Mecp2 Knockdown Reveals DNA Methylation-Independent Gene Repression of Target Genes in Living Cells and a Bias in the Cellular Location of Target Gene Products

Genes Genet. Syst. (2008) 83, p. 199–208 MeCP2 knockdown reveals DNA methylation-independent gene repression of target genes in living cells and a bias in the cellular location of target gene products

Shinya Yakabe1,2, Hidenobu Soejima1*, Hitomi Yatsuki1, Hirotaka Tominaga3, Wei Zhao4, Ken Higashimoto1, Keiichiro Joh1, Shinichi Kudo5, Kohji Miyazaki2 and Tsunehiro Mukai6 1Division of Molecular Genetics and Epigenetics, Department of Biomolecular Sciences, Faculty of Medicine, Saga University, 5-1-1 Nabeshima, Saga 849-8501, Japan 2Division of General Surgery, Department of Surgery, Faculty of Medicine, Saga University, 5-1-1 Nabeshima, Saga 849-8501, Japan 3Section of Clinical Cooperation System, Center for Comprehensive Community Medicine, Faculty of Medicine, Saga University, 5-1-1 Nabeshima, Saga 849-8501, Japan 4Department of Cardiovascular Medicine, Shanghai Shuguang Hospital Affiliated with Shanghai University of T.C.M., No. 185 Pu An Road, Shanghai 200021, R. P. China 5Hokkaido Institute of Public Health, Sapporo 060-0819, Japan 6Saga University, 1 Honjo, Saga, Japan

(Received 9 February 2008, accepted 21 February 2008)

MeCP2, a methyl-CpG binding domain (MBD) protein, is known to bind to methylated CpG sites via a conserved MBD, leading to transcriptional repression. However, studies in cell-free system for gene repression and MeCP2 binding have suggested that DNA methylation-independent repression also occurs in living cells. It has been difficult to characterize the target genes of MeCP2 because a limited number have been identified to date. In this context, we screened for MeCP2 target genes using knockdown (KD) experiments combined with microarray gene expression analyses. Of the 49 genes that showed a more than three- fold increase in expression in two independent KD experiments conducted with different siRNA sets, unexpectedly, half (24 genes) did not contain promoter CpG islands (CGIs). Of seven selected genes that did contain CGIs, only two were methylated at the CGI, bound MeCP2 before KD, and reduced MeCP2 after KD. For three, MeCP2 was observed to bind to the unmethylated CGI before KD, and for one MeCP2 was reduced after KD. Another two genes neither had DNA methylation nor bound MeCP2 before KD. Gene ontology analysis suggested that MeCP2 represses a certain group of genes. These results suggest that in addition to the canonical gene repression function, MeCP2 can repress gene expression by binding to unmethylated DNA in particular genes in living cells.

