This page has been archived and is no longer updated
The fundamental role of epigenetic events in cancer
Author: Peter A. Jones
Keywords
Keywords for this Article
Add keywords to your Content
Save
|
Cancel
Share
|
Cancel
Revoke
|
Cancel
Rate & Certify
Rate Me...
Rate Me
!
Comment
Save
|
Cancel
Flag Inappropriate
The Content is
Objectionable
Explicit
Offensive
Inaccurate
Comment
Flag Content
|
Cancel
Delete Content
Reason
Delete
|
Cancel
Close
Full Screen
"REVIEWS In the mammalian genome, methylation takes place only at cytosine bases that are located 5? to a guanosine in a CpG dinucleotide 1 . This dinucleotide is actually under- represented in much of the genome, but short regions of 0.5?4 kb in length, known as CpG islands, are rich in CpG content 1,2 . Most CpG islands are found in the prox- imal promoter regions of almost half of the genes in the mammalian genome and are, generally, unmethylated in normal cells (FIG. 1). In cancer, however, the hypermethy- lation of these promoter regions is now the most well- categorized epigenetic change to occur in tumours; it is found in virtually every type of human neoplasm and is associated with the inappropriate transcriptional silenc- ing of genes 3,4 . Surprisingly, as shown in FIG. 2, such pro- moter hypermethylation is at least as common as the dis- ruption of classic tumour-suppressor genes in human cancer by mutation and possibly more so. Nearly 50% of the genes that cause familial forms of cancer when mutated in the germ line are known to undergo methy- lation-associated silencing in various sporadic forms of cancer (FIG. 2). Additionally, there is a growing list of can- didate tumour-suppressor genes that are silenced by pro- moter hypermethylation in certain cancers. These genes are predicted to be important for tumorigenesis on the basis of their presumed function, but seem not to be frequently mutated in such cancers. Examples of these genes include O 6 -methylguanine-DNA methyl- transferase (MGMT) 5 , which encodes an important DNA-repair gene; cyclin-dependent kinase inhibitor 2B (CDKN2B), which encodes p15, a cell-cycle regulator 6 ; and RASSF1A 7,8 , which encodes a protein of unknown function that can bind to the RAS oncogene. Promoter hypermethylation is the only mechanism for the loss of function of many of these genes in tumours (FIG. 2). During the past few years, it has become increasingly apparent that aberrant promoter methylation is associ- ated with a loss of gene function that can provide a selective advantage to neoplastic cells, as do mutations. For example, the von Hippel?Lindau syndrome (VHL) 9 , breast cancer 1, early onset (BRCA1) 10 and serine/threo- nine kinase 11 (STK11) 11 genes ? germ-line mutations of which cause familial forms of renal, breast and colon cancer, respectively ? are often epigenetically silenced in the sporadic forms of these tumour types. The importance of epigenetic silencing in the aetiology of non-familial forms of cancer is illustrated by studies of BRCA1. This gene was thought to be important only for familial breast cancer (through BRCA1 germ-line muta- tions). However, it is now apparent that 10?15% of women with the non-familial form of this cancer have tumours in which this gene is hypermethylated 10 . Furthermore, microarray studies indicate that the over- all gene-expression profiles of sporadic breast cancers with hypermethylated BRCA1 are identical to those of the familial cancers in which BRCA1 is mutated and are distinct from those of other breast-cancer types 12,13 . THE FUNDAMENTAL ROLE OF EPIGENETIC EVENTS IN CANCER Peter A. Jones* and Stephen B. Baylin ? Patterns of DNA methylation and chromatin structure are profoundly altered in neoplasia and include genome-wide losses of, and regional gains in, DNA methylation. The recent explosion in our knowledge of how chromatin organization modulates gene transcription has further highlighted the importance of epigenetic mechanisms in the initiation and progression of human cancer. These epigenetic changes ? in particular, aberrant promoter hypermethylation that is associated with inappropriate gene silencing ? affect virtually every step in tumour progression. In this review, we discuss these epigenetic events and the molecular alterations that might cause them and/or underlie altered gene expression in cancer. NATURE REVIEWS | GENETICS VOLUME 3 | JUNE 2002 | 415 *USC/Norris Comprehensive Cancer Center, Departments of Urology, Biochemistry and Molecular Biology, Keck School of Medicine, University of Southern California, 1441 Eastlake Avenue, MS 8302L, Los Angeles, California 90089-9181, USA. ? Sidney Kimmel Comprehensive Cancer Center, Departments of Oncology and Medicine, The Johns Hopkins Medical Institutions, 1650 East Orleans Street, Baltimore, Maryland 21231, USA. Correspondence to P.A.J. and S.B.B. e-mails: jones_p@ccnt.hsc.usc.edu; sbaylin@jhmi.edu doi:10.1038/nrg816 MICROSATELLITE INSTABILITY (min). In diploid tumours, genetic instability that is due to a high mutation rate, primarily in short nucleotide repeats. Cancers with the min phenotype are associated with defects in DNA-mismatch-repair genes. KNUDSON?S TWO-HIT MODEL In 1971, Alfred Knudson proposed that two successive genetic ?hits? are required to turn a normal cell into a tumour cell and that, in familial cancers, one hit was inherited. Two inactivating ?hits? are therefore required to cause the loss of function of tumour- suppressor genes. 416 | JUNE 2002 | VOLUME 3 www.nature.com/reviews/genetics REVIEWS been observed in the apparently normal colonic epithe- lium of patients that have colorectal cancer with microsatellite instability 20 and, in hyperplastic regions, preceding the development of endometrial cancers that develop this type of genetic change 21 . MGMT is another DNA-repair gene that is silenced in association with pro- moter methylation in colon, lung, lymphoid and other tumours 5,22,23 .O 6 -MGMT protein removes carcinogen- induced O 6 -methylguanine adducts from DNA, which result in G ?A transition mutations if left unrepaired. Tumours with silenced MGMT alleles do, indeed, seem to be predisposed to mutation in key genes, such as tumour protein p53 (TP53) 23 and K-RAS 5 . This promoter hyper- methylation, like that at MLH1, seems to precede genetic changes by occurring in pre-malignant polyps that do not yet harbour gene mutations 5,23 . Another important observation with regard to tumour-suppressor genes that are disrupted epigeneti- cally and/or genetically is that they often reside in genomic regions that are characterized by frequent chromosomal deletions. Virtually every chromosomal location that is shown in FIG. 2 is known to be a region that is frequently deleted in human cancer. These dele- tions cause loss of heterozygosity (LOH) and are often used to guide searches for tumour-suppressor genes. Interestingly, there are LOH regions in which epigenetic events, rather than genetic alterations, seem to define The functional significance of the hypermethylation of key tumour-suppressor gene promoters can also be appreciated by examining the consequences of the inac- tivation of individual copies of such genes. KNUDSON?S TWO-HIT MODEL predicts that a phenotypic consequence of tumour-suppressor gene loss is not seen unless both alleles of a gene are inactivated in a tumour 14 . Findings from several studies now clearly show that tumours can stably maintain mutations in one allele of a gene while the other allele is hypermethylated, leading to the func- tional inactivation of the gene 15,16 . In fact, when one of two alleles is mutated in the germ line of a patient with a familial form of cancer, and the resultant tumour retains both alleles of the gene, hypermethylation is commonly seen as the second inactivating change. Moreover, it seems never to be present in the promoter of the mutated gene, but is always associated with the wild- type allele 17 . The importance of epigenetic gene silencing in cancer is also highlighted by the growing awareness that such changes can actually predispose to mutational events during tumour progression. This was first shown for the mismatch-repair gene MLH1 (mutL homologue 1, colon cancer, non-polyposis type 2) 18,19 , which is frequently hypermethylated in sporadic tumours that have microsatellite instability (FIG. 1). Importantly, these changes in the methylation of the 5?region of MLH1 have Promoter methylation Microsatellite instability ? G ? A mutations ? C ? T mutations ? CC ? TT mutations ? G ? T mutations Spontaneous deamination Silencing of repair genes MLH1 MGMT Enhanced UV absorption Enhanced carcinogen binding a b Exon 1 Exon 2 Figure 1 | How epigenetics affects genetics. a | How de novo hypermethylation of CpG islands in the promoters of DNA-repair genes, such as MLH1 and MGMT, can lead to their inactivation. Hypermethylation of the promoter of MLH1 can lead to MICROSATELLITE INSTABILITY, and hypermethylation of the promoter of MGMT leads to increased G ? A mutations. b | Cytosine methylation in the coding region of genes can also increase mutation rates because of the spontaneous hydrolytic deamination of methylated cytosine, which causes C ? T transition mutations at methylated CpG sites. Methylation also changes the absorption wavelength of cytosine, into the range of incident sunlight, resulting in CC ? TT mutations, which commonly occur in skin cancers. Methylated CpGs are also preferred binding sites for benzo(a)pyrene diol epoxide and other carcinogens that are found in tobacco smoke. These cause DNA adducts and G ? T transversion mutations, which are often found in the aerodigestive tumours of smokers. MLH1, mutL homologue 1, colon cancer, non-polyposis type 2; MGMT, O 6 -methylguanine-DNA methyltransferase; UV, ultraviolet. NATURE REVIEWS | GENETICS VOLUME 3 | JUNE 2002 | 417 REVIEWS (HDACs), which alter gene transcription and chromatin configuration (as discussed in more detail later) 30 .Such genome-scanning procedures could prove to be of great use to cancer-biology research and might also have clini- cal applications. For example, in the above microarray approach, silencing by promoter hypermethylation of an entire gene family was discovered. This gene family can normally counteract the WNT (wingless-related) sig- nalling pathway that functions in colon cancer; virtually all colorectal cancers have this change in one or more of the genes in this pathway 30 . Although this review focuses predominantly on the gene-silencing events that are associated with methyla- tion changes in cancer, it is important to recognize that cytosine methylation can influence tumorigenicity by other mechanisms. These occur because 5-methylcyto- sine is itself mutagenic: it can undergo spontaneous hydrolytic deamination to cause C ? T transitions 31 . This