(2001) 20, 3156 ± 3165 ã 2001 Publishing Group All rights reserved 0950 ± 9232/00 $15.00 www.nature.com/onc

DNA , chromatin inheritance, and

Michael R Rountree*,1, Kurtis E Bachman1, James G Herman1 and Stephen B Baylin1

1The Johns Hopkins Oncology Center, Tumor Biology Laboratory, The Johns Hopkins University School of Medicine, Baltimore, Maryland, MD 21231, USA

Cancer is a process driven by the accumulation of sion capacity without DNA sequence alterations, are abnormalities in function. While many of these also central to tumor progression. The epigenetic changes are genetic, epigenetically mediated changes in control of gene function involves the formation of are being increasingly appreciated. This chromatin that modulates gene . The latter process emphasizes the need to understand two key study of this level of gene control has blossomed over components of heritable, but reversible, modulation of the last decade and is a critical link to our under- gene function that are closely tied to one standing of the neoplastic process. another ± formation of chromatin which modulates transcription and establishing patterns of DNA methyla- tion. The link lies ®rst in the recruitment to methylated Chromatin and transcription of a family of methyl-CpG binding domain (MBDs), which are direct transcriptional Eukaryotic cells must accomplish the daunting task of and can complex with transcriptional core- packaging an enormous amount of DNA into their pressors including deacetylases (HDACs). Ad- nucleus, while ensuring the proper expression of a ditionally, the proteins that catalyze DNA methylation, subset of and the silencing of other regions of the DNA (DNMTs), also directly repress . Higher are especially burdened in transcription and associate with HDACs. Regulation of this task due to the large amount of repetitive elements these above chromatin-DNA methylation interactions as (Alu, LINEs, SINEs, etc.) that are scattered through- a function of DNA replication timing is emerging as a out their . In general, the genome is key event in the inheritance of transcriptionally repressed compartmentalized into transcriptionally competent domains of the genome. Importantly, synergy between euchromatin and transcriptionally incompetent hetero- HDAC activity and DNA methylation is operative for a chromatin. Cells accomplish this feat by packaging key epigenetic abnormality in cancer cells, transcrip- DNA into chromatin, whose basic unit is the tional silencing of tumor suppressor genes. This change with *146 bp of DNA wrapped around has now been recognized for genes that are essential for it. The proper transcriptional status is then achieved normal regulation of virtually every major cell function through the interplay of complexes that including cell growth, di€erentiation, apoptosis, DNA associate with, manipulate, and epigenetically modify repair, and cell ± cell, cell ± substratum interaction. this basic unit to either foster or inhibit transcription. Understanding the molecular determinants of both These epigenetic modi®cations, employing combina- normal and abnormal patterns of chromatin formation tions of factors, allow the cell to modulate the and DNA methylation thus holds great promise for our transcriptional activity of given gene promoters. In understanding of cancer and for means to better this manner, a rheostat of transcriptional activity diagnose, prevent, and treat this disease. Oncogene provides a range of gene function from high-level (2001) 20, 3156 ± 3165. expression to complete silencing, as well as setting up gene expression events to react quickly to environ- Keywords: DNA methylation; histone acetylation; mental stimuli. chromatin inheritance; cancer Two epigenetic modi®cations have emerged as critical layers of regulation that participate in this transcriptional rheostat (Figure 1). The ®rst, histone Introduction acetylation appears to be used by all eukaryotes as one layer of transcriptional control (Grunstein, 1997). The The hallmark of cancer is a progressive appearance of acetylation of the amino-terminal tails of H3 malignant cell behavior that is triggered by the and H4 by histone acetyltransferases (HATs) creates an of altered gene function. Many of the gene accessible chromatin con®guration that facilitates changes stem from genetic abnormalities that disrupt transcriptional activity. Removal of these acetyl groups coding regions. However, it is becoming clear that by histone deacetylases (HDACs) facilitates chromatin epigenetic events, or heritable changes in gene expres- compaction that is detrimental to transcription. The cell uses these HATs and HDACs as coactivators and respectively to modulate promoter activ- *Correspondence: MR Rountree ity. DNA methylation, chromatin inheritance and cancer MR Rountree et al 3157 Transcriptionally Transcriptionally Transcriptionally Incompetent Inducible Competent Chromatin Euchromatin

Histone Acetylation DNA Methylation

Inactive X- Environmentally ActiveGenes Silenced Imprinted Genes Responsive Genes Alu, LINEs, SINEs Developmentally Pericentromeric Repeats Responsive Genes

Figure 1 The transcriptional rheostat. DNA methylation and histone acetylation help to establish chromatin states that either foster or inhibit transcription. The shaded bars represent the inverse correlation between these two epigenetic modi®cations. While a cell will use histone acetylation status to modulate gene expression, DNA methylation primarily serves as a transcriptionally repressive `lock'

The second epigenetic modi®cation, DNA methyla- recombination as well (Bird, 1995; Yoder et al., 1997a; tion, has a long-standing relationship with gene Walsh et al., 1998). The methylation of cytosines makes inactivity and has been implicated as a critical layer of them more susceptible to , which has control for enhancing transcriptional silencing (Bird, reduced the overall frequency of CpGs in the bulk 1992; Eden and Cedar, 1994). However, the lack of such genome (Bird, 1980). Conversely, the promoter regions an epigenetic modi®cation in some favorite model of many genes generally remain free of methylation and organisms (e.g. S. cerevisiae, C. elegans) has long raised therefore less susceptible to deamination. Thus, these questions about the importance of DNA methylation promoters have retained the expected frequency of CpG for gene control in eukaryotes. Interestingly, while dinucleotides and are referred to as `CpG islands'. was long thought not to have DNA Studies of the phosphoribosyltransferase (Aprt) methylation, recent reports have identi®ed genes with locus have identi®ed a potential mechanism for how homology to the vertebrate DNA methylation machin- CpG islands remain free of methylation in embryonic ery (Tweedie et al., 1999; Hung et al., 1999; Roder et al., cells. Three groups demonstrated that the presence of 2000; Lyko et al., 2000a). In addition, trace amounts of Sp1 binding sites and presumably the trans-acting factor methylation have been detected in this that binds to these sites protects the CpG island of the organism (Gowher et al., 2000; Lyko et al., 2000b). Aprt gene from methylation (Macleod et al., 1994; In mammals, genomic methylation patterns are Brandeis et al., 1994; Mummaneni et al., 1995, 1998). established during embryogenesis through the interplay Importantly, while most CpG islands remain free of of at least three DNA methyltransferases (Dnmt1, methylation, those associated with transcriptionally Dnmt3a, and Dnmt3b; Figure 2) and presumably the silenced genes on the inactive X-chromosome and factors that associate with these enzymes to target and silenced alleles of imprinted genes are densely methylated regulate their enzymatic activity. All three enzymes are (Brandeis et al., 1993). essential for proper murine development (Li et al., 1992; Okano et al., 1999). It has been proposed that the newly identi®ed Dnmt3a and Dnmt3b enzymes act primarily as DNA methylation and histone deacetylation: partners in de novo methyltransferases to establish methylation transcriptional repression patterns during embryogenesis (Okano et al., 1999). In contrast, Dnmt1 is thought to maintain these methyla- DNA methylation appears capable of directly prevent- tion patterns during DNA replication. However, the ing the binding of some transcription factors to their strict designation of these enzymes as either de novo or DNA binding sites (Tate and Bird, 1993). However, the maintenance methyltransferases is not yet de®nitive. majority of inhibition of transcription in association The vast majority of CpG dinucleotides (*70%) in with DNA methylation appears to occur through mammalian genomes are methylated and reside within complex indirect mechanisms involving changes in repetitive elements (Yoder et al., 1997a). This methyla- chromatin formation. Initial experiments demonstrated tion is a candidate mechanism for helping to transcrip- that in vitro methylated genes transfected into cells tionally silence these elements and thus, has been were transcriptionally inactive and assumed a chroma- proposed to serve as a host defense mechanism to inhibit tin conformation devoid of sensitive sites transposition and perhaps to subdue homologous present in active genes (Keshet et al., 1986). However,

