7) 178–188www.elsevier.com/locate/ygeno

Genomics 89 (200

An integrated map of p53-binding sites and histonemodification in the human ENCODE regions

Kiyofumi Kaneshiro a,1, Shuichi Tsutsumi a,1, Shingo Tsuji a,Katsuhiko Shirahige b, Hiroyuki Aburatani a,⁎

a Genome Science Division, Research Center for Advanced Science and Technology (RCAST), University of Tokyo, 4-6-1 Komaba, Meguro, Tokyo 153-8904, Japanb Division for Gene Research, Center for Biological Resources and Informatics (CBRI), Tokyo Institute of Technology, Kanagawa 226-8501, Japan

Received 5 May 2006; accepted 6 September 2006Available online 3 November 2006

Abstract

TP53 (tumor protein p53; p53) regulates its target genes under various cellular stresses. By combining chromatin immunoprecipitation witholigonucleotide microarrays, we have mapped binding sites of p53 (p53-BS) in the genome of HCT116 human colon carcinoma cells, along withthose of acetylated H3, acetylated H4, and methylated H3-K4. We analyzed a 30-Mb portion of the human genome selected as a representativemodel by the ENCODE Consortium. In the region, we found 37 p53-BS, of which the p53-binding motif was present in 32 (86%). Acetylatedhistone H3 and H4 were detected at 14 (38%) and 33 (89%) of the p53-BS, respectively. A significant portion (58%) of H4 acetylation in thep53-BS was not accompanied by H3 acetylation. Acetyl H3 were preferentially located at the 5′ and 3′ ends of genes, whereas acetyl H4 weredistributed widely across the genome. These results provide novel insights into how p53 binding coordinates with histone modification inhuman.© 2006 Elsevier Inc. All rights reserved.

Keywords: TP53 protein; Histone; Chromatin; Oligonucleotide microarrays; HCT116 cells; Human

Introduction

TP53 (tumor protein p53; Li-Fraumeni syndrome; p53) playsan important role in the cellular response to various stresses suchas DNA damage and hypoxia, and loss of p53 is associated withgenomic instability and tumor development. The activation ofp53 can lead to cell cycle arrest, apoptosis, or senescence,depending on the cellular context [1]. A major function of p53 isthat of a transcription factor to control the expression of its targetgenes, whose functions fall into categories that reflect the role ofp53 as an integrator of diverse cellular signals [2]. Differentialinduction of target genes by p53 is important, particularly indetermining whether a cell is destined for cell cycle arrest orapoptosis [3]. A number of approaches, including geneexpression profiling and genome-wide sequence analysis, haveidentified dozens of p53-regulated genes, including CDKN1A,

⁎ Corresponding author.E-mail address: haburata-tky@umin.ac.jp (H. Aburatani).

1 These two authors contributed equally to this work.

0888-7543/$ - see front matter © 2006 Elsevier Inc. All rights reserved.doi:10.1016/j.ygeno.2006.09.001

MDM2, BAX and BBC3 [4]. However, the mechanism under-lying p53-regulated transcriptional control has not been fullyelucidated [5].

Among many factors, the physical interaction of the p53protein with DNA is central to its role in transcriptionalregulation. Target genes of p53 have one or more p53-bindingsites (p53-BS) very similar to the p53 consensus DNA-bindingmotif that consists of two copies of the 10 basepair (bp)sequence 5′-PuPuPuC(A/T) (T/A)GPyPyPy-3′, separated by 0to 13 bp. However, this rule is not strict and only a few p53-BSare perfect matches for the p53 consensus DNA-binding motif[6,7]. Studies examining the in vivo binding of p53 bychromatin immunoprecipitation (ChIP) have demonstratedthat there are large differences in the kinetics and extent ofp53 binding to target gene promoters [8,9]. There are severalfactors that might contribute to differential binding activity ofp53 to target genes. Posttranslational modifications and co-activators of p53 may regulate its DNA-binding activity [10–13]. Also, individual sequence differences between p53-BS andthe number of p53-BS might contribute to the differential DNA-

Fig. 1. Transactivation of p53 and p53 binding at the promoter of p21. (A)Western blot with protein extracts from HCT116 cells treated with 5FU. Thep53-null cells were derived from parental HCT116 by gene targeting. (B)Chromatin immunoprecipitates from 5FU-treated and untreated HCT116 cellswere analyzed by Q-PCR at the promoter of p21.

179K. Kaneshiro et al. / Genomics 89 (2007) 178–188

binding activity of p53 [14,15]. Recently, a genomic map ofp53-BS along human chromosome 21 and 22 was reported [16].Interestingly, quite a large number of p53-BS were observed inintronic and intergenic regions as well as the promoters ofgenes. How, or even whether, these nongenic sequencescontribute to transcriptional regulation is unclear [17].

Histone tails are thought to be heterogeneously modifiedthroughout the genome and to specify a code to regulate theexpression of genes [18]. p53-dependent changes in histonemodification have been observed at the promoters of a number ofp53 target genes [19–26]. p53 interacts with components ofmultiple different histone remodeling complexes, including CBP/EP300 (CBP/p300), GCN5, PCAF, and SETD7, and may recruitthem to p53-binding sites on target genes [21,27]. These histone-modifying enzymes can regulate chromatin structure and functionin part by recruiting additional proteins [18]. Recently it has beenshown that the promoters of p53 target genes have distinctivepatterns of p53-dependent histone acetylation [28]. In thepromoter of CDKN1A, a distinctive behavior of H3 and H4acetylation has also been reported [29]. These observations raisethe possibility that p53 binding coordinates with histonemodification to regulate transcription. Still, many aspects ofhow p53 associates with histone modification in the humangenome remain obscure.

In this report, we describe a large-scale study of p53 binding,histone modification patterns, and gene expression analysis in5FU-treated HCT116 human cell lines using 30 Mb (1%) ofnonrepetitive human genome as target. For mapping studies, weused an unbiased approach combining ChIP, in vitro transcrip-tion (IVT) amplification, and high-density oligonucleotidetiling arrays (ChIP-chip). The tiling arrays used here cover 44discrete genomic regions selected by the ENCyclopedia Of DNAElements (ENCODE) Consortium as a representative collection ofgenomic features [30]. Thus, our study provides substantialinformation on how p53 associates with histone modification inhuman.

Results

ChIP for p53 and modified histones

Prior to the ChIP experiment, we tested the transactivation ofp53 in wildtype TP53-containing HCT116 colon carcinomacells and TP53-null derivatives. Protein extracts were preparedfrom cells treated with a DNA-damaging agent, 5-fluorouracil(5FU) for 0 to 24 h. The p53 protein level increased after 5FUtreatment for 9 h (Fig. 1A). This increase was accompanied bythe posttranslational modifications of p53 itself and induction ofCDKN1A and MDM2 protein expression, which were not seenin p53-null cells (Fig. 1A). The induction of CDKN1A andMDM2 mRNA expression was also seen in p53 wildtype cellstreated with 5FU for 9 h (data not shown).

