《遗传学》课程教学资源（学科前沿）遗传与表观遗传 Phenotypic plasticity and the epigenetics of human disease

NATUREIVol 447 24 May 2007 do:10.1038/nature05919 INSIGHT REVIEW Phenotypic plasticity and the epigenetics of human disease Andrew P Feinberg It is becoming clear that epigenetic changes are involved in human disease as well as during normal development. a unifying theme of disease epigenetics is defects in phenotypic plasticity- ability to change their behaviour in response to internal or external environmental cues. This model proposes that hereditary disorders of the epigenetic apparatus lead to developmental defects that cancer epigenetics involves disruption of the stem-cell programme, and that common diseases with late-onset phenotypes involve interactions between the epigenome, the genome and the environment Increased understanding of epigenetic disease mechanisms could lead to disease- risk stratification for targeted intervention and to targeted therapies. The original definition of epigenetics by Waddington in 1942(ref. 1) Epigenetic disease genes the idea that that phenotype arises from genotype throu e with-Wiedemann syndrome, which is characterized by prenatal over- of the first class of monogenic epigenetic disease is Beck- biology. The modern definition of epigenetics is information heritable growth, a midline abdominal wall and other malformations, and cancer. during cell division other than the DNA sequence itself. It is becoming Studies of patients with Beckwith-Wiedemann syndrome have taught increasingly clear that there is great overlap between these two defini- us a great deal about the mechanisms of normal imprinting. Patients tions-developmental processes are regulated largely by epigenetics, with Beckwith-Wiedemann syndrome show disrupted imprinting of because different cell types maintain their fate during cell division even either or both of two neighbouring imprinted subdomains on 11p15, though their DNA sequences are essentially the e saine. revealing clustering of imprinted genes(Fig. 2). The first is the H19/ What is epigenetic disease? Genetic lesions-sequence changes, IGF2 (imprinted, maternally expressed, untranslated mRNA/insulin breakages and deletions- can be easily visualized, but what about like growth factor 2)imprinted subdomain, which is regulated by a dif- epigenetic lesions? Several defects in the epigenome are known to lead ferentially methylated region(DMR)that is methylated on the paternal to disease(Fig. 1), including changes in the localized or global density but not the maternal allele. The second subdomain includes p57(a of DNA methylation, and incorrect histone modification. Other defects cyclin-dependent kinase inhibitor), TSSC3 (a pleckstrin homology-like that might cause disease involve altered distribution or function of chro- domain), SLC22AI(an organic cation transporter), KLQTI(a voltage matin-modifying proteins that, in turn, leads to aberrant gene expres- gated potassium channel)and LITI (KCNQI overlapping transcript 1) sion. Another intriguing possibility is the disruption of higher-order with the subdomain regulated by a second DMR, upstream of LITl, that loop structure in disease(Fig. 1). normally methylated on the maternal but not the paternal allele(see Studies of monogenic disorders involving imprinted genes or the ref 2 for a review). In patients with Beckwith-Wiedemann syndrome, late the epigenome. Two decades ago, cancer epigenetics was viewed tions show loss of normal imprinted gene regulation"(Fig. 2).Some with some scepticism, but it is now widely accepted. However, impor- patients with Beckwith-Wiedemann syndrome show loss of imprinting tant questions about the mechanism, timing and consequences of of IGF2, which leads to a double dose of this autocrine factor, result- epigenetic disruption remain. Whether other common diseases have ing in tissue overgrowth and increased cancer risk. The mechanism epigenetic basis is still open to speculation, but if they do, this holds involves aberrant methylation of the maternal H19 DMR(Fig. 2). Other great promise for medicine. patients with this syndrome show localized abnormalities of allele Here I review the epigenetics of single-gene disorders, cancer and specific chromatin modificationaffecting P57 2, a cyclin-dependent ommon complex diseases. I suggest that a common theme to disease kinase inhibitor( Fig. 2). Thus, Beckwith-Wiedemann syndrome illus- epigenetics is the disruption of phenotypic plasticity -the ability of trates hierarchical organization of epigenetic regulation in progressively cells to change their behaviour in response to internal or external envi- larger domains. ronmental cues-an idea that resonates with Waddingtons original A pair of imprinted-gene disorders that are associated with men definition of epigenetics. I also discuss therapeutic implications of dis- tal retardation- Prader-Willi syndrome and Angelman syndrome ease epige involve adjacent reciprocally nted genes, SNRPN(small nuclear ribonucleoprotein polypeptide N)and UBE3A (ubiquitin-protein Imprinted gene disorders ligase E3A), on chromosome 15. Microdeletions in patients with both There are two classes of monogenic epigenetic disease: those involving Prader-Willi syndrome and Angelman syndrome reveal the location genes that are regulated epigenetically, such as imprinted genes, and of the domain that regulates imprinting of both genes those that affect the epigenome as a whole, such as modifiers of DNa Another pair of human disorders caused by different altera- methylation. tions at the same locus are Albright hereditary osteodystrophy and Department of Medicine and Center for Epigenetics, Institute for Basic Biomedical Sciences, Johns Hopkins University School of Medicine, 720 Rutland Avenue, Baltimore, maryland 21205, USA @2007 Nature Publishing Group