Oncogene DNA methylation, chromatin inheritance and cancer MR Rountree et al 3158

DMAP1 P (CXXCXXC)2

Figure 2 Maps of DNMT1, DNMT3A and DNMT3B. C-terminal black regions represent the conserved domains in each protein. `DMAP1' refers to the binding region for this recently de®ned transcriptional co- protein. `P' refers to the PCNA binding site. CXXCXXC2 refers to the cysteine-rich Zn-binding motif of DNMT1 also found in the human trithorax protein and methyl-CpG binding domain protein-1 (MBD1). Regions of DNMT1 involved in Rb and HDAC-associated repression are shown, as well as the region that targets the protein to DNA replication sites. The grey box in DNMT3A and DNMT3B refers to the cysteine-rich, ATRX-like PHD ®nger that mediates transcriptional repression associated with HDAC activity

subsequent transfection assays demonstrated that in (MBD1, MBD2, MBD3, MeCP2) repress transcription vitro methylated genes are initially transcriptionally partially through the action of HDACs. Such an active, but after a few hours the DNA is packaged into association was ®rst shown for MeCP2, which a chromatin con®guration that inhibits transcription complexes with HDAC1 and HDAC2 through a direct (Buschhausen et al., 1987; Kass et al., 1997). These interaction with Sin3A (Nan et al., 1998; Jones et al., results indicate that DNA methylation itself does not 1998). Subsequent work demonstrated that MBD3 interfere with transcription, but rather marks the DNA and/or MBD2 are found in the Mi2/NuRD (nucleo- for establishment of a transcriptionally incompetent some remodeling ) complex that not chromatin state. However, whether DNA methylation only contains HDAC1 and HDAC2 but also Mi-2, a initiates the process in vivo or if methylation is a member of the SWI2/SNF2 family of ATP-dependent secondary e€ect following the formation of transcrip- proteins (Zhang et al., 1999; tionally inactive chromatin is not clear in most cases. Wade et al., 1999). MBD2, along with HDAC1 and For X-chromosome inactivation and the developmental HDAC2, is also a member of the MeCP1 complex, switch in globin gene expression, methylation appears which was ®rst identi®ed as a methyl-binding activity to be a secondary event but is then required for the that required at least 12 symmetrically methylated proper maintenance of the inactive state (Lock et al., CpGs for binding in vitro (Ng et al., 1999). The speci®c 1987; Enver et al., 1988; Sado et al., 2000). HDAC(s) that associate with MBD1 or any other The mechanism(s) behind how methylation mediates protein partners have not been reported to date (Ng et a transcriptionally incompetent chromatin organization al., 2000). The direct linkage between the MBDs, are now starting to be understood (Bird and Wol€e, HDACs and chromatin remodeling machinery has 1999). The seminal ®ndings of and his provided a basis for understanding how DNA colleagues provided the central clue to this under- methylation can mediate a transcriptionally incompe- standing. They have identi®ed a group of proteins, the tent chromatin state. However, as detailed below recent methyl-CpG binding domain proteins (MBDs), which results found for the DNMTs suggest that these preferentially bind to methylated CpGs (Meehan et al., proteins may contribute to transcriptionally incompe- 1989; Lewis et al., 1992; Cross et al., 1997; Hendrich tent chromatin beyond transferring methyl groups to and Bird, 1998). Several of these MBD proteins cytosines.