Next, the ChIP assay was tested for detection of specificTP53 DNA binding at the promoter of CDKN1A. 5FU-treated(9 h) and untreated HCT116 cells were treated with formalde-hyde to form protein–DNA complexes. Chromatin purifiedfrom nuclei was sheared to an average size of 600 bp and

immunoprecipitated with anti-p53 antibody. The cross-linkswere reversed and the DNA was purified and analyzed for theenrichment of a known p53-binding region in the promoter ofCDKN1A, by using quantitative PCR. The amount ofCDKN1A associated p53-binding region increased fourfold in5FU-treated cells (Fig. 1B). Enrichment of the p53-bindingregion was not seen either in TP53-null cells or in negativecontrols in which normal mouse IgG was used for chromatinimmunoprecipitation (Fig. 1B).

To test the kinetics of histone acetylation at the promoter ofCDKN1A, we measured the amount of histone acetylation atthe p53-binding region by ChIP using antibodies recognizingmultiple acetylated forms of H3 or H4. In the absence of 5FU,the level of H4 acetylation was two- to threefold higher inTP53 wildtype cells than in TP53-null cells (Fig. 1B). After5FU treatment, there was no change in H4 acetylation inTP53-null cells, but a five- to sixfold increase in the TP53wildtype cells (Fig. 1B). H3 acetylation showed a patternsimilar to H4 acetylation (Fig. 1B). These results indicate thep53 dependency of histone acetylation at the promoter ofCDKN1A.

180 K. Kaneshiro et al. / Genomics 89 (2007) 178–188

Genomic maps of p53 binding and modified histones

We next generated a genomic map of p53 binding in thehuman ENCODE region, together with H3 and H4 acetylationand H3 Lys4 di- and trimethylation. Cross-linked chromatinfrom the 5FU-treated (9 h) HCT116 cells was immunoprecipi-tated with antibodies against p53, H3 Lys9/14 acetylation, H3Lys4 di- and trimethylation, and H4 Lys5/8/12/16 acetylation.The immunoprecipitated DNA and untreated control DNA(whole-cell extract) were amplified by in vitro transcription andhybridized separately to oligonucleotide arrays that tilenonrepetitive portions of the human ENCODE region at 22-bp intervals. Genomic sites enriched for p53 and modifiedhistone were identified by sliding a 550-bp window across theregions and testing whether signals of oligonucleotide probeswithin the window were significantly greater for the ChIPsample than for the control. After applying a significancethreshold of p<10−5, we excluded those sites composed ofmore than 70% repeat sequences. As shown in SupplementalFigure 1A, most of the identified sites (77%) contained less than10% repeat sequences. The identified sites with more than 70%repeat sequences had enrichment ratios of less than twofold(data not shown). Considering the limitations of the ChIP-chipmethod, it is not surprising that repeat sequences cause false-positives. For p53, we applied a cutoff combining a significancethreshold and a signal ratio of ChIP/control because this cutoffgave lower false-positive rates (Supplemental Fig. 1B). Finally,our analysis identified 40 p53-BS, 512 acetyl H3, 416 acetylH4, 541 dimethyl H3, and 275 trimethyl H3 sites with mediansize of 480, 511, 428, 528, and 658 bp, respectively (Table 1).

Verification of enrichment was performed by quantitativePCR on the preamplified ChIP samples. Relative enrichmentwas quantified for each site in PCRs using 0.5 ng ChIP DNA or0.5 ng nonenriched control DNA as template. Of sites identifiedby array, 37/40 (93%) of p53-BS and 29/31 (94%) of acetyl H4sites were confirmed to have enrichment ratios of at least twofold(Supplemental Fig. 2). These results provide experimentalevidence that very few of the identified sites are false-positives.

As an example, we observed a p53-BS in the first intron ofthe TRIM22 gene, which was reported to be p53 regulated

Table 1Genomic Sites on the ENCODE region Enriched for p53 and Modified Histones

(A) Transcription Start Sites (TSS)

Type # Identified sites Median Size # Sites within 1 kb o

p53 37 (40) 558 (480) bp 2Acetyl H3 (Lys9/14) 512 511 bp 179

416 428 bp 49di-Methyl H3 (Lys4) 541 528 bp 222

275 658 bp 175

Type Motif finder c # Sites # Sites with ≥ 1 moti

p53 p53MH 37 23TRANSFAC 37 29

Acetyl H3 (Lys9/14) p53 512 56a TSS of RefSeq genes in the UCSC genome browser (hg15).b Fold-enrichment to the background estimated using 10,000 regions randomly-sec Two major weight-matrix based programs; p53MH (Hoh, et al., 2002) and TRA

(Fig. 2A). A functional p53-response element had been reportedto lie at the region 827 to 846 bp downstream of the TSS of thegene [31]. Among three p53-binding motifs computationallyfound along the gene, our analysis detected one containing thep53-response element (Fig. 2B), illustrating the advantage ofusing an unbiased genomic tiling array for discovery of in vivobinding sites. Furthermore, in the GeneChip expression analysis,the expression level of TRIM22was observed to be up-regulatedin a p53-mediated manner, which was not seen in a neighborgene, TRIM5 (Fig. 2C). These results indicate that our analysisidentified an authentic p53-BS at high resolution. For furtheranalysis, we used these 37 p53-BS that were confirmed to haveenrichment ratios of at least twofold.

p53-BS lie throughout the human genome

We examined the identified sites for characteristics asso-ciated with the cis promoter of the gene. Of the p53-BS detected,2 (5%), 179 (35%), 49 (12%), 222 (41%), and 175 (64%) of thep53-binding, acetyl H3, acetyl H4, dimethyl H3, and trimethylH3 sites, respectively, were within 1 kb of an annotatedtranscription start site (TSS), and their enrichment levels wereapproximately 4.1-, 26.6-, 9.0-, 31.2-, and 48.5-fold, respec-tively, over what would be expected at random (Table 1). Theseresults indicate that a significant portion of p53-BS wereobserved in the regions outside of the cis promoter, whichhas been studied extensively as a transcriptional regulatoryregion.

We next examined the positions of the identified sitesrelative to annotated exons and introns of RefSeq genes. First,the expected background distribution was generated from2000 randomly selected regions (Fig. 3). Compared to thebackground, the p53-BS showed a slight enrichment at the 5′and 3′ ends of genes (Fig. 3). However, approximately half(52%) of the p53-BS were located at the intergenic region(Fig. 3), indicating the presence of unknown functions of p53binding at the genomic region that have not been studiedwell.