The original definition of epigenetics by Waddington in 1942 (ref. 1) — the idea that that phenotype arises from genotype through pro￾grammed change — is now what is considered to be developmental biology. The modern definition of epigenetics is information heritable during cell division other than the DNA sequence itself. It is becoming increasingly clear that there is great overlap between these two defini￾tions — developmental processes are regulated largely by epigenetics, because different cell types maintain their fate during cell division even though their DNA sequences are essentially the same. What is epigenetic disease? Genetic lesions — sequence changes, breakages and deletions — can be easily visualized, but what about epigenetic lesions? Several defects in the epigenome are known to lead to disease (Fig. 1), including changes in the localized or global density of DNA methylation, and incorrect histone modification. Other defects that might cause disease involve altered distribution or function of chro￾matin-modifying proteins that, in turn, leads to aberrant gene expres￾sion. Another intriguing possibility is the disruption of higher-order loop structure in disease (Fig. 1). Studies of monogenic disorders involving imprinted genes or the epigenetic machinery have revealed a great deal about the nature of the cis-acting regulatory marks and trans-acting factors that modu￾late the epigenome. Two decades ago, cancer epigenetics was viewed with some scepticism, but it is now widely accepted. However, impor￾tant questions about the mechanism, timing and consequences of epigenetic disruption remain. Whether other common diseases have an epigenetic basis is still open to speculation, but if they do, this holds great promise for medicine. Here I review the epigenetics of single-gene disorders, cancer and common complex diseases. I suggest that a common theme to disease epigenetics is the disruption of phenotypic plasticity — the ability of cells to change their behaviour in response to internal or external envi￾ronmental cues — an idea that resonates with Waddington’s original definition of epigenetics. I also discuss therapeutic implications of dis￾ease epigenetics. Imprinted gene disorders There are two classes of monogenic epigenetic disease: those involving genes that are regulated epigenetically, such as imprinted genes, and those that affect the epigenome as a whole, such as modifiers of DNA methylation. Epigenetic disease genes An example of the first class of monogenic epigenetic disease is Beck￾with–Wiedemann syndrome, which is characterized by prenatal over￾growth, a midline abdominal wall and other malformations, and cancer. Studies of patients with Beckwith–Wiedemann syndrome have taught us a great deal about the mechanisms of normal imprinting. Patients with Beckwith–Wiedemann syndrome show disrupted imprinting of either or both of two neighbouring imprinted subdomains on 11p15, revealing clustering of imprinted genes (Fig. 2). The first is the H19/ IGF2 (imprinted, maternally expressed, untranslated mRNA/insulin￾like growth factor 2) imprinted subdomain, which is regulated by a dif￾ferentially methylated region (DMR) that is methylated on the paternal but not the maternal allele. The second subdomain includes p57KIP2 (a cyclin-dependent kinase inhibitor), TSSC3 (a pleckstrin homology-like domain), SLC22A1 (an organic cation transporter), KvLQT1 (a voltage￾gated potassium channel) and LIT1 (KCNQ1 overlapping transcript 1), with the subdomain regulated by a second DMR, upstream of LIT1, that is normally methylated on the maternal but not the paternal allele (see ref. 2 for a review). In patients with Beckwith–Wiedemann syndrome, microdeletions within each imprinted subdomain have confirmed the regulatory role of these sequences, because individuals with these del￾etions show loss of normal imprinted gene regulation3,4 (Fig. 2). Some patients with Beckwith–Wiedemann syndrome show loss of imprinting of IGF2, which leads to a double dose of this autocrine factor, result￾ing in tissue overgrowth and increased cancer risk5 . The mechanism involves aberrant methylation of the maternal H19 DMR (Fig. 2). Other patients with this syndrome show localized abnormalities of allele￾specific chromatin modification6 affecting p57KIP2, a cyclin-dependent kinase inhibitor (Fig. 2). Thus, Beckwith–Wiedemann syndrome illus￾trates hierarchical organization of epigenetic regulation in progressively larger domains. A pair of imprinted-gene disorders that are associated with men￾tal retardation — Prader–Willi syndrome and Angelman syndrome — involve adjacent reciprocally imprinted genes, SNRPN (small nuclear ribonucleoprotein polypeptide N) and UBE3A (ubiquitin–protein ligase E3A), on chromosome 15. Microdeletions in patients with both Prader–Willi syndrome and Angelman syndrome reveal the location of the domain that regulates imprinting of both genes7 . Another pair of human disorders caused by different altera￾tions at the same locus are Albright hereditary osteodystrophy and Phenotypic plasticity and the epigenetics of human disease Andrew P. Feinberg1 It is becoming clear that epigenetic changes are involved in human disease as well as during normal development. A unifying theme of disease epigenetics is defects in phenotypic plasticity — cells’ ability to change their behaviour in response to internal or external environmental cues. This model proposes that hereditary disorders of the epigenetic apparatus lead to developmental defects, that cancer epigenetics involves disruption of the stem-cell programme, and that common diseases with late-onset phenotypes involve interactions between the epigenome, the genome and the environment. Increased understanding of epigenetic￾disease mechanisms could lead to disease-risk stratification for targeted intervention and to targeted therapies. 1 Department of Medicine and Center for Epigenetics, Institute for Basic Biomedical Sciences, Johns Hopkins University School of Medicine, 720 Rutland Avenue, Baltimore, Maryland 21205, USA. 433 NATURE|Vol 447|24 May 2007|doi:10.1038/nature05919 INSIGHT REVIEW ￾

   

INSIGHT REVIEW NATURE Vol 447 24 May 2007 pseudohypoparathyroidism type IA (PHPIA). Albright hereditary system development as well as developmental dysmorphology, which steodystrophy is characterized by short stature and ectopic calcifica- could involve failure of heterochromatin formation and might result tions, and caused by mutational inactivation of guanine nucleotide regu- from DNMT3B having a role in immunoglobulin gene silencing and latory protein(encoded by GNASI) PHPIA is a more severe phenotype reactivation f multiple hormone resistance caused by tissue-specific differential A striking example of developmental disruption caused by mutations imprinting of splice variants of the same gene. It is unlikely that this in a chromatin factor gene is alpha-thalassaemia/mental retardation, omplex pattern of imprinting would have been understood without X-linked(ATRX) syndrome, the gene for which is a helicase involved concomitant clinical studies in chromatin remodelling Mutations lead to defects in psychomotor, urogenital and haematopoietic development, with