Cell, Vol. 70, 5-8,
July 10, 1992, Copyright
0 1992 by Cell Press
The Essentials of DNA Methylation Adrian Bird Institute of Cell and Molecular University of Edinburgh King’s Buildings Edinburgh EH9 3JR Scotland
Biology
An obstacle to progress in understanding eukaryotic DNA methylation has been the lack of a good genetic systemthe most favorable eukaryotes for genetic analysis(yeasts, Caenorhabditis, Drosophila) have no detectable methylation in their genomes. Not only has this hampered progress, but the clear evidence for life without methylation that these organisms provide has raised questions about the relevance of the whole phenomenon. Li et al. (1992) have responded to the need for genetics by creating a methylation mutant and, in doing so, have shown that mice almost certainly have a requirement for DNA methylation. Transgenic approaches were used to obtain mouse embryos homozygous for disruption of the DNA methyltransferase. In a world where gene knockouts can so often be asymptomatic, the effect is notably unambiguous. Homozygous mutant embryos are stunted and die at midgestation. The result is all the more striking since it appears that inactivation of the gene is incomplete (approximately 30% of DNA methyltransferase activity remains in the homozygous disruptants). Evidently, embryogenesis cannot be completed when levels of DNA methyltransferase are reduced. The gene encoding DNA methyltransferase is therefore essential. Death by Leaky Repression? Why is DNA methylation essential? (For the purposes of this article, it is assumed that the mutant phenotype is due to loss of the ability to methylate DNA and not some other as yet unknown function of this protein.) We do not know the answer to this question, but there are some clues. Since DNA methylation can stably repress transcription, it could be that repression is leaky in the absence of DNA methyltransferase, leading to premature death. Repression of what? One possible answer is based on the evidence that cytosine methylation in both eukaryotes and prokaryotes is primarily a system for neutralizing invading DNA (see Bestor, 1990). Failure to repress transposable elements, proviruses, or other potentially damaging sequences could be lethal. Another possibility is that methylation-mediated repression of genes is somehow built into the developmental program. This popular idea could account for the lethality of DNA methyltransferase mutants and is supported by some evidence. The stability of gene repression on the inactive X chromosome in somatic cells of eutherian (but not marsupial) mammals clearly owes much to the methylation of CpG islands (see Riggs and Pfeifer, 1992). More difficult to assess is the wealth of correlative data relating DNA methylation to transcription (e.g., current evidence relating DNA methylation to imprinting), as methylation
Minireview
levels have a known tendency to adapt to the transcriptional status quo. Once they are silenced, genes can subsequently acquire methylation (e.g., Enver et al., 1988, and references therein), and once active, they can subsequently demethylate (e.g., Sullivan et al., 1989). Is methylation dancing to the piper or calling the tune? In one case it may be the latter. Transfection experiments have been used to show that the muscle-specific rat a-actin promoter is repressed by partial methylation in both muscle and nonmuscle cells (Paroush et al., 1990). Activation occurs specifically in myoblasts once demethylation has taken place. An unexplained observation is that constructs in which all CpGs are methylated fail to activate the gene, although they do undergo myoblast-specific demethylation (Yisraeli et al., 1986). 5-Azacytidine May Not Mimic an In Vivo Mechanism Apparent evidence for the involvement of DNA methylation in development came from the early finding that lOTi/ cells differentiate into muscle and other cell types when treated with the methylation inhibitor Qzacytidine (Taylor and Jones, 1979). It seemed that genes were naturally repressed by methylation, and demethylation by 5-azacytidine was activating them. Unfortunately, doubt has been cast on this simple scenario by studies of DNA methylation in cultured cells. In many permanent cell lines, tens of thousands of genes have been stably inactivated, owing to de novo methylation of CpG-island promoters that are normally methylation free (Antequeraet al., 1990). Only genes that are essential for life in culture conditions seem to have been spared (presumably because of selection). Methylation of these promoters is specific to cultured cells; in the animal, their repression (where it occurs) must be due to some other mechanism, since methylation is apparently absent (see below and Figure 2). The effects of 5azacytidine in culture can be explained by removal of the “unnatural” methylation from genes whose expression triggers differentiation. Thus the myogenesis-promoting gene MyoDl has been found to possess a CpG island that is methylated in lOT1/2 cells but demethylates upon differentiation of these cells into muscle under the influence of 5azacytidine (Jones et al., 1990). In tissues of the mouse, the MyoDl CpG island is not methylated as expected, indicating that methylation is not involved in normal regulation of this gene. Many of the best-documented examples of gene reactivation induced by 5-azacytidine appear to involve reversal of unnatural methylation of CpG islands (Gounari et al., 1987; Sneller and Gunter, 1987). (The only exceptions are genes on the inactive X chromosome, whose CpG islands are de novo methylated naturally.) There is no evidence that 5-azacytidine can directly activate naturally methylated, CpGdeficient promoters (see below and Weih et al., 1991). Understanding Methylation-Mediated Repression Uncertainties in accounting for the results of the DNA methyltransferase knockout underline the need for a better understanding of the DNA methylation system. New
Cell 6
PROMOTER
METH.