enhanced mutagenesis is seen in the germ line of all organisms that methylate their DNA. Furthermore, as many as 50% of inactivating point mutations in the coding region of the human TP53 tumour-suppressor gene in somatic cells occur at methylated cytosines 32 . The presence of the methyl group in the CpG dinu- cleotides in the coding region of this gene strongly the important genes. For example, RASSF1A at 3p21 (REFS 7,8) and hypermethylated in cancer 1 (HIC1), which encodes a transcription factor, at 17p13.3 (REF. 24) ? which are two chromosomal regions that are charac- terized by frequent LOH in several tumour types ? are often hypermethylated in many important human can- cers, such as lung, prostate, colon and breast. These now constitute the principal candidate tumour-suppressor genes in these high-frequency LOH regions, where no gene mutations have been found consistently. The wide distribution of hypermethylated genes across the human genome, and the finding of hyperme- thylated candidate tumour-suppressor genes in regions of high-frequency chromosome deletions, has spurred efforts to screen the cancer-cell genome for such genes. There are a growing number of techniques that have been developed for this purpose. Some aim to identify regions of abnormal methylation per se rather than specific genes, whereas others aim to identify genes by discovering aberrantly hypermethylated CpG islands 25?29 (BOX 1). Recently, epigenetically silenced and potentially impor- tant genes for colon cancer have been successfully identi- fied using a microarray approach. This approach was used to assay for the re-expression of silenced genes after the treatment of cancer cells with histone deacetylases CDKN2A Two tumour-suppressor transcripts are encoded by the CDKN2A locus. P16 INK4A inhibits the cyclin-dependent kinases 4 and 6, blocking them from phosphorylating RB1 and so preventing cells from exiting G1. P14 ARF is encoded from an alternative reading frame (arf), helps regulate nuclear location of TP53 and putatively causes cell-cycle arrest at G1 and G2. Loss of heterozygosity of either transcript is associated with cancer. 4 6 7 8 10 X 11 12 13 Y2221181514 201916 17 2 3 5 91 MSH2 VHL MLH1 RASSFIA SMARCA3 FHIT SSI1 STK11 CDH1 BRCA1 NF1 SMAD4 NF2 TP53 HIC1 APC P15 P14,P16 TIMP3 SMARCB1 PTCH PTEN ATM BRCA2 RB GSTP1 TP73 Figure 2 | A map of the human genome. Genes that are frequently mutated and/or hypermethylated in cancer. Only a subset of the genes that are known to be frequently hypermethylated and silenced in one or more cancer types, and genes that are most often mutated in tumours are shown here. Nearly all of these genes are in chromosome regions that commonly show loss of heterozygosity in cancer. Genes for which only genetic mutations have been reported are shown in green, those that have been reported to be only hypermethylated are shown in red and those for which both changes have been reported are shown in purple. In the 9p region, the individual transcripts P16 INK4A and P14 ARF are transcribed from the CDKN2A locus, and are each shown because of the different incidences of hypermethylation and mutation at their promoters. It is noteworthy that as many, if not more, genes are inactivated by promoter hypermethylation and by epigenetic silencing as they are by coding-region mutations. ATM, ataxia telangiectasia mutated; APC, adenomatosis polyposis coli; BRCA1/2, breast cancer1/2, early onset; CDH1, E-cadherin; CDKN2A/B, cyclin-dependent kinase inhibitor 2A/B; FHIT, fragile histidine triad; GSTP1, glutathione S-transferase pi; MLH1, mutL homologue 1, colon cancer, non-polyposis type 2; MSH2, mutS homologue 2, colon cancer, non-polyposis type 1; NF1/2, neurofibromin 1/2; PTCH, patched homologue; PTEN, phosphatase and tensin homologue; RB1, retinoblastoma 1; SMAD4, mothers against decapentaplegic homologue 4; SMARCA3/B1, SWI/SNF-related, matrix-associated, actin-dependent regulator of chromatin, subfamily A, member 3/subfamily B, member 1; STK11, serine/threonine kinase 11; TIMP3, tissue inhibitor of metalloproteinase 3; TP53/73, tumour protein p53/p73; VHL, von Hippel?Lindau syndrome. 418 | JUNE 2002 | VOLUME 3 www.nature.com/reviews/genetics REVIEWS DNA methyltransferase 3?, one of the enzymes that catalyses DNA methylation ? lead to immunodefi- ciency centromeric instability and facial abnormalities (ICF) syndrome 38?40 . ICF patients have a loss of methy- lation at selected centromeric regions and have pro- found chromosomal structural changes. Indeed, many human tumours have similar losses of DNA methyla- tion and chromosomal structural changes in these regions 41 . Increased levels of gene deletion have also been reported in mouse embryonic stem (ES) cells that are deficient in DNA methyltransferase 1 (Dnmt1) 42 , although, in another study, fewer deletions were observed in Dnmt1-deficient ES cells 43 . Therefore, even though it remains an attractive hypothesis that hypomethylation leads to gross chromosomal instability in cancer, the relationship between the two processes is not yet clear. Given the DNA-methylation changes that can occur in cancer, especially those that are associated with tran- scriptional silencing and the loss of function of key genes, it seems timely to review our understanding of the molecular mechanisms that might cause these changes. This review explores how our increased under- standing of the role of chromatin in organizing and increases the rate at which mutations are induced by ultraviolet (UV) light during the development of skin cancers 33 . The biochemical mechanisms of this enhancement are well understood; the methyl group shifts the UV absorption spectrum for cytosine to a region in the spectrum that is prevalent in sunlight. Methylated CpG dinucleotides are also the preferred targets of G ? T transversion mutations, which are induced in mammalian cells by the tobacco carcinogen benzo(a)pyrene diol epoxide 34 . So, the methylation that occurs in the transcribed region of TP53 (REF. 35) increases its susceptibility to spontaneous deamination, UV-induced mutation and hydrocarbon carcinogenesis. The fact that most coding sequences contain abundant CpG methylation indicates that this epigenetic ?mark? might increase the likelihood that this sequence will undergo heritable genetic changes. Finally, it has been known for a long time that tumour cells are hypomethylated in comparison to wild-type cells 36,37 . This has led to the suggestion that the hypomethylation of non-promoter regions of DNA and of structural elements, such as centromeric DNAs, might cause enhanced genomic instability. Indeed, germ-line mutations in DNMT3B ? which encodes Box 1 | Techniques for randomly screening cancer-cell genomes for altered methylation loci Restriction landmark genomic screening (RLGS) In this approach 27 , genomic DNA is cut with methylation-sensitive restriction enzymes, such as Not 1, which recognize large CpG-rich sequences that usually occur in CpG islands. The restricted DNA is electrophoresed in two dimensions to produce a pattern in which a spot will be missing if a particular Not 1 site is methylated and has not, therefore, been cut. Although this technique allows researchers to detect large numbers of CpG islands, these are sometimes not in the promoter regions of genes and are, therefore, probably not involved with transcriptional regulation. MCA?RDA In methylated CpG island amplification?representational difference analysis (MCA?RDA) 26 , DNA is sequentially restricted with two enzymes, which each recognize the same CpG-rich sites that occur predominantly in CpG islands. The first enzyme is methylation sensitive and the second is not. This produces fragments that differ according to the methylation status of the DNA, which will differ between normal and tumour-derived DNA. After PCR amplification of these fragments, the tumour and normal DNA amplicons are subjected to RDA, which exploits the methylation- sensitive restriction-site differences between the normal and tumour-cell DNA to carry out a comparative hybridization subtraction step. The advantages of this approach are very similar to those of RLGS. However, even though many of the CpG islands are associated with genes, defining the start site of a gene and the exact relationship of the island to the transcriptional regulation of a gene can be laborious. MS-AP-PCR Methylation-sensitive arbitrarily primed PCR (MS-AP-PCR) 28 also uses methylation-sensitive restriction enzymes to cut DNA before it is amplified with random CpG-rich primers that target CpG islands. The resulting fragments are displayed on gels, and gel-spot patterns between different cell types are compared, which leads to a rapid identification of CpG islands that are differentially methylated in different tissues. The method suffers from the same limitations as those for RLGS and MCA?RDA. Differential methylation hybridization (DMH) DMH 25 is an array-based method in which genomic DNA is pre-cut with a methylation-insensitive enzyme, such as Mse 1. Linkers are then ligated to the digested DNA before it is cut with a methylation-sensitive enzyme. The resulting digests are amplified by PCR and the products hybridized to an array of immobilized CpG islands. Microarray and gene re-expression approach In this approach, gene re-expression is induced by treating cells with agents that block both promoter hypermethylation and histone deacetylation 30 . Silenced genes that are re-expressed by this treatment are then surveyed by cDNA microarray analysis. The advantage of this approach is that the detection of hypermethylation sites is linked to the transcriptional status of genes, the promoters of which are affected by this change. A disadvantage is that the CpG island that is hypermethylated, and is associated with the gene promoter, is not always easy to identify in genomic databases. NATURE REVIEWS | GENETICS VOLUME 3 | JUNE 2002 | 419 REVIEWS unclear whether methylation is the initial silencing event or whether it is targeted to the region by earlier chromatin-remodelling events. In some systems, the action of an RNA species is required to initiate gene silencing before promoters in the region become densely methylated, as shown by the role of X (inac- tive)-specific transcript (XIST ) in gene silencing on the X chromosome of female mammals 45 . Whatever the sequence of events is at a given promoter, it seems that promoter methylation must be integral to the loss of gene function because drugs that induce DNA de-methylation, such as 5-AZA-2?