Key words: CpG island, DNA methylation, gene ontology, gene repression, MeCP2

2003). In human cancers, a number of genes are known INTRODUCTION to be repressed by epigenetic means, for example aberrant Epigenetic regulation of gene expression plays a critical hypermethylation of the promoter CpG island (CGI). role in development and differentiation, X inactivation, One way in which DNA hypermethylation represses gene and genomic imprinting; it also plays a role in several expression may be inhibition of the binding of transcrip- human diseases, including cancer (Jaenisch and Bird, tion factors to their target sequences. However, it is also widely accepted that methylated CpGs are recognized by Edited by Hiroshi Nojima the methyl-CpG binding proteins, including MeCP2, * Corresponding author. E-mail: [email protected] MBD1, MBD2, MBD3, MBD4 and Kaiso, and that these 200 S. YAKABE et al. proteins recruit protein complexes related to histone mod- (Invitrogen) according to the manufacturer’s protocol. ification and chromatin remodeling (Jaenisch and Bird, The dsRNA sets were as follows: Set A, synthesized 2003; Bienvenu and Chelly, 2006). Recent studies have dsRNA with sequences 5’-CCUAAUGAUUUUGACUU- shown that of these, only MeCP2, MBD1 and MBD2 act CACGGUACAG-3’ and 5’-UUACCGUGAAGUCAAAAU- as methyl-CpG binding proteins in mammals (Lopez- CAUUAGGAU-3’ (Hokkaido System Science, Sapporo, Serra et al., 2006). Japan); Set B, dsRNAs (Stealth Select RNAi) of three dif- MeCP2 is a member of the methyl-CpG binding domain ferent sequences (Invitrogen, Catalog #1299003). A single (MBD) family and plays a pivotal role in DNA methyla- nucleotide at the 5’ terminal of the Set A antisense tion-associated gene repression. MECP2 is a causative sequence was intentionally mismatched to improve the gene for Rett syndrome, a dominant X-linked neurodevel- knockdown effect of the siRNA (Schwarz et al., 2003). At opmental disorder in which affected individuals are usu- 72 hours after transfection, cells were harvested for use ally heterozygous for a de novo mutation in MECP2 in further experiments. (Bienvenu and Chelly, 2006). The canonical gene repression function of MeCP2 involves the molecule binding to RNA extraction and reverse transcription PCR methylated CpG sites via a conserved MBD, leading to Total RNA was extracted using an RNeasy mini kit transcriptional repression, which occurs due to recruit- (Qiagen, Hilden, Germany) with an RNase-free DNase kit ment of Sin3A and histone deacetylases (HDACs) and/or (Qiagen). Total RNA (500 ng) was reverse-transcribed mediation of the methylation of histone H3 lysine 9 with random primers using ReverTra Ace reverse tran- (Bienvenu and Chelly, 2006; Fuks et al., 2003a; Fuks et scriptase (Toyobo, Osaka, Japan). Gene expression was al., 2003b). In cell-free system, MeCP2 can bind to unm- quantitated by real-time PCR on an ABI Prism 7000 with ethylated nucleosomal arrays and repress transcription TaqMan probe (Applied Biosystems, Foster City, CA, from both methylated and unmethylated naked DNA USA) and QuantiTect SYBR Green PCR kits (Qiagen) for (Meehan et al., 1992; Nan et al., 1997; Kaludov and MBDs and other genes, respectively, as shown in Table Wolffe, 2000; Georgel et al., 2003). This naturally sug- 1. The expression level of each gene was normalized gests that MeCP2 may be able to repress gene expression against that of the housekeeping genes GAPDH or β- without DNA methylation in living cells; however, this actin. All quantitative RT-PCRs were performed in trip- hypothesis has not been tested. Given that mutations in licate. MECP2 cause Rett syndrome, attempts have been made to identify the target genes of MeCP2 in the brain, with Western blotting Proteins were extracted from LU65 the result that a limited number of target genes have cells with and without MECP2 KD using SDS lysis buffer been identified (Bienvenu and Chelly, 2006). However, (2% SDS, 50 mM Tris-HCl, pH 7.5). Aliquots (5 μg) of no attempt has been made to perform a genome-wide protein were loaded onto 10% SDS-PAGE gels and elec- screen for MeCP2 target genes in cancer cells, and it trophoresed, then blotted onto PVDF membranes using a remains unknown whether MeCP2 regulates gene expres- semi-dry blotting method. The membranes were probed sion without DNA methylation in living cells. Thus, in with antibodies against MeCP2 (Kudo, 1998), MBD1 this study we screened potential MeCP2 target genes (Abcam, Cambridge, UK, cat. 3753 or 2846), MBD2/3 using knockdown (KD) with siRNA and microarray gene (Millipore, Billerica, MA, USA, cat. 07-199), and β-actin expression analyses, and investigated DNA methylation (Sigma, cat. A5441). An ECL plus western blotting and the binding of MeCP2 to CGIs within promoter regions. We also performed gene ontology (GO) analysis Table 1. Primers for quantitative RT-PCRs for the identified MeCP2 target genes in an effort to Gene name Catalogue number Detection kit determine if they have any features in common. MeCP2 Applied Biosystems, Hs00172845_m1 TaqMan MBD1 Applied Biosystems, Hs00242770_m1 TaqMan MATERIALS AND METHODS MBD2 Applied Biosystems, Hs00187506_m1 TaqMan Cell Lines LU65 cells derived from human lung cancer GAPDH Applied Biosystems, Hs99999905_m1 TaqMan and BEAS-2B cells from human bronchial epithelium DNER Qiagen, QT00045955 SYBR GREEN were cultured in RPMI-1640 medium (Sigma, St. Louis, HAS3 Qiagen, QT00014903 SYBR GREEN MO, USA) supplemented with 10% FCS (Gibco, Invitrogen, IL6R Qiagen, QT00023660 SYBR GREEN Carlsbad, CA, USA) and in MEM-α medium (Gibco) sup- CYP1A1 Qiagen, QT00012341 SYBR GREEN plemented with 10% FCS, respectively. Cells were har- ST3GAL5 Qiagen, QT00054236 SYBR GREEN vested at 70% confluence. FOXA1 Qiagen, QT00212828 SYBR GREEN RNAi experiments We transfected LU65 cells with IFIH1 Qiagen, QT00033789 SYBR GREEN two different dsRNA sets using Lipofectamine 2000 β-actin Qiagen, QT00095431 SYBR GREEN Characterization of MeCP2 target genes via knockdown 201 detection system (GE Healthcare, Buckinghamshire, UK) DNA methylation analyses Genomic DNA extracted was used for detection. from LU65 cells with and without MECP2 KD was sub- jected to sodium bisulfite modification with an EpiTect Microarray expression analysis Total RNA (500 ng) bisulfite kit (Qiagen). Modified DNA was amplified by from LU65 cells with and without MECP2 KD was ampli- PCR with a primer set specific to each gene, followed by fied using a Low RNA Fluorescent Linear Amplification cloning and sequencing. DNA methylation status in Kit (Agilent Technologies, Santa Clara, CA, USA) and LU65 and BEAS-2B cells was also analyzed by combined labeled with Cy-5 and Cy-3, respectively. Labeled cRNAs bisulfite restriction analysis (COBRA). All primers used were co-hybridized with the Whole Human Genome Oligo in this study are shown in Table 2. Microarray (Agilent Technologies), which includes 41,000 human genes and transcripts. Data were extracted Chromatin immunoprecipitation (ChIP) Chro- from the resulting images using Agilent’s Feature matin immunoprecipitation for MeCP2 was performed Extraction Software (Agilent Technologies). Microarray as previously described (Nakagawachi et al., 2003; data were deposited in a public database, CIBEX (http:// Higashimoto et al., 2003). Because the number of cells cibex.nig.ac.jp/) (accession no. CBX52). obtained from the siRNA transfection experiment was