LOW
DENSITY
CpG
HIGH
. OOD
WEAK
WEAK
00
+
DENSITY
CpG
Figure 1. A Model to Explain the DNA Methylation on Transcription
Effects
of
The relevant parameters are: proximity of methyl-CpGs to the promoter, density of methyl-CpGs, and strength of the promoter. MeCPl molecules are shown bound loosely (ovals) or tightly (rectangles). Thin and thick arrows represent transcription from a weak promoter and a strong promoter, respectively. Lollipops indicate CpGs that are methylated (closed) or nonmethylated (open). See text for further explanation.
STRONG
STRONG
+
components of the system have come to light through studies of the mechanism of methylation-associated gene inactivation. Several proteins that bind to methylated DNA are known. Two of these, MeCPl and MeCP2 (for methylCpG-binding proteins), can bind to base-paired methylCpG in any sequence context and may therefore be of general significance. MeCPl binds in vitro to DNA containing at least 12 symmetrically methylated CpGs (Meehan et al., 1989) while MeCP2 can bind to a single methylated CpG pair (Lewis et al., 1992)Studies of MeCPl have implicated it in methylation-associated gene inactivation, most strikingly by showing that cells and extracts with low levels of available MeCPl cannot efficiently repress methylated genes (Boyes and Bird, 1991; Levine et al., 1991). The biological significance of MeCP2 is less clear as yet, but in situ immunofluorescence has shown it to be chromosomally bound and concentrated in heterochromatic regions known to be rich in methyl-CpG (Lewis et al., 1992). In addition to MeCP-mediated repression, transcription can also be repressed by direct interference of site-specific methylation with the binding of transcription factors. Although some factors are known to be blocked in this way (e.g., Watt and Malloy, 1988) it is striking that out of seven fully methylated promoters studied, none are inhibited strongly under conditions where MeCPl is absent (Boyes and Bird, 1991, 1992; Levine et al., 1991). Thus a widespread repression mechanism appears to work via MeCPl . In keeping with this, several studies have shown that transcription is sensitive to any methylation near the promoter, not just to specific methyl groups at key sites (Murray and Grosveld, 1987; see below). The importance of Methyl-CpG Density A crucial determinant of repression is the density of methyl-CpGs near the promoter, as illustrated by studies of weak and strong promoters (Figure 1). Weak promoters are fully repressed by sparse methylation, but when the
promoter is strengthened (by adding an enhancer), transcription is restored without methylation loss (Boyes and Bird, 1992, and references therein). If the density of methylation is further increased, even the enhanced promoter cannot prevail, and repression remains complete. The severity of repression is proportional to affinity for MeCPl (Boyes and Bird, 1992). The system resembles a switch, in which full activity or complete repression depends on the balance between methyl-CpG density (i.e., affinity for MeCPl) and promoter strength (Figure 1). Two Sorts of Promoters Where might the density of CpGs be relevant? There are two sorts of promoters in mammalian genes with respect to CpG (Figure 2): those that are constitutively nonmethylated and CpG rich (CpG-island promoters) and those that are relatively CpG poor and (in most tissues) methylated (see Bird, 1986). The latter are invariably found at tissue-specific genes, suggesting a function for DNA methylation that may explain why DNA methyltransferase is essential. By suppressing basal activity of a gene in inappropriate cells (cell types 1,2,3, and 5 in Figure 2) without affecting high level expression in appropriate cells (cell type 4) methylation could increase the amplitude of the gene activation switch. Loss of methylation in expressing cells would in this case be a passive secondary consequence of expression, as indicated by several timing studies (e.g., Sullivan et al., 1989). Attractive though this model is, it is important to bear in mind a piece of contrary evidence. Part of the promoter of the rat tyrosine aminotransferase gene (a CpGdeficient promoter) is bound to ubiquitous factors in expressing cells but is unoccupied in nonexpressing tissues, even though the factors are still present (Becker et al., 1987). Since the gene is methylated in nonexpressing cells but not when expressed, it was thought that methylation might be responsible for keeping the promoter factor free in inap
