- Email: [email protected]
Cytosine methylation The pros and cons of DNA methylation
TIMOTHY H. BESTOA
The pros and cons of DNA methylation Cytosine methylation is essential for mammalian development but may also contribute to mutagenesis and epigenetic gene inactivation. About 60 % of CpG dinucleotides in vertebrate DNA are methylated at position 5 of the cytosine base. DNA methylation is essential for mammalian development [ 11, but places a heavy mutational load on the mammalian genome. The most frequent mutation in human DNA, replacement of the C at a CpG dinucleotide by a T (a C-T transition), is commonly thought to be so frequent because 5-methylcytosine l(m5C) can be directly converted to T by deamination (Fig. 1). Provocative recent evidence suggests that some of these mutations may involve an active process, and that cytosine methylation may be involved in other pathways of mutagenesis and gene inactivation. C-T transitions are thought to be caused by oxidative deamination at the N4 position of m5C to form T, or by C+U mutations that are fixed as C-T transitions during DNA replication (Fig. 1). These mutations may not always be purely spontaneous. In the absence of the biological methyl donor S-adenosyl L-methionine, abortive transmethylation by a bacterial cytosine methyltransferase, which is closely related to the catalytic domain of mammalian DNA methyltransferase, increases the rate of cytosine deamination by a factor of about lo* [z]. This suggests that DNA methyltransferases act as (presumably stochastic) ,endogenous mutagens under some conditions. Steinberg and Gorman [3] have raised the startling pas sibility that some C+T transitions may be the result of a directed cellular process. They analysed mutant mouse S49 cells known to carry a C+T transition at a CpG dinucleotide in the gene for the regulatory subunit of CAMPdependent protein kinase, and found that 7 out of 102 independent mutant clones also carried a second C-+T transition at a nearby CpG dinucleotide. When both mutations were present they were always on the same allele, affecting CpG dinucleotides that were always on the same DNA strand. Furthermore, the second (3’) mutation was never observed in the absence of the first mutation. These observations are very difficult to reconcile with current views of mutagenesis, and seem to require an endogenous system that actively induces lC+T transitions in response to unknown stimuli. The alteration of bases in polynucleotides by enzymes is not unprecedented. Cs are actively converted to Us during RNA editing [4], and the presence of duplicated DNA sequences in the fungus Neur&ora crassa has been found to induce the conversion of Cs to Ts in both copies of the 384
@
Current
Biology
;-:C,
,
Dearnination?A3
Uridine
Cytosine
DNA replication
Methylation a
A
a
A
m5Cytosine
Thymidine 0 1993current Bdog~
Fig. 1. Conversion from C to T in DNA (a transition). Deamination of C changes the base to U, which is fixed as T if not repaired prior to DN,$ replication. Deamination of msC changes the base directly to T. : .‘
duplicated DNA, with many other Cs being converted to m5C [ 51. In this case, both methylation and the C-+-T transitions occur just prior to meiosis; it has not yet been proven that the transitions involve the deamination of m5C bases. This process, termed FLIP(repeat-induced point mutation) by Eric Selker [5], may have evolved to force the mutation of transposable elements to the point where they are incapable of transposition. Similar reasoning has led to the suggestion that the type of DNA modification observed in mammals may, among other functions, act as a kind of immune system for the genome [6]: by methylating foreign DNA, such as transposable elements or viral DNA, vertebrates may have evolved a mechanism for inactivating such potentially deleterious invasive sequences. This suggestion is supported by the observation that events such as the integration of retroviral DNA provoke efficient de no~o methylation in early mouse embryos and embryonal carcinoma cells 171. Kricker et al. [s] have suggested that a RIP-like phenomenon operates in mammals, driving the divergence of repeated sequences via induced 1993,
Vol
3 NO 6
DISPATCH