-DEOXYCYTIDINE 46 , can partially reactivate silenced genes in cancer cells to restore their function 3,4 . If chromatin is crucial to aberrant gene silencing in cancer, either primarily or secondarily, which processes are involved? Most of our genome is normally pack- aged as transcriptionally repressive chromatin of the type found in PERICENTROMERIC HETEROCHROMATIN regions. This type of chromatin is heavily methylated, and the identities of the chromatin-associated protein com- plexes that might link DNA methylation to transcrip- tional silencing have been discovered over the past few years through studies of such chromatin 47?49 (FIG. 3).The DNA in these transcriptionally silent regions is pack- aged into compacted NUCLEOSOMES that contain deacety- lated histones, in particular deacetylated histone H3 (H3). These histones are extensively deacetylated through the action of HDACs, and this deacetylated histone state helps to maintain nucleosomes in a com- pacted and transcriptionally silent state 47?49 . The tran- scriptional repression status of this chromatin is also facilitated by the participation of key proteins, such as CBX5, the human homologue of the Drosophila CHRO- MODOMAIN protein Hp1? (also known as Suppressor of variegation 205, Su(var)205) 44,46?48 , as discussed in more detail below. DNA methylation itself also seems to be involved crucially in the transcriptionally silent state of pericen- tromeric heterochromatin. Methyl-cytosine-binding proteins (MBPs) associate with methylated cytosines and also with various chromatin-remodelling com- plexes. They have also been shown to act as transcrip- tional repressors in vitro 50?53 . Importantly, these MBPs also reside in complexes that contain HDACs; for example, the methyl-binding proteins methyl-CpG- binding protein 2 (MECP2) and methyl-CpG-binding- domain proteins MBD1 and MBD2 have been found to associate with transcriptional co-repressors, such as SIN3, which are known to bind HDACs directly. This presents a method by which the MBPs might recruit histone deacetylation to methylated DNA in regions of transcriptional silencing 50?53 . By contrast, only a small fraction of the genome is transcriptionally competent. The state of chromatin in these regions must be dynamic to meet the chang- ing transcriptional requirements of a cell. This bal- ance between EUCHROMATIN and heterochromatin ensures that the gene-expression pattern of a given cell type is stably maintained in daughter cells as a heritable state 1 . regulating the transcription of the mammalian genome is enhancing our understanding of the epigenetic changes that occur in cancer. Interpreting methylation signals in cells An important issue in studies of epigenetically medi- ated gene silencing in cancer is to understand how promoter hypermethylation participates in the loss of gene transcription. In experimental systems, it seems that methylation at promoters does not lead to silenced transcription until chromatin proteins are recruited to the region, which mediate gene silenc- ing 44 . In this setting, methylation seems to initiate the process that results in a loss of transcription. However, in some cellular settings, including in abnormal gene-silencing events in cancer, it is 5-AZA-2?-DEOXYCYTIDINE A potent and specific inhibitor of DNA methylation. PERICENTROMERIC HETEROCHROMATIN The late-replicating, gene- sparse, transcriptionally inactive, condensed chromatin regions that are rich in repeated sequence and occur near the centromeres of chromosomes. NUCLEOSOME The fundamental unit into which DNA and histones are packaged in eukaryotic cells. It is the basic structural subunit of chromatin and consists of ~200 bp of DNA and an octamer of histone proteins. CHROMODOMAIN A highly conserved sequence motif that has been identified in various animal and plant species. Chromodomain proteins are often structural components of large macromolecular chromatin complexes or involved in remodelling chromatin structure. Hp1? is a chromodomain-containing protein. EUCHROMATIN The lightly staining regions of the nucleus that generally contain decondensed, transcriptionally active regions of the genome. CA HP1 DNMTs HMT HMT HAT Transcription activator complex TF Methylated CpG Unmethylated CpG Methylated Lys9 HDAC Nucleosome MBP HP1 Figure 3 | Typical chromatin configuration of transcriptionally silent pericentromeric DNA. In heterochromatin, most candidate CpG sites are methylated and are bound by methyl-cytosine-binding proteins (MBPs), which are present in complexes that include histone deacetylases (HDACs). The histones are deacetylated and organized into regularly spaced, tightly compacted nucleosomes. DNA methyltransferases (DNMTs) can access the area, whereas transcription activator complexes ? consisting of a transcription factor (TF), a co-activator protein (CA) and a protein with histone acetyltransferase (HAT) activity ? are excluded. The histone ?mark? of a methylated Lys9 residue on the tail of histone 3 (H3) might help to target DNA methylation to the region and signifies, together with deacetylated histones, that this is transcriptionally repressive chromatin (see text for more details ). H3 Lys9 methylation is maintained by a histone methyltransferase (HMT) that is recruited by the binding of the chromodomain protein HP1? (HP1) to the methylated H3 Lys9. 420 | JUNE 2002 | VOLUME 3 www.nature.com/reviews/genetics REVIEWS promoters, it is not known for how many of them CpG methylation might have a modulatory role in their tran- scription. However, there are examples in which the methylation of individual CpG sites in such promoters can determine the transcriptional status of a gene by blocking the access of certain transcription factors that are sensitive to the methylation of CpG sites in their tar- get binding sites 1,54?56 (FIG. 4). Alternatively, the methyla- tion of sites in these regions might also silence genes by helping to recruit chromatin protein complexes that repress transcription (FIG. 4). In contrast to CpG-poor promoters, methylation does not normally participate in regulating the tran- scription of genes with promoters that contain CpG islands, whether such genes are being actively tran- scribed or not 1,3,4 . These islands reside, especially in active promoters, in chromatin that is composed of widely and irregularly spaced nucleosomes, which contain highly acetylated histones. Such a chromatin conformation and histone-acetylation state is thought to facilitate the accessibility of the promoter to tran- scription-activating complexes (FIG. 5). Unmethylated CpG-island regions are generally flanked by regions of less-CpG-rich DNA, which is heavily methylated 1,3,57,58 , and in a chromatin conformation that, presumably, resembles that of transcriptionally silent pericen- tromeric heterochromatin (FIG. 5). When the transcrip- tion of these CpG-island promoters is downregulated, as required by a given cellular state, the island remains free of methylation 1 . It is likely, in such inactive states, that the promoter is associated with transiently posi- tioned repressor complexes that do not involve methy- lated DNA or MBP complexes. The above separation of unmethylated promoter CpG-island regions from immediately flanking areas of methylation seems to involve a functional boundary. Much remains to be determined about how these boundaries for transcriptionally active versus inactive chromatin are established. However, recent exciting dis- coveries have begun to link a specific histone code to the targeting of methylation to DNA (FIGS 4,5). Specific methylation marks on histone H3 seem to provide a sig- nal that separates regions of transcriptionally active chromatin from regions of transcriptionally inactive chromatin. The methylation of lysine 9 (Lys9) in the tail of histone H3 has been found to be closely associated with transcriptionally repressive chromatin, whereas methylation of lysine 4 (Lys4) on H3 characterizes the transcriptionally active chromatin that immediately flanks the heterochromatin 59?63 . It has been found that the Drosophila Hp1? chromodomain protein, as well as its homologues in other species, is important for estab- lishing transcriptionally repressive chromatin. It does this by binding to the methylated-Lys9 residue through its chromodomain and then interacting with histone methyltransferases to recruit them to sites of Lys9 methylation 59?63 . Different histone methyltransferases mediate the Lys4 methylation. Although this histone methylation pattern is found in organisms such as yeast, which do not methylate their DNA, it has also been associated with the establishment of DNA methylation In terms of DNA methylation, there are two types of gene promoters 1,3,4 (FIGS 4,5). One type accounts for ~50% of the genes in the mammalian genome and con- tains unmethylated CpG islands. The other promoter type is CpG poor in composition, as is the rest of the genome. Among the genes that have CpG-poor Transcription activator complex HP1 HMT Exon 1 Exon 2 Exon 3 a Transcription activator complex Exon 1 Exon 2 Exon 3 b Methylated CpG Unmethylated CpG HDAC MBP MBP MBP CA HAT DNMTs DNMTs TF CA HAT TF TF HMT R1 R2 HP1 HMT R Methylated Lys4 on H3 Methylated Lys9 on H3 Transient repressor complexes Nucleosome with deacetylated histones Nucleosome with acetylated histones SIN3 Co-repressor Figure 4 | A CpG-poor promoter in transcriptionally active and transcriptionally repressed states. a | A CpG-poor promoter in a transcriptionally active state. Individual CpG sites around the transcription start site are unmethylated and are protected from DNA methyltransferases (DNMTs) and transcriptionally repressive complexes that contain histone deacetylases (HDACs) and co-repressor proteins. The nucleosomes around the promoter are more widely spaced than in heterochromatin and contain heavily acetylated histones. The red asterisks depict a histone code ?mark? ? methylation of the Lys4 residue of histone 3 ? that is typical of transcriptionally permissive chromatin, which is methylated by a different histone methyltransferase (HMT) to that shown in FIG. 3. b | The same promoter in a transcriptionally repressed state. Selected CpGs near the transcription start site are methylated. Three possible transcriptionally repressive chromatin changes are shown: methylated CpGs that block the occupation of the target binding site of a transcription factor; a transient repressor complex that might contain SIN3, which is a known transcriptional co-repressor; and transcriptionally repressive MBP complexes that are targeted to methylated CpG regions. CA, co-activator protein; HAT, histone acetyltransferase; HP1, chromodomain protein HP1?; MBP, methyl- cytosine-binding protein; TF, transcription factor. NATURE REVIEWS | GENETICS VOLUME 3 | JUNE 2002 | 421 REVIEWS because they indicate that the methylation of H3 Lys9 might somehow be required to establish sites of DNA methylation in these and, perhaps, other organisms. It is not clear how the methylation of H3 Lys9 directs