PROMOTER:
CGI
CDM
CGI
GENE:
HK
TS
TS
II-
CELL TYPE (1)
NFJ
5kb
Figure
2. Two Sorts
of Promoter
at Mammalian
Genes
The diagram shows DNA sequences that are heavily methylated at CpG (hatched) and sequences that are methylation free (open) at two tissue-specific genes (TS) and one housekeeping gene (HK) in five differentiated cell types of a mammal. CGI, CpG-island promoter; CDM, (&G-deficient, methylated promoter. Arrows indicate transcrip tion in that cell type. CpG-island promoters are not methylated regardless of expression (but see text), while CpG-deficient promoters lose methylation when transcribed. The density of CpGs in CpG island promoters is about an order of magnitude higher than that in CpGdeficient promoters.
cell types. If correct, this idea predicts that removal of methylation with 5azacytidine shduld lead to factor binding and activation. In fact, it does not (Weih at al., 1991). The basal suppression model is not eliminated altogether by these results, but its correctness is evidently not a foregone conclusion. The highest density of CpGs occurs at CpG islands, but these are typically not methylated and do not therefore bind MeCPl. In this condition, CpG islands present a paradigm case of open, active chromatin (Tazi and Bird, 1990). Under special circumstances (inactive X, fragile X syndrome, cell lines), CpG islands can become methylated, however, converting them to high affinity substrates for MeCPl (see Figure 1) and leading to long-term repression. The consequences of methylation of CpG islands at the chromatin level have been studied by comparing the human phosphoglycerate kinase gene promoter on inactive and active X chromosomes (Pfeifer et al., 1990). The active (unmethylated) promoter is heavily occupied by factors, whereas the inactive (methylated) promoter is free of footprints. Since the latter can be reactivated by 5azacytidine, methylation alone is probably responsible for these differences. Footprints attributable to MeCPs were not detected over the methylated island, though this does not prove their absence. propriate
Without MeCPs, it is difficult to explain why methylated DNA in chromatin is much more resistant to Mspl (which cuts at CpG) than to other nucleases, or why MeCPdeficient cells show reduced Mspl resistance (Antequera et al., 1989; Levine et al., 1991). The problem might be partially solved if MeCPs serve to guide methylated DNA into a nuclease-resistant structure without remaining bound, perhaps by ensuring late replication (see Riggs and Pfeifer, 1992). That replication is in fact important in the response to methylation has been demonstrated by Hsieh and Lieber (1992). They show that high density CpG methylation prevents the V(D)J joining reaction in lymphocytes but only after DNA replication has taken place. Resistance to Mspl also increases dramatically following replication, consistent with the idea that methylation ushers the replicating DNA into a heterochromatic structure. The three-way relationship between CpG methylation, chromatin structure, and DNA replication may well reward further study. Compensating for Mutability Finally, it is useful to recall that one of the best-characterized aspects of DNA methylation is its mutability. Over one-third of all point mutations giving rise to human genetic disease are due to mutation from CpG to TpG, despite the rarity of CpG and the existence of a dedicated repair system (see Cooper and Krawczak, 1989; Green et al., 1990; Jones et al., 1992). Most animals avoid this problem by keeping methylation away from their own genes. The vertebrates, however, have allowed methylation to spread throughout the genome, including most genes (see Bird, 1986). Was the advent of gene methylation an evolutionary faux pas for which the vertebrates must forever pay? Or is there some compensating advantage to gene methylation that outweighs the increased mutational load? The evidence that DNA methylation is essential for mouse development makes the second possibility much more attractive. References Antequera,
F., Macleod,
Antequera,
F., Boyes,
Becker, Bestor,
D., and Bird, A. P. (1989). J., and Bird, A. P. (1990).