Table 2. Primers and PCR conditions for methylation analyses and ChIP assay

Bisulfite sequencing Annealing temp Restriction enzymes 5’-primer-3’ and COBRA (PCR cycle) for COBRA DNER TTGGTTATAGTTGTATTTGTT 51, (40) AccII TATATCAAACRTACCCATTA HAS3 TTAGGAAGAGTAGGGTTTGGTT 58, (40) AccII AACCAACCCAACAACCTAAA IL6R ATTAGGTTATTGTTTATTGGTGGT 58, (40) AccII, HinfI RTCTCTAAACCRTCCCAAAATA CYP1A1 AGTAATTTYGGGAAGGAGGTTATT 58, (40) AccII, RsaI, TaqI CCCAAAAACAATACAAAACCCAA ST3GAL5 TATYGTATYGTAGGGTTGATTGGT 58, (40) MluI, TaqI TCTCRCRCCCATTCAACTAA FOXA1 GGYGGTYGTTTTGTAGGTGTATTT 56, (40) AciI CCTCCRCRAAAAATAAACRAACTA IFIH1 TGTTTGATAGAAAGTTGGATTT 53, (40) AciI, HhaI, MboI AAATTATCCACAAAACTCTCAAA

ChIP-PCR DNER CAACTGCCAGGTGAGTGGAA 60 (33) CAAAGTGGGAAAGCCTGTGTC HAS3 TCCCTTGGGTTTTCCGTTCCT 66 (33) GCACTCGCCACGCTTCTCATC IL6R GGTGGCTGATACGCCCTTTTCTC 66 (33) CCAGGGTGCGGTTACGAAAGC CYP1A1 CGCCTGCCGAGTCCTGGTAG 62 (33) TCCCAGGACAGCCCGAAGAGA ST3GAL5 CCACCGCACCGCAGGGCTGAC 68 (33) CCGCCGCTCCCCCGCTCAGAT FOXA1 CCTCCCCATTTCCCTCTTTCC 63 (33) AGACGCTCGCACCTACAAAGC IFIH1 CACACCGCAGAGTCCAAAGTG 66 (33) CTACCGCTCTGTGCCTGACAA Y: C or T, R: A or G. 202 S. YAKABE et al. small, we independently repeated the ChIP assay five times and defined MeCP2 binding as being positive when at least three positive bands were obtained by ChIP- PCR. We defined MeCP2 binding as being negative when ChIP-PCR revealed that at least three bands were absent relative to the input. To quantitatively evaluate the ChIP-PCR products, we used a γ-32P-labeled forward primer and measured the band intensities of the PCR products with a BAS 2000 Bioimaging Analyzer (Fujifilm, Tokyo, Japan).

Gene ontology Gene ontology (GO) analysis was performed using the curated gene annotation database of World Fusion Co. Ltd (Tokyo, Japan). This database contains genes with GO terms that are based on data in the Gene Ontology Database (http://www.geneontology. org/index.shtml) and NCBI database (http://www. Fig. 1. MECP2 knockdown. (a) Quantitative RT-PCR of MECP2. ncbi.nlm.nih.gov/), in particular Entrez Gene (http:// Both siRNA sets effectively reduced the mRNA expression level www.ncbi.nlm.nih.gov/sites/entrez?db=gene). As of May of MECP2. Expression of MECP2 was normalized to that of the 2007, 17,472 human genes were annotated with GO infor- housekeeping gene GAPDH. (b) Western blots of the gene prod- mation. ucts of members of the MBD family. Western blots showed a reduction in MeCP2 but no change in the protein expression of other MBD proteins. Protein expression was normalized to Statistical analysis Differences in GO or GO terms that of β-actin. (c) Quantitative RT-PCR of MBD1 and MBD2. between the genes up-regulated by MECP2 KD and all Quantitative RT-PCR showed that MECP2 KD did not reduce genes in the human genome were assessed using the chi- the mRNA expression of MBD1 or MBD2. mRNA expression of squared test or Fisher’s exact test. Two-sided probability MBDs was normalized to that of the housekeeping gene values of < 0.01 were considered statistically significant. GAPDH. The western blotting and quantitative RT-PCR data indicate the specificity of MECP2 KD. (d) Quantitative RT- PCR of the seven selected genes. The expression of all genes RESULTS was increased by MECP2 KD. Expression of the genes was normalized to that of β-actin. Dark gray bar: control; light gray MECP2 siRNA KD Quantitative RT-PCR following bar: MECP2 KD. MECP2 KD in LU65 cells using two siRNA sets revealed approximately 80% reductions in MECP2 expression using both Set A and Set B, relative to control LU65 cells transcripts, including 49 annotated genes, as shown in without siRNA (Fig. 1a). Western blotting also revealed Table 3, were identified by both siRNA sets. Since a CGI approximately 70% reductions in MeCP2 protein expres- covering the transcription initiation site (promoter CGI) sion relative to control LU65 cells using Set A (Fig. 1b). is usually important for gene repression due to DNA Furthermore, immunostaining revealed that KD had con- methylation (Ushijima, 2005), we searched for the loca- siderably reduced MeCP2 expression in the nuclei (data tions of CGIs in each gene using CpG Island Searcher not shown). Investigation of the mRNA and protein (http://www.cpgislands.com/), applying the criteria of expression levels of MBD1, MBD2, and MBD3 by quanti- Takai and Jones (Takai and Jones, 2002). Unexpectedly, tative RT-PCR and western blotting, respectively, showed we found that only half (23 genes) of the 49 genes no reduction in the expression of any of these MBDs, indi- possessed promoter CGIs. Two other genes had CGIs cating that our KD process was specific for MECP2, and approximately 200 bp upstream, but not covering, the did not affect other members of the MBD family (Fig. 1b transcription initiation site. The other 24 genes did not and 1c). have a CGI in the region from –5 kb to +1 kb relative to the transcription initiation site. None of the 49 up- Microarray gene expression analyses With the aim regulated genes were typical tumor suppressor genes, so of screening for genes repressed by MeCP2, two indepen- we selected 6 genes with promoter CGIs and one with a dent microarray gene expression analyses using RNAs CGI just upstream of the transcription initiation site for from two independent MECP2 KD experiments revealed further analysis. These genes are associated with cell a more than three-fold increase in the expression of 219 growth, differentiation, or transformation. Up-regulation and 190 transcripts for siRNA Set A and Set B, respec- of the seven genes was also confirmed by quantitative RT- tively. To reduce false-positives, we selected transcripts PCR (Fig. 1d). overlapped between the two experiments. A total of 59 Characterization of MeCP2 target genes via knockdown 203