The histone modification sites showed distinctive patternsof distribution. Whereas the acetyl H4 sites were located

f TSS a % Sites within 1 kb of TSS a Fold-enrichment b (vs. background)

5 4.135 26.612 9.041 31.264 48.5

f % Sites with ≥ 1 motif Fold-enrichment b

62 4.978 5.811 0.9

lected from the ENCODE region.NSFAC (Wingerder, et al., 2000).

Fig. 2. A representative p53-binding site. (A) A genomic view of the ChIP-chip result around TRIM5 and TRIM22 (chromosome 11p15.4). Regions enriched for p53,acetyl H3, acetyl H4, dimethyl H3 (Lys4) and trimethyl H3 (Lys4) are indicated as bars in different colors. One p53-binding site and two acetyl H4 sites are located atthe 5′ end of TRIM22. Acetyl H3, dimethyl H3, and trimethyl H3 sites are observed at the 5′ end of TRIM5. (B) p53-binding motifs found around TRIM22. Sequencesmatching the exact consensus are indicated in red. (C) GeneChip expression analysis data for TRIM5 and TRIM 22 in HCT116 (p53+/+) treated with 5FU for 0 to 24 h.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

181K. Kaneshiro et al. / Genomics 89 (2007) 178–188

throughout the genome including the intergenic regions, theacetyl H3 and the di- and trimethyl H3 sites showed asignificant enrichment at the 5′ ends and first introns ofgenes (Fig. 3). This suggests the possibility that there aredifferential roles for histone modifications in the humangenome.

p53-binding motif

The identified p53-BS was searched for the p53-bindingmotif by using the motif finder, p53MH [32]. Of 37 p53-BS, 23(62%) had at least one p53-binding motif meeting a givensignificance threshold (Table 1). This is greater than 4.9-fold

Fig. 3. Classification of regions enriched for p53, acetyl H3, acetyl H4, dimethyl H3 (Lys4), and trimethyl H3 (Lys4). For the background estimation, 2000 randomlyselected regions were used. The sites were classified into six groups: within 5 kb of the 5′UTR, within exons, within the first intron, within the other introns, within5 kb downstream of the 3′UTR of the RefSeq gene, and others (intergenic region).

182 K. Kaneshiro et al. / Genomics 89 (2007) 178–188

enrichment over what would be expected at random. By contrast,the enrichment was not seen for the array-identified acetyl H3sites (Table 1).

Because a motif finder has its unintentional bias originatingfrom the learning data set, we also used another weighted-matrix based motif finder, TRANSFAC [33]. By usingTRANSFAC, 9 other p53-BS were revealed to have at leastone p53-binding motif (Table 1). Thus, in total, 32 (86%) of thep53-BS had at least one p53-binding motif in their sequences. Inthese 32 p53-BS, we found only 1 perfect p53 consensus DNA-binding motif. This is consistent with the fact that mostpreknown p53-BS have a p53-binding motif that is similar butnot identical to the p53 consensus DNA-binding motif.

For five p53-BS lacking an explicit p53-binding motif, wenext searched for other consensus motifs by using theTRANSFAC weighted-matrixes. This analysis revealed theenrichment of TBP-, AP1-, and OCT1-binding motifs in thesefive regions, with enrichment levels that were 6.5-, 6.3-, and 5.0-fold, respectively, over what would be expected at random(Supplemental Table 1).

Furthermore, we compared the DNA-binding activity of p53between the p53-BS with or without a p53-binding motif byquantitative PCR. As shown in Supplemental Figure 3, therewas no obvious correlation between the presence of a p53-binding motif and the fold enrichment of p53.

Histone modification pattern in p53-BS

To elucidate the relationship between the p53-binding andhistone modification, overlap with the histone modification site

was examined for each p53-BS. The examination wasperformed for the region with a fixed length (1.2 kb), spanning0.6 kb from the center of each p53-BS. Of 37 p53-BS identified,14 (38%), 33 (89%), 5 (14%), and 1 (3%) overlapped withacetyl H3 sites, acetyl H4 sites, dimethyl H3 sites, and trimethylH3 sites, respectively (Table 2).

We next examined histone modification patterns of the p53-BS according to their genomic locations relative to annotatedexons and introns of the RefSeq genes. For each genomiclocation, the length of p53-BS overlapping with the histonemodification site was summed up. Histone acetylation of thep53-BS showed distinctive patterns according to the genomiclocation (Fig. 4A). Thus, at the 5′ and 3′ ends of genes, 58 and54% of the p53-BS overlapped with both H3 and H4 acetylationsites (Fig. 4A). At the other genomic locations, acetylation ofboth H3 and H4 was observed for less than 17% of the p53-BS(Fig. 4A). Interestingly, a significant portion (58%) of H4 acety-lation in the p53-BS was not accompanied by H3 acetylation.The H4 acetylation alone of the p53-BS was widely observed atany genomic location, whereas the H3 acetylation without H4acetylation was observed at only a small portion of the p53-BS(Fig. 4A). In contrast, no significant enrichment was observedfor either histone di- or trimethylation of the p53-BS (Fig. 4B).

To explore the p53-dependent kinetics of histone acetylationat the p53-BS, we compared the fold enrichment of acetylatedhistone between TP53 wildtype cells and TP53-null cells byquantitative PCR. In 4 (36%) of the H3-acetylated regions and24 (86%) of the H4-acetylated regions, the fold change ratios ofacetylation (TP53 wildtype / TP53-null) were higher than atleast fourfold (Fig. 5). When hierarchically clustered by using

Table 2List of p53-BS, Histone Modification and Gene Expression Change

ID Chr Start End Foldenrichment a

Acetyl H3 Acetyl H4 di-Me H3 tri-Me H3 p53 cons. motif Genes b within 5 kb ExpressionChange c