P. B., Ruppert, T. H. (1990).
Bird, A. P. (1986).
S., and Schlitz, Phil. Trans.
Nature
G. (1987).
R. Sot.
(Lond.)
Cell 57, 435-443. B 326,
179-187.
327, 209-213.
Boyes,
J., and Bird, A. (1991).
Cell 64, 1123-l
Boyes,
J., and Bird, A. (1992).
EMBO
Cooper,
D. N., and Krawczak,
Enver, T., Zhang, poulos, G. (1988).
Cell 58, 509-517. Cell 62, 503-514.
M. (1989).
Hum. Genet.
J.-w., Papayannopoulou, Genes Dev. 2, 698-706.
Gounari, F., Banks, G. R., Khazaie, (1987). Genes Dev. 7, 899-912.
134.
J. 77, 327-333 83, 181-188.
T., and Stamatoyanno-
K., Jeggo,
P. A., and Holliday,
R.
Green, P. M., Montandon, A. J., Bentley, D. IX, Ljung, R., Nilsson, I. M., and Giannelli, F. (1990). Nucl. Acids Res. 78, 3227-3231. Hsieh,
C.-L.,
and Lieber,
M. R. (1992).
EMBO
J. 77, 315-325.
Jones, P. A., Wolkowicz, M. J., Rideout, W. M., Ill, Gonzales, F. A., Marziasz, C. M., Coetzee, G. A., and Tapscott, S. J. (1990). Proc. Natl. Acad. Sci. USA 87, 6117-6121. Jones, P. A., Rideout, W. M., Shen, Y. C. (1992). Bioessays, 74, 33-36.
J.-C.,
Levine, A., Cantoni, USA 88. 6515-6518.
A. (1991).
G. L., and Razin,
Spruck,
C. H., and Tsai,
Proc. Natl. Acad.
Sci.
Cell 8
Lewis, J. D., Meehan, R. Ft., Henzel, W. J., Maurer-Fogy, P., Klein, F., and Bird, A. (1992). Cell 69, 905-914. Li, E., Bestor,
T. H., and Jaenisch,
R. (1992).
Meehan, Ft. R., Lewis, J. D., McKay, (1989). Cell 58, 499-507. Murray,
E. J., and Grosveld,
Paroush, Z., Keshet, 1229-l 237.
EMBO
Riggs,
A. D., and Pfeifer,
G. P. (1992).
M. C., and Gunter,
K. C. (1987).
Sullivan, C. H., Norman, J. T., Borras, Mol. Cell. Biol. 9, 3132-3135. Taylor,
S. M., and Jones,
P. A. (1979).
Tazi, J., and Bird, A. (1990). Watt,
F., and Molloy,
E. L., and Bird, A. P. J. 6, 2329-2335.
J., and Cedar,
Pfeifer, G. P., Tanguay, R. L., Steigenvald, (1990). Genes Dev. 4, 1277-1287. Sneller,
Cell 69, 915-926.
S., Kleiner,
F. (1987).
I., Yisraeli,
I., Jeppesen,
H. (1990).
Cell 63,
S. D., and Riggs,
Trends
Genet.
J. Immunol.
A. D.
6, 169-174.
7363505-3512.
T., and Grainger,
R. M. (1989).
Cell 17, 771-779.
Cell 60, 909-920.
P. L. (1988).
Genes
Dev. 2, 1136-l
Weih, F., Nitsch, D., Reik, A., SchOtz, EMBO J. 10,2559-2567.
G., and Becker,
Yisraeli, J., Adelstein, R. S., Melloul, Cedar, H. (1986). Cell 46, 409-416.