C+T transitions so as to reduce the frequency of chromosome translocations and deletions caused by homologous recombination between repeats. If this hypothesis is correct, and the mammalian system does closely resemble RIF’ in Neurosporu, genes may be inactivated at some frequency upon. duplication regardless of their nature. There are a number of other known cases where genome perturbations in eukaryotes are apparently a stimulus for DNA methylation. Chromosome translocations in somatic cells can induce changes in the methylation status of sequences near the translocation breakpoints [9]. Strikingly, expansion of the CGG triplet repeats in the human FMR-I gene, thought to be the genetic basis of fragile-X syndrome, appears to provoke methylation of the triplet repeats and nearby promoter sequences; this methylation is apparently the cause of FM&-I inactivation in fragile-X patients [lo]. As many as half of all spontaneous new mutations in Drosophila, which lack modified bases in its DNA, are the result of transposition events (Steven Hennikoff, personal communication), whereas mutations caused by mobile DNA elements are rare in mammals. The relative rarity of these mutations in mammals suggests that lhere is a cellular system that inhibits the activity of transposable elements, as has been documented in plants, where both transgenes and transposable elements have been shown to be controlled by methylation [ 111. Denise B:arlow [ 121 has suggested that endogenous mammalian genes , which show properties in some way characteristic of foreign DNA, are subject to de nouo methylation in the female germline, and that this system is responsible for the phenomenon of genomic imprinting - the differential activity of the maternal and paternal copies of a gene that presumably reflects some differential marking of the gene as it passes through the maternal and paternal germlines. Methylation can also cause epigenetic gene inactivation in the absence of any sequence changes. The methylation of promoter sequences causes their assembly into condensed, inactive chromatin and strongly inhibits transcription both in viva and in vitro [13]. Although methylation patterns are transmitted by clonal inheritance, maintenance methylation i,s not completely efficient and methylation patterns typically show some heterogeneity even in clonal cell populations [ 141. De nova methylation, normally thought to be restricted to certain developmental stages, is an error-prone process that can introduce new methylated sites into the genomes of somatic cells. Nearly all tissue-specific genes and one allele (and sometimes both) of many house-keeping genes are inactivated by methylation in long-established cell lines such as Chinese hamster ovary cells [ 15].1n fact, approximately half of all gene inactivation events in cultured cells are the result of de nova methylation of the affected genes, which usually occurs at sites not methylalted in normal tissues. Ectopic de nova methylation also occurs in viva and may be involved in the development of certain human diseases. There is evidence that ectopic inactivation of tumor suppressor genes by de nova methylation contributes to carcinogenesis: the promoter region of the RB- 1 gene [16]
was found to be methylated in six out of 77 retinoblas tomas, and the 5’ region of the m 2 gene was methylated in two out of 29 Wilms’ tumors [ 171; the methylated genes were otherwise grossly normal. The observed promoter methylation presumably resulted from de ~OYOmethylation events, and would be expected to prevent expfession of the tumor suppressor genes. It has previously been assumed that the inactivation of tumor suppressor genes involves alterations to the base sequence or deletions, but this evidence suggests that it can also be caused by ectopic promoter methylation in some tumors. Cancer incidence is very strongly age-dependent, and the probability of ectopic promoter methylation also increases with age. Careful studies of the methylation patterns in and near promoters in mice and humans have shown that methylated regions gradually encroach on the promoters with age [IS]. This phenomenon is likely to be involved in the gene inactivation events that are thought to contribute to the increased tumor incidence and rising morbidity and mortality rates during aging. -Very recently, two groups [ 19,201 have reported evidence that tumor progression may involve an epigenetic mechanism they have termed ‘relaxation of imprinting’. In mice, the insulin-like growth factor-2 (l&2) gene is expressed only from the paternal allele, and animals containing only maternal alleles are dwarfed. The new results show that, whereas only the paternal IGF2 allele is expressed in normal human tissues, both alleles are active in 2569% of Wilms’ tumors (a childhood carcer derived from embryonal kidney cells). Allele-specific methylation patterns have been observed in all imprinted genes analysed to date, and the prediction that activation of the maternal IGF2 allele in Wiis’ tumor cells is the result of changes in methylation patterns at the maternal allele is readily testable. As mentioned