Oncogene (2014), 1–11 & 2014 Macmillan Publishers Limited All rights reserved 0950-9232/14 www.nature.com/onc
ORIGINAL ARTICLE
PRC2-independent chromatin compaction and transcriptional repression in cancer C Vallot1,2,5, A He´rault1,2,5, S Boyle3, WA Bickmore3,4,6 and F Radvanyi1,2,6 The silencing of large chromosomal regions by epigenetic mechanisms has been reported to occur frequently in cancer. Epigenetic marks, such as histone methylation and acetylation, are altered at these loci. However, the mechanisms of formation of such aberrant gene clusters remain largely unknown. Here, we show that, in cancer cells, the epigenetic remodeling of chromatin into hypoacetylated domains covered with histone H3K27 trimethylation is paralleled by changes in higher-order chromatin structures. Using fluorescence in situ hybridization, we demonstrate that regional epigenetic silencing corresponds to the establishment of compact chromatin domains. We show that gene repression is tightly correlated to the state of chromatin compaction and not to the levels of H3K27me3—its removal through the knockdown of EZH2 does not induce significant gene expression nor chromatin decompaction. Moreover, transcription can occur with intact high-H3K27me3 levels; treatment with histone deacetylase inhibitors can relieve chromatin compaction and gene repression, without altering H3K27me3 levels. Our findings imply that compaction and subsequent repression of large chromatin domains are not direct consequences of PRC2 deregulation in cancer cells. By challenging the role of EZH2 in aberrant gene silencing in cancer, these findings have therapeutical implications, notably for the choice of epigenetic drugs for tumors with multiple regional epigenetic alterations. Oncogene advance online publication, 27 January 2014; doi:10.1038/onc.2013.604 Keywords: cancer; chromatin compaction; gene cluster; histone deacetylase inhibitors; polycomb
INTRODUCTION Transcription of individual genes can be altered in cancer by genetic but also epigenetic lesions. Focal promoter DNA methylation can, for example, modify gene expression. However, studying the regulation of genes in isolation without taking into account their genomic context limits the understanding of regulatory networks in cancer. Many clusters of co-regulated genes have been reported in mammalian genomes,1 such as the HOX clusters or the classical major histocompatibility complex. The regulation of these genomic regions is complex and can result from the interaction of multiple factors including epigenetic modifications,2,3 chromatin conformation and nuclear localization,4 non coding RNAs5 or polycomb complexes.6 PRC2 and PRC1 polycomb complexes take part in the delineation and regulation of gene clusters, such as the HOX or Kcnoqt1 clusters,7 notably through the deposition of H3K27me3 and chromatin compaction mechanisms.8,9 Several studies have addressed the existence of co-regulated gene clusters specific to cancer cells. Regions of expression bias, where genes are coordinately deregulated in tumors compared with normal tissue, have been reported in various tumor types using large-scale expression data sets. Many of these regions were attributable to known DNA copy-number alterations,10 however, a significant portion of them are independent of such genetic alterations.11 Moreover, it has been reported that epigenetic alterations, such as histone modifications with12,13 or without11
aberrant DNA methylation, can affect several neighboring genes in cancer and correspond to the formation of clusters of corepressed genes. Recently, we and others showed that regional epigenetic silencing is frequent in bladder and prostate cancer.14,15 However, little is known about the mechanisms underlying this novel epigenetic alteration. We previously reported in bladder cancer that regional epigenetic silencing occurs in a particularly aggressive group of tumors, the tumors of the carcinoma in situ pathway. This silencing affects simultaneously several regions defining a new epigenetic phenotype that we called multiple regional epigenetic silencing (MRES).15 Whereas superficial bladder tumors were already molecularly well defined, in particular by FGFR3 mutation,16 MRES was one of the first molecular events characterizing tumors of the carcinoma in situ pathway. Such a phenotype suggests that there might be a common mechanism for the formation of the repressed gene clusters. We had previously shown that, among others chromatin marks, H3K27me3, a hallmark of the PRC2 enzyme EZH2, was characteristic of these alterations.15 Although it has been shown that EZH2 overexpression in cancer could be key in altering expression profiles through aberrant silencing mechanisms,17,18 recent data suggest that EZH2 oncogenic properties might not be necessarily linked to gene silencing and polycomb activities.19 Here, we study the involvement of EZH2, and its footprint H3K27me3, in the deregulation of the cancer epigenome