Oncogene DNA methylation, chromatin inheritance and cancer MR Rountree et al 3159 DNMTs: new partners and roles ± tightening the regulatory and/or targeting capacity. The N-termini of silencing link both DNMT3A and DNMT3B contain a cysteine-rich, PHD-like ®nger domain that most resembles a domain Recent developments suggest that the DNMTs possess found in the ATRX protein (Figure 2) (Xie et al., functions in addition to or in cooperation with their 1999). However, this cysteine-rich ATRX-like domain methyltransferase enzymatic activity. All three DNMT does not resemble the cystein-rich CXXC repeats enzymes have C-terminal methytransferase catalytic found in the N-terminus of DNMT1. The portion of domains responsible for transferring methyl groups the DNMT3A and DNMT3B that is N-terminal to the from S-adenosylmethionine to cytosines in CpG ATRX-like domain di€ers between the two, and dinucleotides, and longer non-catalytic N-terminal therefore may provide speci®city for the proteins. portions that presumably function in a regulatory Similar to DNMT1, we have found that the N-terminal capacity (Figure 2) (Bestor, 2000). portions of Dnmt3a and Dnmt3b repress transcription The N-terminus of DNMT1 contains regions in association with HDAC activity (KE Bachman, MR responsible for targeting the enzyme to replication foci Rountree and SB Baylin; manuscript submitted). This (Leonhardt et al., 1992; Chuang et al., 1997; Liu et al., repression and HDAC association is mediated primar- 1998), as well as for discriminating between unmethy- ily through the ATRX-like domain. lated and hemi-methylated DNA (Bestor, 1992). The association of virtually all of the methylation Indeed, this enzyme has a 10 ± 40-fold preference for machinery, DNMTs and MBDs, with HDACs pro- hemimethylated CpGs over unmethylated CpGs (Prad- vides a cooperative linkage in transcriptional silencing han et al., 1997, 1999; Yoder et al., 1997b). This between DNA methylation and histone deacetylation portion of DNMT1 also contains a cysteine-rich (Figure 3). The challenge now is to understand how (CXXC) Zn-binding motif, that resembles motifs found these complexes behave within a cell to establish and in MBD1 and the human Trithorax proteins (Cross et maintain a repressive state. There are a number of al., 1997). Recently, three groups have reported that questions to be answered. Do these complexes work in DNMT1 can repress transcription through its N- a sequence speci®c manner or is there redundancy in terminus, and thus independent of its methylating their actions or perhaps they function in di€erent cell functions (Fuks et al., 2000; Rountree et al., 2000; types? Apart for their anity for a symmetrically Robertson et al., 2000). This repression appears to methylated CpG, the MBDs have not been shown to result, in part, through an interaction with HDAC1 possess sequence speci®city. Likewise, the DNMTs and/or HDAC2. In addition, two of these groups have sequence speci®city for CpGs, but no speci®city identi®ed additional factors that complex with the N- for the context of this dinucleotide. Therefore, if these terminus of DNMT1 to perhaps regulate or target the complexes do have sequence speci®city it must be via transcriptional repressive and/or methyltransferase recruitment by another factor. Another possibility is activities of this enzyme. Robertson et al. (2000) found these complexes work in a coordinated manner to an assortment of proteins that copuri®ed with DNMT1 continuously insure a transcriptionally incompetent including (pRB) and E2F. state through the cell cycle. For example, as discussed DNMT1 binds to pRB and can be recruited to below, the DNMT1 complex appears to play a crucial promoters through and E2F/pRB complex to repress role in the maintenance of chromatin during S-phase. transcription in transient transcription assays. This transcriptional repression did not appear to involve methylation, as methylation was not detected within The inheritance of epigenetic states the promoter. Using the yeast two-hybrid assay, our group identi®ed a novel co-repressor protein, DNMT1 Every turn of the cell cycle requires a cell to duplicate Associated Protein 1 (DMAP1), which binds to the its genome and to recapitulate the chromatin con®g- ®rst 120 amino acids of DNMT1 through a coil-coiled uration that existed prior to the passage of the domain interaction (Rountree et al., 2000). In turn, replication fork in order to retain proper genomic DMAP1 binds Tumor Susceptibility Gene 101 (TSG101), a candidate that is a potent transcriptional co-repressor. Interestingly, through alternative promoter usage the Dnmt1 protein present in oocytes and pre-implantation embryos lack the DMAP1 binding region (Mertineit et al., 1998). In addition, an alternative isoform of Dnmt1 that lacks the DMAP1 binding region is expressed during muscle di€erentiation (Aguirre-Arteta et al., 2000). Further studies will be required to understand the signi®cance of the DNMT1/DMAP1 interaction and the function of these di€erent Dnmt1 isoforms. The N-termini of DNMT3A and DNMT3B remain Figure 3 The connection between DNA methylation and histone relatively uncharacterized in regards to function, but deacetylation. All three of the DNMTs and four of the MBDs presumably like the N-terminus of DNMT1 serve in a repress transcription through an association with HDAC activity

Oncogene DNA methylation, chromatin inheritance and cancer MR Rountree et al 3160 transcriptional properties. This chromatin assembly time a plausible mechanism for how newly synthesized entails the proper inheritance of the genomic methyla- become deacetylated following DNA re- tion pattern, reassembly and positioning of nucleo- plication. Following maintenance of the repressed somes, reestablishment of histone modi®cation status, epigenetic state of chromatin after replication, the and ®nally reassembly of other chromatin factors that MBD/HDAC complexes could then reassociate with ensure the proper transcriptional status of genes (Kass the chromatin and perpetuate the transcriptionally and Wol€e, 1996; Krude, 1999). The cell seems to incompetent state through the rest of the cell cycle. control the proper inheritance of chromatin states by Besides the emerging role of DNMT1 as an controlling the spatial organization and timing of important factor in the propagation of epigenetic states replication of di€erent portions of its genome during S-phase, the presence of this protein may be (Leonhardt et al., 2000). In general, transcriptionally required for ecient and timely replication of DNA. active euchromatin portions of the genome are Inhibition of DNMT1 appears to interfere with DNA typically replicated early in S-phase, while transcrip- replication by blocking the activity of replication tionally inactive heterochromatin tends to replicate origins (Knox et al., 2000). The mechanism for this later in S-phase. interference is not clear but suggests that DNMT1 The proliferating cell nuclear antigen (PCNA), which plays an important role in replication initiation. provides processivity for DNA polymerases during Recently, pRB and E2F were shown to localize to DNA replication, is increasingly recognized as playing the earliest DNA replication foci when a cell starts into a central role in the inheritance of chromatin states S-phase (Kennedy et al., 2000). Perhaps the interaction through the recruitment of key protein complexes to between pRB and DNMT1 occurs at this stage and is the DNA. Recent evidence demonstrating that PCNA required for the proper initiation and timing of S- is critical for the inheritance of heterochromatin regions phase. in yeast supports this notion (Zhang et al., 2000). Shibahara and Stillman (1999) demonstrated that PCNA is left behind on the DNA following replication Disruption of epigenetic states in cancer and speculate that this may provide a `mark' for the epigenetic inheritance of chromatin states. They further Cancer is a disease that results from the accumulation speculate, that this PCNA mark provides a platform to and interplay of genetic and epigenetic changes. The which other proteins bind and reestablish the proper disruption of normal methylation patterns, with both epigenetic states following replication (Shibahara and hypomethylation and hypermethylation events occur- Stillman, 1999). Evidence for such a model comes from ring, is a hallmark of tumorigenesis whose signi®cance two key protein complexes that interact with PCNA is only recently coming to be appreciated (Baylin and and play important roles in chromatin formation and Herman, 2000; Robertson and Wol€e, 2000). epigenetic propagation. Although, we will focus primarily on the hypermethy- DNMT1 is brought to the replication machinery lation changes, the loss of methylation during tumor- through a direct interaction with PCNA (Chuang et al., igenesis may also be extremely important. To date, 1997), and serves to maintain DNA methylation however, the signi®cance of hypomethylation in patterns on the daughter strands following passage of tumorigenesis, which primarily a€ects the normally the replication fork (Leonhardt et al., 1992; Bestor, methylated parasitic and repetitive DNA sequences, 2000). Work conducted in our laboratory has provided has remained elusive. Several important possible evidence that DNMT1 forms a complex at these consequences for this hypomethylation are being replication foci whose constituents change during the explored. One concerns the potential that loss of course of S-phase (Rountree et al., 2000). We reported methylation in could induce unwanted expres- that DMAP1 is recruited through a direct interaction sion of the normally repressed transposons scattered with DNMT1 to replication foci throughout S-phase. throughout the genome leading to deleterious transpo- However, we found that HDAC2 colocalizes with the sition events (Yoder et al., 1997a). There is evidence DNMT1/DMAP1 complex only during late stages of that cancer related hypomethylation results in in- S-phase. This is potentially extremely important since creased expression of some of these elements, but this is when transcriptionally repressive heterochroma- whether this increased expression is harmful to the tin is replicated. genome either directly or through transposition of the The other protein complex that binds to PCNA is elements is not clear (Florl et al., 1999). chromatin assembly factor 1 (CAF-1), which assembles Another area of research is focused on whether newly synthesized histones H3 and H4 onto the hypomethylation leads to genomic instability. It has replicated DNA (Shibahara and Stillman, 1999). Newly been proposed that DNA methylation, in addition to synthesized histones are assembled on the DNA with its repressive e€ects on transcription, may suppress lysine residues 5 and 12 in the acetylated state (Sobel et homologous recombination between repetitive elements al., 1995). The deacetylation of these histones is known to in the genome (Colot and Rossignol, 1999). Work with be an essential step in the maturation of heterochromatin the Ascobolus immersus, provides evidence that (Annunziato and Seale, 1983). Therefore, the positioning DNA methylation can suppress recombination (Colot of HDAC2 with the DNMT1/DMAP1 complex late in S- et al., 1996). Increased genomic instability seen in the phase is signi®cant because it demonstrates for the ®rst hypomethylated genome of embryonic stem cells

