation, frequently associated with epigenetic silencing of tumor suppressor genes, have been subject to intense study for many years. It remains largely unclear how in cancer cells the instructions for depositing these repressive marks at specific locations across the genome are transmitted.
Transcription in eukaryotic genomes is pervasive. Much of the output is unannotated and has reduced protein-coding potential (1). A growing amount of evidence suggests that noncoding RNAs (ncRNAs) can play a functional role in the establishment and maintenance of chromatin states (2–4). Such ncRNAs are often transcribed antisense to protein-coding genes (5). Mouse model studies have shown that the antisense Kcnq1ot1 ncRNA spreads heterochromatin modifications in a target gene-specific manner via RNA–protein interactions between the Kcnq1ot1 transcript and both histone 3 lysine 9 (H3K9) and histone 3 lysine 27 (H3K27) methyltransferases (4,6). In human cells, the KCNQ1OT1 promoter resides within the 11p15.5 imprinting control region (ICR) that regulates the expression of an imprinted gene cluster including cyclin-dependent kinase inhibitor 1C (CDKN1C) (6). The chromatin-interacting ncRNA negatively regulates the paternal CDKN1C allele in normal cells (7,8) and has been shown to stably localize to the CDKN1C gene locus (9).
The maternally expressed CDKN1C is known to encode the CKI Cip/Kip family member p57Kip2 protein, a negative regulator of cell proliferation (10) and positive regulator of apoptosis and tumor suppression(9,10). Loss of CDKN1C expression is implicated in Beckwith–Wiedemann syndrome, an overgrowth syndrome associated with predisposition to embryonal and childhood tumors (11). Expression of the CDKN1C gene is also repressed or lost in an extensive number of sporadic cancers (12–14), and promoter hypermethylation occurs in some solid and hematologic tumors (15–18). Loss of methylation on the maternal 11p15.5 ICR allele occurs in 50% of Beckwith–Wiedemann syndrome patients, resulting in biallelic expression of KCNQ1OT1 (12) and suppression of CDKN1C (11). Loss of methylation at the ICR has been reported in several adult neoplasms (13), concordant with repression of CDKN1C in liver, pancreatic and esophageal tumors and/or cell lines (14–16). Independent of KCNQ1OT1 transcription, the 11p15.5 ICR and its murine homolog KvDMR1 also exhibit chromatin insulator or silencer activity that may be mediated by binding of CCCTC-binding factor (CTCF) in vivo on the unmethylated paternal allele (17,18). While the initiation of estrogen signaling has been shown to repress expression of the Cip/Kip family members p21Cip1 and p27Kip1 in estrogen receptor-alpha (ERα) expressing breast cancer cells (19,20), the role of estrogen in regulating either CDKN1C expression or genomic imprinting is not well understood.
In this study, we first investigated estrogen-mediated epigenetic regulation of CDKN1C expression in breast cancer cells. Activation of estrogen signaling appeared to enhance both expression of the ncRNA KCNQ1OT1 as well as binding of CTCF to the CDKN1C locus and 11p15.5 ICR. We also identified an unannotated antisense transcript in opposite orientation to CDKN1C. We then explored the roles of the CTCF and these noncoding transcripts in repressing CDKN1C. This study leads us to propose three modes of complex regulation by which estrogen-mediated silencing of CDKN1C may occur during breast tumorigenesis.