earlier, the clonal transmission of methylation patterns is an error-prone process, so erroneous demethylation or de nova methylation events may activate the maternal IGF2 allele. The autocrine/paracrine effects of excess IGF2 may be part of the multistep process that results in malignant Wilms’ TVmor, and this epigenetic mechanism may be involved in other types of growth disorder and tumorigenesis. If mammalian genomes d0 indeed have an immune system, why should the development of such a system be favored by natural selection and why does it work so inefficiently (there would be no AIDS if all virus integration events provoked methylation)? The evolution of new functions requires the duplication and divergence of pre-existing genes, yet a genome completely tolerant of additions and duplications would be vulnerable to harmful events such as mutations induced by transposable elements and infection by integrating cytopathic viruses. The evolution of complex higher organisms requires that a balance be struck between the requirement for an expandable genome and the need to minimize the vulnerability to mobile DNA elements and virus integration. As discussed above, the insertion of such sequences does in some cases seem to provoke de nova methylation and gene inacti+ation (and perhaps also directed C-+T transitions), and some relatively innocuous perturbations,
385
386
Current
Biology
1993, Vol
3
No 6
such as triplet repeat expansion in the FAIR-~ gene, may activate the same system. why should unstable methylation patterns sometimes interfere with normal gene expression and contribute to the inactivation ofimportant genes, such as tumor suppressor genes? The evolution of complex organisms with very large genomes, such as vertebrates and vascular plants, was accompanied by the evolution of a system that organized the genome into two compartments: a small, unmethylated compartment that is enriched in regulatory sequences and accessible to diffusible factors, and a much larger compartment that is methylated and maintained in a condensed state [6]. Methylation has been shown to control the degree of condensation of chromatin and may be an important determinant of DNA accessibility to regulatory factors [ 131. The apparent organizing and protective functions of DNA methylation probably represent a relatively recent and still imperfect evolutionary development; this may be reflected in the error-prone nature of genomic methylation patterns, which contributes to the mutation and epigenetic inactivation of normal cellular genes.
8.
9.
10.
11.
12. 13. 14.
15.
17.
3.
4.
5.
LI E, BESTOR TH, JAENI~CH R: Targeted mutation of the DNA methyltransferase gene results in embryonic lethality Cell 1992, 61:91>926. SHEN J-C, RIDEOUT WM, JONES PA: High frequency mutagenesis by a DNA methyltransfemse. Cell 1992, 71:1073-1080. STEINBERG RA, GORMAN KB: Linked spontaneous CG-TA mutations at CpG sites in the gene for protein kinase regulatory subunit Mel Cell Biol 1992, 12:767-772. HODGES P, SCOTT J: Apolipoprotein B mRNA editing: a new tier for the control of gfene expression. Trends Biocbem Sci 1992, 1777-81. SELKER EU: Control of DNA methylation in fungi. In DNA
metbykztion: molecuhr 6.
7.
S, TURC-CAFZL
biology and
biological
18.
19.
20.
P, JANI-SAT
S, SREETANAJAH C,
F, PIERRETTI M, CAXEY
represses
FMR-
ST, sm
1 transcription
in fragile X syndrome. Hum Mol Genet 1992, 1:397-400. MAXZKE M, MATZKIZ AJM: Genomic imprinting in plants: parental effects and tram-inactivation phenomena. Alznu Rev Plant Pbysiol Plant MoL’ Biol 1993, in press. BARLOW DP: DNA methylation and genomic imprinting - &:om immune defense to gene regulation? Science, in press. RIGGS AD, PFEIFFER GD: X chromosome inactivation and cell memory. Trends Genet 1992, 8:169-173. TURKER MS, SWISSHELM K, SMITH AC, MARTIN GM: A partial methy lation profile for a CpG site is stably maintained in mammalian tissues. J Biol &em 1989, 26411632-11636. HOLLIDAY R, Ho T: Evidence for Aelk exclusion in Chinese hamster ovary cells. New Bioiogist 1990, 2:719-726. SAKAI T, TOGUCHIDA J, OHTANI N, YANDELL DRYJA TP: Allele-specific hypermethylation
gene.
DW,
RAPAPORT FM,
of the
Am J Hum
retinoblas1991,
Genet
B, SCHNEIDER S: Wilms’ tumor-specific methylation pattern in 11~13 detected by PFGE. Genes Chromosomes Cancer 1992, 5132-140. ONO T, YAMAMOTO S, KURISHITA A, YAMAMOTO K, YAMAMOTO Y, UJENO Y, SAGISAKA K, FUKLJI Y, MNAMOTO M, ‘TAWA R, HIROSE S, OKADA S: Comparison of age-associated changes of c-myc methylation in liver between man and mouse. Mut Res 1990, 237:23+246. RAWIER S, JOHNSON IA, D~BRY CJ, PING AJ, GRWY PE, FEWBERG AP: Relaxation of imprinted genes in human cancer. Nature 1993, 362~747-749.