maturational defects Single-gene disorders of the epigenetic machinery in erythroid precursors resembling those of alpha-thalassaemia The other class of monogenic epigenetic disease involves genes that Rubinstein-Taybi syndrome involves the CREb(cyclic-AMP respon ncode comphese genes causes developmental disorders. For example, acetyltransferase activity, and mutations in CBP lead to skeletal and nents of the machinery that regulates the epigenome. sive-element-binding protein )-binding protein CBP, which has histone lutation of Rett syndrome involves mutations of the methyl CpG-binding protein 2 cardiac malformations, as well as neurodevelopmental malformations MeCP2)gene, which encodes a protein that binds to methylated dNa and loss of neural plasticity A common theme of these disorders is that sequences In Rett syndrome, DNA methylation proceeds normally mutations in epigenome regulators cause developmental disruption and but epigenetic silencing is impaired because of a failure to properly often cause ph henotypic changes in multiple organ systems recognize this mark(Fig 3). What is striking about the phenotype of this disorder is that prenatal and early infant development is normal, DNA methylation in cancer and erosion of neurodevelopmental milestones is not seen until later Cancer is commonly characterized as showing global hypomethylation childhood and site-specific gene hypermethylation, but a more accurate description Epigenetically disrupted development can occur in various biological is that cancer involves both global and gene-specific hypomethylation pathways or systems. Immunodeficiency/centromeric instability/facial and hypermethylation, as well as widespread chromatin modifications anomalies(ICF)syndrome, for example, affects the immune system (Fig 3). The first epigenetic change described in tumours was gene and involves mutations of the DNA methyltransferase gene DNMT3B, hypomethylation, and we now know that many growth-promoting ich is responsible for de novo DNA methylation during develop- genes are activated through hypomethylation in tumours, including ment. Patients wih ICF syndrome show failure of normal immune HRAS, cyclin D2 and maspin in gastric cancer, carbonic anhydrase IXin a normal b Epigenetic lesions Figure 1 The nature of epigenetic lesions. Although the nature of switch its epigenotype through the silencing of normally active genes or netic lesions is well understood ctivation of lly silent genes, with the attendant changes in DNA difficult to define. Here we depict known and possible defects in the pigenome that could lead to disease. a, X is a transcriptionally active the epigenetic lesion could include a change in the number or density ne with sparse DNA methylation(brown circles), an open chromatin of heterochromatin proteins in gene X(such as EZH2 in cancer)or tructure, interaction with euchromatin proteins(green prote euchromatic proteins in gene Y (such as trithorax in leukaemia). There complex)and histone modifications such as H3K9 acetylation and H3K4 may also be an abnormally dense pattern of methylation in gene promoters thylation(green circles ). Y is a transcriptionally silent gene with shown in gene X), and an overall reduction in DNA methylate dense DNA methylation, a closed chromatin structure, interaction with in cancer. The insets show that the higher-order heterochromatin proteins(red protein complex)and histone modifications configuratio be altered, although such structures are currently only uch as H3K27 methylation(pink circles). b, The abnormal cell could beginning to be understood. @2007 Nature Publishing Group

pseudohypoparathyroidism type IA (PHPIA). Albright hereditary osteodystrophy is characterized by short stature and ectopic calcifica￾tions, and caused by mutational inactivation of guanine nucleotide regu￾latory protein (encoded by GNAS1). PHPIA is a more severe phenotype of multiple hormone resistance caused by tissue-specific differential imprinting of splice variants of the same gene8 . It is unlikely that this complex pattern of imprinting would have been understood without concomitant clinical studies. Single-gene disorders of the epigenetic machinery The other class of monogenic epigenetic disease involves genes that encode components of the machinery that regulates the epigenome. Mutation of these genes causes developmental disorders. For example, Rett syndrome involves mutations of the methyl CpG-binding protein 2 (MeCP2) gene, which encodes a protein that binds to methylated DNA sequences9 . In Rett syndrome, DNA methylation proceeds normally but epigenetic silencing is impaired because of a failure to properly recognize this mark10 (Fig. 3). What is striking about the phenotype of this disorder is that prenatal and early infant development is normal, and erosion of neurodevelopmental milestones is not seen until later childhood. Epigenetically disrupted development can occur in various biological pathways or systems. Immunodeficiency/centromeric instability/facial anomalies (ICF) syndrome, for example, affects the immune system and involves mutations of the DNA methyltransferase gene DNMT3B, which is responsible for de novo DNA methylation during develop￾ment11. Patients wih ICF syndrome show failure of normal immune system development as well as developmental dysmorphology, which could involve failure of heterochromatin formation and might result from DNMT3B having a role in immunoglobulin gene silencing and reactivation12. A striking example of developmental disruption caused by mutations in a chromatin factor gene is alpha-thalassaemia/mental retardation, X-linked (ATRX) syndrome, the gene for which is a helicase involved in chromatin remodelling. Mutations lead to defects in psychomotor, urogenital and haematopoietic development, with maturational defects in erythroid precursors resembling those of alpha-thalassaemia13. Rubinstein–Taybi syndrome involves the CREB (cyclic-AMP respon￾sive-element-binding protein)-binding protein CBP, which has histone acetyltransferase activity, and mutations in CBP lead to skeletal and cardiac malformations, as well as neurodevelopmental malformations and loss of neural plasticity14. A common theme of these disorders is that mutations in epigenome regulators cause developmental disruption and often cause phenotypic changes in multiple organ systems. DNA methylation in cancer Cancer is commonly characterized as showing global hypomethylation and site-specific gene hypermethylation, but a more accurate