DNA methylation. It has been suggested that it might bring about the binding of DNA methyltransferases to the methylated lysine or to the key proteins that are required for histone methylation, such as Hp1? or the histone methyltransferases themselves. Intriguingly, some histone methyltransferases contain MBP-like domains, which indicate that there might be an addi- tional interaction between the histone and the DNA methylation process 63 . In the light of these findings, how does the gene silencing that is associated with aberrant CpG-island methylation in cancer occur? Although much remains to be proven experimentally, hypermethylated promot- ers are now known to have certain features that are typi- cal of pericentromeric heterochromatin (FIG. 5).For example, they have a closed chromatin configuration, which is typical of chromatin that is composed of tightly compacted, highly deacetylated nucleosomes 66 . Interestingly, the specific association of MBPs to such promoters is beginning to be described 67?69 .However, studies that have tried to determine which MBP family members, and their associated chromatin-remodelling proteins and HDACs, associate with specific hyperme- thylated promoters, such as that of the cyclin-dependent kinase inhibitor 2A (p16 INK4a ; also known as CDKN2A) transcript, differ so far 67?69 . Therefore, this important question will require further characterization. The pos- sibility that a specific histone methylation code, as described above, might be required to initiate and/or maintain the aberrant hypermethylation of gene pro- moters in cancer must now also be considered and will be a particularly exciting area of future research. An important characteristic of the interaction between DNA methylation and HDAC activity, par- ticularly in maintaining the aberrant silencing of hypermethylated genes in cancer, is that the methyla- tion seems to function as the dominant event that seals transcriptional repression. This finding stems from studies that have found that the inhibition of HDAC activity alone, by potent and specific drugs such as trichostatin (TSA), does not result in the reactivation of aberrantly silenced and hypermethy- lated genes in tumour cells 66 . However, TSA can reac- tivate these same genes if the cells are first treated with demethylating drugs, such as 5-aza-cytidine, which demethylate promoters 66 . This finding not only is important for understanding how DNA methylation works to silence transcription, but also has clinical implications, as discussed in more detail below. A molecular explanation for the above interaction between DNA methylation and histone deacetylation awaits clarification. However, in a recent study of MDR1, which encodes a multidrug transport resistance protein and which is hypermethylated in leukaemia cells, it was found that, although HDAC inhibition alone leads to histone acetylation in the promoter, the in two organisms ? Neurospora crassa and Arabidopsis thaliana. In these organisms, mutations in a Lys9 his- tone methyltransferase eliminate all or some DNA methylation 64,65 . These findings are extremely important DNMTs DNMTs Transcription activator complex Transcription activator complex Exon 1 a b Exon 1 Exon 2 Exon 3 Exon 2 Exon 3 CA HAT TF CA HAT TF HMT HP1 HMT HP1 HMT HP1 HMT HP1 HMT HMT HP1 Methylated CpG Unmethylated CpG HDAC Methylated Lys4 on H3 Methylated Lys9 on H3 Nucleosome with deacetylated histones Nucleosome with acetylated histones Co-repressor MBP Figure 5 | A CpG-rich promoter in transcriptionally active and transcriptionally repressed states. a | A typical, unmethylated and transcribed promoter that contains a CpG island. The island resides in a domain that is protected from the spread of methylated chromatin that flanks it. Around the promoter, there are widely spaced nucleosomes, which contain acetylated histones in which the Lys4 residue of histone H3 is methylated by a Lys4 methyltransferase (HMT). This is in contrast to the flanking transcriptionally repressive chromatin. This chromatin has methylated CpG sites, methyl-cytosine-binding protein (MBP)- and histone deacetylase (HDAC)-containing complexes, is tightly compacted, and has regularly spaced nucleosomes with de-acetylated histones that contain methylated H3 Lys9, as maintained by an HP1? (HP1)?Lys9 HMT complex. DNA methyltransferase (DNMT) complexes are excluded from the promoter. b | The same promoter in an aberrantly hypermethylated state that is associated with transcriptional repression in a cancer cell. The promoter region is accessible to DNMT complexes and transcriptionally repressive chromatin occupies the promoter, which is transcriptionally silenced. CA, co-activator protein; HAT, histone acetyltransferase; HP1, chromodomain protein HP1?; TF, transcription factor. 422 | JUNE 2002 | VOLUME 3 www.nature.com/reviews/genetics REVIEWS acetylated histones must be de-acetylated during the formation of late-replicating, transcriptionally repressed and heavily methylated DNA 80 . Several recent investigations indicate that the three active DNMTs, especially in cancer cells, might cooper- ate, in ways that are not predicted by mouse knockout studies of the Dnmt genes, to maintain the methylated state and transcriptional silencing of genomic regions. In mouse, human and monkey cells, DNMT1 has been found to localize to DNA-replication foci throughout S phase, whereas DNMT3A and DNMT3B co-localize to these sites only during late S phase 78 . Furthermore, in mouse cells, Dnmt3a and Dnmt3b can also be found in non-dividing cells in large pericentromeric hete- rochromatin-like complexes that contain the mouse homologue of the important silencing protein Hp1? and also, intriguingly, the methyl-cytosine-binding pro- teins, MECP2 and MBD1 (REF. 78). This cell-cycle- dependent localization seems ideal to facilitate a recently defined pathway in which the error-prone methylating activities of DNMT1 are backed up by DNMT3A and DNMT3B, which might fill in missed methylation sites 81 . In cancer cells, the re-methylation of the pro- moter CpG island of p16 INK4A after demethylation induced by 5-aza-2?-deoxycytidine seems to be depen- dent on active DNA replication, whereas other sites that are not located in CpG islands re-methylate without DNA synthesis 82 . This might indicate a previously unex- pected de novo role for DNMT1 in methylating CpG islands in cancer cells given its tight association with S-phase activity. Finally, two genetic changes in cancer cells, one naturally occurring and one experimentally induced, highlight the potential cooperativity between the DNMTs. First, the translocation that produces the PML?RAR? (promyelocytic leukaemia?retinoic acid receptor-?) fusion protein in PML cells induces the transcriptional repression and the aberrant methyla- tion of the promoter of a candidate RAR? transcrip- tional target, another retinoic acid receptor gene, retinoic acid receptor-? (RARB). This methylation occurs concomitantly with the appearance of com- plexes that contain each of the DNMTs at the RARB promoter 83 . Second, the somatic inactivation of DNMT1 in human colon cancer cells led to the sur- prising finding that, although loss of DNMT1 func- tion reduces overall DNMT activity to only 5% that of the wild type, there is surprisingly little effect on the overall DNA methylation of the genome of the cell 84 . Furthermore, the abnormal promoter hyper- methylation and the silencing of genes, such as CDKN2A, persists in these cells 84 . When DNMT3B is inactivated in colon cancer cells, both overall DNMT activity and genomic methylation change very little and abnormal promoter methylation persists 85 . However, when both DNMT1 and DNMT3B are simultaneously deleted, the cells lose most of their DNMT activity and, also, 95% of total genome-wide methylation. Furthermore, promoter hypermethyla- tion and transcriptional silencing were reversed for some genes, such as CDKN2A and tissue inhibitor of re-expression of the gene did not occur unless the demethylation of its promoter first reduced the pres- ence of MECP2 (REF. 69). Theoretically, the loss of this protein might remove a direct transcriptional repressor of MDR1, which might then be recruited by methylated CpGs and/or a complex that contains HDACs. Future studies will be required to verify this possibility and to elucidate the molecular mechanisms that underlie the interplay between DNA methylation and histone acety- lation in gene silencing. Initiating and maintaining methylation Our ideas of how DNA methylation is established in the mammalian genome are changing as rapidly as is our understanding of how chromatin organization modu- lates gene expression. Several years ago, only one mam- malian DNA methyltransferase (DNMT) enzyme, DNMT1, was thought to be responsible for maintaining DNA-methylation patterns in adult cells, and the mech- anism for establishing de novo methylation was not known 1 . Now, two other biologically active DNMTs, DNMT3A and DNMT3B have been identified. Targeted disruptions of these Dnmt genes in mice, alone and together, cause embryonic lethality, and studies of these mutant embryos have led to a model in which Dnmt3a and Dnmt3b are responsible for establishing de novo methylation patterns, which are then maintained by Dnmt1 (REF. 39). All three DNMTs, at the mRNA and/or protein level, are modestly overexpressed in many types of tumour cell 70,71 . Furthermore, modest overexpression of exogenous mouse Dnmt1 in NIH 3T3 cells 72 or the induction of Dnmt1 expression in these cells by the overexpression of FOS (v-fos FBJ mouse osteosar- coma viral oncogene homologue) 73 can promote cellu- lar transformation. This induction of tumorigenic properties by Dnmt1 in cell culture, as well as the fact that the genetic inactivation of Dnmt1 in mice also decreases the development of gastrointestinal tumours 74 in the min mouse model of gastrointestinal cancer, indicate a possible role for at least one DNMT in tumorigenesis. However, the mechanisms that underlie such a role in cancer are still not defined. Our understanding of aberrant gene silencing in cancer has been expanded recently by studies that indi- cate that all three biologically active DNMTs are much more complex proteins than has been appreciated. In experimental cell-culture systems, each protein can directly repress transcription in reporter-gene systems by interacting with HDACs and by binding to other proteins with transcriptional-repression activities 75?79 . So, DNMTs could participate in gene silencing with, or without, accompanying DNA methylation, and this has potential ramifications for the abnormal methylation of CpG islands in cancer, especially with respect to the question of which comes first, gene silencing or methy- lation. The direct link