Trichostatin A (TSA), 5-aza-2′-deoxycytidine (DAC), 17β-estradiol (E2), 4-hydroxytamoxifen, actinomycin D, cycloxheximide, MG132 and bisphenol A were purchased from Sigma–Aldrich (St. Louis, MO). ON-TARGETplus non-targeting pool, CTCF SMARTpool small interfering RNA (siRNA) and Dharmafect 4 transfection reagent were purchased from Dharmacon (Lafayette, CO). Antibodies were obtained from the following vendors: acetylated histone H3, H3K4me2, HDAC1, CTCF (Millipore, Billerica, MA); H3K9me2 (Abcam, Cambridge, MA); H3K27me3 (Diagenode, Denville, NJ); RNA Pol II largest subunit (Santa Cruz Biotechnology, Santa Cruz, CA) and IgG (BD Biosciences, Franklin Lakes, NJ). The pIRES2-EGFP vector (Clontech, Mountain View, CA) was used to create the p57-S, p57-AS and p57-ASΔ expression constructs. Polymerase chain reaction (PCR) primer sequences are available by request.
Hormone-deprived MCF7, MDA-MB-453, T47D breast cancer cells and normal mammospheres-derived epithelial cells (MDECs) were stimulated with 10 nM E2 or vehicle. In other experiments, MCF7 cells were pretreated with DAC (5 μM) and/or TSA (300 nM). Alternatively, MCF7 cells were pretreated with 4-hydroxytamoxifen (100 nM), actinomycin D (0.25 μg/ml), cycloheximide (5 μg/ml) or MG132 (1 μM) 1 h prior to estrogen stimulation. For siRNA knockdown experiments, MCF7 cells were transfected with 100 nM non-targeting siRNA control or 100 nM CTCF siRNA for 72 h under hormone-deprived conditions. Breast cancer cell lines were obtained from the American Type Culture Collection (Manassas, VA) for in vitro studies.
Breast tissues, obtained from individuals undergoing mastectomy or reduction mammoplasty, were collected in accordance with the protocols approved by the Institutional Review Boards of the Ohio State University and the National Taiwan University Hospital. Breast progenitor cells were isolated from noncancerous tissues and grown as mammospheres in ultra-low attachment dishes (Corning, Lowell, MA) containing serum-free mammary epithelial growth medium (Cambrex, East Rutherford, NJ) as described previously (21). These mammospheres were exposed to bisphenol A (4 nM) or dimethyl sulfoxide for 3 weeks. After the exposure, mammospheres were differentiated into epithelial cells by culturing on a collagen-coated dish (BD Biosciences) containing phenol red-free mediumfor 3 weeks.
Total RNA was isolated using Trizol (Invitrogen, Carlsbad, CA) and followed by RNeasy column clean up (Qiagen, Valencia, CA). Reverse transcription (RT) reactions were performed on total RNA with Superscript III (Invitrogen) for oligo dT- or random hexamer-primed reactions. Strand-specific RTs were performed with gene-specific primers using Thermoscript (Invitrogen).
Chromatin immunoprecipitation (ChIP) was performed on hormone-deprived MCF7 cells treated with 10 nM E2 or vehicle for 12 h with antibodies against IgG, acetylated histone H3, HDAC1, CTCF, RNA Pol II largest subunit, H3K27me3, H3K4me2 and H3K9me2. For ChIP, E2 or vehicle-treated cells were formaldehyde cross-linked and immunoprecipitated with the above antibodies using magnet Dynabeads (Invitrogen). Samples were eluted, subjected to sequential digests with RNAse A and Proteinase K and then column purification (Qiagen).
Template complementary DNAs were generated by strand-specific RT for CDKN1C-S and CDKN1C-AS quantitative reverse transcriptase–polymerase chain reaction (qRT–PCR) assays. Oligo dT-primed RT reactions were used for the assaying of total CDKN1C, TBP, 36B4, GAPDH, HPRT1, GFP, NEOR, CCNG2 and NRIP1; random hexamers were used for KCNQ1OT1. Expression copy number of CDKN1C-S, CDKN1C-AS, CDKN1C and TBP were determined by absolute standard curves of cloned PCR products (six points, dilutions ranged from 125 000 to 40 copies). Relative expression of 36B4, GAPDH, HPRT1, GFP, NEOR, CCNG2, NRIP1 and KCNQ1OT1 were determined by the ΔΔCt method unless otherwise specified. All PCRs were assayed in triplicate and included melting curve analysis. Reactions were performed on an ABI 7500 using Power SYBR Green (Applied Biosystems, Carlsbad, CA).