description is that cancer involves both global and gene-specific hypomethylation and hypermethylation, as well as widespread chromatin modifications (Fig. 3). The first epigenetic change described in tumours was gene hypomethylation15, and we now know that many growth-promoting genes are activated through hypomethylation in tumours, including HRAS, cyclin D2 and maspin in gastric cancer, carbonic anhydrase IXin Gene X Gene Y Gene X Gene Y a Normal b Epigenetic lesions Figure 1 | The nature of epigenetic lesions. Although the nature of genetic lesions is well understood, epigenetic lesions have been more difficult to define. Here we depict known and possible defects in the epigenome that could lead to disease. a, X is a transcriptionally active gene with sparse DNA methylation (brown circles), an open chromatin structure, interaction with euchromatin proteins (green protein complex) and histone modifications such as H3K9 acetylation and H3K4 methylation (green circles). Y is a transcriptionally silent gene with dense DNA methylation, a closed chromatin structure, interaction with heterochromatin proteins (red protein complex) and histone modifications such as H3K27 methylation (pink circles). b, The abnormal cell could switch its epigenotype through the silencing of normally active genes or activation of normally silent genes, with the attendant changes in DNA methylation, histone modification and chromatin proteins. In addition, the epigenetic lesion could include a change in the number or density of heterochromatin proteins in gene X (such as EZH2 in cancer) or euchromatic proteins in gene Y (such as trithorax in leukaemia). There may also be an abnormally dense pattern of methylation in gene promoters (shown in gene X), and an overall reduction in DNA methylation (shown in gene Y) in cancer. The insets show that the higher-order loop configuration may be altered, although such structures are currently only beginning to be understood. 434 INSIGHT REVIEW NATURE|Vol 447|24 May 2007 ￾

   

NATUREIVol 447 24 May 2007 INSIGHT REVIEW Figure 2 Beckwith-Wiedemann a Normal Enlarged kidneys, overgrowth syndrome as an example of a monogenic Wilms tumou disease that reveals mechanisms of normal epigenetic regulation. Depicted are a pair of normal chromosomes(a) maternal chromosomes are n:(b-f); al illustrative lesie (Me). Imprinted genes are depicted, not b UPD scale or in their entirety, with green representing active and red representing silent alleles, respectively. Patients wit uniparental disomy(UPD, b)have complete genetic replacement of the maternal allele region with a second mal copy(dashed enclosure). Loss of imprinting(LOf)of IG F2(c)causes a C LOI IGF2 switch in epigenotype of the IGF2/H19 subdomain(dashed enclosure). LOI of LITI(d) of the p57/K,LQTI/LITI subdomain. Some patients show LOI of the entire domain in the absence UPD(e). Other patients show localized chromatin disruption(small d LOI LITI yellow circle, f)sil imprinting is organized hierarchically ning two smaller subdomains In addit patients show microdeletions in either of the two domains(black crosses), revealing the location ofimprinting control centres. The domain organization e loi domain similarly reveals the contiguous gene ndrome nature of the disease. patients with involvement(genetic or epigenetic of the IGF2/H19 domain have enlarg kidneys and wilms' tumours Patients with involvement of the P57/K,LQT1/ I domain show somatic overgrowth, an enlarged tongue and omphalocele(in f p57 silencing which abdominal organs protrude from the navel). And children with involvement of both domains show both phenotypes. d It (see refs 16, 17 for reviews). In addition, many C/T(cancer/testis)genes should also be noted that as many genes are silenced as are activated in that are expressed normally in the healthy testis are activated in other tumours by both drug-induced hypomethylation and by knockdown of cells by hypomethylation in cancer, including the melanoma-associated DNA methyltransferases", thus both hypomethylation and hypermethyl antigen(MAGE)gene family, which has antigenic and immunothera- ation can lead to gene activation and ancer eutic value in melanoma and glioblastoma.and the oncogenic micro RNAlet-7a-3(ref. 18). Activation of the human papilloma virus HPV16 Loss of imprinting in cancer by hypomethylation is a major mechanism affecting tumour latency in The earliest clue that genomic imprinting might be involved in can- cervical cancer. 7. Recently, oestrogen-and tamoxifen-induced activa- cer came from two rare types of tumour: hydatidiform moles, which tion of PAX2 and endometrial proliferation, leading to cell proliferation, are malignant trophoblastic tumours caused by a pregnancy arising was found to be cancer-specific because of PAX2 promoter hypomethyl- from two complete sets of the paternal genome, and ovarian teratomas, which are benign tumours with many tissue types that arise from two By contrast, tumour suppressor gene silencing has been linked to pro- complete sets of the maternal genome. Molecular evidence for a role of moter hypermethylation, first described for RB, the gene associated with genomic imprinting in cancer emerged from studies showing a universal tinoblastoma and many other tumour suppressor genes, including loss of the maternal allele in wilms tumours and embryonal rhabdo- P16, VHL (von Hippel-Lindau), MLH1, APC (adenomatosis polyposis myosarcoma, with loss of heterozygosity(LOH)of 11p15, implying coli)and E-cadherin(see ref 21 for a review). Recent high-throughput that normally only the maternal allele of an as yet unidentified tumour approaches have been used to identify other candidate genes. An exci- suppressor gene might be expressed". So it was surprising that the first ting demonstration of domain-wide silencinginvolves an entire chromo- molecular evidence for a role of genomic imprinting in cancer was loss somal band, suggesting a disturbance ofhigher-order chromatin( Fig. 1). of imprinting(LOD), causing abnormal activation of the normally silent However, studies focused on loss of DNA methylation in cancer may copy of IGF2, an important autocrine growth factor, leading to patho have overlooked hypomethylation of tissue-specific methylation marks logical biallelic expression of IGF2 in Wilms'tumours, the most common at CpG islands. Indeed, whole-genome analysis suggests that CpG island childhood solid tumour.. LOI refers either to aberrant activation of the 435 @2007 Nature Publishing Group