ROYER-POKORA
OGAWA
0,
ECCLES MR,
SZETO J, MCNOE
SMITH PJ, REEVE AE: Relaxation II geqe imprinting implicated 362:749-751.
signzjicance.
Edited by Jost JP, Saluz HP. Geneva: Bid&user; 1992. BESTOR TH: DNA methylation: evolution of a bacterial immune function into a regulatar of gene expression and genome structure in higher eukauyotes. Phil Tralzs R Sot Lond [Biolj 1990, 326:179-187. J-R D, JAENISCH R: DNA methylation in early mammalian development. Jn DNA Metfrykztion. Edited by Razin A, Ceder H,
C, DAL Cw
SLITC~JFFE JS, NEISON DL, ZHANG D, WARREN ST: DNA methylation
toma tumor-suppressor 48:88&888.
References
2.
PALJ~N
.,
LEONG SP, VOGELSTEIN B, KINZLER KW, SANDBERG A& GERM: Myxoid liposarcoma with (t12;16)(q13;11) contains sitespecific differences in metbylation patterns Surrounding a zinc-finger gene mapped to the breakpoint region on chro. mosome 12. Cant Res 1990, 50:7902-7907.
16.
1.
R&s A. Berlin: Springer-Verlag; 1984. KRICKER MC, DRAKE JW, RADMAN M: Duplication-targeted DNA:, methylation and mutagenesis in the evolution of eukaryotic chromosomes. Proc Nat1 Acad Sci USA 1992, 89:1075-1079,
I&
YUN
K, MAW
of insulin-like growth in Wilms’ tumor. Nature
MA,
factor 1993,
I
Timothy H. Bestor and Angela Coxon, Harvard Medical School, 45 Shattuck Street, Boston, Massachusetts 02115, USA.
66
Triple(t)protection
The pros and cons of DNA methylation Cytosine methylation is essential for mammalian development but may also contribute to mutagenesis and epigenetic gene inactivation. About 60 % of CpG dinucleotides in vertebrate DNA are methylated at position 5 of the cytosine base. DNA methylation is essential for mammalian development [ 11, but places a heavy mutational load on the mammalian genome. The most frequent mutation in human DNA, replacement of the C at a CpG dinucleotide by a T (a C-T transition), is commonly thought to be so frequent because 5-methylcytosine l(m5C) can be directly converted to T by deamination (Fig. 1). Provocative recent evidence suggests that some of these mutations may involve an active process, and that cytosine methylation may be involved in other pathways of mutagenesis and gene inactivation. C-T transitions are thought to be caused by oxidative deamination at the N4 position of m5C to form T, or by C+U mutations that are fixed as C-T transitions during DNA replication (Fig. 1). These mutations may not always be purely spontaneous. In the absence of the biological methyl donor S-adenosyl L-methionine, abortive transmethylation by a bacterial cytosine methyltransferase, which is closely related to the catalytic domain of mammalian DNA methyltransferase, increases the rate of cytosine deamination by a factor of about lo* [z]. This suggests that DNA methyltransferases act as (presumably stochastic) ,endogenous mutagens under some conditions. Steinberg and Gorman [3] have raised the startling pas sibility that some C+T transitions may be the result of a directed cellular process. They analysed mutant mouse S49 cells known to carry a C+T transition at a CpG dinucleotide in the gene for the regulatory subunit of CAMPdependent protein kinase, and found that 7 out of 102 independent mutant clones also carried a second C-+T transition at a nearby CpG dinucleotide. When both mutations were present they were always on the same allele, affecting CpG dinucleotides that were always on the same DNA strand. Furthermore, the second (3’) mutation was never observed in the absence of the first mutation. These observations are very difficult to reconcile with current views of mutagenesis, and seem to require an endogenous system that actively induces lC+T transitions in response to unknown stimuli. The alteration of bases in polynucleotides by enzymes is not unprecedented. Cs are actively converted to Us during RNA editing [4], and the presence of duplicated DNA sequences in the fungus Neur&ora crassa has been found to induce the conversion of Cs to Ts in both copies of the 384
@
Current
Biology
;-:C,
,
Dearnination?A3
Uridine
Cytosine
DNA replication
Methylation a
A
a
A
m5Cytosine
Thymidine 0 1993current Bdog~
Fig. 1. Conversion from C to T in DNA (a transition). Deamination of C changes the base to U, which is fixed as T if not repaired prior to DN,$ replication. Deamination of msC changes the base directly to T. : .‘