to HDACs provides another mechanism by which they can be targeted to sites of transcriptional repression (FIGS 4,5). Interestingly, DNMT1 co-localizes with HDAC2 at replication foci during late S phase 77 , at a time when newly arriving NATURE REVIEWS | GENETICS VOLUME 3 | JUNE 2002 | 423 REVIEWS Many of the above observations were unexpected given that knockout studies in mice have indicated that Dnmt1 might predominantly maintain DNA methyla- tion after early embryonic development 39 and given that ES cells from Dnmt1 knockout mice survive in culture only if kept undifferentiated 86 . How, then, could a cancer cell live without this protein and why should coopera- tivity between DNMTs be apparent in this setting? The explanation might be that the genetic alterations that metalloproteinase 3 (TIMP3) 85 . This finding not only highlights the potential cooperativity and comple- mentary activities of the DNMTs, but also provides genetic evidence of the importance of promoter hypermethylation for the silencing of key tumour- suppressor genes. In fact, growth of the double- knockout cells is markedly slowed, except for one clone in which the P16 transcript of CDKN2A remained hypermethylated 85 . Box 2 | The dynamics of genetic and epigenetic gene silencing In cancer, the dynamics of genetically and epigenetically mediated loss of gene function are very different. A somatic genetic mutation (red cross in panel a) that occurs in one round of DNA replication is associated with an immediate block in the production of a functional protein from the mutant allele (see panel a). If a selective advantage is conferred to cells that carry this mutation, they can expand clonally to give rise to a tumour in which all cells lack the capacity to produce the functional protein. By contrast, epigenetically mediated silencing of the same gene might begin gradually, starting in the earliest phases of tumour progression. This process might start with a subtle decrease in the level of transcription of a gene (see panel b), which leads to a minor decrease in protein production. This decreased transcription might then foster a decrease in protection (shown as red vertical bars) of the CpG island from the spread of flanking heterochromatin and methylation into the island (black angled arrows). This loss results in gradual increases in the methylation of individual CpG sites in the island, the extent of which might vary between copies of the same gene in different cells. In an emerging clone of tumour cells, the result of this progressive and heterogeneous methylation is an increasing degree of transcriptional loss and a variable decrease in protein production in individual cells of the tumour clone. The degree of loss of protein production is, therefore, not uniform throughout the tumour-cell population, unlike that which is produced by the genetic changes shown in panel a. This situation is an ideal one for mediating the dynamic heterogeneity that characterizes important tumour properties, such as metastasis, as discussed in the text. Filled ovals, methylated CpGs; open ovals, unmethylated CpGs. Functional protein Genetic mutation No functional protein No functional protein Protein Protein Protein Protein Tumour clone Protein a b Tumour clone 424 | JUNE 2002 | VOLUME 3 www.nature.com/reviews/genetics REVIEWS Genetic alterations in cancer can lead to immediate disruptions of protein function, the phenotypic effects of which might not be instantly detectable. However, if a subsequent cell abnormality complements and ampli- fies the consequences of a given gene mutation, it can provide the cell with a selective tumorigenic advantage. All subsequent clones at that stage of tumour develop- ment will harbour that mutation. By contrast, experi- mental data indicate that promoter CpG-island hyper- methylation is a more gradual and progressive process that spreads from the lateral edges of heavily methylated DNA that flanks a CpG island and moves centrally towards the transcription start site of the gene 58,89 .The extent of this spreading in genes ? such as in CDKN2B in leukaemia 93 and CDH1 (which encodes E-cadherin) in breast and other tumours ? even in long-established cell cultures 58 , seems to vary between individual DNA strands and between cells 58,93 . As the degree of transcrip- tional silencing of CpG-containing promoters is usually dependent on the number of CpG sites in the island that are methylated, this methylation heterogeneity from one DNA strand, and from one cell to another, can lead to heterogeneous populations of cells in a tumour, with respect to their gene-expression levels 58 (BOX 2). How does such cellular heterogeneity for the density of promoter methylation and its timing affect the can- cer phenotype? Tumours consist of cell populations that are functionally heterogeneous for key neoplastic properties 94,95 . How promoter hypermethylation could underlie such processes is outlined in BOX 2. So, the loss of function of the key tumour-suppressor transcript, P16, in association with promoter hypermethylation is a cellularly heterogeneous process that can begin early in the neoplastic process, in pre-malignant lesions, and progress to involve virtually all cells in the resulting cancer 96 . The early loss of p16 by this process is proba- bly a fundamental event in the progression of many types of cancer 97,98 . One of the most dynamic, cellularly heterogeneous processes in tumour progression is metastasis. Cells with the most metastatic properties exist as a subpopu- lation in a heterogeneous cell population in the primary tumour. Once these cells metastasize, they regenerate, at the distant site, the cellular heterogeneity of the metasta- tic subpopulation 92,99 . This type of cellular diversity is unlikely to be mediated by permanent genetic muta- tions. However, epigenetically mediated gene silencing is an excellent candidate to support such cellular dynam- ics. Indeed, the heterogeneous promoter status of a key gene that is related to cell invasion, CDH1, supports this idea. Loss of CDH1 function favours tumour cells that acquire the invasive properties of metastatic tumour cells 58,100,101 . As has been shown, cell?cell heterogeneity in the hypermethylation status of the promoter of this gene can underlie the cellular heterogeneity of CDH1 expression in both primary and metastatic tumours and in cell cultures 58 . Established cultures of prostate and breast cancer cells provide an experimental model in which to study the implications of this heterogeneity for cell invasion. In vitro assays for cell invasion actually select for those cells with the most densely methylated evolve during tumour progression produce such a state of interdependence between DNMTs. For example, mouse Dnmt1-knockout cells, which have profoundly decreased genome-wide DNA methylation levels, can live longer in culture, even in a differentiated state, when they are also null for transformation-related protein 53 (Trp53) 87 . Although the colon cancer cells used in the above knockout studies are wild type for TP53,other genetic and epigenetic abnormalities might produce a state in which DNA-methylation patterns are produced and/or maintained by mechanisms that differ from those seen in normal cells. In summary, recent experimental data are changing our view of the molecular mechanisms that underlie the establishment and maintenance of DNA methylation. Defining the mechanisms that determine the composi- tion and function of the complexes that link DNMTs in their role in normal and neoplastic cells, and that deter- mine the DNA sites that they affect, are important research priorities. Finding answers to such questions will help to explain how altered chromatin organization and abnormal DNA methylation mediate aberrant gene silencing in cancer. Spreading of aberrant methylation in cancer Transcriptionally repressive heterochromatin can spread into adjacent regions of the genome unless ?boundary ele- ments? are present to maintain active and silenced domains 1,88 . One explanation for the widespread de novo methylation of CpG islands in cancer might be that the compartmentalization of the genome into euchromatin and heterochromatin, and into unmethylated and methy- lated components, breaks down during carcinogenesis to allow the spread of heterochromatin to occur 89 . For some genes, such as the oestrogen receptor (ESR1) gene, this process seems to be initiated during the course of ageing 90 . The mechanisms that underlie the process are unknown, but they seem to reflect the vulnerability of the protective barriers to CpG-island methylation, which perhaps occurs over the course of cell renewal and seems to affect some genes more than others 91 . Although the molecular link between promoter hypermethylation and ageing awaits further clarification, it is noteworthy that ageing might be the single most important risk factor for cancer 92 ; so, age-dependent promoter methylation could explain the association between cancer and ageing. Despite what we know about how transcriptionally active versus repressive chromatin is formed, we have yet to define the mechanisms that allow normal cells to pro- tect most of their CpG islands from methylation and gene silencing. We are beginning to recognize, however, that the loss of this protection, during tumour progression, is a time-dependent process. During this process, the extent of methylation at any given locus can change dynamically from one cell to another, from one DNA strand to another and even from one CpG site in a CpG island to another. As such, epigenetically mediated silencing might have quite different dynamics to that associated with the loss of protein function caused by genetic mutations and might, therefore, have different phenotypic consequences during tumour progression (BOX 2). NATURE REVIEWS | GENETICS VOLUME 3 | JUNE 2002 | 425 REVIEWS Their uses for restoring gene function to treated cells in culture has indicated that they might also be useful for treating patients with malignant disease 103 . The use of aza-nucleoside analogues to treat human diseases is complicated by the fact that the drugs are unstable in aqueous solution, which requires their frequent preparation in fresh solutions for them to remain active. They are also MYELOSUPPRESSIVE, partic- ularly when used at high doses, as is often the case in clinical trials. The efficiency with which genes in cells treated with 5-aza analogues are reactivated is not enhanced necessarily by increasing the dose of these drugs beyond their optimally effective concentration 46 . So, low-dose trials might be more efficacious. In addi- tion, the synergy between demethylating drugs and histone deacetylating agents 66 makes clinical trials that test demethylating agents in