renal-cell cancer, and S100 calcium-binding protein A4 in colon cancer (see refs 16, 17 for reviews). In addition, many C/T (cancer/testis) genes that are expressed normally in the healthy testis are activated in other cells by hypomethylation in cancer, including the melanoma-associated antigen (MAGE) gene family, which has antigenic and immunothera￾peutic value in melanoma and glioblastoma16,17 and the oncogenic micro RNA let-7a-3 (ref. 18). Activation of the human papilloma virus HPV16 by hypomethylation is a major mechanism affecting tumour latency in cervical cancer16,17. Recently, oestrogen- and tamoxifen-induced activa￾tion of PAX2 and endometrial proliferation, leading to cell proliferation, was found to be cancer-specific because of PAX2 promoter hypomethyl￾ation in the tumours19. By contrast, tumour suppressor gene silencing has been linked to pro￾moter hypermethylation, first described for RB, the gene associated with retinoblastoma20, and many other tumour suppressor genes, including p16, VHL (von Hippel–Lindau), MLH1, APC (adenomatosis polyposis coli) and E-cadherin (see ref. 21 for a review). Recent high-throughput approaches have been used to identify other candidate genes22,23. An exci￾ting demonstration of domain-wide silencing involves an entire chromo￾somal band24, suggesting a disturbance of higher-order chromatin (Fig. 1). However, studies focused on loss of DNA methylation in cancer may have overlooked hypomethylation of tissue-specific methylation marks at CpG islands25. Indeed, whole-genome analysis suggests that CpG island hypermethylation may be less widespread than had been suspected26. It should also be noted that as many genes are silenced as are activated in tumours by both drug-induced hypomethylation and by knockdown of DNA methyltransferases27, thus both hypomethylation and hypermethyl￾ation can lead to gene activation and gene silencing in cancer. Loss of imprinting in cancer The earliest clue that genomic imprinting might be involved in can￾cer came from two rare types of tumour: hydatidiform moles, which are malignant trophoblastic tumours caused by a pregnancy arising from two complete sets of the paternal genome, and ovarian teratomas, which are benign tumours with many tissue types that arise from two complete sets of the maternal genome. Molecular evidence for a role of genomic imprinting in cancer emerged from studies showing a universal loss of the maternal allele in Wilms’ tumours and embryonal rhabdo￾myosarcoma, with loss of heterozygosity (LOH) of 11p15, implying that normally only the maternal allele of an as yet unidentified tumour suppressor gene might be expressed28. So it was surprising that the first molecular evidence for a role of genomic imprinting in cancer was loss of imprinting (LOI), causing abnormal activation of the normally silent copy of IGF2, an important autocrine growth factor, leading to patho￾logical biallelic expression of IGF2 in Wilms’ tumours, the most common childhood solid tumour29,30. LOI refers either to aberrant activation of the a Normal b UPD p57K1P2 LIT1 IGF2 H19 c LOI IGF2 d LOI LIT1 e LOI domain f p57 silencing Enlarged kidneys, Wilm’s tumour Omphalocele, overgrowth, macroglosia Me Me Me Me Me Me Me Me Me Me Me Me Figure 2 | Beckwith–Wiedemann syndrome as an example of a monogenic disease that reveals mechanisms of normal epigenetic regulation. Depicted are a pair of normal chromosomes (a) and several illustrative lesions (b–f); maternal chromosomes are pink and paternal blue, and DMRs are indicated (Me). Imprinted genes are depicted, not to scale or in their entirety, with green representing active and red representing silent alleles, respectively. Patients with uniparental disomy (UPD, b) have complete genetic replacement of the maternal allele region with a second paternal copy (dashed enclosure). Loss of imprinting (LOI) of IGF2 (c) causes a switch in epigenotype of the IGF2/H19 subdomain (dashed enclosure). LOI of LIT1 (d) causes a switch in epigenotype of the p57/KVLQT1/LIT1 subdomain. Some patients show LOI of the entire imprinted gene domain in the absence of UPD (e). Other patients show localized chromatin disruption (small yellow circle, f) silencing p57KIP2. Thus, imprinting is organized hierarchically into a large domain containing two smaller subdomains. In addition, some patients show microdeletions in either of the two domains (black crosses), revealing the location of imprinting control centres. The domain organization similarly reveals the contiguous gene syndrome nature of the disease. Patients with involvement (genetic or epigenetic) of the IGF2/H19 domain have enlarged kidneys and Wilms’ tumours. Patients with involvement of the p57/KVLQT1/ LIT1 domain show somatic overgrowth, an enlarged tongue and omphalocele (in which abdominal organs protrude from the navel). And children with involvement of both domains show both phenotypes. 435 NATURE|Vol 447|24 May 2007 INSIGHT REVIEW ￾

   

INSIGHT REVIEW NATURE Vol 447 24 May 2007 Figure 3 Phenotypic plasticity and the epigenetics Rett syndrome of human disease and ageing. A common feature of epigenetic lesions in human disease is that they affect Subo goma Johe Gene A a cells ability to change its phenotype disorders such as Rett syndrome, a defect in the lf impedes normal development. DNA methylation(brown circles on he dNA) proceeds normally but is not recognize he absence of the MeCP2-methylatio Gene A er Gene A interacting protein(large red oval). This leads t failure to completely silence genes appropriately during development(dashed arrow). b, Cancer involves nany epigenetic lesions that could affect a pluripotent Cancer ading to an incorrect distribution of differentiated cell lineages(indicated by the bivalent euchromatin ns shown in the upper left and activation of gene B after differentiation(lower left panel). Examples of epigenetic lesions found in cancer Gene B include changes caused by increased expression of MLLI in leukaemia representing HOX genes), leading to aberrant HOX expression in differentiated cancer lineages(lower right 她颜m In cancer expression of EZH2 (upp g uppressor genes), leading to aberrant silencing of these genes in differentiated cancer lineages (lower right panel). c, Ageing involves a loss of the normal plasticity Gene B Gene b of to internal and external environmenta signals. The epigenome could have an important role such signals. For example, a gene(at this point Ageing pathetically) showing increased H3K9-methylatie red circles on nucleos or dna ENVIRONM