D., Nudel,
143. P. 8. (1991).
Lt., Yaffe,
D., and
July 10, 1992, Copyright
0 1992 by Cell Press
The Essentials of DNA Methylation Adrian Bird Institute of Cell and Molecular University of Edinburgh King’s Buildings Edinburgh EH9 3JR Scotland
Biology
An obstacle to progress in understanding eukaryotic DNA methylation has been the lack of a good genetic systemthe most favorable eukaryotes for genetic analysis(yeasts, Caenorhabditis, Drosophila) have no detectable methylation in their genomes. Not only has this hampered progress, but the clear evidence for life without methylation that these organisms provide has raised questions about the relevance of the whole phenomenon. Li et al. (1992) have responded to the need for genetics by creating a methylation mutant and, in doing so, have shown that mice almost certainly have a requirement for DNA methylation. Transgenic approaches were used to obtain mouse embryos homozygous for disruption of the DNA methyltransferase. In a world where gene knockouts can so often be asymptomatic, the effect is notably unambiguous. Homozygous mutant embryos are stunted and die at midgestation. The result is all the more striking since it appears that inactivation of the gene is incomplete (approximately 30% of DNA methyltransferase activity remains in the homozygous disruptants). Evidently, embryogenesis cannot be completed when levels of DNA methyltransferase are reduced. The gene encoding DNA methyltransferase is therefore essential. Death by Leaky Repression? Why is DNA methylation essential? (For the purposes of this article, it is assumed that the mutant phenotype is due to loss of the ability to methylate DNA and not some other as yet unknown function of this protein.) We do not know the answer to this question, but there are some clues. Since DNA methylation can stably repress transcription, it could be that repression is leaky in the absence of DNA methyltransferase, leading to premature death. Repression of what? One possible answer is based on the evidence that cytosine methylation in both eukaryotes and prokaryotes is primarily a system for neutralizing invading DNA (see Bestor, 1990). Failure to repress transposable elements, proviruses, or other potentially damaging sequences could be lethal. Another possibility is that methylation-mediated repression of genes is somehow built into the developmental program. This popular idea could account for the lethality of DNA methyltransferase mutants and is supported by some evidence. The stability of gene repression on the inactive X chromosome in somatic cells of eutherian (but not marsupial) mammals clearly owes much to the methylation of CpG islands (see Riggs and Pfeifer, 1992). More difficult to assess is the wealth of correlative data relating DNA methylation to transcription (e.g., current evidence relating DNA methylation to imprinting), as methylation
Minireview
levels have a known tendency to adapt to the transcriptional status quo. Once they are silenced, genes can subsequently acquire methylation (e.g., Enver et al., 1988, and references therein), and once active, they can subsequently demethylate (e.g., Sullivan et al., 1989). Is methylation dancing to the piper or calling the tune? In one case it may be the latter. Transfection experiments have been used to show that the muscle-specific rat a-actin promoter is repressed by partial methylation in both muscle and nonmuscle cells (Paroush et al., 1990). Activation occurs specifically in myoblasts once demethylation has taken place. An unexplained observation is that constructs in which all CpGs are methylated fail to activate the gene, although they do undergo myoblast-specific demethylation (Yisraeli et al., 1986). 5-Azacytidine May Not Mimic an In Vivo Mechanism Apparent evidence for the involvement of DNA methylation in development came from the early finding that lOTi/ cells differentiate into muscle and other cell types when treated with the methylation inhibitor Qzacytidine (Taylor and Jones, 1979). It seemed that genes were naturally repressed by methylation, and demethylation by 5-azacytidine was activating them. Unfortunately, doubt has been cast on this simple scenario by studies of DNA methylation in cultured cells. In many permanent cell lines, tens of thousands of genes have been stably inactivated, owing to de novo methylation of CpG-island promoters that are normally methylation free (Antequeraet al., 1990). Only genes that are essential for life in culture conditions seem to have been spared (presumably because of selection). Methylation of these promoters is specific to cultured cells; in the animal, their repression (where it occurs) must be due to some other mechanism, since methylation is apparently absent (see below and Figure 2). The effects of 5azacytidine in culture can be explained by removal of the “unnatural” methylation from genes whose expression triggers differentiation. Thus the myogenesis-promoting gene MyoDl has been found to possess a CpG island that is methylated in lOT1/2 cells but demethylates upon differentiation of these cells into muscle under the influence of 5azacytidine (Jones et al., 1990). In tissues of the mouse, the MyoDl CpG island is not methylated as expected, indicating that methylation is not involved in normal regulation of this gene. Many of the best-documented examples of gene reactivation induced by 5-azacytidine appear to involve reversal of unnatural methylation of CpG islands (Gounari et al., 1987; Sneller and Gunter, 1987). (The only exceptions are genes on the inactive X chromosome, whose CpG islands are de novo methylated naturally.) There is no evidence that 5-azacytidine can directly activate naturally methylated, CpGdeficient promoters (see below and Weih et al., 1991). Understanding Methylation-Mediated Repression Uncertainties in accounting for the results of the DNA methyltransferase knockout underline the need for a better understanding of the DNA methylation system. New
Cell 6
PROMOTER
METH.