1 chr1 147942358 147943557 4.3 + + + + - PIP5K1A I2 chr2 220512668 220513867 10.5 + + + - + - -3 chr2 220561677 220562876 16.2 + + - - + - -4 chr2 234512404 234513603 2 - + - - + - -5 chr5 55943111 55944310 8.8 - + - - + - -6 chr5 56086704 56087903 9.8 - + - - + - -7 chr5 56127634 56128833 4.1 - + - - + - -8 chr5 131459911 131461110 11.4 - + - - + - -9 chr5 132165535 132166734 2.6 - - - - + - -10 chr5 141930795 141931994 8.5 - + - - + - -11 chr6 41737459 41738658 9.8 - + - - + SNT-2 NC12 chr7 115440293 115441492 6.3 + + + - + TES I13 chr7 115464783 115465982 10.3 + + - - + - -14 chr7 116810368 116811567 10.1 + + - - + CFTR A15 chr7 116811165 116812364 90.5 + + - - + CFTR NC16 chr7 126248970 126250169 9.9 - - - - + GRM8 NC17 chr11 914308 915507 21 - + - - - PYPAF5 A18 chr11 5669418 5670617 8.6 - + - - - TRIM22 I19 chr11 5670561 5671760 61.6 - + - - + TRIM22 I20 chr12 40138437 40139636 19.5 - + - - - SLC2A13 NC21 chr12 40268988 40270187 6.3 - + - - + SLC2A13 NC22 chr12 40519961 40521160 26.6 - + - - + - -23 chr14 52010609 52011808 6.6 - + - - + - -24 chr14 97787751 97788950 17.3 - + - - + - -25 chr19 59376460 59377659 9.4 + + - - + LENG5, PRS9 D, I26 chr20 34563868 34565067 26 + + - - + C20orf44 I27 chr20 34706768 34707967 47.5 + + - - + CEP2 NC28 chr20 34707501 34708700 12.3 + + - - + CEP2 NC29 chr20 34882371 34883570 5.6 - - - - + CPNE1 D30 chr21 32823527 32824726 8.6 - + - - + - -31 chr21 34268301 34269500 6.3 - + + - + - -32 chr21 39285599 39286798 9.6 + + + - + - -33 chr22 30403435 30404634 8.3 - + - - + MGC50372 NC34 chr22 30826933 30828132 7.6 - - - - - SLC5A1 A35 chrX 120904031 120905230 8.9 - + - - + GRIA3 A36 chrX 151339909 151341108 14 + + - - + G6PD I37 chrX 151340582 151341781 20.7 + + - - + G6PD I

I: Increased D: Decreased NC: No change A: Absent.a Fold enrichment quantified by Q-PCR using ChIP DNA and non-enriched input DNA as templates.b The RefSeq gene in the UCSC genome browser (hg15).c Change call calculated using a comparative analysis program in the Affymetrix MAS 5.0 package.

183K. Kaneshiro et al. / Genomics 89 (2007) 178–188

the enrichment folds of acetyl histone, roughly three patterns ofp53 dependency were observed among the p53-BS (Fig. 5). Inpattern A, p53-dependent acetylation was observed for both H3and H4. In pattern B, p53-dependent acetylation was observedonly for H4. In pattern C, there was no significant change inacetylation for either H3 or H4.

Expression change of adjacent gene

To see the relationship between p53 binding and transcrip-tional regulation, we examined the expression changes of geneswithin 5 kb of the p53-BS. GeneChip expression analysis wasperformed for RNA prepared from TP53 wildtype cells treatedwith 5FU for 0 to 24 h. The expression change was examinedusing the comparative analysis program of the Affymetrix MAS5.0. There was no significant correlation between p53 binding

and the expression change, probably due to a limited number ofgenes examined (Table 2). Among 17 genes within 5 kb of thep53-BS, 6 genes showed an increase and 2 genes showed adecrease in expression level after the 5FU treatment. There wereno expression changes observed in 6 genes. No expression wasdetected at any time point for the rest of genes.

Discussion

Novel p53-BS in the human genome

In a 30-Mb portion of the human genome, we found 37 p53-BS. Extrapolating our results to the whole genome scalepredicts a total of 3700 p53-BS. In the previous ChIP-chip studyof p53 sites along human chromosomes 21 and 22, Cawley et al.

Fig. 4. Histone modification patterns of the p53-BS classified according to theirgenomic locations. (A) Histone acetylation pattern. A portion of the p53-BSobserved for both acetyl H3 and acetyl H4 sites is shown in yellow. A portionobserved for only acetyl H3 sites is shown in blue. A portion observed for onlyacetyl H4 sites is shown in red. A portion observed for neither acetyl H3 nor H4is shown in gray. (B) Histone methylation pattern. (For interpretation of thereferences to colour in this figure legend, the reader is referred to the web versionof this article.)

Fig. 5. Differential patterns of the p53 dependency of histone acetylation at thep53-BS. For each p53-BS, a fold enrichment against the input DNA wasmeasured by quantitative PCR for acetyl H3 and acetyl H4 in HCT116 (p53+/+)and HCT116 (p53−/−). The fold enrichment was scaled with dark and lightcoloring. Those sites not observed for acetylation by arrays were indicated as N.D. (not detected). Nodes of the tree are designated A, B, and C.

184 K. Kaneshiro et al. / Genomics 89 (2007) 178–188

predicted a total of 1600 p53-BS [16]. This discordance mightbe attributable to the different cellular stresses used for p53activation. We used 5FU as a DNA-damaging agent, whereasCawley et al. used bleomycin. As with Cawley et al.’s report,the p53-BS identified in this study lie throughout the humangenome, including exons and introns of annotated potentialgene targets and uncharacterized intergenic regions. Recently,the p53-BS was determined by a method combining ChIP andpaired-end di-tag (PET) sequencing [34]. Comparing our datawith the PET-2 data released at the University of California atSanta Cruz (UCSC) database, 9 (24%) p53-BS were concordantwith our results. The other p53-BS have not been describedpreviously.

Of 37 p53-BS, 30 (81%) had at least one p53-bindingconsensus motif in their sequences, whereas 7 (19%) had noexplicit p53-binding motif. Those TF-BS lacking a consensusbinding motif were also reported in the previous genome-widestudies for other TFs, such as RELA, GATA1 and CREB1[35–37]. Interestingly, in the 7 p53-BS lacking a p53-bindingmotif, we found enrichment of other TF-binding motifs, suchas the EGR family, TBP, and CREB1. Physical interaction ofp53 with EGR1 and TBP has been reported [38,39]. p53 hasbeen reported to interact with CREB1 via binding to the KIXdomain of EP300 [40]. These data raise the possibility that

other DNA-binding proteins may mediate indirect interactionsbetween p53 and DNA.

We examined the DNA-binding activity at each p53-BS bymeasuring a relative enrichment of the identified site withquantitative PCR using ChIP DNA and nonenriched input DNAas a template. Enrichment levels were variable and seemed tohave no correlation with the presence of a p53-binding motif.Also, there was no significant correlation with the genomiclocation of p53-binding (data not shown), suggesting that theDNA-binding activity of p53 in vivo is determined by factorsother than the number or location of the p53-BS.

Although p53 is typically described as a transcriptionalregulator, many genes within 5 kb of p53-BS did not show anysignificant change in their expression levels, suggesting thatthere must be additional factors required for transcriptionalregulation, such as cofactors, histone modification, anddemethylation of DNA.