combination with deacetylase inhibitors, in both leukaemias and solid tumours, an attractive goal. The results of such clinical investigations will be important for assessing the effi- cacy of re-activating epigenetically silenced genes as a future therapeutic option for treating cancer. Although analogues that are more stable in aqueous solution might be of value, another drug that inhibits methylation is procanamide, which is used for the treatment of cardiac arrhythmias. This drug is a non- competitive inhibitor of the methyltransferase enzymes and can reactivate the transcription of the P16 tran- script from a methylation-silenced promoter in prostate cancer cells that grow in nude mice 104 .Given the therapeutic promise of reversing methylation changes in cancer, there are certain to be more searches to identify better inhibitors of DNA methylation to use in clinical studies. As a result, several companies are pursuing this important goal. One concern about the use of methylation inhibitors in patients is whether the drugs will cause the inappropriate activation of genes in normal cells. Although important, this might not be an insur- mountable problem. This is because transcription from promoter-containing CpG islands does not seem to be commonly controlled by methylation in normal cells. In cases in which the methylation of promoter- associated CpG islands is important, such as in X-chromosome inactivation in mammals, the exis- tence of many levels of gene silencing 1,50 means that these genes cannot be routinely and easily activated by inhibiting DNA methylation in normal cells. So, the inactive X chromosome remains inactive in normal human fibroblasts after 5-azacytidine treatment 105 , whereas it can be readily reactivated in rodent?human somatic cells 106 , which indicates that some higher level of control is absent in these hybrids. The issue of whether transposable elements, which are normally methylated 107 , might be activated by demethylation and cause deleterious effects in normal tissue is a valid concern that requires further study. However, patients who have received 5-aza nucleosides for malignant and non-malignant diseases have generally not shown massive toxicity owing to inappropriate gene activa- tion. Nevertheless, these concerns will need to be promoters and the lowest expression of CDH1 (REF. 58). Moreover, a less densely methylated CDH1 promoter and the re-expression of CDH1 in these cells occurs concomitantly with their growth in cell clusters 58 ,an assay that mimics growth at distant metastatic sites. Reactivating silenced genes The finding that many genes controlling normal cellular homeostasis can be silenced inappropriately by structural chromatin changes that involve DNA methylation has encouraged a search for agents that might reverse these changes and, therefore, restore principal cellular path- ways. The demethylating agent 5-azacytidine and its deoxy derivative 5-aza-2?-deoxycytidine were first synthe- sized in Czechoslovakia as potential chemotherapeutic agents for cancer 102 . These agents are powerful inhibitors of DNA methylation, as they are incorporated into the nucleic acids of dividing cells, where they act as mecha- nism-based inhibitors of DNA methyltransferases 46 . MYELOSUPPRESSION The depressed production of blood cells that are derived from the myeloid lineage, including platelets, some leukocytes and erythrocytes. Myelosuppression is a common side effect of many anticancer drugs as they suppress the growth or proliferation of rapidly dividing cells. Methylated alleles in lumen or blood Ageing Cancer Hypomethylating drug Figure 6 | Progressive methylation changes in epithelial carcinogenesis. CpG islands in epithelial cells are generally unmethylated but undergo some de novo methylation during ageing. Cancer cells often show focal CpG-island methylation, which occurs coincidently with genome- wide demethylation. This methylation can silence tumour-related genes and have a causal role in carcinogenesis. As tumour cells die and metastasize, methylated alleles are shed into the lumen of the organ or into the plasma, where they can be detected with high sensitivity. This offers opportunities for cancer diagnosis, detection and prognosis. Methylated alleles can be reactivated by drugs such as 5-aza-2?-deoxyazacytidine or procanamide, which leads to gene re-expression and reversion of some aspects of the transformed state. Filled circles, methylated CpGs; open circles, unmethylated CpGs. 426 | JUNE 2002 | VOLUME 3 www.nature.com/reviews/genetics REVIEWS methylated in lymphoid, but not in solid, cancers 6 ,and the glutathione S-transferase (GSTP1) gene is methy- lated in prostate, breast and hepatic cancers, but rarely in others 104 . The use of small panels of hypermethyla- tion markers that almost cover the genomes of individ- ual tumour types has been shown 30,112,113 . These might lead to the development of one of the most powerful molecular strategies for cancer detection. Conclusions The contribution of epigenetic silencing to the inactiva- tion of key pathways that are involved in carcinogenesis is only now being realized by the scientific and medical communities. Despite what we know, there are undoubtedly more components in these pathways that are subject to this kind of inactivation. Many of these will be uncovered by the application of candidate-gene approaches and genome-wide screens. Further develop- ments using array-based technologies coupled to infor- matic approaches are likely to speed the identification of new genes that are silenced by hypermethylation in can- cer and to classify the known genes into biologically and clinically relevant hierarchies. There is tremendous ferment in the fields of chro- matin structure and DNA methylation that is just beginning to link the ?histone code? to the ?cytosine methylation code?. It seems almost certain that the methylation of H3 Lys9 and the DNMTs are inexorably linked in mammals. This means that we shall soon begin to understand how cytosine-methylation patterns are established. Knowing this is essential to deciphering what happens during ageing and carcinogenesis to reshuffle the epigenetic deck. The existence of altered cytosine-methylation patterns that are tissue and tumour specific, and that can be readily detected by sen- sitive PCR methods, holds considerable promise for the detection and diagnosis of cancer and for providing bet- ter prognostic information. Unlike mutagenic events, epigenetic events in cancer can also be reversed to restore the function of key control pathways in malig- nant and premalignant cells. addressed by future studies before these agents can be used routinely. Furthermore, demethylating drugs can activate genes that either do not have CpG islands in their promoters 108 (FIG. 4) or are unmethylated 30,109 , and little is known about the spectrum of genes that might be activated in normal cells after exposure to such agents. Another impediment to the use of demethyla- tion therapy is the tendency of methylation to spread back into the demethylated CpG island after the inhibitor is removed. This process has been shown to lead to de novo methylation and to the re-silencing of the target gene 110 . Detection and prognosis The de novo methylation of CpG islands occurs early in the process of carcinogenesis and can even be detected in the apparently normal epithelium of patients ? a process that is associated with ageing and inflamma- tion 99 . As methylated alleles can be detected with a very high degree of sensitivity, there is great interest in using methylation as a potential early detection system for cancer (FIG. 6). Hypermethylated CpG islands have been detected in cancer patients, in DNA derived from spu- tum, serum and urine samples (in which the DNA is released presumably after apoptosis or necrosis of the cancer). The results of some of these studies indicate that this approach might offer some promise for early diagnosis and/or risk assessment 111 .However,whether these technologies are of practical use has not yet been shown and might be compromised by the fact that some methylation changes occur in apparently normal epithelial cells. As such, they might indicate a risk for cancer rather than show the presence of cancer in an individual. There also remains the problem of knowing what type of cancer might be releasing methylated DNA into the circulation, where it can be detected against a background of normally methylated lymphocyte DNAs and other DNAs. Nevertheless, although some methyla- tion changes seem to be common to many kinds of can- cer, there are clearly tissue-specific differences that exist. For example, the CDKN2B gene promoter is frequently 1. Bird, A. DNA methylation patterns and epigenetic memory. Genes Dev. 16, 6?21 (2002). An excellent review of mammalian DNA methylation. 2. Takai, D. & Jones, P. A. Comprehensive analysis of CpG islands in human chromosomes 21 and 22. Proc. Natl Acad. Sci. USA 99, 3740?3745 (2002). 3. Baylin, S. B. & Herman, J. G. DNA hypermethylation in tumorigenesis: epigenetics joins genetics. Trends Genet. 16, 168?174 (2000). 4. Jones, P. A. & Laird, P. W. Cancer epigenetics comes of age. Nature Genet. 21, 163?167 (1999). 5. Esteller, M. et al. Inactivation of the DNA repair gene O 6 -methylguanine-DNA methyltransferase by promoter hypermethylation is associated with G to A mutations in K-ras in colorectal tumorigenesis. Cancer Res. 60, 2368?2371 (2000). 6. Herman, J. G. et al. Distinct patterns of inactivation of p15 INK4B and p16 INK4A characterize the major types of hematological malignancies. Cancer Res. 57, 837?841 (1997). 7. Burbee, D. G. et al. Epigenetic inactivation of RASSF1A in lung and breast cancers and malignant phenotype suppression. J. Natl Cancer Inst. 93, 691?699 (2001). 8. Dammann, R. et al. Epigenetic inactivation of a RAS association domain family protein from the lung tumour suppressor locus 3p21.3. Nature Genet. 25, 315?319 (2000). 9. Herman, J. G. et al. Silencing of the VHL tumor- suppressor gene by DNA methylation in renal carcinoma. Proc. Natl Acad. Sci. USA 91, 9700?9704 (1994). 10. Esteller, M. et al. Promoter hypermethylation and BRCA1 inactivation in sporadic breast and ovarian tumors. J. Natl Cancer Inst. 92, 564?569 (2000). 11. Esteller, M. et al. Epigenetic inactivation of LKB1 in primary tumors associated with the Peutz?Jeghers syndrome. Oncogene 19, 164?168 (2000). 12. Hedenfalk, I. et al. Gene-expression profiles in hereditary breast cancer. N. Engl. J. Med. 344, 539?548 (2001). 13. Van?t Veer, L. J. et al. Gene expression profiling predicts clinical outcome of breast cancer. Nature 415, 530?536 (2002). 14. Knudson, A. G. Chasing the cancer demon. Annu. Rev. Genet. 34, 1?19 (2000). 