Table 3. List of annotated genes with increased expression after MeCP2 knockdown

Gene GenBank siRNA siRNA Gene Name CGI Symbol accession no. Set A Set B AHNAK2 AL512802 AHNAK nucleoprotein 2 (C14orf78; KIAA2019) 3.84 3.19 promoter AQP3 BC013566 aquaporin 3 3.09 3.41 promoter CPM AF368463 carboxypeptidase M 4.88 7.28 promoter CPNE5 AB046819 copine V 3.97 3.88 promoter DNER BC024766 delta-notch-like EGF repeat-containing transmembrane 5.72 4.34 promoter DYSF AF075575 dysferlin, limb girdle muscular dystrophy 2B 5.24 4.60 promoter FBP1 BC012927 fructose-1,6-bisphosphatase 1 3.13 3.38 promoter FGD6 AC011598 FYVE, RhoGEF and PH domain containing 6 3.20 3.09 promoter FOXA1 BC033890 forkhead box A1 3.81 7.66 promoter GDF15 AB000584 growth differentiation factor 15 6.41 3.80 promoter GGTLA4 AA931238 gamma-glutamyltransferase-like activity 4 6.97 3.88 promoter HAS3 AF232772 hyaluronan synthase 3 3.08 3.01 promoter IFIH1 AF095844 interferon induced with helicase C domain 1 3.00 3.32 promoter IL6R BM911928 interleukin 6 receptor 5.83 4.00 promoter ITGA2 AC008966 integrin, alpha 2 4.03 4.64 promoter LOXL2 AC090197 lysyl oxidase-like 2 4.08 3.18 promoter NTSR1 AL035669 neurotensin receptor 1 8.77 3.30 promoter PLXNA2 AB007932 plexin A2 7.47 4.70 promoter RPS6KA2 AA588877 ribosomal protein S6 kinase, 90kDa, polypeptide 2 4.92 3.90 promoter SEC24D AB018298 SEC24 related gene family, member D (S. cerevisiae) 3.34 3.31 promoter SOX7 AK055556 SRY (sex determining region Y)-box 7 3.46 4.21 promoter ST3GAL5 AY152815 ST3 beta-galactoside alpha-2,3-sialyltransferase 5 4.16 4.92 promoter SULF2 AB033073 sulfatase 2 4.93 4.28 promoter CYP1A1 X02612 cytochrome P450, family 1, subfamily A, polypeptide 1 3.15 3.74 up FKBP10 AB045981 FK506 binding protein 10, 65 kDa 4.86 3.81 up AKR1C1 BC040210 aldo-keto reductase family 1, member C1 8.54 8.42 none AKR1C3 AV761799 aldo-keto reductase family 1, member C3 8.29 6.55 none ASB2 AB056723 ankyrin repeat and SOCS box-containing 2 6.62 3.31 none BATF AF016898 basic leucine zipper transcription factor, ATF-like 5.36 3.52 none C1QTNF6 BC020551 C1q and tumor necrosis factor related protein 6 4.24 5.10 none CDA BC048284 cytidine deaminase 4.54 3.04 none CEACAM6 M18216 carcinoembryonic antigen-related cell adhesion olecule 6 10.71 9.00 none COL17A1 AL138761 collagen, type XVII, alpha 1 3.51 3.84 none DEFB1 BC047677 defensin, beta 1 3.73 9.01 none EVI2B AA251185 ecotropic viral integration site 2B 4.66 3.73 none GPR110 AK026337 G protein-coupled receptor 110 4.83 3.60 none IFI44L AB000115 interferon-induced protein 44-like 4.14 12.29 none IL1B M15330 interleukin 1, beta 5.64 3.56 none IL32 BC009401 interleukin 32 3.08 3.36 none MMP1 AJ002550 matrix metalloproteinase 1 22.24 6.14 none MMP7 BC003635 matrix metalloproteinase 7 4.15 3.82 none MMP9 AW468235 matrix metalloproteinase 9 9.42 3.58 none MUC5AC XP_001134429 mucin 5AC 3.40 5.47 none NXF3 AJ277527 nuclear RNA export factor 3 7.00 4.28 none OVGP1 AA909093 oviductal glycoprotein 1, 120kDa (mucin 9, oviductin) 4.24 3.07 none PECAM1 BC022512 platelet/endothelial cell adhesion molecule (CD31 antigen) 4.33 5.20 none PPP1R1C AC009962 protein phosphatase 1, regulatory (inhibitor) subunit 1C 4.61 5.26 none SRPX2 AF060567 sushi-repeat-containing protein, X-linked 2 3.84 4.20 none VNN1 AJ132099 vanin 1 4.91 3.04 none Genes selected for further investigation are indicated by bolded font. Fold increase expression for each microarray analysis following KD is given for both siRNA sets. CGI: promoter, CGI covering the transcription initiation site; up, CGI approximately 200 bp upstream, but not covering, transcription initiation site; none, no CGI around transcription initiation site. 204 S. YAKABE et al.