185K. Kaneshiro et al. / Genomics 89 (2007) 178–188

Discrete pathways for histone acetylation and histonemethylation

Histone tails are the target of a variety of covalent modi-fications. Among the many possible modifications of histonetails, di- and trimethylation of histone H3-K4 and acetylation ofhistone H3-K9/14 and H4-K5/8/12/16 are well documented tobe correlated with actively transcribed regions [41]. In thisstudy, we focused on the relationship between p53 binding andthese histone modifications.

Although p53 forms protein complexes containing bothhistone acetyltransferases and methyltransferases, p53 bindingwas strongly associated with histone acetylation, rather thanhistone methylation. Among several protein methyltransferases,SETD7 is known to interact with p53 and target H3-K4 in vitro[42]. However, a recent study has revealed that purified SETD7has no catalytic activity for histone H3-K4 assembled intonucleosomes, which are thought to be the physiologicallyrelevant substrates for chromatin-modifying enzymes [43].Collectively, the evidence suggests that there are other p53-independent factors necessary for histone methylation in vivo,and p53 itself mainly mediates histone acetylation viarecruitment of histone acetyltransferases to target sites.

Distinctive patterns of histone acetylation in the p53-BSaccording to the genomic locations relative to the genestructure

When we classified the p53-BS into several groupsaccording to the location relative to annotated exons andintrons of RefSeq genes, differential patterns of histoneacetylation were observed among the groups. The acetylationof H3 was often observed in the 5′ ends and 3′ ends of genes,whereas the acetylation of H4 was observed at any genomiclocation. Recent comparative genomic studies have reportedthat noncoding sequences located at the 5′ and 3′ ends of genesare often conserved among species and have some roles intranscriptional regulation [44]. The biased distribution ofhistone acetylation suggests the generation of noncodingtranscripts in these genomic regions. However, it is still unclearwhether those transcripts participate in p53-regulated functions.

One possible explanation for the differential pattern in theacetylation of H3 and H4 is the presence of variants in H3. Fourdifferent isoforms of H3 have been reported: H3.1, H3.2, H3.3,and CENPA [45]. In Drosophila cells, H3.3 is known to bepresent at transcriptionally active loci [46]. The deposition ofH3.3 occurs replication-independently, and some histonemodifications have been shown to be restricted to nucleosomescontaining H3.3 [47]. Such a nucleosome variant mightaccount for the distinctive distribution of histone acetylation.

Although it is still unclear whether these histone acetylationscan be set prior to p53 binding, we provide evidence for the p53dependency of these changes in the chromatin state. In TP53-null cells, histone acetylation was observed at only a few of thep53-binding regions detected in TP53 wildtype cells. However,it remains unknown how p53 can recognize its bindingsequence in condensed chromatin.

Conclusion

Here we have presented an integrated map of p53 bindingand histone modification along 30 Mb of the human genome athigh resolution. Novel p53-BS were found throughout thegenome and showed various features with respect to density,genic location, and the presence or absence of consensus p53-binding motifs. Our analysis also revealed distinctive patterns ofhistone modification in the p53-BS according to locationrelative to gene structures. The combination of these variousfactors might account for the great diversity of p53-binding sitesin the genome and provide a physical basis for the broad rangeof p53-regulated transcriptional programs in humans. On theother hand, we did not observe a significant correlation betweenp53 binding and the expression of adjacent genes, probably dueto a limited number of targets examined. To better elucidate themechanisms underlying the complex transcriptional regulationby p53, further large-scale genomic studies are required.

Materials and methods

Cell culture, cell extracts, and Western blotting

HCT116 cells (TP53-wild-type+/+) and TP53-null derivatives were dis-tributed from Professor Bert Vogelstein (Johns Hopkins University), grown inMacCoy’s 5A medium supplemented with 10% FBS at 37, in a humidifiedatmosphere with 5% CO2. p53 was induced by treatment with 375 μM5FU for 0to 24 h. The cells were washed in Dulbecco’s phosphate-buffered saline andharvested in lysis buffer [50 mMTris (pH 8.0), 150 mMNaCl, 1 mM EDTA, 1%Triton X-100, 0.1% SDS, 0.1% sodium deoxycholate, 10 mM NaF, and 2 mMNa3VO4] containing protease inhibitors (Sigma). Lysates were maintained at 4for 30 min followed by spin centrifugation for 15 min at 15,000g. Supernatantswere collected and boiled with 2× SDS sample buffer. Immunoblotting wasperformed as previously described [48]. Antibodies used were p53 (FL-393),p21 (C-19), and MDM2 (SMP-14) from Santa Cruz (California, USA) andphosphorylated p53 (Ser15), phosphorylated p53 (Ser20), and acetylated p53(Lys382) from Professor Youichi Taya (National Cancer Center, Japan).

RNA extraction and microarray expression analysis

Total RNA was extracted with the use of Trizol reagent (Invitrogen)according to the manufacturer’s instructions. The experimental procedures forGeneChip (Affymetrix, Santa Clara, CA) were performed according to theAffymetrix GeneChip expression analysis technical manual. Briefly, 10 μg oftotal RNA was used to synthesize biotin-labeled cRNA, which was thenhybridized to a GeneChip Human U133 plus 2.0 oligonucleotide array(Affymetrix). After being washed, the arrays were stained with streptavidin–phycoerythrin and analyzed on a Affymetrix scanner to collect the image data.Microarray Suite software 5.0 (Affymetrix) was used to calculate the averagedifference (AD) for each gene probe set, which was shown as an intensityvalue of the gene expression. The AD values were normalized for each array sothat the average of all AD values was 100. For expression change calls, weused the comparative analysis (Wilcoxon signed-rank test-based analysis)program in the Microarray Suite 5.0 software. The analysis was performedwith the default parameters, where the p value for significant difference wasbelow 0.0025. For details, please refer to Statistical Algorithms ReferenceGuide available at the website http://www.affymetrix.com/support/technical/technotes/statistical_reference_guide.pdf. Raw data are available at the websitehttp://www.genome.rcast.u-tokyo.ac.jp/Tiling_Array/.