MENT ENVIRONMENT methylation(brown circles on DNA), might be relatively refractory to environmentally induced activation(lowe Gene a right panel)than if the gene had not undergone age- Gene A Gene A ormally silent allele of an imprinted growth-promoting gene, or aber- development. For example, in the fungus Neurospora DNa methylation rant silencing of the normally expressed copy of an imprinted tumour depends on H3K9 methylation", and in mice DNA methylation of suppressor gene, such as the still unidentified locus on 1lp15(ref. 31). homeobox(Hox) genes depends on a full length Mll (myeloid/lymph Subsequently, LOI of IGF2 has been found to be common in lung can- oid or mixed-lineage leukaemia) gene. DNMTI interacts with the cer", breast cancer ovarian cancer and glioma" LOI of other genes H3K9 methyltransferases G9a and SUV39H1, which are needed for ARHI in breast cancer DLKI/GTL2 in pheochromocytoma, normal replication-dependent DNA methylation". Some chromosomal stoma and wilms tumour, and PEGI(also known as MEST) rearrangements and, less commonly, mutations in cancer act by caus ing widespread chromatin disruption. MLLI, which is rearranged and activated in acute lymphoblastic leukaemia, methylates H3K4 to activate Chromatin and cancer ne expression and interacts with integrase integrator 1(INII)in the It has become increasingly clear that in cancer chromatin modifica- SWI/SNF chromatin remodelling complex". INII is mutated in rhab- tions are at least as widespread and important as alterations in dNa doid tumour, a deadly soft-tissue malignancy. Sotos syndrome, which ethylation. For example, overexpression of the polycomb group protein is characterized by tissue overgrowth, leukaemia and wilms tumour, is EZH2, a H3 lysine-27(H3K27)histone methyltransferase, is found in caused by mutations in NSDI, an H3K36/H4K20 methyltransf metastatic prostate cancer and may lead to widespread transcriptional Thus, a strong argument can be made for chromatin modifications driv- repression"(Fig 3). Generalized loss of H4 acetylated Lys-16 (H4K16ac) ing epigenetic disruptions during cancer developmen and trimethylated Lys-20(H4K20me3)is found in lymphoma and olorectal cancer, which could also lead to transcriptional silencing The argument for causality Fig 3). It is not surprising that both DNA methylation and histone One problem with the idea that alterations in DNA methylation under modification are altered in cancer, given their interdependence in normal lie cancer is that no mutations in either the methylation modification or 436 @2007 Nature Publishing Group

normally silent allele of an imprinted growth-promoting gene, or aber￾rant silencing of the normally expressed copy of an imprinted tumour suppressor gene, such as the still unidentified locus on 11p15 (ref. 31). Subsequently, LOI of IGF2 has been found to be common in lung can￾cer32, breast cancer33, ovarian cancer34 and glioma35. LOI of other genes include ARHI in breast cancer36, DLK1/GTL2 in pheochromocytoma, neuroblastoma and Wilms’ tumour37, and PEG1 (also known as MEST) in breast cancer38. Chromatin and cancer It has become increasingly clear that in cancer chromatin modifica￾tions are at least as widespread and important as alterations in DNA methylation. For example, overexpression of the polycomb group protein EZH2, a H3 lysine-27 (H3K27) histone methyltransferase, is found in metastatic prostate cancer and may lead to widespread transcriptional repression39 (Fig. 3). Generalized loss of H4 acetylated Lys-16 (H4K16ac) and trimethylated Lys-20 (H4K20me3) is found in lymphoma and colorectal cancer, which could also lead to transcriptional silencing40 (Fig. 3). It is not surprising that both DNA methylation and histone modification are altered in cancer, given their interdependence in normal development. For example, in the fungus Neurospora DNA methylation depends on H3K9 methylation41, and in mice DNA methylation of homeobox (Hox) genes depends on a full length Mll (myeloid/lymph￾oid or mixed-lineage leukaemia) gene42. DNMT1 interacts with the H3K9 methyltransferases G9a and SUV39H1, which are needed for normal replication-dependent DNA methylation43. Some chromosomal rearrangements and, less commonly, mutations in cancer act by caus￾ing widespread chromatin disruption. MLL1, which is rearranged and activated in acute lymphoblastic leukaemia, methylates H3K4 to activate gene expression and interacts with integrase integrator 1 (INI1) in the SWI/SNF chromatin remodelling complex44. INI1 is mutated in rhab￾doid tumour, a deadly soft-tissue malignancy45. Sotos syndrome, which is characterized by tissue overgrowth, leukaemia and Wilms’ tumour, is caused by mutations in NSD1, an H3K36/H4K20 methyltransferase46. Thus, a strong argument can be made for chromatin modifications driv￾ing epigenetic disruptions during cancer development. The argument for causality One problem with the idea that alterations in DNA methylation under￾lie cancer is that no mutations in either the methylation modification or Gene A Gene A Gene A Rett syndrome Cancer Gene A Gene A Gene B Gene A Gene A Normal Disease and ageing Gene B Gene A Gene A Gene B a b c Gene B Gene A Gene A Gene A ENVIRONMENT Ageing ENVIRONMENT Figure 3 | Phenotypic plasticity and the epigenetics of human disease and ageing. A common feature of epigenetic lesions in human disease is that they affect a cell’s ability to change its phenotype. a, In monogenic disorders such as Rett syndrome, a defect in the normal epigenetic apparatus itself impedes normal development. DNA methylation (brown circles on the DNA) proceeds normally but is not recognized owing to the absence of the MeCP2-methylation￾interacting protein (large red oval). This leads to failure to completely silence genes appropriately during development (dashed arrow). b, Cancer involves many epigenetic lesions that could affect a pluripotent programme in tissue-specific stem cells, possibly leading to an incorrect distribution of differentiated cell lineages (indicated by the bivalent euchromatin and heterochromatin proteins shown in the upper left panel) and normal tissue-specific silencing of gene A and activation of gene B after differentiation (lower left panel). Examples of epigenetic lesions found in cancer include changes in chromatin proteins in stem cells caused by increased expression of MLL1 in leukaemia (upper right panel, green complex above gene A representing HOX genes), leading to aberrant HOX expression in differentiated cancer lineages (lower right panel). Another epigenetic lesion found in cancer is increased expression of EZH2 (upper right panel, red complex above gene B, representing diverse tumour suppressor genes), leading to aberrant silencing of these genes in differentiated cancer lineages (lower right panel). c, Ageing involves a loss of the normal plasticity of response to internal and external environmental signals. The epigenome could have an important role in ageing if the aged epigenome is less responsive to such signals. For example, a gene (at this point hypothetically) showing increased H3K9-methylation (upper right panel, red circles on nucleosomes) or DNA methylation (brown circles on DNA), might be relatively refractory to environmentally induced activation (lower right panel) than if the gene had not undergone age￾dependent epigenetic change (left panels). 436 INSIGHT REVIEW NATURE|Vol 447|24 May 2007 ￾

   