ChIP quantitative–polymerase chain reaction (ChIP–qPCR) profiling interrogated seven regions of the CDKN1C locus, spanning −976 to +2857 bp relative to the transcriptional start site (TSS), as well as the telomeric 11p15.5 ICR. A previously reported E2 off-target control region in the FKBP6 gene was also interrogated (22). Copy number of immunoprecipitated and the corresponding input DNAs for each individual sample were determined by absolute standard curves of cloned PCR products (six points, dilutions ranged from 125 000 to 40 copies). DNA enrichment was calculated as fraction of input, the ratio of IP:Input copy number for each sample. PCR assays were performed as described above.
Genome content differences between breast cancer cells and normal tissues were determined by qPCR assays of DNA copy number of the distal ICR (identical primers used ChIP–qPCR) and a Line-1 reference sequence were quantified from absolute standard curves of cloned PCR products (six points, dilutions ranged from 125 000 to 40 copies). Copy number ratios were obtained by calibrating results to the average value of disease-free reduction normal breast tissues (n = 15). PCR assays were performed as described above.
Approximately 300 ng of genomic DNA was bisulfite modified with EZ DNA Methylation Kit (Zymo Research, Irivine, CA) according to the manufacturer’s protocol. Methylation of the 11p15.5 ICR was assayed by Bisulfite PCR Pyrosequencing (Qiagen) in breast cancer cell lines (n = 52), primary tumors (n = 306) and normal tissues (n = 19).
Statistical tests were performed in Minitab 15 unless otherwise specified. Differences between groups were assessed by two-sample t-test. Comparisons of more than two sets of experiments were corrected for multiple hypothesis testing. Simple and multiple linear regression analyses of 11p15.5 ICR methylation level, 11p15.5 ICR DNA copy number ratio and relative CDKN1C ΔCt expression in breast cancer cell lines were performed using Sigmaplot 11. In graphs of qRT–PCR, ChIP–qPCR and DNA copy experiments, data are presented as mean ± standard deviation. All tests were two-tailed and P-values of <0.05 were considered significant.
Expanded methods and any associated references are included in the supplementary materials, available at Carcinogenesis Online.
We investigated the effects of E2 stimulation on the expression of CDKN1C in MCF7 breast cancer cells treated with 10 nM E2. qRT–PCR results showed a 3-fold reduction of CDKN1C expression after 12 h treatment (P < 0.001) (Figure 1A). Treatment with E2 had no effect on CDKN1C expression in hormone-unresponsive control MDECs lacking expression of ESR1 (ERα) (Figure 1B). To confirm whether this downregulation was mediated through ERα signaling, MCF7 cells were treated with ER antagonist 4-hydroxytamoxifen prior to 12 h E2 stimulation. Competitive inhibition of ERα signaling abrogated the E2-mediated suppression of CDKN1C (Figure 1C, lanes 3–4). We next determined whether this E2-mediated regulation occurred at the transcriptional or posttranscriptional level. MCF7 cells were treated with actinomycin D (RNA synthesis inhibitor, 0.25 μg/ml) or cycloheximide (protein synthesis inhibitor, 5 μg/ml) prior to 12 h E2 stimulation. Both inhibitor treatments resulted in the abolishment of E2-mediated repression (lanes 5–8). This is consistent with previous reports that late downregulation of target genes by ligand-bound ERα often requires the earlier synthesis of nuclear receptor corepressors (22). Treatment of MCF7 cells with MG132 (proteasome inhibitor, 1 μM) showed little or no effect on the E2-mediated regulation of CDKN1C (lanes 9–10). Taken together, these results suggest that estrogen-mediated repression of CDKN1C takes place at the transcriptional level in breast cancer cells. This transcriptional control was additionally observed for two known ERα targets, CCNG2 and NRIP1 (supplementary Figure S1A–B is available at Carcinogenesis Online).