Oncogene DNA methylation, chromatin inheritance and cancer MR Rountree et al 3161 lacking Dnmt1 provides evidence for a genome cells results in silencing? The second, is how aberrant protective capacity for methylation (Chen et al., CpG island methylation arises during tumorigenesis? 1998). Furthermore, the pericentromeric areas of many are particularly heavily methylated in the constitutive heterochromatin and loss of this pericen- Mechanisms of promoter silencing by CpG island tromeric methylation both in a genetic disorder, hypermethylation immunode®ciency-centromeric instability-facial anoma- lies (ICF) syndrome, and through treatment with the The underlying mechanism(s) by which aberrant CpG demethylating agent 5-aza-2'- (5aza-dC) island methylation contributes to transcriptional re- is associated with increased chromosome translocations pression in tumor cells will almost certainly involve the (Ji et al., 1997; Hernandez et al., 1997; Tuck-Muller et MBD/HDAC and DNMT/HDAC complexes discussed al., 2000). In ICF syndrome, patients harbor mutations above in this review. Evidence for the involvement of in DNMT3B (Keshet et al., 1986; Xu et al., 1999; both DNA methylation and histone deacetylation in Hansen et al., 1999; Okano et al., 1999). Similar silencing tumor suppressor genes in cancer come from patterns of pericentromeric hypomethylation are also studies from our group using inhibitors against HDAC seen in multiple types of human cancer (Narayan et al., and DNMT activity (Cameron et al., 1999). Cameron 1998; Qu et al., 1999). et al. (1999) demonstrated a synergistic e€ect of an HDAC inhibitor (trichostatin A) and DNMT inhibitor (5-aza-dC) on reactivation of the aberrantly methylated Promoter silencing by CpG island hypermethylation genes , TIMP3, and MLH1 in a colon cancer cell line. In this study, treatment with trichostatin A alone The most emphasized alteration of DNA methylation did not reactivate the genes, rather the cells ®rst had to in cancer is the aberrant hypermethylation of CpG be exposed to 5-aza-dC suggesting that the DNA islands surrounding gene promoter regions (Baylin and methylation is dominant. This combined treatment Herman, 2000). The critical issue is that these must now be considered not only in respect to methylation changes are associated with transcriptional inhibiting the MBD/HDAC complexes, but now also silencing of the involved genes. To date, aberrant CpG more directly on the DNMT/HDAC complexes. island methylation has been shown to be associated The critical evidence for the involvement of methyla- with the silencing of a number of classic tumor tion and protein complexes attracted to this modi®ca- suppressor genes. In fact, this aberrant silencing can tion for the transcriptional silencing of hyper- disrupt the expression of genes involved in the methylated genes in cancer lies in demonstrating which fundamental pathways that lead to cancer (Table 1). speci®c complexes are responsible for the silencing of Besides these classic tumor suppressor genes, there is selected gene promoters. This identi®cation may lead to an ever increasing list of genes whose corresponding more elegant and targeted therapeutic approaches to CpG island shows aberrant methylation in cancer achieve reactivation of speci®c genes. (Herman, 1999). The signi®cance of inactivating some of these genes for tumor progression is not clear, but such events may simply re¯ect a methylation abnorm- Establishment of aberrant CpG island methylation ality of cancer cells. While the sites of hypermethyla- during tumorigenesis tion in cancer already have translational implications for molecular approaches to tumor detection and for The mechanism(s) leading to the development of therapeutic possibilities (Baylin and Herman, 2000), we altered genomic methylation patterns in cancer is not will stress here two major mechanistic questions understood. It is not clear whether global hypomethy- currently at the forefront of this ®eld. The ®rst, is lation and regional CpG island hypermethylation are how this aberrant CpG island methylation in tumor mechanistically linked. Another unknown is whether aberrant CpG island methylation is the initiating event in or if the promoter is silenced by Table 1 Fundamental pathways altered by DNA methylation in another mechanism and methylation is a secondary cancer `lock-down' event. There are probably multiple ways Pathway affected Genes silenced by CpG island methylation that CpG islands become hypermethylated and silenced Cell cycle control Rb, p16(INK4a), p15, p14(ARF), p73 during tumorigenesis, but let us consider a model(s) DNA damage repair MLH1, O6MGMT, GSTp, BRCA1 (Figure 4) and evidence implicating speci®c events that Inhibiting apoptosis DAP-kinase, Caspase-8, TMS1 are potentially involved. Invasion tumor E-cadherin, VHL, APC, LKB1, TIMP3, architecture Thrombospondin1 Growth factor ER, RARb, Androgen Receptor, Endothelin Loss of a CpG island protection factor response B Receptor, RASSF1A As mentioned previously, in embryonic cells the CpG Methylation of most of these genes recently reviewed in (Herman, island of the Aprt gene remains methylation free through 1999; Baylin and Herman, 2000). References for the genes not covered by these reviews include; (Damman et al., 2000; Teitz et al., the binding of a yet unidenti®ed protective trans-acting 2000; Zheng et al., 2000; Bianco et al., 2000; Conway et al., 2000; factor to Sp1 sites near the edge of the island (Macleod et Esteller et al., 2000a,b,c,d) al., 1994; Brandeis et al., 1994; Mummaneni et al., 1995).

