DNA methylation affects the formation of active chromatin

Cell, Vol. 44, 535-543,

February

28, 1988, Copyright

0 1986 by Cell Press

DNA Methylation Affects the Formation of Active Chromatin llana Keshet, Judy Lieman-Hurwitz, and Howard Cedar The Department of Cellular Biochemistry and the Department of Molecular Biology The Hebrew University Jerusalem, Israel

To study the mechanism of gene repression by DNA methylation, Ml3 gene constructs were methylated to completion and inserted into mouse L cells by DNAmediated gene transfer. All unmethylated sequences, regardless of their source, integrated into the DNA in a potentially active DNAase l-sensitive conformation. Total CpG methylation prevented the formation of this structure and rendered these sequences DNAase l-insensitive over the entire methylated domain. Whereas unmethylated DNA demonstrated additional conformational features of active genes, such as DNAase I hypersensitivity and restriction endonuclease-sensitive segments, these markers were not present when methylated DNA was used for transfection. The use of micrococcal nuclease to probe for active or inactive supranucleosome particles also showed that DNA methylation directs DNA into an inactive type of structure. The results suggest that DNA methylation may exert its effect on gane transcription by altering both specific and nonspecific interactions between DNA and nuclear proteins. Introduction Several lines of evidence suggest that DNA methylation may play a role in the regulation of gene activity in animal cells. In addition to a strong in vivo correlation between gene expression and DNA undermethylation in specific cell types, in vitro methylation of many gene sequences causes an inhibition of their expression when measured either in fibroblasts or in Xenopus oocytes (Cedar, 1984). Furthermore, in some cells, gene functions can be activated following treatment with 5aza-cytidine, a potent hypomethylating agent (Jones, 1984). Despite the recognition that DNA methylation can inhibit gene expression, the mechanism of action of this modification is not understood, but it is usually assumed that these nucleotide modifications alter the specific binding of either regulatory proteins or other elements of the transcriptional machinery. Consistent with this concept are the experiments showing that some genes are influenced by DNA methylation in their 5’ regulatory regions but remain unaffected by modification in the body of the gene, which is in many cases probably devoid of specific protein-interactive sequences (Busslinger et al., 1983; Kruczek and Doerfler, 1983; Langner et al., 1984). In contrast to this, massive methylation of the coding sequences of the Herpes tk

gene had a profound effect on transcription in mouse fibroblasts even though the clearly defined promoter region of this gene was left unmodified (Keshet et al., 1985). This experiment suggested that DNA methylation may also influence RNA synthesis by interfering with factors not related to the specific protein-DNA interactions that take place in the promoter region. DNA-mediated gene transfer into mouse L cells represents an excellent system for studying the effect of DNA methylation on chromatin structure. Presumably because of the selection process required to obtain stable transformants, almost all gene sequences introduced into this system are actively expressed, albeit occasionally at very low levels. Furthermore, whenever examined, these genes and even other nonexpressible transfected sequences have always been found in an active chromatin structure as determined by their nuclear sensitivity to DNAase I or other nucleases (Weintraub, 1983; Sweet et al., 1982; Davies et al., 1982). We reasoned that if DNA methylation mediates its inhibitory effect through specific protein-DNA interactions at certain critical sequences in the gene domain, a methylated construct should be present in transfected cells in a DNAase l-sensitive conformation similar to other nonexpressed, nonmethylated sequences inserted into cells in the same manner (Weintraub, 1983). If, on the other hand, DNA methylation also affects nonspecific interactions involving the creation of the active chromosomal state, such methylated molecules may demonstrate an altered, inactive nuclease-insensitive structure. Using a combination of different types of methylated gene constructs, we demonstrate that the presence of this modification on transfected DNA alters its overall DNAase I sensitivity and inhibits the formation of DNAase I and restriction enzyme-hypersensitive sites, creating a structure similar to that seen when a gene is in its inactive state. These studies suggest that methylation may be one of the important elements necessary for casting the correct chromosomal structure throughout the genome. Results To obtain transfectants that contain integrated DNA molecules methylated at every CpG residue, it was necessary to insert the appropriate gene sequences into M13. To this end, the recombinant Ml3 a-CAT moleculecontaining the rat a-actin promoter region linked to the CAT gene and including both pBR322 and SV40 sequences was used as a template in vitro for complementary strand synthesis in the presence of either dCTP or dmCTP Resulting doublestranded Ml3 molecules were then inserted into mouse L cells by DNA-mediated gene transfer using the Herpes tk gene to select for stable colonies, which were then pooled to yield a large average population of transfected cells. Because of the maintenance methylation machinery of these cells, in vitro hemimethylated molecules are converted in vivo to DNA methylated exclusively at every CpG residue