1 CNRS, UMR 144 - Cell Biology Department, Institut Curie, Paris, France; 2Institut Curie, Centre de Recherche, Paris, France; 3Chromosome and Gene Expression Section, MRC Human Genetics Unit, MRC Institute of Genetics and Molecular Medicine at the University of Edinburgh, Scotland, UK and 4Breakthrough Breast Cancer Research Unit, University of Edinburgh, Scotland, UK. Correspondence, Current address: Dr C Vallot, UMR7216, Universite´ Paris Diderot, 35 rue He´le`ne Brion, Paris 75205, France or Dr F Radvanyi, CNRS, UMR 144, Institut Curie, Paris 75248, France. E-mail: [email protected] or [email protected] 5 These authors contributed equally to this work. 6 The last two authors contributed equally to this work. Received 9 July 2013; revised 10 December 2013; accepted 20 December 2013
PRC2-independent silencing in cancer C Vallot et al
2 through the remodeling of large chromatin domains in bladder cancer cells. We specifically assess the link between H3K27me3, chromatin compaction and gene repression. We show that gene repression is tightly correlated to the state of chromatin compaction but not with H3K27me3 levels in cancer cells. RESULTS Regional epigenetic silencing events correspond to higher-order chromatin alterations We had previously identified seven gene clusters recurrently epigenetically repressed in bladder cancer; these chromosomal regions are concomitantly silenced in a subgroup of aggressive tumors, defining a multiple regional epigenetic phenotype (Figure 1a).15 Here, we studied more extensively the chromatin landscape and conformation at these clusters to get insight into the regulation of such epigenetic alterations. We first used chromatin immunoprecipitation (ChIP) coupled with quantitative PCR (qPCR) to analyze the levels of H3K27me3 and global H3 acetylation (H3ac), not only on promoters but also within and between the genes across the entire loci. We analyzed two of the seven clusters: 2q31 (HOXD8, HOXD4, HOXD3 and HOXD1) and 3p22 (VILL, PLCD1, DLEC1 and ACAA1) in normal human urothelial (NHU) cells in culture20 as well as in two bladder cancer cell lines, CL1207 and RT112. CL1207 is derived from a high-grade tumor with the MRES phenotype, RT112 is from a well-differentiated tumor without the MRES phenotype.15 In CL1207 cells, but not in RT112, genes belonging to both clusters of interest are epigenetically repressed relative to normal bladder cells.15 High levels of the polycomb-mediated histone modification H3K27me3 are not restricted to gene promoters in CL1207 cells (Figure 1b): all positions assayed in the clusters display high H3K27me3 levels in comparison with the ubiquitously expressed GAPDH in the same cells, and compared with NHU cells and RT112 cells. Conversely, H3 was hypoacetylated across the two gene clusters in CL1207 cells relative to levels in NHU and RT112 cells (Figure 1b) and compared with GAPDH. Therefore, the repressed clusters in MRES cancer cells correspond to the formation of relatively homogeneous chromatin domains characterized by high levels of H3K27me3 and low levels of H3ac. Histone acetylation has been shown to directly affect secondary chromatin structures: it can neutralize charges in the histone N-terminal tails and thereby nucleosome-DNA or nucleosome– nucleosome interactions.21,22 Moreover, the polycomb repressive complexes, especially PRC1, have also been shown to be involved in chromatin compaction in vitro23 and in vivo.8 Given the observed histone modification profiles in CL1207 cells, we examined whether there was a corresponding chromatin compaction at the two MRES loci in bladder cancer cells. We used fluorescence in situ hybridation (FISH) with fosmid probe pairs spanning the 2q31 and 3p22 clusters to assess the state of chromatin compaction of these loci in the MRES þ cancer cell line