duplicated DNA, with many other Cs being converted to m5C [ 51. In this case, both methylation and the C-+-T transitions occur just prior to meiosis; it has not yet been proven that the transitions involve the deamination of m5C bases. This process, termed FLIP(repeat-induced point mutation) by Eric Selker [5], may have evolved to force the mutation of transposable elements to the point where they are incapable of transposition. Similar reasoning has led to the suggestion that the type of DNA modification observed in mammals may, among other functions, act as a kind of immune system for the genome [6]: by methylating foreign DNA, such as transposable elements or viral DNA, vertebrates may have evolved a mechanism for inactivating such potentially deleterious invasive sequences. This suggestion is supported by the observation that events such as the integration of retroviral DNA provoke efficient de no~o methylation in early mouse embryos and embryonal carcinoma cells 171. Kricker et al. [s] have suggested that a RIP-like phenomenon operates in mammals, driving the divergence of repeated sequences via induced 1993,
Vol
3 NO 6
DISPATCH
C+T transitions so as to reduce the frequency of chromosome translocations and deletions caused by homologous recombination between repeats. If this hypothesis is correct, and the mammalian system does closely resemble RIF’ in Neurosporu, genes may be inactivated at some frequency upon. duplication regardless of their nature. There are a number of other known cases where genome perturbations in eukaryotes are apparently a stimulus for DNA methylation. Chromosome translocations in somatic cells can induce changes in the methylation status of sequences near the translocation breakpoints [9]. Strikingly, expansion of the CGG triplet repeats in the human FMR-I gene, thought to be the genetic basis of fragile-X syndrome, appears to provoke methylation of the triplet repeats and nearby promoter sequences; this methylation is apparently the cause of FM&-I inactivation in fragile-X patients [lo]. As many as half of all spontaneous new mutations in Drosophila, which lack modified bases in its DNA, are the result of transposition events (Steven Hennikoff, personal communication), whereas mutations caused by mobile DNA elements are rare in mammals. The relative rarity of these mutations in mammals suggests that lhere is a cellular system that inhibits the activity of transposable elements, as has been documented in plants, where both transgenes and transposable elements have been shown to be controlled by methylation [ 111. Denise B:arlow [ 121 has suggested that endogenous mammalian genes , which show properties in some way characteristic of foreign DNA, are subject to de nouo methylation in the female germline, and that this system is responsible for the phenomenon of genomic imprinting - the differential activity of the maternal and paternal copies of a gene that presumably reflects some differential marking of the gene as it passes through the maternal and paternal germlines. Methylation can also cause epigenetic gene inactivation in the absence of any sequence changes. The methylation of promoter sequences causes their assembly into condensed, inactive chromatin and strongly inhibits transcription both in viva and in vitro [13]. Although methylation patterns are transmitted by clonal inheritance, maintenance methylation i,s not completely efficient and methylation patterns typically show some heterogeneity even in clonal cell populations [ 141. De nova methylation, normally thought to be restricted to certain developmental stages, is an error-prone process that can introduce new methylated sites into the genomes of somatic cells. Nearly all tissue-specific genes and one allele (and sometimes both) of many house-keeping genes are inactivated by methylation in long-established cell lines such as Chinese hamster ovary cells [ 15].1n fact, approximately half of all gene inactivation events in cultured cells are the result of de nova methylation of the affected genes, which usually occurs at sites not methylalted in normal tissues. Ectopic de nova methylation also occurs in viva and may be involved in the development of certain human diseases. There is evidence that ectopic inactivation of tumor suppressor genes by de nova methylation contributes to carcinogenesis: the promoter region of the RB- 1 gene [16]