Fig. 2. Bisulfite sequences of the seven selected genes. Unexpectedly, only two, DNER and HAS3, were methylated on the promoter CGI. The other five genes were mostly unmethylated. Methylation status was not changed by MECP2 KD, indicating that MeCP2 has no effect on DNA methylation. Black circles: methylated CpGs; white circles: unmethylated CpGs. Arrows in the upper and lower rows indicate the regions that were amplified for ChIP-PCR (ChIP) and bisulfite-PCR (BS) for sequencing and COBRA, respectively. Gray bars indicate CpG islands (CGI). Broken arrows indicate transcription initiation sites.

Fig. 3. Results of COBRA. In LU65 cancer cells, the promoter CGIs of DNER and HAS3 were methylated irrespective of MECP2 KD, which was consistent with the results of bisulfite sequencing. However, in BEAS-2B bronchial epithelium cells, methylation was not observed, suggesting cancer-specific methylation of DNER and HAS3. The CGI of IL6R was unmethylated in both LU65 and BEAS-2B cells. All other genes were unmethylated in both BEAS-2B and LU65 cells. me: control methylated DNA. Characterization of MeCP2 target genes via knockdown 205

DNA Methylation analysis Since a core region within a promoter CGI is usually important for gene repression due to DNA methylation (Ushijima, 2005), we performed bisulfite sequencing of regions surrounding the initiation site for the seven selected genes (Fig. 2). Our primer sets for bisulfite sequencing mostly covered the entire CGI region, including the core region. Unexpectedly, the CGIs of five of the analyzed genes were mostly unmethylated before MECP2 KD. Only two genes (DNER and HAS3) harbored methylated CGIs: DNER was densely hypermethylated, whereas HAS3 was approximately half methylated. Further, MECP2 KD did not change the methylation status of any CGI, and these results were confirmed by COBRA (Fig. 3). Our findings indicate that MeCP2 has no effect on DNA methylation. Further examination of the methylation status of all CGIs in BEAS-2B bronchial epithelium cells using COBRA showed that no CGIs were methylated, suggesting that DNER and HAS3 were specifically methylated in LU65 cancer cells (Fig. 3).

MeCP2 binding Using five independent ChIP-PCR assays, we determined whether MeCP2 bound to the CGIs of the seven selected genes, as described in the Materials and Methods. Before KD, we found that five (DNER, HAS3, IL6R, CYP1A, and ST3GAL5) of the seven genes were bound by MeCP2, while two (FOXA1 and IFIH1) were not (Fig. 4). For the five CGIs that bound to MeCP2, after MECP2 KD, three genes (DNER, HAS3, and IL6R) had reduced MeCP2 binding, while contradictory or inconsistent results were obtained for two (CYP1A and ST3GAL5), over five independent experiments. We Fig. 4. Representative results of ChIP-PCR analysis. DNER, HAS3, IL6R, CYP1A1, and ST3GAL5 were bound by MeCP2 speculated that specific chromatin conformation of the before KD, and MeCP2 binding was reduced for DNER, HAS3, two genes might influence ChIP procedure, leading to and IL6R after KD. Contradictory or inconsistent results were such contradictory or inconsistent results. In summary, noted for CYP1A1 and ST3GAL5 in different ChIP assays after MeCP2 constitutively bound to the CGIs of five of the KD. Since FOXA1 and IFIH1 were not bound by MeCP2 before seven selected genes before KD, three of which showed KD, we did not perform ChIP assays for these genes after KD. Five independent ChIP assays were performed. We defined reduced MeCP2 binding after KD. Two of the selected MeCP2 binding as positive when at least three positive bands genes were not bound by MeCP2. were obtained by ChIP-PCR with a γ-32P-labeled forward primer. Thus, among the seven selected genes, only for two MeCP2 binding was defined as negative when at least three genes, DNER and HAS3, were DNA methylation and con- bands were absent relative to the input. The relative intensity stitutive MeCP2 binding before and reduced MeCP2 bind- of MeCP2 to input is shown to the right of each gel. ing after KD all observed (Table 4). Although the CGIs of IL6R, CYP1A1, and ST3GAL5 were not methylated, that used to describe the biology of a gene product. There MeCP2 constitutively bound to the CGIs of these genes are three independent sets of vocabularies, or ontologies, before KD. After KD, however, only IL6R underwent an which describe the molecular function of a gene product, obvious reduction in MeCP2 binding. The CGIs of the the biological process in which the gene product partici- two remaining genes (FOXA1, and IFIH1) were also unm- pates, and the cellular component where the gene product ethylated, and were not bound by MeCP2 at all. can be found. GO data comprises “terms”, each of which is assigned to one of the three ontologies. Thus, a gene Gene ontology analysis of the genes up-regulated product might be associated with not only one or more by MECP2 KD To characterize the genes up-regulated ontologies but also several terms. by MECP2 KD (putative MeCP2 target genes), we Ontologies of biological process, cellular component, searched for gene ontology (GO) data relating to these and molecular function were associated with 43, 41, and genes. The GO Project provides a controlled vocabulary 42 of the genes upregulated by MeCP2, respectively 206 S. YAKABE et al.