Chromatin immunoprecipitation

HCT116 cells (p53+/+, p53−/−) were cultured in 15-cm plates to approximately70 to 80% confluence, and 5 plates were used per immunoprecipitation. For p53

186 K. Kaneshiro et al. / Genomics 89 (2007) 178–188

induction, cells were treated for 9 h with 375 μM 5FU dissolved in dimethylsulfoxide. Cells were fixed with 1% formaldehyde for 10 min at room temperature,with occasional swirling. Glycine was added to a final concentration of 0.2 M andthe incubation was continued for an additional 5 min. Cells were washed with ice-cold PBS, harvested by scraping, pelleted, and resuspended in SDS lysis buffer[10 mM Tris-Cl (pH 8.0), 150 mM NaCl, 1% SDS, 1 mM EDTA]. Samples weresonicated with TomyUD-201 for 5 cycles of 1 min (60% duty, output level 2) withan interval of 1 min. Then, samples were centrifuged at 13,000g at 4 for 10 min anddiluted fivefold in ChIP dilution buffer [20 mM Tris-Cl (pH 8.0), 150 mM NaCl,1mMEDTA, 1%TritonX-100, and protease inhibitors]. After removal of a controlaliquot (whole-cell extract), samples were incubated at 4 overnight with 50 μlprotein A–Sepharose beads that had been incubated overnight with 5–10 μgantibodies against p53 (FL393, Santa Cruz), acetyl H3 Lys9/14 (Upstate), andacetyl H4 Lys5/8/12/16 (Upstate). After being washed, the immunoprecipitatedmaterials were eluted from the beads by heating for 30 min at 65 in elution buffer[25 mM Tris-Cl (pH 7.5), 10 mM EDTA, 0.5% SDS]. The cross-links were thenreversed with incubation with 1.5 μg/ml Pronase at 42 for 2 h followed byheating at 65 overnight. The DNA was extracted with a PCR purification kit(Qiagen).

Quantitative PCR

Quantitative PCR was performed using an iCycler iQ Detection System(Bio-Rad, Hercules, CA). Reaction mixtures contained SYBR Green Ifluorescence (BMA, Rockland, ME) and primers as follows: for CDKN1A,GTGGCTCTGATTGGCTTTCTG (forward) and CTGAAAACAGGCAGCC-CAAG (reverse). PCR conditions were as follows: 1 cycle of 94 for 3 minfollowed by 40 cycles at 94 for 30 s, 65 for 30 s, 72 for 1 min, and detection (85for 10 s). All the samples were run in triplicate and the results were averaged.

DNA amplification and array hybridization

ChIP DNA samples were amplified by two cycles of IVT. The first cycle ofthe IVT amplification was performed as described [23]. The amplified RNAwasconverted into single-strand cDNA with random primers, annealed with T7primer followed by second-strand cDNA synthesis, and amplified again by invitro transcription using the T7 Megascript Kit (Ambion) according to themanufacturer’s instructions. A second amplified RNA was converted intodouble-strand cDNAwith random primers, fragmented with DNase I, and end-labeled with biotin as described [51]. Two microgram samples of ChIP andcontrol (whole-cell extract) were hybridized on the separate AffymetrixENCODE 01 oligonucleotide arrays. Arrays were hybridized for 16 h at 45,washed using the antibody amplification protocol described in the AffymetrixExpression Analysis Technical Manual, and scanned using an Affymetrix GeneArray scanner. For p53, acetyl H3 and acetyl H4, two biologically independentChIP samples were amplified in triplicate and hybridized individually. Fordi- and trimethyl H3, three biologically independent ChIP samples wereindividually amplified and hybridized. Raw data are available at the websitehttp://www.genome.rcast.u-tokyo.ac.jp/Tiling_Array/.

Analysis of tiling array data

Raw array data were quantile normalized within enriched and controlreplicate groups [49]. (PM, MM) intensity pairs were mapped to the genomeusing exact 25-mer matching hs.NCBIv33. For each genomic position to whicha probe pair mapped, a data set was generated consisting of all (PM, MM) pairsmapping within a window of +/− 275 bp. A Wilcoxon ranked-sum test wasapplied to the transformation log2[max(PM-MM, 1)] for data from the sixtreatment and six control DNA arrays, testing the null hypothesis that treatmentand control data come from the same probability distribution. For each window,a signal ratio was also estimated by the Hedges-Lehmann method computing themedian of fold enrichment among the probe sets within the window.

Sequence analysis

The sequences of human RefSeq transcripts (hg16) were downloaded fromthe UCSC genome browsers (http://www.genome.ucsc.edu/). For each array-

identified site, overlap with the transcripts was judged at a base to base level.Enrichment analysis of consensus binding motifs was performed with theposition-weighted matrix (PWM) method. PWM matrixes were retrieved fromTRANSFAC and the p53MH program. The threshold score of each motif wascalculated with p<10−4 in the background genome, where the motif was foundapproximately once in 10 kb.

Hierarchical clustering

The hierarchical clustering was performed using Eisen’s cluster program[50]. The fold enrichment of each acetylated histone was log-transformed andclustered with an algorithm of correlation (uncentered).

Acknowledgments

We thank Bert Vogelstein and Yoichi Taya for providingHCT116 cells (p53+/+, p53−/−) and antibodies of modified p53,respectively. We sincerely thank Yoshiyuki Sakaki for hishelpful suggestion, Hiroshi Yoshikawa and Vincent Stanton forcritically reviewing the article. Special thanks are extended toHiroko Meguro for array hybridizations and Yuri Kamase forquantitative PCR analysis. This study wasmainly supported by agrant from the Genome Network Project from the Ministry ofEducation, Culture, Sports, Science, and Technology (MEXT),Japan, and partly supported by Grants-in-Aid for ScientificResearch (S) 16101006 (HA) and Grant-in-Aid for YoungScientists (B) 17790692 (ST) from the Ministry of Education,Culture, Sports, Science and Technology, and Special Coordina-tion Fund for the MEXT.

Appendix A. Supplementary data

Supplementary data associated with this article can be found,in the online version, at doi:10.1016/j.ygeno.2006.09.001.

References

[1] B. Vogelstein, D. Lane, A.J. Levine, Surfing the p53 network, Nature 408(2000) 307–310.

[2] R. Zhao, K. Gish, M. Murphy, Y. Yin, D. Notterman, W.H. Hoffman, E.Tom, D.H. Mack, A.J. Levine, The transcriptional program followingp53 activation, Cold Spring Harbor Symp. Quant. Biol. 65 (2000)475–482.

[3] K. Oda, H. Arakawa, T. Tanaka, K. Matsuda, C. Tanikawa, T. Mori, H.Nishimori, K. Tamai, T. Tokino, Y. Nakamura, Y. Taya, p53AIP1, apotential mediator of p53-dependent apoptosis, and its regulation bySer-46-phosphorylated p53, Cell 102 (2000) 849–862.

[4] K. Kannan, N. Amariglio, G. Rechavi, J. Jakob-Hirsch, I. Kela, N.Kaminski, G. Getz, E. Domany, D. Givol, DNA microarrays identificationof primary and secondary target genes regulated by p53, Oncogene 20(2001) 2225–2234.