In a previous study, we showed that critical genes controlling tumor suppression can be downregulated on prolonged estrogen exposure to breast progenitor cells (21). Therefore, we investigated whether chronic exposure of progenitor-containing mammospheres to low-dose (4 nM) ERα agonist bisphenol A would later affect CDKN1C expression in their MDEC progeny. Expression of CDKN1C was assayed by qRT–PCR in 10 primary MDEC samples differentiated from mammospheres exposed to bisphenol A for 3 weeks (Figure 1D, left). While we noticed individual variations in response to this preexposure, MDECs usually showed lower expression of CDKN1C compared with vehicle (P < 0.05) (Figure 1D, right). Since the repression could be transmitted from progenitor cells to MDECs, we suggest that epigenetic mechanisms are probably used to reinforce this exposure memory in the progeny.
To further determine whether epigenetic modifications contribute to the decreased expression of CDKN1C, MCF7 cells were subjected to pharmacological inhibition of DNA methylation (DAC, 5 μM) and/or histone deacetylation (TSA, 300 nM). Coadministration of DAC and TSA induced expression 14-fold, whereas DAC alone enhanced expression 1.7-fold (Figure 2A). Treatment with TSA alone also induced CDKN1C expression (supplementary Figure S1C is available at Carcinogenesis Online). Consistent with the prior observation in Figure 1, E2-induced repression of CDKN1C was maintained in cells treated with both DAC and TSA (Figure 2A).
The above studies suggest that estrogen stimulation impacts the chromatin state of the CDKN1C locus. We then conducted chromatin profiling in seven regions of the CDKN1C locus, spanning −976 to +2857 bp relative to TSS (Figure 2B; supplementary Figure S2 is available at Carcinogenesis Online) and a previously reported E2 off-target control region in the FKBP6 gene (22). Treatment of MCF7 cells with E2 significantly reduced the levels of active histone modifications (i.e. H3Ac and H3K4me2) at the first exon and first intron, respectively. These changes were concomitant with increased recruitment of the chromatin-dependent transcriptional regulators HDAC1 and CTCF in wider regions, including downstream regions at the 3′-end. E2 stimulation also resulted in increased levels of repressive H3K27me3 and H3K9me2 in the CDKN1C locus at the firstst exon and TSS, respectively. Increased RNA Pol II occupancy was observed at the first exon and 3′ downstream regions also bound by CTCF. These results suggest that upon activation of estrogen signaling, repressive chromatin marks are established in the 5′ regulatory region of CDKN1C.
The above result led us to investigate the role of a noncoding transcript,KCNQ1OT1, in repressing CDKN1C. This chromatin-interacting RNA, known to be transcribed in the 11p15.5 imprinting center, negatively regulates the paternal CDKN1C allele in normal cells (7). To determine whether E2 induces the expression of this transcript, KCNQ1OT1 RNA levels were measured by qRT–PCR in MCF7 cells treated with 10 nM E2 or vehicle in the presence or absence of 4-hydroxytamoxifen or actinomycin D. After 12 h, E2 treatment induced the expression of KCNQ1OT1 1.75-fold (P < 0.05) (Figure 3A). Importantly, the induction of KCNQ1OT1 was abolished by ER antagonist 4-hydroxytamoxifen as well as transcriptional inhibitor actinomycin D, demonstrating that KCNQ1OT1 expression is regulated by estrogen signaling.