Oncogene DNA methylation, chromatin inheritance and cancer MR Rountree et al 3162

Figure 4 Model for the Epigenetic Silencing of CpG Islands in Cancer. A transcriptionally active CpG island promoter is depicted with positioned nucleosomes, consisting of acetylated (Ac) histone tails, and unmethylated CpG residues (white circles on the black DNA strand). At each end of the CpG island a putative trans-acting factor (Stop signs) protects the CpG island from encroaching methylation. In this model, one or more of the DNMT/HDAC (HD) complexes is proposed to initiate the epigenetic silencing of the CpG island. (1) DNMT/HDAC complexes gain access to the CpG island and begin to transcriptionally silence (X) the region by methylating CpG sites (grey circles) and deacetylating the histones. Events leading to this possibly include one but probably more of the following: loss of the trans-acting protective factor, elevation in DNMT(s) levels, and/or disregulation of the DNMT(s) causing mistargeting of the DNMT/HDAC complex either through loss of a interacting factor (e.g. Rb or DMAP1 for DNMT1) or gain of a targeting factor through illegitimate expression (e.g. embryonic factor). (2) This initial silencing may then predispose this CpG island to replicate later in S-phase where the DNMT1/HDAC2/DMAP1 heterochromatin inheritance machinery will further facilitate this conversion. The DNMT1/HDAC2/DMAP1 complex is bound to PCNA near the replication fork with only one strand of newly synthesized chromatin is illustrated for simplicity (dashed line represents the other). (3) Alternatively, the potential deregulation of replication during tumorigenesis (perhaps through loss of the Rb pathway) exposes the transcriptionally active CpG island to the DNMT1 complex either through slippage to a late S-phase replication timing or deregulation of the machinery leading to methylation and histone deacetylation events that silence transcription. (4) The progressive methylation of the CpG island will then allow MBD complexes (including HDACs and chromatin remodeling machinery `CR') to bind and assist in the progression to a late replicating hypermethylated/hypoacetylated transcriptionally incompetent state

Thus, one realistic model that has been proposed to cells. This loss and the resulting diminished transcrip- account for aberrant CpG island methylation in cancer is tional capability perhaps in concert with some of the the breakdown in the protection of these islands through changes discussed below may make this CpG island loss of such a trans-acting factor. However, direct more susceptible to becoming methylated. evidence for this type of mechanism in cancer has been lacking. Cancer cell lines harboring a hypermethylated DNMT over-expression in cancer E-cadherin (E-cad) promoter express an exogenously introduced unmethylated E-cad promoter construct less There is considerable evidence indicating an up eciently than cancer cell lines with its endogenous E- regulation of DNMT1 in cancer (Belinsky et al., 1996; cad promoter hypomethylated (Gra€ et al., 1995). This Baylin et al., 1998). However, other reports indicate that result suggests that trans-acting factors needed for full there is no correlation between CpG island hypermethy- expression of the E-cad promoter are lost in some cancer lation and DNMT1 levels (Nass et al., 1999; Eads et al.,