was found to be methylated in six out of 77 retinoblas tomas, and the 5’ region of the m 2 gene was methylated in two out of 29 Wilms’ tumors [ 171; the methylated genes were otherwise grossly normal. The observed promoter methylation presumably resulted from de ~OYOmethylation events, and would be expected to prevent expfession of the tumor suppressor genes. It has previously been assumed that the inactivation of tumor suppressor genes involves alterations to the base sequence or deletions, but this evidence suggests that it can also be caused by ectopic promoter methylation in some tumors. Cancer incidence is very strongly age-dependent, and the probability of ectopic promoter methylation also increases with age. Careful studies of the methylation patterns in and near promoters in mice and humans have shown that methylated regions gradually encroach on the promoters with age [IS]. This phenomenon is likely to be involved in the gene inactivation events that are thought to contribute to the increased tumor incidence and rising morbidity and mortality rates during aging. -Very recently, two groups [ 19,201 have reported evidence that tumor progression may involve an epigenetic mechanism they have termed ‘relaxation of imprinting’. In mice, the insulin-like growth factor-2 (l&2) gene is expressed only from the paternal allele, and animals containing only maternal alleles are dwarfed. The new results show that, whereas only the paternal IGF2 allele is expressed in normal human tissues, both alleles are active in 2569% of Wilms’ tumors (a childhood carcer derived from embryonal kidney cells). Allele-specific methylation patterns have been observed in all imprinted genes analysed to date, and the prediction that activation of the maternal IGF2 allele in Wiis’ tumor cells is the result of changes in methylation patterns at the maternal allele is readily testable. As mentioned earlier, the clonal transmission of methylation patterns is an error-prone process, so erroneous demethylation or de nova methylation events may activate the maternal IGF2 allele. The autocrine/paracrine effects of excess IGF2 may be part of the multistep process that results in malignant Wilms’ TVmor, and this epigenetic mechanism may be involved in other types of growth disorder and tumorigenesis. If mammalian genomes d0 indeed have an immune system, why should the development of such a system be favored by natural selection and why does it work so inefficiently (there would be no AIDS if all virus integration events provoked methylation)? The evolution of new functions requires the duplication and divergence of pre-existing genes, yet a genome completely tolerant of additions and duplications would be vulnerable to harmful events such as mutations induced by transposable elements and infection by integrating cytopathic viruses. The evolution of complex higher organisms requires that a balance be struck between the requirement for an expandable genome and the need to minimize the vulnerability to mobile DNA elements and virus integration. As discussed above, the insertion of such sequences does in some cases seem to provoke de nova methylation and gene inacti+ation (and perhaps also directed C-+T transitions), and some relatively innocuous perturbations,
385
386
Current
Biology
1993, Vol
3
No 6
such as triplet repeat expansion in the FAIR-~ gene, may activate the same system. why should unstable methylation patterns sometimes interfere with normal gene expression and contribute to the inactivation ofimportant genes, such as tumor suppressor genes? The evolution of complex organisms with very large genomes, such as vertebrates and vascular plants, was accompanied by the evolution of a system that organized the genome into two compartments: a small, unmethylated compartment that is enriched in regulatory sequences and accessible to diffusible factors, and a much larger compartment that is methylated and maintained in a condensed state [6]. Methylation has been shown to control the degree of condensation of chromatin and may be an important determinant of DNA accessibility to regulatory factors [ 131. The apparent organizing and protective functions of DNA methylation probably represent a relatively recent and still imperfect evolutionary development; this may be reflected in the error-prone nature of genomic methylation patterns, which contributes to the mutation and epigenetic inactivation of normal cellular genes.
8.
9.
10.
11.
12. 13. 14.
15.
17.
3.
4.
5.
LI E, BESTOR TH, JAENI~CH R: Targeted mutation of the DNA methyltransferase gene results in embryonic lethality Cell 1992, 61:91>926. SHEN J-C, RIDEOUT WM, JONES PA: High frequency mutagenesis by a DNA methyltransfemse. Cell 1992, 71:1073-1080. STEINBERG RA, GORMAN KB: Linked spontaneous CG-TA mutations at CpG sites in the gene for protein kinase regulatory subunit Mel Cell Biol 1992, 12:767-772. HODGES P, SCOTT J: Apolipoprotein B mRNA editing: a new tier for the control of gfene expression. Trends Biocbem Sci 1992, 1777-81. SELKER EU: Control of DNA methylation in fungi. In DNA
metbykztion: molecuhr 6.