15. Myohanen, S. K., Baylin, S. B. & Herman, J. G. Hypermethylation can selectively silence individual p16 ink4A alleles in neoplasia. Cancer Res. 58, 591?593 (1998). 16. Grady, W. M. et al. Methylation of the CDH1 promoter as the second genetic hit in hereditary diffuse gastric cancer. Nature Genet. 26, 16?17 (2000). 17. Esteller, M. et al. DNA methylation patterns in hereditary human cancers mimic sporadic tumorigenesis. Hum. Mol. Genet. 10, 3001?3007 (2001). 18. Kane, M. F. et al. Methylation of the hMLH1 promoter correlates with lack of expression of hMLH1 in sporadic colon tumors and mismatch repair-defective human tumor cell lines. Cancer Res. 57, 808?811 (1997). 19. Herman, J. G. et al. Incidence and functional consequences of hMLH1 promoter hypermethylation in colorectal carcinoma. Proc. Natl Acad. Sci. USA 95, 6870?6875 (1998). References 18 and 19 provided an important link between epigenetic and genetic changes in colon cancer. Reference 19 illustrates how demethylation of an aberrantly silenced gene in cancer can restore gene expression and function. 20. Nakagawa, H. et al. Age-related hypermethylation of the 5? region of MLH1 in normal colonic mucosa is associated with microsatellite-unstable colorectal cancer development. Cancer Res. 61, 6991?6995 (2001). 21. Esteller, M. et al. hMLH1 promoter hypermethylation is an NATURE REVIEWS | GENETICS VOLUME 3 | JUNE 2002 | 427 REVIEWS early event in human endometrial tumorigenesis. Am. J. Pathol. 155, 1767?1772 (1999). 22. Esteller, M., Hamilton, S. R., Burger, P. C., Baylin, S. B. & Herman, J. G. Inactivation of the DNA repair gene O 6 -methylguanine-DNA methyltransferase by promoter hypermethylation is a common event in primary human neoplasia. Cancer Res. 59, 793?797 (1999). 23. Esteller, M. et al. Promoter hypermethylation of the DNA repair gene O(6)-methylguanine-DNA methyltransferase is associated with the presence of G:C to A:T transition mutations in p53 in human colorectal tumorigenesis. Cancer Res. 61, 4689?4692 (2001). 24. Wales, M. M. et al. p53 activates expression of HIC-1, a new candidate tumour suppressor gene on 17p13.3. Nature Med. 1, 570?577 (1995). 25. Huang, T. H., Perry, M. R. & Laux, D. E. Methylation profiling of CpG islands in human breast cancer cells. Hum. Mol. Genet. 8, 459?470 (1999). 26. Toyota, M. et al. Identification of differentially methylated sequences in colorectal cancer by methylated CpG island amplification. Cancer Res. 59, 2307?2312 (1999). 27. Costello, J. F. et al. Aberrant CpG-island methylation has non-random and tumour-type-specific patterns. Nature Genet. 24, 132?138 (2000). 28. Gonzalgo, M. L. et al. Identification and characterization of differentially methylated regions of genomic DNA by methylation-sensitive arbitrarily primed PCR. Cancer Res. 57, 594?599 (1997). 29. Gitan, R. S., Shi, H., Chen, C. M., Yan, P. S. & Huang, T. H. Methylation-specific oligonucleotide microarray: a new potential for high-throughput methylation analysis. Genome Res. 12, 158?164 (2002). 30. Suzuki, H. et al. A genomic screen for genes upregulated by demethylation and HDAC inhibition in human colorectal cancer. Nature Genet. May 6 [epub ahead of print] 2002. 31. Coulondre, C., Miller, J. H., Farabaugh, P. J. & Gilbert, W. Molecular basis of base substitution hotspots in Escherichia coli. Nature 274, 775?780 (1978). 32. Rideout, W. M. III, Coetzee, G. A., Olumi, A. F. & Jones, P. A. 5-Methylcytosine as an endogenous mutagen in the human LDL receptor and p53 genes. Science 249, 1288?1290 (1990). This paper provides an example of how the base methylcytosine provides an endogenous site of mutagenic potential in the coding region of important genes. 33. Pfeifer, G. P., Tang, M. & Denissenko, M. F. Mutation hotspots and DNA methylation. Curr. Top. Microbiol. Immunol. 249, 1?19 (2000). 34. Yoon, J. H. et al. Methylated CpG dinucleotides are the preferential targets for G-to-T transversion mutations induced by benzo[a]pyrene diol epoxide in mammalian cells: similarities with the p53 mutation spectrum in smoking- associated lung cancers. Cancer Res. 61, 7110?7117 (2001). 35. Magewu, A. N. & Jones, P. A. Ubiquitous and tenacious methylation of the CpG site in codon 248 of the p53 gene may explain its frequent appearance as a mutational hotspot in human cancer. Mol. Cell. Biol. 14, 4225?4232 (1994). 36. Feinberg, A. P. & Vogelstein, B. Hypomethylation distinguishes genes of some human cancers from their normal counterparts. Nature 301, 89?92 (1983). 37. Feinberg, A. P., Gehrke, C. W., Kuo, K. C. & Ehrlich, M. Reduced genomic 5-methylcytosine content in human colonic neoplasia. Cancer Res. 48, 1159?1161 (1988). These two papers outline the presence of DNA hypomethylation in human tumours. 38. Xu, G. L. et al. Chromosome instability and immunodeficiency syndrome caused by mutations in a DNA methyltransferase gene. Nature 402, 187?191 (1999). 39. Okano, M., Bell, D. W., Haber, D. A. & Li, E. DNA methyltransferases Dnmt3a and Dnmt3b are essential for de novo methylation and mammalian development. Cell 99, 247?257 (1999). An important paper that discusses the function of the mammalian DNMTs in establishing and maintaining DNA methylation patterns. 40. Hansen, R. S. et al. The DNMT3B DNA methyltransferase gene is mutated in the ICF immunodeficiency syndrome. Proc. Natl Acad. Sci. USA 96, 14412?14417 (1999). 41. Qu, G. Z., Grundy, P. E., Narayan, A. & Ehrlich, M. Frequent hypomethylation in Wilms tumors of pericentromeric DNA in chromosomes 1 and 16. Cancer Genet. Cytogenet. 109, 34?39 (1999). 42. Chen, R. Z., Pettersson, U., Beard, C., Jackson-Grusby, L. & Jaenisch, R. DNA hypomethylation leads to elevated mutation rates. Nature 395, 89?93 (1998). 43. Chan, M. F. et al. Reduced rates of gene loss, gene silencing, and gene mutation in Dnmt1-deficient embryonic stem cells. Mol. Cell. Biol. 21, 7587?7600 (2001). 44. Kass, S. U., Landsberger, N. & Wolffe, A. P. DNA methylation directs a time-dependent repression of transcription initiation. Curr. Biol. 7, 157?165 (1997). 45. Wutz, A., Rasmussen, T. P. & Jaenisch, R. Chromosomal silencing and localization are mediated by different domains of Xist RNA. Nature Genet. 30, 167?174 (2002). 46. Jones, P. A. & Taylor, S. M. Cellular differentiation, cytidine analogs and DNA methylation. Cell 20, 85?93 (1980). This paper was the first to show that 5-azacytidine can induce demethylation of DNA. 47. Murzina, N., Verreault, A., Laue, E. & Stillman, B. Heterochromatin dynamics in mouse cells: interaction between chromatin assembly factor 1 and HP1 proteins. Mol. Cell 4, 529?540 (1999). 48. Struhl, K. Histone acetylation and transcriptional regulatory mechanisms. Genes Dev. 12, 599?606 (1998). 49. Taddei, A., Maison, C., Roche, D. & Almouzni, G. Reversible disruption of pericentric heterochromatin and centromere function by inhibiting deacetylases. Nature Cell Biol. 3, 114?120 (2001). 50. Bird, A. P. & Wolffe, A. P. Methylation-induced repression ? belts, braces, and chromatin. Cell 99, 451?454 (1999). 51. Jones, P. L. et al. Methylated DNA and MeCP2 recruit histone deacetylase to repress transcription. Nature Genet. 19, 187?191 (1998). 52. Ng, H. H. et al. MBD2 is a transcriptional repressor belonging to the MeCP1 histone deacetylase complex. Nature Genet. 23, 58?61 (1999). 53. Wade, P. A. et al. Mi-2 complex couples DNA methylation to chromatin remodelling and histone deacetylation. Nature Genet. 23, 62?66 (1999). References 51?53 were the first to link DNA methylation with gene silencing by identifying the chromatin-remodelling proteins with which methylcytosine-binding proteins associate. 54. Bell, A. C. & Felsenfeld, G. Methylation of a CTCF- dependent boundary controls imprinted expression of the Igf2 gene. Nature 405, 482?485 (2000). 55. Hark, A. T. et al. CTCF mediates methylation-sensitive enhancer-blocking activity at the H19/Igf2 locus. Nature 405, 486?489 (2000). 56. Mikovits, J. A. et al. Infection with human immunodeficiency virus type 1 upregulates DNA methyltransferase, resulting in de novo methylation of the gamma interferon (IFN-?) promoter and subsequent downregulation of IFN-? production. Mol. Cell. Biol. 18, 5166?5177 (1998). 57. Graff, J. R., Herman, J. G., Myohanen, S., Baylin, S. B. & Vertino, P. M. Mapping patterns of CpG island methylation in normal and neoplastic cells implicates both upstream and downstream regions in de novo methylation. J. Biol. Chem. 272, 22322?22329 (1997). 58. Graff, J. R., Gabrielson, E., Fujii, H., Baylin, S. B. & Herman, J. G. Methylation patterns of the E-cadherin 5? CpG island are unstable and reflect the dynamic, heterogeneous loss of E-cadherin expression during metastatic progression. J. Biol. Chem. 275, 2727?2732 (2000). 59. Jenuwein, T. & Allis, C. D. Translating the histone code. Science 293, 1074?1080 (2001). An important review of the latest thinking on the histone code, which marks and/or establishes regions of transcriptionally repressive versus transcriptionally active chromatin. 60. Bannister, A. J. et al. Selective recognition of methylated lysine 9 on histone H3 by the HP1 chromo domain. Nature 410, 120?124 (2001). 61. Nakayama, J., Rice, J. C., Strahl, B. D., Allis, C. D. & Grewal, S. I. Role of histone H3 lysine 9 methylation in epigenetic control of heterochromatin assembly. Science 292, 110?113 (2001). 62. Noma, K., Allis, C. D. & Grewal, S. I. Transitions in distinct histone H3 methylation patterns at the heterochromatin domain boundaries. Science 293, 1150?1155 (2001). 63. Zhang, Y. & Reinberg, D. Transcription regulation by histone methylation: interplay between different covalent modifications of the core histone tails. Genes Dev. 15, 2343?2360 (2001). 64. Tamaru, H. & Selker, E. U. A histone H3 methyltransferase controls DNA methylation in Neurospora crassa. Nature 414, 277?283 (2001). 65. Jackson, J. P., Lindroth, A. M., Cao, X. & Jacobsen, S. E. Control of CpNpG DNA methylation by the KRYPTONITE histone H3 methyltransferase. Nature 416, 556?560 (2002). References 64 and 65 provide the exciting concept that a specific site of histone methylation might specify where DNA methylation occurs. 66. Cameron, E. E., Bachman, K. E., Myohanen, S., Herman, J. G. & Baylin, S. B. Synergy of demethylation and histone deacetylase inhibition in the re-expression of genes silenced in cancer. Nature Genet. 21, 103?107 (1999). This paper illustrates the synergy between DNA methylation and histone deacetylation in the aberrant silencing of tumour-suppressor genes and the dominant role of DNA methylation. 67. Nguyen, C. T., Gonzales, F. A. & Jones, P. A. Altered chromatin structure associated with methylation-induced gene silencing in cancer cells: correlation of accessibility, methylation, MeCP2 binding and acetylation. Nucleic Acids Res. 29, 4598?4606 (2001). 