Oncogene DNA methylation, chromatin inheritance and cancer MR Rountree et al 3163 1999). Experimental evidence indicates that forced over- expression experiments described above were conducted expression of the murine Dnmt1 gene in NIH3T3 cells with a form of DNMT1 which lacks the ®rst 118 amino results in cellular transformation (Wu et al., 1993). In acid found normally in somatic cells (Wu et al., 1993; human ®broblasts, sustained over-expression of Vertino et al., 1996). This missing portion contains the DNMT1 leads to the processive time-dependent DMAP1 binding region and, therefore, the exogenously hypermethylation of a number of CpG islands (Vertino expressed DNMT1 could not have interacted with et al., 1996). The DNMT1 gene is upregulated in DMAP1 and this fact must now be considered as response to fos over-expression and appears to play a contributing to these experimental results. role in the fos induced cell transformation (Bakin and Curran, 1999). Conversely, reduction of DNMT1 levels Disruption of Replication Timing appears to have protective e€ects. Mice predisposed to colonic polyp formation (Min mice) develop fewer As described early in our review, duplication of the polyps in a Dnmt1 heterozygous background and also genome occurs through spatial and temporal patterns when treated with 5aza-dC (Laird and Jaenisch, 1996). during the course of S-phase progression. The DNMT1 Reduction of DNMT1 through an antisense approach complex that is modulated di€erentially during di€erent blocked tumorigenesis (MacLeod and Szyf, 1995; portions of S-phase progression is emerging as an Ramchandani et al., 1997). important component in the perpetuation of chromatin Interestingly, deletion of the DNMT1 gene in a states during this process (Rountree et al., 2000). Thus, colon cancer cell line (HCT116), while resulting in we should consider cancer related methylation abnorm- slower growth, diminished genomic methylation levels alities in this context. The global hypomethylation seen only modestly (*20%) (Rhee et al., 2000). In in many cancers begs for a mechanism that is global in particular, aberrant CpG island methylation was nature. Could S-phase abnormalities a€ect methylation retained suggesting that another methyltransferase(s) patterns by altering the replication timing of sequences? maintains this methylation. Like DNMT1, the more Alteration in the spatial and temporal patterns of recently described DNMT3A and DNMT3B enzymes replication could expose sequences to replication and also appear to be modestly over-expressed in cancer chromatin formation machinery with which they do not and therefore, the balance of all three enzymes and normally interact. For example, certain regions of the their accumulative and coordinated e€ects must be genome might fail to become normally methylated and studied (Xie et al., 1999; Robertson et al., 1999). transcriptionally repressed because of replication timing errors that prevent the proper targeting of DNA methylation and chromatin formation complexes. Disruption or mistargeting of a DNMT complex Similarly, and critical for the emphasis of this review, Another potential factor, perhaps in conjunction with CpG islands might become aberrantly methylated via the modest increase in DNMT levels, is the deregulation replication timing errors that shift these regions to late or mistargeting of the enzymatic or repression proper- S-phase and allow exposure to DNA methylation and ties of these enzymes. The deregulation could be due to chromatin formation machinery that promotes tran- the loss of one or more of the factors that complex with scriptionally incompetent chromatin formation. the DNMTs (e.g. pRb, DMAP1, TSG101). The cell To form credible hypotheses regarding the above cycle control Rb pathway is disrupted in almost every replication events, we must ®rst ask whether cancer cancer; therefore, we must consider the implications of cells have the types of S-phase abnormalities that may the disruption of the DNMT1/pRB interaction on the be involved. Kennedy et al. (2000) recently demon- methylation abnormalities that arise in cancer. In strated that the earliest observed replication foci in addition, the Dmap1 gene maps to a methylation normal cells surround the nucleolus and contain pRB modi®er locus (MEMO1) on the short arm of and its family members as well as E2F. In addition, (MR Rountree, KE Bachman and SB recent studies indicate that a pRB-SWI/SNF interac- Baylin, unpublished results) in a region of frequent tion is necessary for proper S-phase progression LOH in a number of cancers. LOH of the MEMO1 (Harbour and Dean, 2000a,b). Furthermore, a very locus correlates with methylation abnormalities at the signi®cant observation made by Kennedy et al. (2000) class I major histocompatibility complex gene cluster in was that these earliest replication foci are not present neuroblastoma cell lines (Cheng et al., 1996a,b). We in immortalized cells. Thus, the immortilization have found additional methylation abnormalities, both process, perhaps through altered targeting of pRB, hypo- and hypermethylation events, associated with may result in changes in the spatial organization of Neuroblastoma cell lines with LOH of this region (MR replication. It is likely that cancer cells have similar Rountree, KE Bachman and SB Baylin, unpublished defects, if not even more severe ones. Evidence for results). It is not clear whether DMAP1 is the MEMO1 replication timing errors in cancer comes from studies locus, but it has to be considered the top candidate. demonstrating a more asynchronous replication timing Many other cancers have LOH of this region and it will for homologous loci in malignant cells compared to be interesting to see if certain methylation abnormalities their normal counterparts (Amiel et al., 1998a,b, 1999). correlate with this loss. In addition, because additional Certainly, further work will be required to test whether 5' coding sequences were only recently identi®ed (Yoder replication abnormalities can initiate and propagate et al., 1996; Gaudet et al., 1998), the DNMT1 over- methylation abnormalities during tumorigenesis. How-

Oncogene DNA methylation, chromatin inheritance and cancer MR Rountree et al 3164 ever, this area seems a rich one to study for potentially methylation, is a critical aspect of maintaining normal understanding the genesis of abnormal DNA methyla- domains of transcriptional repression throughout the tion and chromatin patterns in cancer cells. genome of mammalian cells. Critical to this process is our understanding of how these events are coordinated with timing of DNA replication and cell cycle events. Summary Alterations in patterns of both chromatin formation and DNA methylation are fundamental to abnormal- In this review, we have discussed the rapidly emerging ities of key gene expression in cancer. The potentially recognition that synergy between heritable formation reversible nature of these cancer changes makes an of chromatin, especially the role of histone deacetyla- ever-deepening understanding of the basic mechanisms tion in this process, and the establishment of DNA underlying them an important goal of future research.