LOW
DENSITY
CpG
HIGH
. OOD
WEAK
WEAK
00
+
DENSITY
CpG
Figure 1. A Model to Explain the DNA Methylation on Transcription
Effects
of
The relevant parameters are: proximity of methyl-CpGs to the promoter, density of methyl-CpGs, and strength of the promoter. MeCPl molecules are shown bound loosely (ovals) or tightly (rectangles). Thin and thick arrows represent transcription from a weak promoter and a strong promoter, respectively. Lollipops indicate CpGs that are methylated (closed) or nonmethylated (open). See text for further explanation.
STRONG
STRONG
+
components of the system have come to light through studies of the mechanism of methylation-associated gene inactivation. Several proteins that bind to methylated DNA are known. Two of these, MeCPl and MeCP2 (for methylCpG-binding proteins), can bind to base-paired methylCpG in any sequence context and may therefore be of general significance. MeCPl binds in vitro to DNA containing at least 12 symmetrically methylated CpGs (Meehan et al., 1989) while MeCP2 can bind to a single methylated CpG pair (Lewis et al., 1992)Studies of MeCPl have implicated it in methylation-associated gene inactivation, most strikingly by showing that cells and extracts with low levels of available MeCPl cannot efficiently repress methylated genes (Boyes and Bird, 1991; Levine et al., 1991). The biological significance of MeCP2 is less clear as yet, but in situ immunofluorescence has shown it to be chromosomally bound and concentrated in heterochromatic regions known to be rich in methyl-CpG (Lewis et al., 1992). In addition to MeCP-mediated repression, transcription can also be repressed by direct interference of site-specific methylation with the binding of transcription factors. Although some factors are known to be blocked in this way (e.g., Watt and Malloy, 1988) it is striking that out of seven fully methylated promoters studied, none are inhibited strongly under conditions where MeCPl is absent (Boyes and Bird, 1991, 1992; Levine et al., 1991). Thus a widespread repression mechanism appears to work via MeCPl . In keeping with this, several studies have shown that transcription is sensitive to any methylation near the promoter, not just to specific methyl groups at key sites (Murray and Grosveld, 1987; see below). The importance of Methyl-CpG Density A crucial determinant of repression is the density of methyl-CpGs near the promoter, as illustrated by studies of weak and strong promoters (Figure 1). Weak promoters are fully repressed by sparse methylation, but when the
promoter is strengthened (by adding an enhancer), transcription is restored without methylation loss (Boyes and Bird, 1992, and references therein). If the density of methylation is further increased, even the enhanced promoter cannot prevail, and repression remains complete. The severity of repression is proportional to affinity for MeCPl (Boyes and Bird, 1992). The system resembles a switch, in which full activity or complete repression depends on the balance between methyl-CpG density (i.e., affinity for MeCPl) and promoter strength (Figure 1). Two Sorts of Promoters Where might the density of CpGs be relevant? There are two sorts of promoters in mammalian genes with respect to CpG (Figure 2): those that are constitutively nonmethylated and CpG rich (CpG-island promoters) and those that are relatively CpG poor and (in most tissues) methylated (see Bird, 1986). The latter are invariably found at tissue-specific genes, suggesting a function for DNA methylation that may explain why DNA methyltransferase is essential. By suppressing basal activity of a gene in inappropriate cells (cell types 1,2,3, and 5 in Figure 2) without affecting high level expression in appropriate cells (cell type 4) methylation could increase the amplitude of the gene activation switch. Loss of methylation in expressing cells would in this case be a passive secondary consequence of expression, as indicated by several timing studies (e.g., Sullivan et al., 1989). Attractive though this model is, it is important to bear in mind a piece of contrary evidence. Part of the promoter of the rat tyrosine aminotransferase gene (a CpGdeficient promoter) is bound to ubiquitous factors in expressing cells but is unoccupied in nonexpressing tissues, even though the factors are still present (Becker et al., 1987). Since the gene is methylated in nonexpressing cells but not when expressed, it was thought that methylation might be responsible for keeping the promoter factor free in inap