Table 4. DNA methylation and MeCP2 binding of 7 selected MeCP2 preferentially represses a certain group of genes, genes the products of which exist in the extracellular space. MeCP2 binding Other terms, including “integral to plasma membrane”, Gene DNA methylation Before KD After KD “cell adhesion”, “endoplasmic reticulum”, and “nucleus”, which were related to the cellular component, and “cal- DNER + + reduced cium ion binding” and “extracellular region”, which were HAS3 + + reduced related to the molecular function, showed a tendency to IL6R – + reduced be higher for the up-regulated genes than for all human CYP1A1 – + ± genes, with p values of less than 0.05. ST3GAL5 – + ± FOXA1 – – n.t. DISCUSSION IFIH1 – – n.t. In this study, we found that half of the 49 genes up-reg- ± : inconclusive ; n.d.: not tested. ulated by MECP2 KD did not have a promoter CGI (Table 3), which is believed to be important for epigenetic gene (Table 5). The frequencies of each GO for the group of 49 repression. Furthermore, the promoter CGIs in some up-regulated genes did not differ significantly from the genes were not always methylated or bound by MeCP2 frequencies for all human genes. A total of 223 terms (Table 4). A recent study using a human neuronal cell were obtained for the 49 genes. Only the 14 terms that line showed that only 5.9% of MeCP2-bound sites over- were associated with more than five genes are shown in lapped with CGIs and that only 6.0% of MeCP2-bound Table 5. The frequency of “extracellular space” as a cel- promoters were highly methylated, being consistent with lular component, including GDF15, IL1B, IL32, LOXL2, our results (Yasui et al., 2007). We also found that, in MMP1, MMP7, MMP9, OVGP1, and SULF2, was signifi- living cells, MeCP2 bound to the CGIs of three of the cantly higher for the up-regulated genes than for all seven genes that were up-regulated by MECP2 KD, and human genes (18.4% vs. 2.7%). This suggests that that these CGIs were mostly unmethylated. Expression of the genes downstream of these CGIs increased after MECP2 KD, in particular expression of IL6R, for which Table 5. Ontology of genes up-regulated by MeCP2 knockdown MeCP2 binding was completely lost after KD, and a 5.3- The up-regulated Whole human Gene ontology or term p value fold increase in expression was found (Figs. 1d and 4, genes (49) gene (17,472)* Table 4). This indicates that non-canonical, DNA GO methylation-independent gene repression is being effected Biological process 43 (87.8%) 14,143 (80.1%) 0.5581 (χ2) by MeCP2, which differs from the DNA methylation- Cellular component 41 (83.7%) 14,237 (81.5%) 0.8338 (χ2) dependent means of canonical MeCP2 gene repression. Molecular function 42 (85.7%) 15,607 (89.3%) 0.3030 (χ2) It has been reported that MeCP2 can bind to unmethy- Term lated nucleosomal arrays and condense them in cell-free membrane (CC) 13 (26.5%) 4,764 (27.3%) 0.9640 (χ2) system, depending on the region(s) of the protein other extracellular space (CC) 9 (18.4%) 465 (2.7%) < 0.00001 (F) than the MBD (Georgel et al., 2003). The affinity of protein binding (BP) 8 (16.3%) 4,564 (26.1%) 0.1626 (χ2) MeCP2 for methylated DNA is only up to 3.3-fold stronger zinc ion binding (MF) 8 (16.3%) 2,253 (12.9%) 0.6156 (χ2) than that for unmethylated DNA (Ballestar et al., 2000; Fraga et al., 2003). Furthermore, a recent study showed calcium ion binding (MF) 7 (14.3%) 892 ( 5.1%) 0.0120 (F) that 28% of unmethylated promoter CGIs were bound by integral to plasma 7 (14.3%) 1,049 ( 6.0%) 0.0266 (F) membrane (CC) MBD proteins in cancer cell lines, some of which were signal transduction (BP) 7 (14.3%) 1,729 ( 9.9%) 0.3316 (F) bound by MeCP2 (Lopez-Serra et al., 2006). These integral to membrane (CC) 6 (12.2%) 3,544 (20.3%) 0.2224 (χ2) reports support our finding that MeCP2 binds to unmeth-