68. Magdinier, F. & Wolffe, A. P. Selective association of the methyl-CpG binding protein MBD2 with the silent p14/p16 locus in human neoplasia. Proc. Natl Acad. Sci. USA 98, 4990?4995 (2001). 69. El-Osta, A., Kantharidis, P., Zalcberg, J. R. & Wolffe, A. P. Precipitous release of methyl-CpG binding protein 2 and histone deacetylase 1 from the methylated human multidrug resistance gene (MDR1) on activation. Mol. Cell. Biol. 22, 1844?1857 (2002). 70. Robertson, K. D. et al. The human DNA methyltransferases (DNMTs) 1, 3a and 3b: coordinate mRNA expression in normal tissues and overexpression in tumors. Nucleic Acids Res. 27, 2291?2298 (1999). 71. De Marzo, A. M. et al. Abnormal regulation of DNA methyltransferase expression during colorectal carcinogenesis. Cancer Res. 59, 3855?3860 (1999). 72. Wu, J. et al. Expression of an exogenous eukaryotic DNA methyltransferase gene induces transformation of NIH 3T3 cells. Proc. Natl Acad. Sci. USA 90, 8891?8895 (1993). 73. Bakin, A. V. & Curran, T. Role of DNA 5-methylcytosine transferase in cell transformation by fos. Science 283, 387?390 (1999). 74. Laird, P. W. et al. Suppression of intestinal neoplasia by DNA hypomethylation. Cell 81, 197?205 (1995). This paper presented a role for DNMT1 in tumorigenesis and the concept that decreasing the action of this enzyme could be a therapeutic goal for treating cancer. 75. Fuks, F., Burgers, W. A., Brehm, A., Hughes-Davies, L. & Kouzarides, T. DNA methyltransferase Dnmt1 associates with histone deacetylase activity. Nature Genet. 24, 88?91 (2000). 76. Robertson, K. D. et al. DNMT1 forms a complex with Rb, E2F1 and HDAC1 and represses transcription from E2F- responsive promoters. Nature Genet. 25, 338?342 (2000). 77. Rountree, M. R., Bachman, K. E. & Baylin, S. B. DNMT1 binds HDAC2 and a new co-repressor, DMAP1, to form a complex at replication foci. Nature Genet. 25, 269?277 (2000). 78. Fuks, F., Burgers, W. A., Godin, N., Kasai, M. & Kouzarides, T. Dnmt3a binds deacetylases and is recruited by a sequence-specific repressor to silence transcription. EMBO J. 20, 2536?2544 (2001). 79. Bachman, K. E., Rountree, M. R. & Baylin, S. B. Dnmt3a and Dnmt3b are transcriptional repressors that exhibit unique localization properties to heterochromatin. J. Biol. Chem. 276, 32282?32287 (2001). References 75?79 report that mammalian DNMTs probably have additional roles in transcriptional silencing by recruiting histone deacetylases, acting as transcriptional repressors and serving as binding platforms for transcriptional co-repressors. 80. Taddei, A., Roche, D., Sibarita, J. B., Turner, B. M. & Almouzni, G. Duplication and maintenance of heterochromatin domains. J. Cell Biol. 147, 1153?1166 (1999). 81. Liang, G. et al. Cooperativity between DNA methyltransferases in the maintenance methylation of repetitive elements. Mol. Cell. Biol. 22, 480?491 (2002). 82. Velicescu, M. et al. Cell division is required for de novo methylation of CpG islands in bladder cancer cells. Cancer Res. 62, 2378?2384 (2002). 83. Di Croce, L. et al. Methyltransferase recruitment and DNA hypermethylation of target promoters by an oncogenic transcription factor. Science 295, 1079?1082 (2002). 84. Rhee, I. et al. CpG methylation is maintained in human cancer cells lacking DNMT1. Nature 404, 1003?1007 (2000). 85. Rhee, I. et al. DNMT1 and DNMT3b cooperate to silence genes in human cancer cells. Nature 416, 552?556 (2002). References 83?85 report that the DNMTs might act cooperatively in maintaining aberrant gene silencing. Reference 85 provides evidence for epigenetics in the silencing of the tumour-suppressor gene CDKN2A. 86. Panning, B. & Jaenisch, R. DNA hypomethylation can activate Xist expression and silence X-linked genes. Genes Dev. 10, 1991?2002 (1996). 87. Jackson-Grusby, L. et al. Loss of genomic methylation causes p53-dependent apoptosis and epigenetic deregulation. Nature Genet. 27, 31?39 (2001). 88. Sun, F. L. & Elgin, S. C. Putting boundaries on silence. Cell 99, 459?462 (1999). 428 | JUNE 2002 | VOLUME 3 www.nature.com/reviews/genetics REVIEWS 89. Turker, M. S. & Bestor, T. H. Formation of methylation patterns in the mammalian genome. Mutat. Res. 386, 119?130 (1997). 90. Issa, J. P. et al. Methylation of the oestrogen receptor CpG island links ageing and neoplasia in human colon. Nature Genet. 7, 536?540 (1994). 91. Issa, J. P. Aging, DNA methylation and cancer. Crit. Rev. Oncol. Hematol. 32, 31?43 (1999). Papers 90 and 91 discuss a possible link between aberrant DNA methylation, ageing and cancer. 92. Cairns, J. Cancer: Science and Society (W. H. Freeman & Co., San Francisco, California, 1978). 93. Cameron, E. E., Baylin, S. B. & Herman, J. G. p15(INK4B) CpG island methylation in primary acute leukemia is heterogeneous and suggests density as a critical factor for transcriptional silencing. Blood 94, 2445?2451 (1999). 94. Greller, L. D., Tobin, F. L. & Poste, G. Tumor heterogeneity and progression: conceptual foundations for modeling. Invasion Metastasis 16, 177?208 (1996). 95. Fidler, I. J. The biology of cancer metastasis, or ?you cannot fix it if you do not know how it works?. Bioessays 13, 551?554 (1991). 96. Nuovo, G. J., Plaia, T. W., Belinsky, S. A., Baylin, S. B. & Herman, J. G. In situ detection of the hypermethylation- induced inactivation of the p16 gene as an early event in oncogenesis. Proc. Natl Acad. Sci. USA 96, 12754?12759 (1999). 97. Wong, D. J., Foster, S. A., Galloway, D. A. & Reid, B. J. Progressive region-specific de novo methylation of the p16 CpG island in primary human mammary epithelial cell strains during escape from M(0) growth arrest. Mol. Cell. Biol. 19, 5642?5651 (1999). 98. Kiyono, T. et al. Both Rb/p16 INK4a inactivation and telomerase activity are required to immortalize human epithelial cells. Nature 396, 84?88 (1998). 99. Issa, J. P., Ahuja, N., Toyota, M., Bronner, M. P. & Brentnall, T. A. Accelerated age-related CpG island methylation in ulcerative colitis. Cancer Res. 61, 3573?3577 (2001). 100. Comijn, J. et al. The two-handed E box binding zinc finger protein SIP1 downregulates E-cadherin and induces invasion. Mol. Cell 7, 1267?1278 (2001). 101. Mareel, M., Bracke, M. & Van Roy, F. Cancer metastasis: negative regulation by an invasion-suppressor complex. Cancer Detect. Prev. 19, 451?464 (1995). 102. Cihak, A. Biological effects of 5-azacytidine in eukaryotes. Oncology 30, 405?422 (1974). 103. Lubbert, M. DNA methylation inhibitors in the treatment of leukemias, myelodysplastic syndromes and hemoglobinopathies: clinical results and possible mechanisms of action. Curr. Top. Microbiol. Immunol. 249, 135?164 (2000). 104. Lin, X. et al. Reversal of GSTP1 CpG island hypermethylation and reactivation of pi-class glutathione S-transferase (GSTP1) expression in human prostate cancer cells by treatment with procainamide. Cancer Res. 61, 8611?8616 (2001). 105. Wolf, S. F. & Migeon, B. R. Studies of X chromosome DNA methylation in normal human cells. Nature 295, 667?671 (1982). 106. Mohandas, T., Sparkes, R. S. & Shapiro, L. J. Reactivation of an inactive X human chromosome: evidence for X inactivation by DNA methylation. Science 211, 393?396 (1981). 107. Jaenisch, R., Schnieke, A. & Harbers, K. Treatment of mice with 5-azacytidine efficiently activates silent retroviral genomes in different tissues. Proc. Natl Acad. Sci. USA 82, 1451?1455 (1985). 108. Liang, G., Gonzales, F. A., Jones, P. A., Orntoft, T. F. & Thykjaer, T. Analysis of gene induction in human fibroblasts and bladder cancer cells exposed to the methylation inhibitor 5-aza-2?-deoxycytidine. Cancer Res. 62, 961?966 (2002). 109. Soengas, M. S. et al. Inactivation of the apoptosis effector Apaf-1 in malignant melanoma. Nature 409, 207?211 (2001). 110. Bender, C. M. et al. Roles of cell division and gene transcription in the methylation of CpG islands. Mol. Cell. Biol. 19, 6690?6698 (1999). 111. Sidransky, D. Emerging molecular markers of cancer. Nature Rev. Cancer 2, 210?219 (2002). 112. Esteller, M., Corn, P. G., Baylin, S. B. & Herman, J. G. A gene hypermethylation profile of human cancer. Cancer Res. 61, 3225?3229 (2001). 113. Adorjan, P. et al. Tumour class prediction and discovery by microarray-based DNA methylation analysis. Nucleic Acids Res. 30, e21 (2002). Acknowledgements P.A.J. thanks C. Nguyen for help with some of the figures. P.A.J. and S.B.B. are supported by grants from the National Cancer Institute and the National Institute of Environmental Health Sciences. Peter A. Jones is a shareholder and consultant for Epigenomics AG, a company with interests in the development of methylation-based diagnosis and detection technologies. Stephen B. Baylin is a consultant for Tibotech-Virco, which is developing methylation-based diagnosis technologies and is eligible for royalties from the sales of any products that are marketed. Online links DATABASES The following terms in this article are linked online to: CancerNet: http://www.nci.nih.gov breast cancer | colon cancer | leukaemia | lung cancer | prostate cancer LocusLink: http://www.ncbi.nlm.nih.gov/LocusLink ATM | APC | BRCA1 | BRCA2 | CBX5 | CDH1 | CDKN2A | CDKN2B | DNMT1 | Dnmt1 | DNMT3A | Dnmt3a | DNMT3B | Dnmt3b | ESR1 | FHIT | FOS | GSTP1 | HDAC2 | HIC1 | Hp1? | MBD1 | MBD2 | MDR1 | MECP2 | MGMT | MLH1 | MSH2 | NF1 | NF2 | PML | PTCH | PTEN | RAR? | RARB | RB1 | SMAD4 | SMARCA3 | SMARCB1 | STK11 | TIMP3 | TP53 | TP73 | Trp53 | VHL | XIST Medscape DrugInfo: http://promini.medscape.com/drugdb procanamide Mouse Genome Informatics: http://www.informatics.jax.org min OMIM: http://www.ncbi.nlm.nih.gov/Omim ICF syndrome | promyelocytic leukaemia FURTHER INFORMATION Cancer Genetics Web: http://www.cancerindex.org/geneweb/index.htm Encyclopedia of Life Sciences: http://www.els.net Cancer | DNA methylation Guide to Internet Resources for Cancer: http://www.cancerindex.org/clinks1.htm Access to this interactive links box is free online. "
Add Content to Group
|
Bookmark
|
Keywords
|
Flag Inappropriate
share
Close
Digg
Facebook
MySpace
Google+
Comments
Close
Please Post Your Comment
*
The Comment you have entered exceeds the maximum length.
Submit
|
Cancel
*
Required
Comments
Please Post Your Comment
No comments yet.
Save Note
Note
View
Public
Private
Friends & Groups
Friends
Groups
Save
|
Cancel
|
Delete
Please provide your notes.
Next
|
Prev
|
Close
|
Edit
|
Delete
Genetics
Gene Inheritance and Transmission
Gene Expression and Regulation
Nucleic Acid Structure and Function
Chromosomes and Cytogenetics
Evolutionary Genetics
Population and Quantitative Genetics
Genomics
Genes and Disease
Genetics and Society
Cell Biology
Cell Origins and Metabolism
Proteins and Gene Expression
Subcellular Compartments
Cell Communication
Cell Cycle and Cell Division
Scientific Communication
Career Planning
Loading ...
Scitable Chat
Register
|
Sign In
Visual Browse
Close
Comments
CloseComments
Please Post Your Comment