The active status of the 11p15.5 ICR was confirmed by ChIP–qPCR (Figure 3B). Upon E2 stimulation, we observed significantly increased recruitment of CTCF, the binding of which is known to mediate long-range epigenetic silencing (23,24). To test the hypothesis that CTCF plays a functional role in the regulation of CDKN1C expression, we transiently transfected MCF7 cells with CTCF siRNA oligonucleotides. After 72 h, siRNA knockdown reduced the expression of CTCF 3-fold (P < 0.001) (Figure 3C, top). Targeting of CTCF had no effect on expression of control genes GAPDH and HPRT1 (supplementary Figure S3A–B is available at Carcinogenesis Online). CTCF depletion significantly increased CDKN1C messenger RNA (mRNA) levels (1.6-fold, P < 0.001) (Figure 3C, bottom). We next determined the effect of CTCF inhibition on the regulation of CDKN1C by E2 signaling. Depletion of CTCF significantly reduced the estrogen-mediated repression of CDKN1C (Figure 3D; supplementary Figure S3C is available at Carcinogenesis Online). Taken together, these results suggest that repression of CDKN1C by estrogen signaling may involve the long-range binding activity of CTCF in the 11p15.5 imprinted domain.
To determine whether there is existing evidence that an ERα regulatory complex may be recruited to the 11p15.5 imprinted domain, we analyzed published genome-wide MCF7 ERα and FOXA1 cistromes determined by ChIP–chip (25). One high-confidence distal ERα-binding site (4.15-fold change, P < 5−17, false discovery rate 0%) was detected 300 and 114 kb upstream of the KCNQ1OT1 TSS and CDKN1C TSS, respectively. The distal site also shares overlapping binding with the nuclear hormone receptor cofactor FOXA1 (2.28-fold change, P < 5−08, false discovery rate 0.75%). These results suggest that ERα signaling may repress CDKN1C through long-range transactivation of the 11p15.5 ICR-KCNQ1OT1 domain.
The absence of a correlation between decreased CDKN1C expression and ERα status in patients suggests that ERα signaling may be sufficient but not required for the repression of CDKN1C in breast cancer (26). While multiple regions of 11p are subject to loss of heterozygosity in primary breast tumors (27), loss of KCNQ1OT1 promoter CpG island methylation at the 11p15.5 ICR has yet to be investigated in a large breast cancer patient population or by modern high-resolution quantitative techniques. We therefore determined the methylation status of this CpG island in 52 breast cancer cell lines, 306 primary tumors and 20 normal tissues by bisulfite PCR pyrosequencing (Figure 4A). The assay demonstrated high specificity. Individual values for normal tissues ranged from 45.6 to 54.8% (average 49.2 ± 2.2%), approaching the theoretical value of 50% (supplementary Figure S4A–B is available at Carcinogenesis Online) expected at the differentially methylated KCNQ1OT1 locus (12). The methylated maternal allele is known to release the repression of CDKN1C in cis, whereas the unmethylated paternal allele acts in cis to repress CDKN1C (see illustration in Figure 6C). Following previous protocols for pyrosequencing analysis of 11p15 differentially methylated domains (28), we applied the cutoff value of ≤ 35% to classify samples as having decreased or hypomethylation. Decreased methylation of KCNQ1OT1 was observed in 50% (26) of 52 breast cancer cells lines and 15% (46) of 306 primary tumors examined [data summarized in Figure 4A, population frequency of methylation values plotted in supplementary Figure S4A–B (available at Carcinogenesis Online)]. This initial result suggests that hypomethylation of the imprinting center occurs in breast cancer. Alternatively, this methylation imbalance can be attributed to genetic loss of the methylated maternal allele of KCNQ1OT1.