[5] D. Lane, How cells choose to die, Nature 414 (2001) 25, 27.[6] W.S. el-Deiry, S.E. Kern, J.A. Pietenpol, K.W. Kinzler, B. Vogelstein,

Definition of a consensus binding site for p53, Nat. Genet. 1 (1992) 45–49.[7] H. Qian, T. Wang, L. Naumovski, C.D. Lopez, R.K. Brachmann, Groups of

p53 target genes involved in specific p53 downstream effects cluster intodifferent classes of DNA binding sites, Oncogene 21 (2002) 7901–7911.

[8] M.D. Kaeser, R.D. Iggo, Chromatin immunoprecipitation analysis fails tosupport the latency model for regulation of p53 DNA binding activity invivo, Proc. Natl. Acad. Sci. USA 99 (2002) 95–100.

[9] S.T. Szak, D. Mays, J.A. Pietenpol, Kinetics of p53 binding to promotersites in vivo, Mol. Cell. Biol. 21 (2001) 3375–3386.

187K. Kaneshiro et al. / Genomics 89 (2007) 178–188

[10] K. Sakaguchi, J.E. Herrera, S. Saito, T. Miki, M. Bustin, A. Vassilev, C.W.Anderson, E. Appella, DNA damage activates p53 through a phosphory-lation-acetylation cascade, Genes Dev. 12 (1998) 2831–2841.

[11] W. Gu, R.G. Roeder, Activation of p53 sequence-specific DNA binding byacetylation of the p53 C-terminal domain, Cell 90 (1997) 595–606.

[12] Y. Samuels-Lev, D.J. O'Connor, D. Bergamaschi, G. Trigiante, J.K. Hsieh,S. Zhong, I. Campargue, L. Naumovski, T. Crook, X. Lu, ASPP proteinsspecifically stimulate the apoptotic function of p53, Mol. Cell 8 (2001)781–794.

[13] E.R. Flores, K.Y. Tsai, D. Crowley, S. Sengupta, A. Yang, F. McKeon, T.Jacks, p63 and p73 are required for p53-dependent apoptosis in response toDNA damage, Nature 416 (2002) 560–564.

[14] L. Resnick-Silverman, S. St. Clair, M. Maurer, K. Zhao, J.J. Manfredi,Identification of a novel class of genomic DNA-binding sites suggests amechanism for selectivity in target gene activation by the tumor suppressorprotein p53, Genes Dev. 12 (1998) 2102–2107.

[15] W.S. el-Deiry, Regulation of p53 downstream genes, Semin. Cancer Biol.8 (1998) 345–357.

[16] S. Cawley, S. Bekiranov, H.H. Ng, P. Kapranov, E.A. Sekinger, D.Kampa, A. Piccolboni, V. Sementchenko, J. Cheng, A.J. Williams, R.Wheeler, B. Wong, J. Drenkow, M. Yamanaka, S. Patel, S. Brubaker, H.Tammana, G. Helt, K. Struhl, T.R. Gingeras, Unbiased mapping oftranscription factor binding sites along human chromosomes 21 and 22points to widespread regulation of noncoding RNAs, Cell 116 (2004)499–509.

[17] E.T. Dermitzakis, A. Reymond, R. Lyle, N. Scamuffa, C. Ucla, S. Deutsch,B.J. Stevenson, V. Flegel, P. Bucher, C.V. Jongeneel, S.E. Antonarakis,Numerous potentially functional but non-genic conserved sequences onhuman chromosome 21, Nature 420 (2002) 578–582.

[18] T. Jenuwein, C.D. Allis, Translating the histone code, Science 293 (2001)1074–1080.

[19] M. Murphy, J. Ahn, K.K. Walker, W.H. Hoffman, R.M. Evans, A.J.Levine, D.L. George, Transcriptional repression by wild-type p53 utilizeshistone deacetylases, mediated by interaction with mSin3a, Genes Dev. 13(1999) 2490–2501.

[20] J.M. Espinosa, B.M. Emerson, Transcriptional regulation by p53 throughintrinsic DNA/chromatin binding and site-directed cofactor recruitment,Mol. Cell 8 (2001) 57–69.

[21] N.A. Barlev, L. Liu, N.H. Chehab, K. Mansfield, K.G. Harris, T.D.Halazonetis, S.L. Berger, Acetylation of p53 activates transcriptionthrough recruitment of coactivators/histone acetyltransferases, Mol. Cell8 (2001) 1243–1254.

[22] P.G. Ard, C. Chatterjee, S. Kunjibettu, L.R. Adside, L.E. Gralinski, S.B.McMahon, Transcriptional regulation of the mdm2 oncogene by p53requires TRRAP acetyltransferase complexes, Mol. Cell. Biol. 22 (2002)5650–5661.

[23] G. Liu, T. Xia, X. Chen, The activation domains, the proline-rich domain,and the C-terminal basic domain in p53 are necessary for acetylation ofhistones on the proximal p21 promoter and interaction with p300/CREB-binding protein, J. Biol. Chem. 278 (2003) 17557–17565.

[24] G. Lagger, A. Doetzlhofer, B. Schuettengruber, E. Haidweger, E.Simboeck, J. Tischler, S. Chiocca, G. Suske, H. Rotheneder, E.Wintersberger, C. Seiser, The tumor suppressor p53 and histonedeacetylase 1 are antagonistic regulators of the cyclin-dependent kinaseinhibitor p21/WAF1/CIP1 gene, Mol. Cell. Biol. 23 (2003) 2669–2679.

[25] L.J. Juan, W.J. Shia, M.H. Chen, W.M. Yang, E. Seto, Y.S. Lin, C.W. Wu,Histone deacetylases specifically down-regulate p53-dependent geneactivation, J. Biol. Chem. 275 (2000) 20436–20443.

[26] B. Shan, J. Xu, Y. Zhuo, C.A. Morris, G.F. Morris, Induction of p53-dependent activation of the human proliferating cell nuclear antigen gene inchromatin by ionizing radiation, J. Biol. Chem. 278 (2003) 44009–44017.

[27] R. Candau, D.M. Scolnick, P. Darpino, C.Y. Ying, T.D. Halazonetis, S.L.Berger, Two tandem and independent sub-activation domains in the aminoterminus of p53 require the adaptor complex for activity, Oncogene 15(1997) 807–816.

[28] M.D. Kaeser, R.D. Iggo, Promoter-specific p53-dependent histoneacetylation following DNA damage, Oncogene 23 (2004) 4007–4013.

[29] Y. Mitani, N. Oue, Y. Hamai, P.P. Aung, S. Matsumura, H. Nakayama, N.

Kamata, W. Yasui, Histone H3 acetylation is associated with reduced p21(WAF1/CIP1) expression by gastric carcinoma, J. Pathol. 205 (2005) 65–73.