Cdl

536

A

C

B

6

12345

12345

6 (/

I*i

1

234

5

C

C e-7

a-l.2

)c-,

MC

% MC .. :

i

e7

+1.2

-I

e4.2 +a

t

1

234

F

5

a

1

2

3

4

5

C

C

-4

-4.4 t b

Figure 1. DNAase I Sensitivity

of Methylated

and Unmethylated

Transfected

DNA

L cell nuclei (@/ml) containing the unmethylated (C)or methylated (W) Ml3 a-actin construct or Ml3 human p-globin construct were digested with increasing concentrations of DNAase I (lane 1, no enzyme; lane 2, 0.3 &ml; lane 3. 1 pglml; lane 4, 3 Kg/ml; and lane 5, 10 &/ml) for 20 min at 3pC in a volume of 300 ~1. Deproteinized DNA was then subjected to restriction enzyme analysis, gel electrophoresis on 1% agarose gels, and blot hybridization using appropriate probes. In all cases digestion of total DNA for both cell types was determined to be identical for each DNAase I concentration by visual analysis of the ethidium bromide-stained gels. (A) DNA was digested with Pst I and hybridized with probe a. The marker shown in lane 6 represents a mixture of p a-CAT DNA digested with Pst I and Pst I-Sac I. The bulk of the hybridizable material indeed appears at the appropriate size of 4.2 kb, but a minor fraction consistently appears at about 2 kb. (B) Digestion with Pst I and hybridization with probe b (the entire Ml3 sequence). The marker in lane 6 was produced by linearization of M13mp8 RF (C)The W40 sequences were analyzed by cleavage with Pst I-Hinf I and hybridization with probe c. (D) DNA was digested with Pvu II and probed with the HW-tk0.8 kb Bgl II-Hinf I fragment @&Knight, 1980). The marker in lane 6 shows a Pvu II digestion of cellular DNA. (E) DNA was digested with Pst I and probes with the 1 kb Barn HI-&o RI genomic fragment of human Bglobin. This blot shows the appearance of the expected 4.4 kb globin fragment. (F) DNA was digested with Pst I and probed (a) with the 0.9 kb Pst I fragment of a cDNA clone of mouse DHFR (Stein et al., 1983) and (b) with the 0.8 kb Pst I-Hind Ill fragment of the mouse P-globin gene (Hofer and Darnell, 1981). DNA samples used for parts A, B, C, D, and F were taken from one experiment in which nuclei containing either methylated orunmethylated o-actin CAT DNA were digested with various concentrations of DNAase 1. Thus, the quantities of DNA in all lanes are equal. DNA samples used for part E were taken from DNAase l-treated nuclei obtained from cells containing either the methylated or unmethylated human Sglobin. The map indicates the locations of cleavage of Pst I (P), Hinf I (H), and Sac I (S) and the various probes used in the blot hybridizations of the a-actin construct. The large triangle marks the location of a strong hypersensitive site in the actin promoter region.

DNA Methylation 537

and Active Chromatin

Figure 2.. Quantitative

Formation

Analysis of DNAase I Sensitivity

The autoradiograms shown in Figure 1 were scanned by optical densitometry, and the percent DNAase I resistance of each sequence was digitalized and graphed as a function of the DNAase I concentration. The digestion patterns from cells containing methylated DNA are represented by filled-in symbols, while those from control cells are depicted by open symbols. The DNA fragments analyzed include the o-CAT-containing Pst I fragment (0, H) (see Figure lA), the M13containing Pst I fragment (0, 0) (see Figure lB), the tSV40containing Hinf I-Pst I fragment (+) (see Figure 1C) from Ml3 o-actin CAT and the Pvu II fragment of the t/r gene (A, A) (see Figure ID). Naked total DNA from cells containing either the methylated (v) or unmethylated (v) constructs was also subjected to DNAase I digestion with low concentrations of enzyme, and the resulting DNA was subjected to Pst I digestion and blot hybridization using probe a. The axis of the plot shows the concentration of DNAase I in micrograms per milliliter, and the ordinate shows percent DNAase I resistance.