CL1207 and in the MRES cancer cell line RT112 as well as in NHU cells. As a control, we chose a fosmid probe pair at a 1q21 locus that was rarely affected by DNA alterations or expression changes in our bladder cancer data set (data not shown). We measured the distance (d) between the hybridization signals from these probe pairs in nuclei of the different cell lines. At probe separation of o2 Mb, there is a linear relationship between the mean-square interphase distances between two loci (d2) and their separation in kilobases.24 Such analysis has been used to show that different parts of the human genome display various levels of chromatin compaction,25 and that the same genomic region can change its compaction state in differentiation, development4,26 or when the activity of chromatin modifiers is ablated.8 In the latter cases, this included analyses of the HoxD locus, the murine homolog of one of the repressed gene clusters (2q31) studied here in human cancer cells. At the 3p22 cluster (Figure 1c, left panel), the interphase separation of the probe signals (127 kb apart) in the nuclei of CL1207 cells (mean d2 ¼ 0.17 mm2) is significantly smaller than in RT112 (d2 ¼ 0.33 mm2, P ¼ 0.005) or NHU (d2 ¼ 0.53 mm2, Po10 4) cells. This indicates that this genomic region is in a more compact chromatin state in CL1207 than in RT112 and NHU cells, whereas the chromatin compaction does not differ significantly between RT112 and NHU cells (P ¼ 0.2). A similarly more compact chromatin state in CL1207 cells was seen for probes spanning 104 kb at the HOXD cluster at 2q31 (d2 ¼ 0.20 mm2) in comparison with RT112 cancer cells (d2 ¼ 0.27 mm2, P ¼ 0.01) or normal NHU cells (d2 ¼ 0.27 mm2, P ¼ 0.005) (Figure 1c, central panel). This compaction is not a general feature of the chromatin structure throughout the genome in CL1207 cells, as shown by the similar level of chromatin compaction between all three cell lines at the control 1q21 locus (Figure 1c, right panel). We also confirmed the link between clustered gene repression and chromatin compaction in two other bladder cancer cell lines for 2q31 (Supplementary Figure 1). In the MRES þ TCCSUP cancer cell line, as in CL1207, genes of the 2q31 locus are repressed, whereas they are not in the MRES- MGHU3 cancer cell line.15 Compared with NHU cells (d2 ¼ 0.27 mm2) the chromatin at 2q31 is more condensed in TCCSUP (d2 ¼ 0.14 mm2, P ¼ 0.01), but not in MGHU3 (d2 ¼ 0.32 mm2, P ¼ 0.93). EZH2 is overexpressed in tumors with multiple regional epigenetic alterations As we found that the regions of epigenetic alterations are covered with abnormal H3K27me3 in cancer cells (Figure 1), we chose to investigate the role of histone methyltransferases capable of catalyzing the addition of a trimethyl group on H3K27: EZH127,28 and EZH2.29 Although a potential role of EZH1 in cancer has not yet been addressed, the deregulated expression and activity of EZH2 has been described. EZH2 is highly expressed in various tumor types,30 including breast,31 prostate,32 bladder33 or liver
Figure 1. Repressive chromatin domains form in cancer cells. (a) Location of the seven recurrent regional epigenetic alterations characteristic of the MRES phenotype. (b) Study of the 3p22 and 2q31 clusters: ChIP assays were performed for all promoters, for one position in the middle of every gene and between every pair of neighboring genes with an antibody against H3K27me3 (upper panel) and H3ac (lower panel). The graph shows the amount of immunoprecipitated target DNA expressed as a ratio of total input DNA measured by qPCR. Error bars indicate the variation between the means of two independent experiments. The map (not to scale) under each graph indicates the position of the primer pairs used for each point on the graph. The amount of immunoprecipitated DNA for the GAPDH promoter is shown as a reference. Data from the MRES bladder cancer cell line CL1207 is plotted in green; the non-MRES bladder cancer cell line RT112 in red; and normal human urothelial (NHU) cells in blue. (c) Two-dimensional FISH with a probe pair spanning the 3p22 cluster (left panel), the 2q31 cluster (middle panel) and a control locus in 1q21 (right panel). Above images, the map (to scale) shows the probe pair position (in bp) in the UCSC human genome browser (March 2006 Assembly, hg18). For the images, scale bar ¼ 5 mm. Below images, box plots showing the distribution of interprobe distances2 (d2) of each set of data, obtained from 50 nuclei. The horizontal line in the middle box represents the median, the two hinges of the middle box represent the interquartile range of the data and the empty circles indicate outliers. The whiskers extend to the most extreme data point that is no more than 1.5 times the interquartile range from the box. The statistical significance of differences was examined by Mann–Whitney U-tests (Materials and methods). Oncogene (2014) 1 – 11