[30] T.E.P. Consortium, The ENCODE (ENCyclopedia Of DNA Elements)Project, Science 306 (2004) 636–640.

[31] S. Obad, H. Brunnstrom, J. Vallon-Christersson, A. Borg, K. Drott, U.Gullberg, Staf50 is a novel p53 target gene conferring reducedclonogenic growth of leukemic U-937 cells, Oncogene 23 (2004)4050–4059.

[32] J. Hoh, S. Jin, T. Parrado, J. Edington, A.J. Levine, J. Ott, The p53MHalgorithm and its application in detecting p53-responsive genes, Proc. Natl.Acad. Sci. USA 99 (2002) 8467–8472.

[33] E. Wingender, X. Chen, R. Hehl, H. Karas, I. Liebich, V. Matys, T.Meinhardt, M. Pruss, I. Reuter, F. Schacherer, TRANSFAC: An integratedsystem for gene expression regulation, Nucleic Acids Res. 28 (2000)316–319.

[34] C.L. Wei, Q. Wu, V.B. Vega, K.P. Chiu, P. Ng, T. Zhang, A. Shahab, H.C.Yong, Y. Fu, Z. Weng, J. Liu, X.D. Zhao, J.L. Chew, Y.L. Lee, V.A.Kuznetsov, W.K. Sung, L.D. Miller, B. Lim, E.T. Liu, Q. Yu, H.H. Ng, Y.Ruan, A global map of p53 transcription-factor binding sites in the humangenome, Cell 124 (2006) 207–219.

[35] R. Martone, G. Euskirchen, P. Bertone, S. Hartman, T.E. Royce, N.M.Luscombe, J.L. Rinn, F.K. Nelson, P. Miller, M. Gerstein, S. Weissman, M.Snyder, Distribution of NF-κB-binding sites across human chromosome22, Proc. Natl. Acad. Sci. USA 100 (2003) 12247–12252.

[36] C.E. Horak, M.C. Mahajan, N.M. Luscombe, M. Gerstein, S.M.Weissman, M. Snyder, GATA-1 binding sites mapped in the beta-globinlocus by using mammalian chIp-chip analysis, Proc. Natl. Acad. Sci.USA 99 (2002) 2924–2929.

[37] G. Euskirchen, T.E. Royce, P. Bertone, R. Martone, J.L. Rinn, F.K. Nelson,F. Sayward, N.M. Luscombe, P. Miller, M. Gerstein, S. Weissman, M.Snyder, CREB binds to multiple loci on human chromosome 22, Mol. Cell.Biol. 24 (2004) 3804–3814.

[38] D.W. Martin, R.M. Munoz, M.A. Subler, S. Deb, p53 binds to the TATA-binding protein-TATA complex, J. Biol. Chem. 268 (1993) 13062–13067.

[39] J. Liu, L. Grogan, M.M. Nau, C.J. Allegra, E. Chu, J.J. Wright, Physicalinteraction between p53 and primary response gene Egr-1, Int. J. Oncol. 18(2001) 863–870.

[40] K. Van Orden, H.A. Giebler, I. Lemasson, M. Gonzales, J.K. Nyborg,Binding of p53 to the KIX domain of CREB binding protein. Apotential link to human T-cell leukemia virus, type I-associated leuke-mogenesis, J. Biol. Chem. 274 (1999) 26321–26328.

[41] D. Schubeler, D.M. MacAlpine, D. Scalzo, C. Wirbelauer, C. Kooperberg,F. van Leeuwen, D.E. Gottschling, L.P. O'Neill, B.M. Turner, J. Delrow,S.P. Bell, M. Groudine, The histone modification pattern of active genesrevealed through genome-wide chromatin analysis of a higher eukaryote,Genes Dev. 18 (2004) 1263–1271.

[42] H. Wang, R. Cao, L. Xia, H. Erdjument-Bromage, C. Borchers, P. Tempst,Y. Zhang, Purification and functional characterization of a histone H3-lysine 4-specific methyltransferase, Mol. Cell 8 (2001) 1207–1217.

[43] S. Chuikov, J.K. Kurash, J.R. Wilson, B. Xiao, N. Justin, G.S. Ivanov, K.McKinney, P. Tempst, C. Prives, S.J. Gamblin, N.A. Barlev, D. Reinberg,Regulation of p53 activity through lysine methylation, Nature 432 (2004)353–360.

[44] X. Xie, J. Lu, E.J. Kulbokas, T.R. Golub, V. Mootha, K. Lindblad-Toh,E.S. Lander, M. Kellis, Systematic discovery of regulatory motifs inhuman promoters and 3′ UTRs by comparison of several mammals,Nature 434 (2005) 338–345.

[45] K. Sarma, D. Reinberg, Histone variants meet their match, Nat. Rev. Mol.Cell. Biol. 6 (2005) 139–149.

[46] K. Ahmad, S. Henikoff, The histone variant H3.3 marks active chromatin byreplication-independent nucleosome assembly, Mol. Cell 9 (2002)1191–1200.

[47] E. McKittrick, P.R. Gafken, K. Ahmad, S. Henikoff, Histone H3.3 isenriched in covalent modifications associated with active chromatin, Proc.Natl. Acad. Sci. USA 101 (2004) 1525–1530.

[48] Y. Hippo, K. Watanabe, A. Watanabe, Y. Midorikawa, S. Yamamoto, S.Ihara, S. Tokita, H. Iwanari, Y. Ito, K. Nakano, J. Nezu, H. Tsunoda, T.Yoshino, I. Ohizumi, M. Tsuchiya, S. Ohnishi, M. Makuuchi, T.

188 K. Kaneshiro et al. / Genomics 89 (2007) 178–188

Hamakubo, T. Kodama, H. Aburatani, Identification of soluble NH2-terminal fragment of glypican-3 as a serological marker for early-stagehepatocellular carcinoma, Cancer Res. 64 (2004) 2418–2423.

[49] B.M. Bolstad, R.A. Irizarry, M. Astrand, T.P. Speed, A comparison ofnormalization methods for high density oligonucleotide array data basedon variance and bias, Bioinformatics 19 (2003) 185–193.

[50] M.B. Ersen, P.T. Spellman, P.O. Brown, D. Botstein, Cluster analysis anddisplay of genome-wide expression paterns, Proc. Natl. Acad. Sci. USA 95(1998) 14863–14868.

[51] P. Kapranov, S.E. Cawley, J. Drenkow, S. Bekiranov, R.L. Strausberg, S.P.Fodor, T.R. Gingeras, Large-scale transcriptional activity in chromosomes21 and 22, Science 296 (2002) 916–919.