cell adhesion (CC) 5 (10.2%) 509 ( 2.9%) 0.0141 (F) ylated promoter CGIs and represses gene expression in living cells, representing a methylation-independent form endoplasmic reticulum (CC) 5 (10.2%) 646 ( 3.7%) 0.0347 (F) of repression. Although the number of genes analyzed extracellular region (MF) 5 (10.2%) 681 ( 3.9%) 0.0419 (F) was not large, the data we gathered suggested that at nucleus (CC) 5 (10.2%) 4,133 (23.7%) 0.0408 (χ2) least some of the target genes of MeCP2 are regulated in χ2 metal ion binding (CC) 5 (10.2%) 2,187 (12.5%) 0.7852 ( ) this non-canonical manner. Whether this non-canonical receptor activity (CC) 5 (10.2%) 1,412 ( 8.1%) 0.5944 (F) gene repression is associated with histone modification, Only GO terms annotated with more than five genes among the such as deacetylation of H3/H4, methylation of H3K9, or upregulated 49 genes are shown. GO associated with each term are shown in parentheses. *: number of whole human genes with GO based other chromatin remodeling factors, remains to be clari- on data from the Gene Ontology webpage and NCBI in May 2007. BP: 2 fied. It is also unclear whether other MBD proteins biological process; CC: cellular component; MF: molecular function. χ : chi-square test; F: Fisher’s exact test. repress gene expression in a similar way. Interestingly, Characterization of MeCP2 target genes via knockdown 207 we found two genes (FOXA1 and IFIH1) in which the Promotion of Science, a Grant for Child Health and Development CGIs were not methylated and not bound by MeCP2, but from the National Center for Child Health and Development, and for which expression was elevated by MECP2 KD. by grants provided by the Takeda Science Foundation, the Nakayama Science Foundation, the Promotion of Science and FOXA1 had three other CGIs within –5 kb to +5 kb rela- Public Trust Surgery Research Fund, and the Ichiro Kanehara tive to the transcription initiation site, but IFIH1 did not Foundation. have any other CGI within the region. The other CGIs might be responsible for MeCP2 mediated repression of REFERENCES FOXA1. Or non-CGI region might be involved in repression of these genes. As another possibility, MeCP2 Baek, S. J., Kim, K. S., Nixon, J. B., Wilson, L. C., and Eling, T. might function via an indirect repression mechanism. E. (2001) Cyclooxygenase inhibitors regulate the expression Two of the selected genes that were up-regulated by of a TGF-beta superfamily member that has proapoptotic and antitumorigenic activities. Mol. Pharmacol. 59, 901– MECP2 knockdown, DNER and HAS3, were methylated 908. and bound by MeCP2 in LU65 cancer cells, indicating Ballestar, E., Yusufzai, T. M., and Wolffe, A. P. (2000) Effects of that these genes are repressed in the canonical DNA Rett syndrome mutations of the methyl-CpG binding methylation-dependent manner. Since these genes were domain of the transcriptional repressor MeCP2 on selectiv- specifically methylated in LU65 cells but not in BEAS-2B ity for association with methylated DNA. Biochemistry 39, 7100–7106. bronchial epithelium cells, it seems likely that aberrantly Bienvenu, T., and Chelly, J. (2006) Molecular genetics of Rett methylated CGIs might be preferentially targeted in the syndrome: when DNA methylation goes unrecognized. Nat. canonical manner by MeCP2. Because methylation was Rev. Genet. 7, 415–426. not altered by MECP2 KD, irrespective of gene expres- Dai, Y., Yang, Y., MacLeod, V., Yue, X., Rapraeger, A. C., sion, we conclude that CGI methylation alone is not Shriver, Z., Venkataraman, G., Sasisekharan, R., and Sanderson, R. D. (2005) HSulf-1 and HSulf-2 are potent sufficient to repress gene expression. Heterochromatini- inhibitors of myeloma tumor growth in vivo. J. Biol. Chem. zation due to the recruitment of MBD proteins and chro- 280, 40066–40073. matin modification factor(s) is needed for repression, as Fuks, F., Hurd, P. J., Deplus, R., and Kouzarides, T. (2003a) The shown in our previous study (Zhao et al., 2005). DNA methyltransferases associate with HP1 and the To date, there have been attempts to identify MeCP2 SUV39H1 histone methyltransferase. Nucleic Acids Res. 31, 2305–2312. target genes in mice and humans (Bienvenu and Chelly, Fuks, F., Hurd, P. J., Wolf, D., Nan, X., Bird, A. P., and 2006). The number of target genes identified that are Kouzarides, T. (2003b) The methyl-CpG-binding protein expressed in neurons or brain tissue, however, was too MeCP2 links DNA methylation to histone methylation. J. small to permit GO analysis. We performed GO analysis Biol. Chem. 278, 4035–4040. for the 49 putative MeCP2 target genes identified. This Fraga, M. F., Ballestar, E., Montoya, G., Taysavang, P., Wade, P. A., and Esteller, M. (2003) The affinity of different MBD is the first GO analysis of MeCP2 target genes. The proteins for a specific methylated locus depends on their findings suggest that MeCP2 represses specific groups of intrinsic binding properties. Nucleic Acids Res. 31, 1765– genes, those that express products that exist in the extra- 1774. cellular space, in human LU65 cells. Some of them have Georgel, P. T., Horowitz-Scherer, R. A., Adkins, N., Woodcock, C. been reported to possess anti-tumor functions; GDF15 L., Wade, P. A., and Hansen, J. C. (2003) Chromatin com- paction by human MeCP2. Assembly of novel secondary codes an antitumorigenic and proapoptotic protein (Baek chromatin structures in the absence of DNA methylation. et al., 2001; Lee et al., 2005; Jang et al., 2006). Over J. Biol. Chem. 278, 32181–32188. expression of IL1B occurs cell apoptosis in MCF7 cancer Higashimoto, K., Urano, T., Sugiura, K., Yatsuki, H., Joh, K., cell (Roy et al., 2006). LOXL2 was markedly reduced in Zhao, W., Iwakawa, M., Ohashi, H., Oshimura, M., Niikawa, malignantly transformed cell lines (Rost et al., 2003). N., Mukai, T., and Soejima, H. (2003) Loss of CpG methylation is strongly correlated with loss of histone H3 lysine 9 SULF2 is a potent inhibitor of myeloma tumor growth in methylation at DMR-LIT1 in patients with Beckwith-Wiede- vivo (Dai et al., 2005). The expression of certain MMPs mann syndrome. Am. J. Hum. Genet. 73, 948–956. including MMP9 provides a protective effect in cancer Jaenisch, R., and Bird, A. (2003) Epigenetic regulation of gene progression though MMPs are generally believed to stim- expression: how the genome integrates intrinsic and envi- ulate tumor growth, invasion, and metastasis (Martin ronmental signals. Nat. Genet. 33 Suppl., 245–254. Jang, T. J., Kim, N. I., and Lee, C. H. (2006) Proapoptotic activ- and Matrisian, 2007). Thus, the repression of these ity of NAG-1 is cell type specific and not related to COX-2 genes by MeCP2 could be beneficial to LU65 cancer cell. expression. Apoptosis 11, 1131–1138. More target genes must be identified from different cell Kaludov, N. K., and Wolffe, A. P. (2000) MeCP2 driven tran- types or tissues to allow a more precise characterization scriptional repression in vitro: selectivity for methylated of the MeCP2 target genes and to permit a better under- DNA, action at a distance and contacts with the basal transcription machinery. Nucleic Acids Res. 28, 1921–1928. standing of the biological roles of MeCP2. Kudo, S. (1998) Methyl-CpG-binding protein MeCP2 represses Sp1-activated transcription of the human leukosialin gene This study was supported in part by a Grant-in-Aid for Scien- when the promoter is methylated. Mol. Cell. Biol. 18, tific Research (C) (No. 18590313) from the Japan Society for the 5492–5499. 208 S. YAKABE et al.