To this end, we analyzed the combined 11p15.5 methylation status, 11p15.5 DNA copy number and CDKN1C expression levels of 20 breast cancer cell lines by bisulfite PCR pyrosequencing, DNA copy number qPCR and qRT–PCR, respectively (results summarized in supplementary Table SI, available at Carcinogenesis Online). Linear regression analysis showed that CDKN1C expression decreased with genetic loss of 11p15.5 ICR (R2 = 0.523, P < 0.001) (supplementary Figure S5A is available at Carcinogenesis Online). There was a similarly strong relationship between decreased CDKN1C expression and loss but not gain of 11p15.5 methylation (R2 = 0.533, P < 0.001) (supplementary Figure S5B is available at Carcinogenesis Online). Loss of methylation occurred in conjunction with genetic loss of 11p15.5 ICR (R2 = 0.583, P < 0.001) (supplementary Figure S5C is available at Carcinogenesis Online). Multiple linear regression analysis showed that the expression of CDKN1C could be most strongly predicted from the combined genetic and epigenetic status of the 11p15.5 ICR locus (R2 = 0.612, P < 0.001) (Figure 4B). While instances of 11p15.5 ICR hypermethylation (cutoff value ≥ 65%) were observed in primary tumors and breast cancer lines, expression of CDKN1C was stochastic in the four hypermethylated cell lines. We investigated the effects of E2 stimulation on CDKN1C expression in hypermethylated T47D and MDA-MB-453 cells. Treatment with E2 had no statistically significant effect on the expression of CDKN1C in these cell lines (Figure 4C). Taken together, these results suggest that the observed hypomethylation event in breast cancer is actually derived from genetic deletion of the methylated 11p15.5 ICR allele, leaving only the presence of the active KCNQ1OT1-expressing allele that represses CDKN1C.
While natural antisense transcripts potentially constitute a large class of regulatory noncoding RNAs, the functional relevance of antisense transcription remains largely speculative for the majority of bidirectionally transcribed genes (29). Genomic alignments of expressed sequence tags and Long-SAGE tags in opposite orientation to the CDKN1C locus (supplementary Tables SII–SIV are available at Carcinogenesis Online) as well as more recent data generated by asymmetric strand-specific analysis of gene expression (30) suggest the presence of a cis-encoded unannotated antisense transcript expressed in a variety of human tissues. The presence of a novel antisense transcript paired with CDKN1C was determined by a series of strand-specific RT–PCR assays over a region >2.5 kb in length in MCF7 cells treated with E2 (Figure 5A–B; supplementary Figure S6 is available at Carcinogenesis Online). MCF7 cells expressed approximately one copy of antisense transcript for every 59 copies of sense transcript (compare vehicle treatment; supplementary Figure S6B and C is available at Carcinogenesis Online). Similar results were observed in a second breast cancer cell line, BT474 (supplementary Figure S6D is available at Carcinogenesis Online). We confirmed E2 repressed sense CDKN1C expression 3-fold (P < 0.001) (supplementary Figure S6B is available at Carcinogenesis Online). In contrast, expression of CDKN1C-AS was significantly induced by 12 h of E2 treatment in MCF7 cells subjected to pharmacologic inhibition of DNA methylation and histone deacetylation (P < 0.01) (Figure 5B). Interestingly, analysis of published Affymetrix microarray datasets with probes in opposite orientation to CDKN1C showed that increased expression of CDKN1C-AS was associated with poor bone metastasis-free survival in breast cancer patients (P = 0.006, Log rank test) (supplementary Figure S7 is available at Carcinogenesis Online).
We then conducted a reporter assay to determine whether CDKN1C-AS might play a role in repressing CDKN1C expression in MCF7 cells. Three reporter constructs were engineered: (i) p57-S contains the CDKN1C promoter and first exon and a green fluorescent protein (GFP) reporter gene, (ii) p57-AS, with the same features but containing a portion of CDKN1C-AS driven by a cytomegalovirus promoter located between the CDKN1C promoter and the GFP gene and (iii) p57-ASΔ contains a stop sequence downstream of the cytomegalovirus promoter that prevents CDKN1C-AS transcription (Figure 5C). These constructs were transfected into MCF7 cells, selected for 4 weeks and evaluated for GFP activity by qRT–PCR analysis of GFP mRNA expression. The exogenous antisense transcript failed to repress expression of GFP in cis (Figure 5D, left). Expression of the antisense transcript by p57-AS was elevated >35-fold compared with endogenous antisense CDKN1C (p57-S) and not elevated in p57-ASΔ (middle). This result led us to suggest that the antisense transcript might repress CDKN1C expression in trans. To explore this possibility, we measured endogenous CDKN1C mRNA levels in cells expressing the three constructs. Endogenous CDKN1C was significantly repressed (P < 0.01) in p57-AS-expressing cells but not p57-ASΔ-expressing cells (right).