NATUREIVol 447 24 May 2007 INSIGHT REVIEW the recognition machinery have yet been identified in human cancer. ch modifications -for exam CGAAC arry an increased cancer risk, in contrast to the chromatin-modifying Gene disruptions described above. In a consequence of altered gene expression rather than causal; it has been known since the 1980s that numerous genes are aberrantly expressed in tumours". Furthermore, activation of tumour suppressor genes by 5-aza-2-deoxycytidine or DNMTI knockout may not be stable, as has been shown for both MLHI (ref. 48)and p16 (ref. 49), suggesting that the altered methylation might be a consequence rather than a cause of So how can a convincing causal argument be made? Good evidence Radiation Methionine would be constitutional epigenetic alterations linked to cancer risk. The first such example was Beckwith-Wiedemann syndrome, which leads to an 800-fold increased risk of embryonal tumours -that is, those involving residual fetal tissues, such as wilms tumour of the kidney and rhabdomyosarcoma". LOI of IGF2 is specifically associated with increased cancer risk in children with Beckwith-Wiedemann syn- drome, even though it occurs in only a fraction of the affected individu- als(Fig. 2). Thus, the epigenetic change precedes cancer and confers ATACGTAG in adults, at a frequency of 5-10%51, and is associated with a fivefold Gene A increased frequency of benign and malignant colorectal neoplasms d 20-fold for cancer), as well as an increased family history of can cer, consistent with a causal role in cancer predisposition". Another example of epigenetic alterations in normal tissue is the hypermethyl Protein a ation of p16 that occurs with ageing and in the normal tissue of women with breast cancer), although neither case has yet been linked Experimental data in mice further support a causal role for epigenetic Chaperone changes in cancer. When DNMTI hypomorphs are crossed with Min ultiple intestinal neoplasia)mice with an Apc mutation, they show an increased frequency of intestinal neoplasia and liver cancers ypomethylation also causes increased chromosomal instability, lead GTAC A GTCA ing to aggressive T-cell lymphomas ", as well as an increase in sarco- Genec mas in mice with p53 and neurofibromin 1(NFI)mutations. DNA hypermethylation is also important, as DNMTI hypomorphs also Figure 4/ The epigenome at the intersection between environment agenetic hypomethylation is more important in the earliest stages of carcino- (CDGE) hypothesis, the epigenome may modulate the effect of genetic variation(example shown is the large nucleotide in gene A, which could genesis, whereas hypermethylation has a greater role during tumour be C or G), either by affecting the gene s expression through the action progression. In addition, engineered loss of one allele of HICl leads to an increased number of late-onset tumours with epigenetic silencing folding of the gene product of the variant locus or chromatin protein.The the frequency of adenomas in mice caused by mutations in Apc. More- encoding chromatin or chaperone proteins(genes B and C, respectively) over,engineered global LOI leads to intestinal and hepatic tumours in Environmental factors(such as toxins, growth factors, dietary methyl aeric mice donors and hormones)can affect the genome and the epigenome Cancer epigenetics and the stem-cell hypothesis an increased propensity to form melanomas, many of the tumour prop Although epigenetic alterations are co only looked on as surrogates erties must be epigenetic in origin and some cells within the tumour are for genetic change in cancer, they are probably also critical first steps in pluripotent neoplastic progression, disrupting the normal stem-or progenitor-cell In breast cancer, widespread epigenetic alterations are found in programme, for example by stimulating stem-cell proliferation outside tumour cells, stromal cells and the myoepithelium, suggesting that ir normal microenvironment. This 'epigenetic progenitor model, in the entire tumour microenvironment, including apparently normal which originates in stem or progenitor cells after epigenetic altera- cells, is the target of epigenetic disruption". Cancers also seem to tions, is supported by the ubiquitous early nature of epigenetic changes in show increased epigenetic plasticity. This epigenetic plasticity may be cancer,discussed above, as well as the demonstration of altered progen- an inherent property of the stem cells from which cancers arise, for in the intestine of mice harbouring an Apc mutation, and increas. nt trample the bivalent nature of adjacent H3K4 and H3K27 methylation tor cells in normal tissues of patients cancer LOI of IGF2 leads to an expanded progenitor-cell compartm nat is seen at many genomic sites in stem cells but not in somatic cells ith LOI of IGF2 and increased risk of colon cancer. Similarly, LOI cell differentiation%.. Polycomb group genes might also be tumour of IG F2 in Beckwith-Wiedemann syndrome is specifically associated progenitors, as they are overexpressed in cancer, as noted earlier, and with cancer risk and leads to the expansion of nephrogenic progenitor they repress developmental regulators in embryonic stem cells". Thus, cells. Further support for the model comes from the fact that mouse epigenetics seems to be central to plasticity both in development and in melanoma and medulloblastoma nuclei can be cloned to form blasto- tumour cells, and epigenetic discovery will be critical to understanding cysts or chimaeric mice. Although mice derived from the former show these 437 @2007 Nature Publishing Group