PROMOTER:
CGI
CDM
CGI
GENE:
HK
TS
TS
II-
CELL TYPE (1)
NFJ
5kb
Figure
2. Two Sorts
of Promoter
at Mammalian
Genes
The diagram shows DNA sequences that are heavily methylated at CpG (hatched) and sequences that are methylation free (open) at two tissue-specific genes (TS) and one housekeeping gene (HK) in five differentiated cell types of a mammal. CGI, CpG-island promoter; CDM, (&G-deficient, methylated promoter. Arrows indicate transcrip tion in that cell type. CpG-island promoters are not methylated regardless of expression (but see text), while CpG-deficient promoters lose methylation when transcribed. The density of CpGs in CpG island promoters is about an order of magnitude higher than that in CpGdeficient promoters.
cell types. If correct, this idea predicts that removal of methylation with 5azacytidine shduld lead to factor binding and activation. In fact, it does not (Weih at al., 1991). The basal suppression model is not eliminated altogether by these results, but its correctness is evidently not a foregone conclusion. The highest density of CpGs occurs at CpG islands, but these are typically not methylated and do not therefore bind MeCPl. In this condition, CpG islands present a paradigm case of open, active chromatin (Tazi and Bird, 1990). Under special circumstances (inactive X, fragile X syndrome, cell lines), CpG islands can become methylated, however, converting them to high affinity substrates for MeCPl (see Figure 1) and leading to long-term repression. The consequences of methylation of CpG islands at the chromatin level have been studied by comparing the human phosphoglycerate kinase gene promoter on inactive and active X chromosomes (Pfeifer et al., 1990). The active (unmethylated) promoter is heavily occupied by factors, whereas the inactive (methylated) promoter is free of footprints. Since the latter can be reactivated by 5azacytidine, methylation alone is probably responsible for these differences. Footprints attributable to MeCPs were not detected over the methylated island, though this does not prove their absence. propriate
Without MeCPs, it is difficult to explain why methylated DNA in chromatin is much more resistant to Mspl (which cuts at CpG) than to other nucleases, or why MeCPdeficient cells show reduced Mspl resistance (Antequera et al., 1989; Levine et al., 1991). The problem might be partially solved if MeCPs serve to guide methylated DNA into a nuclease-resistant structure without remaining bound, perhaps by ensuring late replication (see Riggs and Pfeifer, 1992). That replication is in fact important in the response to methylation has been demonstrated by Hsieh and Lieber (1992). They show that high density CpG methylation prevents the V(D)J joining reaction in lymphocytes but only after DNA replication has taken place. Resistance to Mspl also increases dramatically following replication, consistent with the idea that methylation ushers the replicating DNA into a heterochromatic structure. The three-way relationship between CpG methylation, chromatin structure, and DNA replication may well reward further study. Compensating for Mutability Finally, it is useful to recall that one of the best-characterized aspects of DNA methylation is its mutability. Over one-third of all point mutations giving rise to human genetic disease are due to mutation from CpG to TpG, despite the rarity of CpG and the existence of a dedicated repair system (see Cooper and Krawczak, 1989; Green et al., 1990; Jones et al., 1992). Most animals avoid this problem by keeping methylation away from their own genes. The vertebrates, however, have allowed methylation to spread throughout the genome, including most genes (see Bird, 1986). Was the advent of gene methylation an evolutionary faux pas for which the vertebrates must forever pay? Or is there some compensating advantage to gene methylation that outweighs the increased mutational load? The evidence that DNA methylation is essential for mouse development makes the second possibility much more attractive. References Antequera,
F., Macleod,
Antequera,
F., Boyes,
Becker, Bestor,
D., and Bird, A. P. (1989). J., and Bird, A. P. (1990).