& 2014 Macmillan Publishers Limited
PRC2-independent silencing in cancer C Vallot et al
3 cancer.34 Moreover, gain of function mutations of EZH2 have been reported in lymphomas.35,36 To understand if deregelulation of EZH2 expression might be involved in the appearance of the MRES phenotype, we compared EZH2 expression levels using Affymetrix arrays in a wide tumor set (n ¼ 150) between invasive tumors with (n ¼ 74) and without (n ¼ 29) MRES phenotype and normal urothelium samples (n ¼ 4). Tumors with the MRES phenotype were identified using Affymetrix expression data and cluster
analysis.15 As the MRES phenotype is found in invasive bladder tumors, we limited our analysis to invasive tumors, so that differences in EZH2 expression levels between MRES þ and MRES tumors would not only be due to the heterogeneity of tumor stages between each group. EZH2 is significantly more highly expressed in invasive tumors (T2–T4) with MRES phenotype than in invasive tumors without the MRES phenotype (P ¼ 0.018, t-test) and in normal samples (Figure 2a). This was validated by
* * 1
2
3
4
5
6
7
8
9
10
12
11
13
14
15
16
17
18
21
22
Anti-H3K27me3
0.5
0.6
0.4
0.5
0.3
0.4 0.3
0.2
0.2 0.1
0.1
0
PLCD1 DLEC1 ACAA1
NHU
HOXD8 HOXD4 HOXD3 HOXD1
RT112 CL1207
Anti-H3ac
GAPDH
0
GAPDH
VILL
Anti -H3ac
0.3 Ratio Bound over Input +/- SD
20
0.4 0.3
0.2
0.2 0.1 0.1 0 GAPDH
VILL
0
PLCD1 DLEC1 ACAA1
GAPDH
Ratio Bound over Input +/- SD
Anti-H3K27me3
19
HOXD8 HOXD4 HOXD3 HOXD1
3p22 cluster
2q31 cluster
Control locus 1q21
G248P80945A9 - G248P82821D6 (127 kb)
G248P89002G8 - G248P82958B3 (104 kb)
G248P80671B3 - G248P87792E10 (221 kb)
38000000
38100000
38200000
176700000
176800000
149900000
150000000
150100000
CL1207 CL1207 EVX2 PLCD1
HOXD9
HOXD13
CTDSPL
ACAA1
DLEC1
HOXD12
OXSR1 HOXD11
VILL
Distance2 (μm2)
NHU
MYD88
RT112
3
CL1207
p
ORIGINAL ARTICLE
PRC2-independent chromatin compaction and transcriptional repression in cancer C Vallot1,2,5, A He´rault1,2,5, S Boyle3, WA Bickmore3,4,6 and F Radvanyi1,2,6 The silencing of large chromosomal regions by epigenetic mechanisms has been reported to occur frequently in cancer. Epigenetic marks, such as histone methylation and acetylation, are altered at these loci. However, the mechanisms of formation of such aberrant gene clusters remain largely unknown. Here, we show that, in cancer cells, the epigenetic remodeling of chromatin into hypoacetylated domains covered with histone H3K27 trimethylation is paralleled by changes in higher-order chromatin structures. Using fluorescence in situ hybridization, we demonstrate that regional epigenetic silencing corresponds to the establishment of compact chromatin domains. We show that gene repression is tightly correlated to the state of chromatin compaction and not to the levels of H3K27me3—its removal through the knockdown of EZH2 does not induce significant gene expression nor chromatin decompaction. Moreover, transcription can occur with intact high-H3K27me3 levels; treatment with histone deacetylase inhibitors can relieve chromatin compaction and gene repression, without altering H3K27me3 levels. Our findings imply that compaction and subsequent repression of large chromatin domains are not direct consequences of PRC2 deregulation in cancer cells. By challenging the role of EZH2 in aberrant gene silencing in cancer, these findings have therapeutical implications, notably for the choice of epigenetic drugs for tumors with multiple regional epigenetic alterations. Oncogene advance online publication, 27 January 2014; doi:10.1038/onc.2013.604 Keywords: cancer; chromatin compaction; gene cluster; histone deacetylase inhibitors; polycomb