References

Aguirre-Arteta AM, Grunewald I, Cardoso MC and Conway KE, McConnell BB, Bowring CE, Donald CD, Leonhardt H. (2000). Cell Growth Di€er., 11, 551 ± 559. Warren ST and Vertino PM. (2000). Cancer Res., 60, Amiel A, Kirgner I, Gaber E, Manor Y, Fejgin M and 6236 ± 6242. Lishner M. (1999). Cancer Genet. Cytogenet., 108, 32 ± 37. Cross SH, Meehan RR, Nan X and Bird A. (1997). Nat. Amiel A, Kolodizner T, Fishman A, Gaber E, Klein Z, Beyth Genet., 16, 256 ± 259. Y and Fejgin MD. (1998b). Cancer, 83, 1966 ± 1971. Dammann R, Li C, Yoon JH, Chin PL, Bates S and Pfeifer AmielA,LitmanovitchT,LishnerM,MorA,GaberE, GP. (2000). Nat. Genet., 25, 315 ± 319. Tangi I, Fejgin M and Avivi L. (1998a). Genes Chromo- Eads CA, Danenberg KD, Kawakami K, Saltz LB, somes. Cancer, 22, 225 ± 231. Danenberg PV and Laird PW. (1999). Cancer Res., 59, Annunziato AT and Seale RL. (1983). J. Biol. Chem., 258, 2302 ± 2306. 12675 ± 12684. Eden S and Cedar H. (1994). Curr. Opin. Genet. Dev., 4, Bakin AV and Curran T. (1999). Science, 283, 387 ± 390. 255 ± 259. Baylin SB and Herman JG. (2000). and Loss of Enver T, Zhang JW, Papayannopoulou T and Stamatoyan- Gene Function in Cancer. In DNA Alterations in Cancer, nopoulos G. (1988). Genes Dev., 2, 698 ± 706. Melanie Erlich, (ed) Natick, MA: Eaton Publishing, pp. Esteller M, Avizienyte E, Corn PG, Lothe RA, Baylin SB, 293 ± 309. Aaltonen LA and Herman JG. (2000d). Oncogene, 19, Baylin SB, Herman JG, Gra€ JR, Vertino PM and Issa JP. 164 ± 168. (1998). Adv. Cancer Res., 72, 141 ± 196. Esteller M, Silva JM, Dominguez G, Bonilla F, Matias-Guiu Belinsky SA, Nikula KJ, Baylin SB and Issa JP. (1996). Proc. X, Lerma E, Bussaglia E, Prat J, Harkes IC, Repasky EA, Natl. Acad. Sci. USA, 93, 4045 ± 4050. Gabrielson E, Schutte M, Baylin SB and Herman JG. Bestor TH. (1992). EMBO J., 11, 2611 ± 2617. (2000c). J. Natl. Cancer Inst., 92, 564 ± 569. Bestor TH. (2000). Hum. Mol. Genet., 9, 2395 ± 2402. EstellerM,SparksA,ToyotaM,Sanchez-CespedesM, Bianco T, Chenevix-Trench G, Walsh DC, Cooper JE and Capella G, Peinado MA, Gonzalez S, Tarafa G, Sidransky Dobrovic A. (2000). , 21, 147 ± 151. D, Meltzer SJ, Baylin SB and Herman JG. (2000b). Cancer Bird A. (1992). Cell, 70, 5±8. Res., 60, 4366 ± 4371. Bird AP. (1980). Nucleic Acids Res., 8, 1499 ± 1504. Esteller M, Tortola S, Toyota M, Capella G, Peinado MA, Bird AP. (1995). Trends. Genet., 11, 94 ± 100. Baylin SB and Herman JG. (2000a). Cancer Res., 60, 129 ± Bird AP and Wol€e AP. (1999). Cell, 99, 451 ± 454. 133. Brandeis M, Ariel M and Cedar H. (1993). Bioessays, 15, Florl AR, Lower R, Schmitz-Drager BJ and Schulz WA. 709 ± 713. (1999). Br. J. Cancer, 80, 1312 ± 1321. Brandeis M, Frank D, Keshet I, Siegfried Z, Mendelsohn M, Fuks F, Burgers WA, Brehm A, Hughes-Davies L and Nemes A, Temper V, Razin A and Cedar H. (1994). Kouzarides T. (2000). Nat. Genet., 24, 88 ± 91. Nature, 371, 435 ± 438. Gaudet F, Talbot D, Leonhardt H and Jaenisch R. (1998). J. Buschhausen G, Wittig B, Graessmann M and Graess- Biol. Chem., 273, 32725 ± 32729. mann A. (1987). Proc. Natl. Acad. Sci. USA, 84, 1177 ± Gowher H, Leismann O and Jeltsch A. (2000). EMBO J., 19, 1181. 6918 ± 6923. CameronEE,BachmanKE,MyohanenS,HermanJGand Gra€ JR, Herman JG, Lapidus RG, Chopra H, Xu R, Baylin SB. (1999). Nat. Genet., 21, 103 ± 107. Jarrard DF, Isaacs WB, Pitha PM, Davidson NE and Chen RZ, Pettersson U, Beard C, Jackson-Grusby L and Baylin SB. (1995). Cancer Res., 55, 5195 ± 5199. Jaenisch R. (1998). Nature, 395, 89 ± 93. Grunstein M. (1997). Nature, 389, 349 ± 352. Cheng NC, Beitsma M, Chan A, Op DC, Westerveld A, Hansen RS, Wijmenga C, Luo P, Stanek AM, Can®eld TK, Pronk J and Versteeg R. (1996a). Oncogene, 13, 1737 ± Weemaes CM and Gartler SM. (1999). Proc. Natl. Acad. 1744. Sci. USA, 96, 14412 ± 14417. Cheng NC, Chan AJ, Beitsma MM, Speleman F, Westerveld Harbour JW and Dean DC. (2000a). Curr. Opin. Cell Biol., A and Versteeg R. (1996b). Hum. Mol. Genet., 5, 309 ± 317. 12, 685 ± 689. ChuangLS,IanHI,KohTW,NgHH,XuGandLiBF. Harbour JW and Dean DC. (2000b). The Rb/E2F pathway: (1997). Science, 277, 1996 ± 2000. expanding roles and emerging paradigms. Genes Dev., 14, Colot V, Maloisel L and Rossignol JL. (1996). Cell, 86, 855 ± 2393 ± 2409. 864. Hendrich B and Bird A. (1998). Mol. Cell Biol., 18, 6538 ± Colot V and Rossignol JL. (1999). Bioessays, 21, 402 ± 411. 6547.