Lee, S. H., Kim, J. S., Yamaguchi, K., Eling, T. E., and Baek, S. Werner, J. A. (2003) Reduction of LOX- and LOXL2-mRNA J. (2005) Indole-3-carbinol and 3,3’-diindolylmethane induce expression in head and neck squamous cell carcinomas. expression of NAG-1 in a p53-independent manner. Bio- Anticancer Res. 23, 1565–1573. chem. Biophys. Res. Commun. 328, 63–69. Roy, D., Sarkar, S., and Felty, Q. (2006) Levels of IL-1 beta con- Lopez-Serra, L., Ballestar, E., Fraga, M. F., Alaminos, M., trol stimulatory/inhibitory growth of cancer cells. Front. Setien, F., and Esteller, M. (2006) A profile of methyl-CpG Biosci. 11, 889–898. binding domain protein occupancy of hypermethylated pro- Schwarz, D. S., Hutvagner, G., Du, T., Xu, Z., Aronin, N., and moter CpG islands of tumor suppressor genes in human Zamore, P. D. (2003) Asymmetry in the assembly of the cancer. Cancer Res. 66, 8342–8346. RNAi enzyme complex. Cell 115, 199–208. Martin, M. D., and Matrisian, L. M. (2007) The other side of Takai, D., and Jones, P. A. (2002) Comprehensive analysis of MMPs: protective roles in tumor progression. Cancer CpG islands in human chromosomes 21 and 22. Proc. Natl. Metastasis Rev. 26, 717–724. Acad. Sci. USA. 99, 3740–3745. Meehan, R. R., Lewis, J. D., and Bird, A. P. (1992) Characteriza- Ushijima, T. (2005) Detection and interpretation of altered tion of MeCP2, a vertebrate DNA binding protein with affin- methylation patterns in cancer cells. Nat. Rev. Cancer 5, ity for methylated DNA. Nucleic Acids Res. 20, 5085–5092. 223–231. Nakagawachi, T., Soejima, H., Urano, T., Zhao, W., Higashimoto, Yasui, D. H., Peddada, S., Bieda, M. C., Vallero, R. O., Hogart, K., Satoh, Y., Matsukura, S., Kudo, S., Kitajima, Y., A., Nagarajan, R. P., Thatcher, K. N., Farnham, P. J., and Harada, H., Furukawa, K., Matsuzaki, H., Emi, M., Lasalle, J. M. (2007) Integrated epigenomic analyses of neu- Nakabeppu, Y., Miyazaki, K., Sekiguchi, M., and Mukai, T. ronal MeCP2 reveal a role for long-range interaction with (2003) Silencing effect of CpG island hypermethylation and active genes. Proc. Natl. Acad. Sci. USA. 104, 19416– histone modifications on O6-methylguanine-DNA methyl- 19421. transferase (MGMT) gene expression in human cancer. Zhao, W., Soejima, H., Higashimoto, K., Nakagawachi, T., Oncogene 22, 8835–8844. Urano, T., Kudo, S., Matsukura, S., Matsuo, S., Joh, K., and Nan, X., Campoy, F. J., and Bird, A. (1997) MeCP2 is a tran- Mukai, T. (2005) The essential role of histone H3 Lys9 di- scriptional repressor with abundant binding sites in methylation and MeCP2 binding in MGMT silencing with genomic chromatin. Cell 88, 471–481. poor DNA methylation of the promoter CpG island. J. Bio- Rost, T., Pyritz, V., Rathcke, I. O., Gorogh, T., Dunne, A. A., and chem. (Tokyo) 137, 431–440.