(Stein et al., 1982). This level of modification had a profound effect on the activity of the a-actin promoter, since CAT activity in cells containing the methylated constructs was about 30-fold less than that observed for its unmethylated counterpart (Yisraeli and Adelstein, unpublished results). The DNA methyl moieties were found to be faithfully transmitted during cell division, as shown by the fact that Hpa II (CCGG) and Hha I (GCGC) sites remained greater than 95% methylated following over 70 generations of growth in selective medium (data not shown). Nuclei from L cell cultures made up of pooled clones of unmethylated transfectants were treated with increasing concentrations of DNAase I, and the resulting DNA was analyzed by restriction digestion and blot hybridization (Figure 1). The large Pst I fragment that includes the entire CAT gene and its promoter as well as pBR322 and SV40 sequences was clearly sensitive to DNAase I at a level of about 1 rglml of enzyme. In contrast, in nuclei containing the methylated construct, the identical fragment was still resistant to DNAase I digestion at concentrations 10 times greater. These data suggest that there is a pronounced difference in sensitivity to nuclease between unmethylated and methylated DNA. The decreased nuclease sensitivity associated with the methylated construct was not restricted to the immediate region of the gene promoter, since the entire Ml3 sequence encompassing 7 kb of DNA was also found to be resistant to DNAase I treatment. Furthermore, other Ml3 methylated constructs, including the chicken adult /@lobin gene and the Herpes tk gene, behaved similarly in independent transfection experi-

ments. The effect of methylation on the DNAase I sensitivity of the human fi-globin gene is illustrated in Figure 1. It should be noted that DNAase I sensitivity of transfected DNA was not restricted to actively transcribed regions, since both Ml3 and pBR322 sequences were found to be nuclease-sensitive, despite the fact that dot blots revealed only marginal levels of RNA. The effect of DNA methylation in these experiments was limited to the immediate locality of the methylated regions themselves. The tk gene, which was used as the selective vector for DNA-mediated gene transfer, was inserted unmethylated in conjunction with either the methylated or unmethylated Ml3 construct. The 2 kb Pvu II fragment containing this gene was found to be DNAase l-sensitive in both cases, consistent with the fact that this region is unmethylated. It should be noted that for all of the unmethylated exogenous sequences used in this study, the range of DNAase I sensitivity (0.5-3 rglml) was similar to that observed for active endogenous genes from various specific cell types (Carmon et al., 1982; Sheffery et al., 1982; McGhee et al., 1981). As shown in Figure 1, the endogenous active DHFR gene was sensitive to DNAase I to the same degree as the transfected unmethylated constructs. On the other hand, the kinetics of digestion of the inactive mouse globin gene was similar to that for the methylated exogenous DNA. We also studied the digestion of a 1.2 kb fragment containing SV40 coding sequences, which fortuitously lacks any CpG dinucleotide and is therefore necessarily unmethylated even in the methylated construct. In this case there was a slight (2fold) difference in sensitivity between nuclei containing the unmethylated and methylated construct, but it is clear from Figure 2 that methyl moieties surrounding this fragment do not force it into a totally DNAase-insensitive conformation. On the other hand, the small observed difference in sensitivity does suggest some spreading of the effect, which might be concentrated at the edges of the 1.2 kb fragment closest to the heavily methylated DNA. In any event, the range of the methylation effect appears limited to less than a few hundred nucleotides. It is not likely that DNA methylation has an inhibiting effect beyond this range since, in at least the case of the hamster aprt gene, methylation flanking the promoter region does not inhibit its activity either in vivo (Stein et al., 1983) or in vitro (Keshet et al., 1985). One possible explanation for the difference in nuclease sensitivity between methylated and unmethylated substrates is that methylated DNA itself inhibits the nuclease reaction or serves as a poor substrate for this enzyme. To control for this possibility, total cellular DNA was isolated from both types of pooled clones, digested with very low concentrations of DNAase I, and analyzed by blot hybridization in a manner analogous to nuclear preparations. Despite the marked difference in the overall kinetics of nuclease digestion between chromatin and naked DNA (Felber et al., 1981) it is clear that methylation does not influence the specificity or extent of this reaction (Figure 2). In any event, methylated DNA is digested by DNAase I at levels of enzyme 2 orders of magnitude less than that required to digest either unmethylated or methylated DNA