P. B., Ruppert, T. H. (1990).
Bird, A. P. (1986).
S., and Schlitz, Phil. Trans.
Nature
G. (1987).
R. Sot.
(Lond.)
Cell 57, 435-443. B 326,
179-187.
327, 209-213.
Boyes,
J., and Bird, A. (1991).
Cell 64, 1123-l
Boyes,
J., and Bird, A. (1992).
EMBO
Cooper,
D. N., and Krawczak,
Enver, T., Zhang, poulos, G. (1988).
Cell 58, 509-517. Cell 62, 503-514.
M. (1989).
Hum. Genet.
J.-w., Papayannopoulou, Genes Dev. 2, 698-706.
Gounari, F., Banks, G. R., Khazaie, (1987). Genes Dev. 7, 899-912.
134.
J. 77, 327-333 83, 181-188.
T., and Stamatoyanno-
K., Jeggo,
P. A., and Holliday,
R.
Green, P. M., Montandon, A. J., Bentley, D. IX, Ljung, R., Nilsson, I. M., and Giannelli, F. (1990). Nucl. Acids Res. 78, 3227-3231. Hsieh,
C.-L.,
and Lieber,
M. R. (1992).
EMBO
J. 77, 315-325.
Jones, P. A., Wolkowicz, M. J., Rideout, W. M., Ill, Gonzales, F. A., Marziasz, C. M., Coetzee, G. A., and Tapscott, S. J. (1990). Proc. Natl. Acad. Sci. USA 87, 6117-6121. Jones, P. A., Rideout, W. M., Shen, Y. C. (1992). Bioessays, 74, 33-36.
J.-C.,
Levine, A., Cantoni, USA 88. 6515-6518.
A. (1991).
G. L., and Razin,
Spruck,
C. H., and Tsai,
Proc. Natl. Acad.
Sci.
Cell 8
Lewis, J. D., Meehan, R. Ft., Henzel, W. J., Maurer-Fogy, P., Klein, F., and Bird, A. (1992). Cell 69, 905-914. Li, E., Bestor,
T. H., and Jaenisch,
R. (1992).
Meehan, Ft. R., Lewis, J. D., McKay, (1989). Cell 58, 499-507. Murray,
E. J., and Grosveld,
Paroush, Z., Keshet, 1229-l 237.
EMBO
Riggs,
A. D., and Pfeifer,
G. P. (1992).
M. C., and Gunter,
K. C. (1987).
Sullivan, C. H., Norman, J. T., Borras, Mol. Cell. Biol. 9, 3132-3135. Taylor,
S. M., and Jones,
P. A. (1979).
Tazi, J., and Bird, A. (1990). Watt,
F., and Molloy,
E. L., and Bird, A. P. J. 6, 2329-2335.
J., and Cedar,
Pfeifer, G. P., Tanguay, R. L., Steigenvald, (1990). Genes Dev. 4, 1277-1287. Sneller,
Cell 69, 915-926.
S., Kleiner,
F. (1987).
I., Yisraeli,
I., Jeppesen,
H. (1990).
Cell 63,
S. D., and Riggs,
Trends
Genet.
J. Immunol.
A. D.
6, 169-174.
7363505-3512.
T., and Grainger,
R. M. (1989).
Cell 17, 771-779.
Cell 60, 909-920.
P. L. (1988).
Genes
Dev. 2, 1136-l
Weih, F., Nitsch, D., Reik, A., SchOtz, EMBO J. 10,2559-2567.
G., and Becker,
Yisraeli, J., Adelstein, R. S., Melloul, Cedar, H. (1986). Cell 46, 409-416.
D., Nudel,
143. P. 8. (1991).
Lt., Yaffe,
D., and