Oncogene DNA methylation, chromatin inheritance and cancer MR Rountree et al 3165 Herman JG. (1999). Semin. Cancer Biol., 9, 359 ± 367. Okano M, Bell DW, Haber DA and Li E. (1999). Cell, 99, Hernandez R, Frady A, Zhang XY, Varela M and Ehrlich M. 247 ± 257. (1997). Cytogenet. Cell Genet., 76, 196 ± 201. Pradhan S, Bacolla A, Wells RD and Roberts RJ. (1999). J. Hung MS, Karthikeyan N, Huang B, Koo HC, Kiger J and Biol. Chem., 274, 33002 ± 33010. Shen CJ. (1999). Proc. Natl. Acad. Sci. USA, 96, 11940 ± Pradhan S, Talbot D, Sha M, Benner J, Hornstra L, Li E, 11945. Jaenisch R and Roberts RJ. (1997). Nucleic. Acids. Res., Ji W, Hernandez R, Zhang XY, Qu GZ, Frady A, Varela M 25, 4666 ± 4673. and Ehrlich M. (1997). Mutat. Res., 379, 33 ± 41. Qu G, Dubeau L, Narayan A, Yu MC and Ehrlich M. (1999). Jones PL, Veenstra GJ, Wade PA, Vermaak D, Kass SU, Mutat. Res., 423, 91 ± 101. Landsberger N, Strouboulis J and Wol€e AP. (1998). Nat. Ramchandani S, MacLeod AR, Pinard M, von Hofe E and Genet., 19, 187 ± 191. Szyf M. (1997). Proc. Natl. Acad. Sci. USA, 94, 684 ± 689. Kass SU, Landsberger N and Wol€e AP. (1997). Curr. Biol., Rhee I, Jair KW, Yen RW, Lengauer C, Herman JG, Kinzler 7, 157 ± 165. KW, Vogelstein B, Baylin SB and Schuebel KE. (2000). Kass SU and Wol€e AP. (1996). Histones, histone modi®ca- Nature, 404, 1003 ± 1007. tions, and the inheritance of chromatin structure. In Robertson KD, Ait SA, Yokochi T, Wade PA, Jones PL and Epigenetic Mechanisms of Gene Regulation, Russo V.E.A., Wol€e AP. (2000). Nat. Genet., 25, 338 ± 342. Martienssen R.A. and Riggs A.D. (eds.) New York: Cold Robertson KD, Uzvolgyi E, Liang G, Talmadge C, Sumegi J, Spring Harbor Laboratory Press, pp. 529 ± 546. Gonzales FA and Jones PA. (1999). Nucleic. Acids. Res., Kennedy BK, Barbie DA, Classon M, Dyson N and Harlow 27, 2291 ± 2298. E. (2000). Genes Dev., 14, 2855 ± 2868. Robertson KD and Wol€e AP. (2000). Nat. Rev. Genet., 1, Keshet I, Lieman-Hurwitz J and Cedar H. (1986). Cell, 44, 11 ± 19. 535 ± 543. Roder K, Hung MS, Lee TL, Lin TY, Xiao H, Isobe KI, KnoxJD,AraujoFD,BigeyP,SlackAD,PriceGB,Zannis- Juang JL and Shen CJ. (2000). Mol. Cell Biol., 20, 7401 ± Hadjopoulos M and Szyf M. (2000). J. Biol. Chem., 275, 7409. 17986 ± 17990. Rountree MR, Bachman KE and Baylin SB. (2000). Nat. Krude T. (1999). Curr. Biol., 9, 394 ± 396. Genet., 25, 269 ± 277. Laird PW and Jaenisch R. (1996). Annu.Rev.Genet.,30, Sado T, Fenner MH, Tan SS, Tam P, Shioda T and Li E. 441 ± 464. (2000). Dev. Biol., 225, 294 ± 303. Leonhardt H, Page AW, Weier HU and Bestor TH. (1992). Shibahara K and Stillman B. (1999). Cell, 96, 575 ± 585. Cell, 71, 865 ± 873. Sobel RE, Cook RG, Perry CA, Annunziato AT and Allis LeonhardtH,RahnHP,WeinzierlP,SporbertA,CremerT, CD. (1995). Proc. Natl. Acad. Sci. USA, 92, 1237 ± 1241. Zink D and Cardoso MC. (2000). J. Cell Biol., 149, 271 ± Tate PH and Bird AP. (1993). Curr. Opin. Genet. Dev., 3, 280. 226 ± 231. Lewis JD, Meehan RR, Henzel WJ, Maurer-Fogy I, Teitz T, Wei T, Valentine MB, Vanin EF, Grenet J, Valentine Jeppesen P, Klein F and Bird A. (1992). Cell, 69, 905 ± 914. VA, Behm FG, Look AT, Lahti JM and Kidd VJ. (2000). Li E, Bestor TH and Jaenisch R. (1992). Cell, 69, 915 ± 926. Nat. Med., 6, 529 ± 535. Liu Y, Oakeley EJ, Sun L and Jost JP. (1998). Nucleic. Acids. Tuck-Muller CM, Narayan A, Tsien F, Smeets DF, Sawyer Res., 26, 1038 ± 1045. J, Fiala ES, Sohn OS and Ehrlich M. (2000). Cytogenet. Lock LF, Takagi N and Martin GR. (1987). Cell, 48, 39 ± 46. Cell Genet., 89, 121 ± 128. Lyko F, Ramsahoye BH and Jaenisch R. (2000b). Nature, Tweedie S, Ng HH, Barlow AL, Turner BM, Hendrich B and 408, 538 ± 540. Bird A. (1999). Nat. Genet., 23, 389 ± 390. Lyko F, Whittaker AJ, Orr-Weaver TL and Jaenisch R. Vertino PM, Yen RW, Gao J and Baylin SB. (1996). Mol. (2000a). Mech. Dev., 95, 215 ± 217. Cell Biol., 16, 4555 ± 4565. MacLeod AR and Szyf M. (1995). J. Biol. Chem., 270, 8037 ± WadePA,GegonneA,JonesPL,BallestarE,AubryFand 8043. Wol€e AP. (1999). Nat. Genet., 23, 62 ± 66. Macleod D, Charlton J, Mullins J and Bird AP. (1994). Genes Walsh CP, Chaillet JR and Bestor TH. (1998). Nat. Genet., Dev., 8, 2282 ± 2292. 20, 116 ± 117. Meehan RR, Lewis JD, McKay S, Kleiner EL and Bird AP. Wu J, Issa JP, Herman J, Bassett DEJ, Nelkin BD and Baylin (1989). Cell, 58, 499 ± 507. SB. (1993). Proc. Natl. Acad. Sci. USA, 90, 8891 ± 8895. MertineitC,YoderJA,TaketoT,LairdDW,TraslerJMand XieS,WangZ,OkanoM,NogamiM,LiY,HeWW, Bestor TH. (1998). Development, 125, 889 ± 897. Okumura K and Li E. (1999). Gene, 236, 87 ± 95. Mummaneni P, Walker KA, Bishop PL and Turker MS. Xu GL, Bestor TH, Bourchis D, Hsieh CL, Tommerup N, (1995). J. Biol. Chem., 270, 788 ± 792. Bugge M, Hulten M, Qu X, Russo JJ and Viegas- Mummaneni P, Yates P, Simpson J, Rose J and Turker MS. Pequignot E. (1999). Nature, 402, 187 ± 191. (1998). Nucleic. Acids. Res., 26, 5163 ± 5169. Yoder JA, Soman NS, Verdine GL and Bestor TH. (1997b). Nan X, Ng HH, Johnson CA, Laherty CD, Turner BM, J. Mol. Biol., 270, 385 ± 395. Eisenman RN and Bird A. (1998). Nature, 393, 386 ± Yoder JA, Walsh CP and Bestor TH. (1997a). Trends. Genet., 389. 13, 335 ± 340. Narayan A, Ji W, Zhang XY, Marrogi A, Gra€ JR, Baylin Yoder JA, Yen RC, Vertino PM, Bestor TH and Baylin SB. SB and Ehrlich M. (1998). Int. J. Cancer, 77, 833 ± 838. (1996). J. Biol. Chem., 271, 31092 ± 31097. Nass SJ, Ferguson AT, El Ashry D, Nelson WG and Zhang Y, Ng HH, Erdjument-Bromage H, Tempst P, Bird A Davidson NE. (1999). Oncogene, 18, 7453 ± 7461. and Reinberg D. (1999). Genes. Dev., 13, 1924 ± 1935. Ng HH, Jeppesen P and Bird A. (2000). Mol. Cell Biol., 20, Zhang Z, Shibahara K and Stillman B. (2000). Nature, 408, 1394 ± 1406. 221 ± 225. NgHH,ZhangY,HendrichB,JohnsonCA,TurnerBM, Zheng S, Chen P, McMillan A, Lafuente A, Lafuente MJ, Erdjument-Bromage H, Tempst P, Reinberg D and Bird Ballesta A, Trias M and Wiencke JK. (2000). Carcinogen- A. (1999). Nat. Genet., 23, 58 ± 61. esis, 21, 2057 ± 2064.

Oncogene