INTRODUCTION Transcription of individual genes can be altered in cancer by genetic but also epigenetic lesions. Focal promoter DNA methylation can, for example, modify gene expression. However, studying the regulation of genes in isolation without taking into account their genomic context limits the understanding of regulatory networks in cancer. Many clusters of co-regulated genes have been reported in mammalian genomes,1 such as the HOX clusters or the classical major histocompatibility complex. The regulation of these genomic regions is complex and can result from the interaction of multiple factors including epigenetic modifications,2,3 chromatin conformation and nuclear localization,4 non coding RNAs5 or polycomb complexes.6 PRC2 and PRC1 polycomb complexes take part in the delineation and regulation of gene clusters, such as the HOX or Kcnoqt1 clusters,7 notably through the deposition of H3K27me3 and chromatin compaction mechanisms.8,9 Several studies have addressed the existence of co-regulated gene clusters specific to cancer cells. Regions of expression bias, where genes are coordinately deregulated in tumors compared with normal tissue, have been reported in various tumor types using large-scale expression data sets. Many of these regions were attributable to known DNA copy-number alterations,10 however, a significant portion of them are independent of such genetic alterations.11 Moreover, it has been reported that epigenetic alterations, such as histone modifications with12,13 or without11
aberrant DNA methylation, can affect several neighboring genes in cancer and correspond to the formation of clusters of corepressed genes. Recently, we and others showed that regional epigenetic silencing is frequent in bladder and prostate cancer.14,15 However, little is known about the mechanisms underlying this novel epigenetic alteration. We previously reported in bladder cancer that regional epigenetic silencing occurs in a particularly aggressive group of tumors, the tumors of the carcinoma in situ pathway. This silencing affects simultaneously several regions defining a new epigenetic phenotype that we called multiple regional epigenetic silencing (MRES).15 Whereas superficial bladder tumors were already molecularly well defined, in particular by FGFR3 mutation,16 MRES was one of the first molecular events characterizing tumors of the carcinoma in situ pathway. Such a phenotype suggests that there might be a common mechanism for the formation of the repressed gene clusters. We had previously shown that, among others chromatin marks, H3K27me3, a hallmark of the PRC2 enzyme EZH2, was characteristic of these alterations.15 Although it has been shown that EZH2 overexpression in cancer could be key in altering expression profiles through aberrant silencing mechanisms,17,18 recent data suggest that EZH2 oncogenic properties might not be necessarily linked to gene silencing and polycomb activities.19 Here, we study the involvement of EZH2, and its footprint H3K27me3, in the deregulation of the cancer epigenome
1 CNRS, UMR 144 - Cell Biology Department, Institut Curie, Paris, France; 2Institut Curie, Centre de Recherche, Paris, France; 3Chromosome and Gene Expression Section, MRC Human Genetics Unit, MRC Institute of Genetics and Molecular Medicine at the University of Edinburgh, Scotland, UK and 4Breakthrough Breast Cancer Research Unit, University of Edinburgh, Scotland, UK. Correspondence, Current address: Dr C Vallot, UMR7216, Universite´ Paris Diderot, 35 rue He´le`ne Brion, Paris 75205, France or Dr F Radvanyi, CNRS, UMR 144, Institut Curie, Paris 75248, France. E-mail: [email protected] or [email protected] 5 These authors contributed equally to this work. 6 The last two authors contributed equally to this work. Received 9 July 2013; revised 10 December 2013; accepted 20 December 2013
PRC2-independent silencing in cancer C Vallot et al
2 through the remodeling of large chromatin domains in bladder cancer cells. We specifically assess the link between H3K27me3, chromatin compaction and gene repression. We show that gene repression is tightly correlated to the state of chromatin compaction but not with H3K27me3 levels in cancer cells. RESULTS Regional epigenetic silencing events correspond to higher-order chromatin alterations We had previously identified seven gene clusters recurrently epigenetically repressed in bladder cancer; these chromosomal regions are concomitantly silenced in a subgroup of aggressive tumors, defining a multiple regional epigenetic phenotype (Figure 1a).15 Here, we studied more extensively the chromatin landscape and conformation at these clusters to get insight into the regulation of such epigenetic alterations. We first used chromatin immunoprecipitation (ChIP) coupled with quantitative PCR (qPCR) to analyze the levels of H3K27me3 and global H3 acetylation (H3ac), not only on promoters but also within and between the genes across the entire loci. We analyzed two of the seven clusters: 2q31 (HOXD8, HOXD4, HOXD3 and HOXD1) and 3p22 (VILL, PLCD1, DLEC1 and ACAA1) in normal human urothelial (NHU) cells in culture20 as well as in two bladder cancer cell lines, CL1207 and RT112. CL1207 is derived from a high-grade tumor with the MRES phenotype, RT112 is from a well-differentiated tumor without the MRES phenotype.15 In CL1207 cells, but not in RT112, genes belonging to both clusters of interest are epigenetically repressed relative to normal bladder cells.15 High levels of the polycomb-mediated histone modification H3K27me3 are not restricted to gene promoters in CL1207 cells (Figure 1b): all positions assayed in the clusters display high H3K27me3 levels in comparison with the ubiquitously expressed GAPDH in the same cells, and compared with NHU cells and RT112 cells. Conversely, H3 was hypoacetylated across the two gene clusters in CL1207 cells relative to levels in NHU and RT112 cells (Figure 1b) and compared with GAPDH. Therefore, the repressed clusters in MRES cancer cells correspond to the formation of relatively homogeneous chromatin domains characterized by high levels of H3K27me3 and low levels of H3ac. Histone acetylation has been shown to directly affect secondary chromatin structures: it can neutralize charges in the histone N-terminal tails and thereby nucleosome-DNA or nucleosome– nucleosome interactions.21,22 Moreover, the polycomb repressive complexes, especially PRC1, have also been shown to be involved in chromatin compaction in vitro23 and in vivo.8 Given the observed histone modification profiles in CL1207 cells, we examined whether there was a corresponding chromatin compaction at the two MRES loci in bladder cancer cells. We used fluorescence in situ hybridation (FISH) with fosmid probe pairs spanning the 2q31 and 3p22 clusters to assess the state of chromatin compaction of these loci in the MRES þ cancer cell line
CL1207 and in the MRES cancer cell line RT112 as well as in NHU cells. As a control, we chose a fosmid probe pair at a 1q21 locus that was rarely affected by DNA alterations or expression changes in our bladder cancer data set (data not shown). We measured the distance (d) between the hybridization signals from these probe pairs in nuclei of the different cell lines. At probe separation of o2 Mb, there is a linear relationship between the mean-square interphase distances between two loci (d2) and their separation in kilobases.24 Such analysis has been used to show that different parts of the human genome display various levels of chromatin compaction,25 and that the same genomic region can change its compaction state in differentiation, development4,26 or when the activity of chromatin modifiers is ablated.8 In the latter cases, this included analyses of the HoxD locus, the murine homolog of one of the repressed gene clusters (2q31) studied here in human cancer cells. At the 3p22 cluster (Figure 1c, left panel), the interphase separation of the probe signals (127 kb apart) in the nuclei of CL1207 cells (mean d2 ¼ 0.17 mm2) is significantly smaller than in RT112 (d2 ¼ 0.33 mm2, P ¼ 0.005) or NHU (d2 ¼ 0.53 mm2, Po10 4) cells. This indicates that this genomic region is in a more compact chromatin state in CL1207 than in RT112 and NHU cells, whereas the chromatin compaction does not differ significantly between RT112 and NHU cells (P ¼ 0.2). A similarly more compact chromatin state in CL1207 cells was seen for probes spanning 104 kb at the HOXD cluster at 2q31 (d2 ¼ 0.20 mm2) in comparison with RT112 cancer cells (d2 ¼ 0.27 mm2, P ¼ 0.01) or normal NHU cells (d2 ¼ 0.27 mm2, P ¼ 0.005) (Figure 1c, central panel). This compaction is not a general feature of the chromatin structure throughout the genome in CL1207 cells, as shown by the similar level of chromatin compaction between all three cell lines at the control 1q21 locus (Figure 1c, right panel). We also confirmed the link between clustered gene repression and chromatin compaction in two other bladder cancer cell lines for 2q31 (Supplementary Figure 1). In the MRES þ TCCSUP cancer cell line, as in CL1207, genes of the 2q31 locus are repressed, whereas they are not in the MRES- MGHU3 cancer cell line.15 Compared with