Epigenetic lesions such as repressive histone modification and CpG island hypermethylation are frequently associated with the silencing of tumor suppressor genes. Yet the underlying mechanisms for transmitting instructions to deposit these repressive epigenetic marks at specific locations across the genome are poorly understood. This study has demonstrated that expression of the imprinted gene CDKN1C is suppressed by ERα signaling in hormone-responsive breast cancer cells through epigenetic mechanisms probably involving the chromatin-interacting ncRNA KCNQ1OT1 and the chromatin-dependent transcriptional regulator CTCF. Several lines of evidence suggest that this epigenetic repression is mediated through long-range transactivation of the 11p15.5 ICR-KCNQ1OT1 domain. First, our experiments show that activation of estrogen signaling led to the concomitant induction of KCNQ1OT1, transcriptional repression of CDKN1C and establishment of repressive chromatin marks at the 5′ regulatory region of CDKN1C in MCF7 cells. Consistent with our results, CDKN1C is reported to be a direct target of polycomb group protein EZH2 in breast cancer cells (31). Second, upon E2 stimulation, we also observed significantly increased recruitment of HDAC1 and CTCF to the CDKN1C locus and distal 11p15.5 ICR region. Using chromosome conformation capture (3C) methodology, Vu et al. (32) recently demonstrated that the 11p15.5 ICR physically interacts with the CDKN1C promoter in MCF7 cells. Knockout studies have demonstrated that HDAC1 regulates the expression of Cdkn1c and other imprinted genes in the syntenic region of mouse chromosome 7 (33). The binding of CTCF plays a critical role in the epigenetic regulation of higher-order chromatin structure and long-range gene silencing in this region as well (23,24). Third, several regions of the CDKN1C locus were co-enriched for both CTCF and Pol II following estrogen treatment. CTCF can interact with and recruit the largest subunit of Pol II to CTCF-binding sites genome wide (34). Fourth, transient depletion of CTCF by siRNA increased CDKN1C expression and significantly reduced the estrogen-mediated repression of CDKN1C. Transgenic studies have shown KvDMR1, the mouse homolog of 11p15.5 ICR, can silence Cdkn1c independent of Kcnq1ot1 transcription (35). Biochemical and in vivo evidence suggest that this intrinsic silencer activity involves the binding of CTCF (17,18). Fifth, breast cancer cell lines harboring 11p15.5 ICR hypermethylation showed limited or no capacity to regulate CDKN1C through activation of estrogen signaling. If this regulation is dependent on CTCF binding of 11p15.5 ICR, binding should not occur if both alleles are methylated (3,18). Finally, analysis of published MCF7 genome-wide ChIP–chip datasets showed ERα and nuclear hormone receptor cofactor FOXA1 are both recruited to a distal site in the 11p15.5 imprinted domain upstream of KCNQ1OT1 and CDKN1C (25). Based on our experiments, established models of long-range chromatin interactions (23,24,32), and our recent finding that estrogen signaling is capable of triggering complex loop formation between cis-regulatory regions and promoters (36), we suggest a possible looping mechanism for repressing the paternal CDKN1C allele through long-distance interactions as depicted in Figure 6A.