the recognition machinery have yet been identified in human cancer. Indeed, congenital disorders involving such modifications — for exam￾ple, Rett syndrome (MeCP2) and ICF syndrome (DNMT3B) — do not carry an increased cancer risk, in contrast to the chromatin-modifying disruptions described above. In addition, epigenetic changes might be a consequence of altered gene expression rather than causal; it has been known since the 1980s that numerous genes are aberrantly expressed in tumours47. Furthermore, activation of tumour suppressor genes by 5-aza-2ʹ-deoxycytidine or DNMT1 knockout may not be stable, as has been shown for both MLH1 (ref. 48) and p16 (ref. 49), suggesting that the altered methylation might be a consequence rather than a cause of gene silencing. So how can a convincing causal argument be made? Good evidence would be constitutional epigenetic alterations linked to cancer risk. The first such example was Beckwith–Wiedemann syndrome, which leads to an 800-fold increased risk of embryonal tumours — that is, those involving residual fetal tissues, such as Wilms’ tumour of the kidney and rhabdomyosarcoma50. LOI of IGF2 is specifically associated with increased cancer risk in children with Beckwith–Wiedemann syn￾drome, even though it occurs in only a fraction of the affected individu￾als5 (Fig. 2). Thus, the epigenetic change precedes cancer and confers risk for cancer, a strong argument for causality. LOI of IGF2 was found in adults, at a frequency of 5–10%51, and is associated with a fivefold increased frequency of benign and malignant colorectal neoplasms (and 20-fold for cancer), as well as an increased family history of can￾cer, consistent with a causal role in cancer predisposition52,53. Another example of epigenetic alterations in normal tissue is the hypermethyl￾ation of p16 that occurs with ageing54 and in the normal tissue of women with breast cancer55, although neither case has yet been linked to cancer risk. Experimental data in mice further support a causal role for epigenetic changes in cancer. When DNMT1 hypomorphs are crossed with Min (multiple intestinal neoplasia) mice with an Apc mutation, they show an increased frequency of intestinal neoplasia and liver cancers56. Hypomethylation also causes increased chromosomal instability, lead￾ing to aggressive T-cell lymphomas57, as well as an increase in sarco￾mas in mice with p53 and neurofibromin 1 (NF1) mutations58. DNA hypermethylation is also important, as DNMT1 hypomorphs also show delayed progression of adenomas in Min mice56,59, suggesting that hypomethylation is more important in the earliest stages of carcino￾genesis, whereas hypermethylation has a greater role during tumour progression. In addition, engineered loss of one allele of HIC1 leads to an increased number of late-onset tumours with epigenetic silencing of the remaining allele60. Genetically induced LOI of IGF2 increases the frequency of adenomas in mice caused by mutations in Apc61. More￾over, engineered global LOI leads to intestinal and hepatic tumours in chimaeric mice62. Cancer epigenetics and the stem-cell hypothesis Although epigenetic alterations are commonly looked on as surrogates for genetic change in cancer, they are probably also critical first steps in neoplastic progression, disrupting the normal stem- or progenitor-cell programme, for example by stimulating stem-cell proliferation outside their normal microenvironment63. This ‘epigenetic progenitor model’, in which cancer originates in stem or progenitor cells after epigenetic altera￾tions, is supported by the ubiquitous early nature of epigenetic changes in cancer, discussed above, as well as the demonstration of altered progeni￾tor cells in normal tissues of patients with cancer. LOI of IGF2 leads to an expanded progenitor-cell compartment in the intestine of mice harbouring an Apc mutation, and increased expression of progenitor-cell markers61,64, a feature also seen in humans with LOI of IGF2 and increased risk of colon cancer61. Similarly, LOI of IGF2 in Beckwith–Wiedemann syndrome is specifically associated with cancer risk and leads to the expansion of nephrogenic progenitor cells65. Further support for the model comes from the fact that mouse melanoma and medulloblastoma nuclei can be cloned to form blasto￾cysts or chimaeric mice66. Although mice derived from the former show an increased propensity to form melanomas, many of the tumour prop￾erties must be epigenetic in origin and some cells within the tumour are pluripotent66. In breast cancer, widespread epigenetic alterations are found in tumour cells, stromal cells and the myoepithelium, suggesting that the entire tumour microenvironment, including apparently normal cells, is the target of epigenetic disruption67. Cancers also seem to show increased epigenetic plasticity. This epigenetic plasticity may be an inherent property of the stem cells from which cancers arise, for example the bivalent nature of adjacent H3K4 and H3K27 methylation that is seen at many genomic sites in stem cells but not in somatic cells after differentiation68. The gene MLL1, which is rearranged and acti￾vated in childhood leukaemias, is also a key regulator of normal stem￾cell differentiation69,70. Polycomb group genes might also be tumour progenitors, as they are overexpressed in cancer, as noted earlier, and they repress developmental regulators in embryonic stem cells71. Thus, epigenetics seems to be central to plasticity both in development and in tumour cells, and epigenetic discovery will be critical to understanding these. Gene B Gene A Chaperone protein Protein A RNA Chromatin protein Gene C ...CGAAGTCTA ...AGTACGTAG... ...GTACAGTCA... ENVIRONMENT IGF2 Radiation Methionine Oestrogen Figure 4 | The epigenome at the intersection between environment and genetic variation. According to the common disease genetic and epigenetic (CDGE) hypothesis, the epigenome may modulate the effect of genetic variation (example shown is the large nucleotide in gene A, which could be C or G), either by affecting the gene’s expression through the action of chromatin proteins or DNA methylation, or by modulating protein folding of the gene product of the variant locus or chromatin protein. The epigenome may, in turn, be affected by sequence variation in the genes encoding chromatin or chaperone proteins (genes B and C, respectively). Environmental factors (such as toxins, growth factors, dietary methyl donors and hormones) can affect the genome and the epigenome. 437 NATURE|Vol 447|24 May 2007 INSIGHT REVIEW ￾