NHU cells (d2 ¼ 0.27 mm2) the chromatin at 2q31 is more condensed in TCCSUP (d2 ¼ 0.14 mm2, P ¼ 0.01), but not in MGHU3 (d2 ¼ 0.32 mm2, P ¼ 0.93). EZH2 is overexpressed in tumors with multiple regional epigenetic alterations As we found that the regions of epigenetic alterations are covered with abnormal H3K27me3 in cancer cells (Figure 1), we chose to investigate the role of histone methyltransferases capable of catalyzing the addition of a trimethyl group on H3K27: EZH127,28 and EZH2.29 Although a potential role of EZH1 in cancer has not yet been addressed, the deregulated expression and activity of EZH2 has been described. EZH2 is highly expressed in various tumor types,30 including breast,31 prostate,32 bladder33 or liver
Figure 1. Repressive chromatin domains form in cancer cells. (a) Location of the seven recurrent regional epigenetic alterations characteristic of the MRES phenotype. (b) Study of the 3p22 and 2q31 clusters: ChIP assays were performed for all promoters, for one position in the middle of every gene and between every pair of neighboring genes with an antibody against H3K27me3 (upper panel) and H3ac (lower panel). The graph shows the amount of immunoprecipitated target DNA expressed as a ratio of total input DNA measured by qPCR. Error bars indicate the variation between the means of two independent experiments. The map (not to scale) under each graph indicates the position of the primer pairs used for each point on the graph. The amount of immunoprecipitated DNA for the GAPDH promoter is shown as a reference. Data from the MRES bladder cancer cell line CL1207 is plotted in green; the non-MRES bladder cancer cell line RT112 in red; and normal human urothelial (NHU) cells in blue. (c) Two-dimensional FISH with a probe pair spanning the 3p22 cluster (left panel), the 2q31 cluster (middle panel) and a control locus in 1q21 (right panel). Above images, the map (to scale) shows the probe pair position (in bp) in the UCSC human genome browser (March 2006 Assembly, hg18). For the images, scale bar ¼ 5 mm. Below images, box plots showing the distribution of interprobe distances2 (d2) of each set of data, obtained from 50 nuclei. The horizontal line in the middle box represents the median, the two hinges of the middle box represent the interquartile range of the data and the empty circles indicate outliers. The whiskers extend to the most extreme data point that is no more than 1.5 times the interquartile range from the box. The statistical significance of differences was examined by Mann–Whitney U-tests (Materials and methods). Oncogene (2014) 1 – 11
& 2014 Macmillan Publishers Limited
PRC2-independent silencing in cancer C Vallot et al
3 cancer.34 Moreover, gain of function mutations of EZH2 have been reported in lymphomas.35,36 To understand if deregelulation of EZH2 expression might be involved in the appearance of the MRES phenotype, we compared EZH2 expression levels using Affymetrix arrays in a wide tumor set (n ¼ 150) between invasive tumors with (n ¼ 74) and without (n ¼ 29) MRES phenotype and normal urothelium samples (n ¼ 4). Tumors with the MRES phenotype were identified using Affymetrix expression data and cluster
analysis.15 As the MRES phenotype is found in invasive bladder tumors, we limited our analysis to invasive tumors, so that differences in EZH2 expression levels between MRES þ and MRES tumors would not only be due to the heterogeneity of tumor stages between each group. EZH2 is significantly more highly expressed in invasive tumors (T2–T4) with MRES phenotype than in invasive tumors without the MRES phenotype (P ¼ 0.018, t-test) and in normal samples (Figure 2a). This was validated by
* * 1
2
3
4
5
6
7
8
9
10
12
11
13
14
15
16
17
18
21
22
Anti-H3K27me3
0.5
0.6
0.4
0.5
0.3
0.4 0.3
0.2
0.2 0.1
0.1
0
PLCD1 DLEC1 ACAA1
NHU
HOXD8 HOXD4 HOXD3 HOXD1
RT112 CL1207
Anti-H3ac
GAPDH
0
GAPDH
VILL
Anti -H3ac
0.3 Ratio Bound over Input +/- SD
20
0.4 0.3
0.2
0.2 0.1 0.1 0 GAPDH
VILL
0
PLCD1 DLEC1 ACAA1
GAPDH
Ratio Bound over Input +/- SD
Anti-H3K27me3
19
HOXD8 HOXD4 HOXD3 HOXD1
3p22 cluster
2q31 cluster
Control locus 1q21
G248P80945A9 - G248P82821D6 (127 kb)
G248P89002G8 - G248P82958B3 (104 kb)
G248P80671B3 - G248P87792E10 (221 kb)
38000000
38100000
38200000
176700000
176800000
149900000
150000000
150100000
CL1207 CL1207 EVX2 PLCD1
HOXD9
HOXD13
CTDSPL
ACAA1
DLEC1
HOXD12
OXSR1 HOXD11
VILL
Distance2 (μm2)
NHU
MYD88
RT112
3
CL1207
p