7.
S, TURC-CAFZL
biology and
biological
18.
19.
20.
P, JANI-SAT
S, SREETANAJAH C,
F, PIERRETTI M, CAXEY
represses
FMR-
ST, sm
1 transcription
in fragile X syndrome. Hum Mol Genet 1992, 1:397-400. MAXZKE M, MATZKIZ AJM: Genomic imprinting in plants: parental effects and tram-inactivation phenomena. Alznu Rev Plant Pbysiol Plant MoL’ Biol 1993, in press. BARLOW DP: DNA methylation and genomic imprinting - &:om immune defense to gene regulation? Science, in press. RIGGS AD, PFEIFFER GD: X chromosome inactivation and cell memory. Trends Genet 1992, 8:169-173. TURKER MS, SWISSHELM K, SMITH AC, MARTIN GM: A partial methy lation profile for a CpG site is stably maintained in mammalian tissues. J Biol &em 1989, 26411632-11636. HOLLIDAY R, Ho T: Evidence for Aelk exclusion in Chinese hamster ovary cells. New Bioiogist 1990, 2:719-726. SAKAI T, TOGUCHIDA J, OHTANI N, YANDELL DRYJA TP: Allele-specific hypermethylation
gene.
DW,
RAPAPORT FM,
of the
Am J Hum
retinoblas1991,
Genet
B, SCHNEIDER S: Wilms’ tumor-specific methylation pattern in 11~13 detected by PFGE. Genes Chromosomes Cancer 1992, 5132-140. ONO T, YAMAMOTO S, KURISHITA A, YAMAMOTO K, YAMAMOTO Y, UJENO Y, SAGISAKA K, FUKLJI Y, MNAMOTO M, ‘TAWA R, HIROSE S, OKADA S: Comparison of age-associated changes of c-myc methylation in liver between man and mouse. Mut Res 1990, 237:23+246. RAWIER S, JOHNSON IA, D~BRY CJ, PING AJ, GRWY PE, FEWBERG AP: Relaxation of imprinted genes in human cancer. Nature 1993, 362~747-749.
ROYER-POKORA
OGAWA
0,
ECCLES MR,
SZETO J, MCNOE
SMITH PJ, REEVE AE: Relaxation II geqe imprinting implicated 362:749-751.
signzjicance.
Edited by Jost JP, Saluz HP. Geneva: Bid&user; 1992. BESTOR TH: DNA methylation: evolution of a bacterial immune function into a regulatar of gene expression and genome structure in higher eukauyotes. Phil Tralzs R Sot Lond [Biolj 1990, 326:179-187. J-R D, JAENISCH R: DNA methylation in early mammalian development. Jn DNA Metfrykztion. Edited by Razin A, Ceder H,
C, DAL Cw
SLITC~JFFE JS, NEISON DL, ZHANG D, WARREN ST: DNA methylation
toma tumor-suppressor 48:88&888.
References
2.
PALJ~N
.,
LEONG SP, VOGELSTEIN B, KINZLER KW, SANDBERG A& GERM: Myxoid liposarcoma with (t12;16)(q13;11) contains sitespecific differences in metbylation patterns Surrounding a zinc-finger gene mapped to the breakpoint region on chro. mosome 12. Cant Res 1990, 50:7902-7907.
16.
1.
R&s A. Berlin: Springer-Verlag; 1984. KRICKER MC, DRAKE JW, RADMAN M: Duplication-targeted DNA:, methylation and mutagenesis in the evolution of eukaryotic chromosomes. Proc Nat1 Acad Sci USA 1992, 89:1075-1079,
I&
YUN
K, MAW
of insulin-like growth in Wilms’ tumor. Nature
MA,
factor 1993,
I
Timothy H. Bestor and Angela Coxon, Harvard Medical School, 45 Shattuck Street, Boston, Massachusetts 02115, USA.
66
Triple(t)protection