Using high-resolution bisulfite pyrosequencing, we observed that loss of KCNQ1OT1 promoter CpG island (11p15.5 ICR) methylation occurs frequently in breast cancer cells and primary tumors. While loss of heterozygosity in regions of 11p were first reported in breast carcinomas more than two decades ago (27), loss of 11p15.5 ICR methylation in adult cancers has been attributed to hypomethylation (13,14). Our linear regression analyses of CDKN1C mRNA levels, 11p15.5 ICR DNA methylation and copy number in breast cancer cell lines suggest that DNA demethylation resulting from genetic loss of the methylated KCNQ1OT1-ICR allele leads to repression of CDKN1C expression. On the other hand, deletion of the unmethylated allele could in principle lead to induction of CDKN1C, possibly through disruption of the proposed chromatin looping mechanism. This speculation, however, warrants further investigation.
We identified a cis-encoded unannotated antisense transcript inversely expressed with CDKN1C. This inverse ratio of antisense to sense transcription is in agreement with other studies (37,38). Moreover, inversely expressed sense–antisense pairs have striking conservation throughout evolution (39). Expression of endogenous CDKN1C-AS was significantly induced by E2 treatment in MCF7 cells subjected to pharmacologic inhibition of DNA methylation and histone deacetylation. Inhibition of DNA methyltransferase and histone deacetylase activity in cultured cells causes relaxation of imprinting, leading to the biallelic expression of imprinted genes (40–42). Given that mechanistic differences in imprinted gene silencing exist among various embryonic lineages (35), it is possible that estrogen may regulate CDKN1C-AS in a tissue-specific manner. Alternatively, CDKN1C-AS may play a role in establishing rather than maintaining the aberrant epigenetic silencing of CDKN1C. Strand-specific qRT–PCR analysis demonstrated that expression of both CDKN1C and CDKN1C-AS were significantly induced following co-treatment with DAC and TSA. This suggests the transcriptional potential of CDKN1C-AS is reduced as the locus becomes epigenetically repressed. Using a reporter assay, we confirmed in transfected MCF7 cells that the antisense transcript represses CDKN1C in trans. Based on this in vitro study, we suggest a regulatory mechanism of this noncoding transcript as depicted in Figure 6B. Repression may occur through the formation of double-stranded RNAs, which negatively regulate stability, transport and/or translation of the sense transcript (29,43,44). However, this speculation merits further investigation in the future.
This study demonstrated that expression of the imprinted gene CDKN1C is suppressed by estrogen signaling in hormone-responsive breast cancer cells through epigenetic mechanisms involving the ncRNA KCNQ1OT1 and CTCF. Whereas this repression is frequently linked to DNA hypermethylation of the CDKN1C promoter CpG island in hematologic cancers (45,46) and other solid tumors (47,48), we and other investigators have not observed this aberrant event in breast cancer (31) (Rodriguez, B.A.T. and Huang, T.H.M, unpublished results). Based on our present findings and previous imprinting studies, we propose three models by which repression of CDKN1C may occur in breast cancer (Figure 6C). First, expression may be repressed in trans by CDKN1C-AS possibly through a double-stranded RNA mechanism triggered by formation of a sense–antisense RNA duplex. Under certain cellular conditions, repression in trans may be induced by estrogen-mediated upregulation of CDKN1C-AS. Second, CDKN1C expression may be lost due to genetic deletion of the 11p15.5 ICR. In this model, DNA demethylation resulting from genetic loss of the methylated allele leads to aberrant domain silencer activity mediated by CTCF recruitment and KCNQ1OT1 transcription, repressing CDKN1C. Third, expression may be repressed by estrogen-mediated transactivation of the 11p15.5 ICR. Estrogen induces transcription of KCNQ1OT1 and CTCF recruitment to mediate ICR silencer activity, which in turn directs the epigenetic repression of the CDKN1C locus. Based on our findings, we predict that in addition to promoter hypermethylation, epigenetic repression of tumor suppressor genes by CTCF and ncRNAs could be more common and important than previously understood.
National Institutes of Health; Ohio State University Comprehensive Cancer Center (U01ES015986, U54CA113001, R01CA069065, R01ES017594).
Conflict of Interest Statement: None declared.