Cell 530

C +

4.2

c2

5

c4.2 c2 Figure 3. Detection of DNAase I Hypersensitive

Sites

Nuclei containing the unmethylated (c) or methylated (‘-‘C) Ml3 U-CAT were treated with DNAase I (lane 1, no enzyme; lane 2,0.3 rglml; lane 3,l &ml; lane 4.3 wgglml; lane 5,lO pglml) for 20 min at 37%. Samples containing 20 rg were deproteinized and digested with Pst I, run on 1% agarose gels, blotted and hybridized to the CAT fragment (probe d in Figure 1). The marker shown in lane 6 is identical with that included in Figure IA.

in nuclei, suggesting that the effect of DNA methylation is indeed on the protein-DNA conformation and not at the level of the DNA itself. Since active genes are generally undermethylated in comparison to the same genes in inactive cell types, methylation effects at the level of DNA might well have offered an explanation for the difference in DNAase I sensitivity between active and inactive genes in nuclei from different tissues. Our experiments on selected sequences in L cells provide an important control for these types of studies by showing that differences in DNAase I sensitivity are determined exclusively at the level of chromatin conformation. Generalized DNAase I sensitivity is just one of the conformational characteristics of active DNA. It was also of interest to investigate the effect of DNA methylation on other structural markers that are usually associated with the active state. DNAase I hypersensitive sites have been observed, usually at the 5’ end, in many active tissuespecific and constitutive genes, sites that are usually absent in cell types in which these genes are inactive (Weisbrod, 1982). In the experiments shown in Figure 1, no hypersensitive site was observed in the actin promoter region when blots were hybridized using the full Pst I fragment as probe. This is consistent with the fact that hypersensitive sites were not observed in normal rat muscle nuclei (Carmon et al., 1982). Since the appearance of subbands may have been hidden by the presence of an unexplained minor 2 kb band observed in Figure 1, similar blots were assayed by hybridization with alternate probes representing the 3’end of the insert (probed). At relatively low concentrations of DNAase I, a hypersensitive site indeed appears within the domain of the unmethylated &AT construct in mouse L cells, yielding a sub-band of about 2 kb, and restriction mapping suggests that this site

is within the cr-actin promoter region, not far from the Sac I site. In the methylated construct, however, this sub-band does not appear, even following DNAase I digestion with large excesses of enzyme at which the original 4.2 kb Pst I band is digested almost to completion (Figure 3). Similar results were obtained with methylated and unmethylated Ml3 constructs containing the chicken adult globin gene, which contains a strong hypersensitive site in its promoter region (McGhee et al., 1981). In addition to DNAase I hypersensitive sites, many active genes have a small region at their 5’ regulatory end in which restriction sites are accessible to enzyme cleavage even within nuclei, suggesting that these areas may contain nucleosome gaps. In the adult chicken fl-globin gene, for example, a 114 bp region has been shown to be sensitive to digestion by several restriction enzymes in chicken blood cells and their precursors, but not in cells that do not express the globin phenotype (McGhee et al., 1981). On the other hand, when inserted into L cells as part of a globin promoter-thymidine kinase construct by DNA-mediated gene transfer, this sequence adopted an active conformation including the correct restriction enzyme-sensitive segment (G. Felsenfeld, personal communication). A DNA fragment containing this region was cloned into Ml3 and introduced, either in the modified or unmodified form, into mouse t/r- fibroblasts by DNAmediated gene transfer. Nuclei from these cells were treated with the enzymes Hinf I or Msp I, and the DNA was analyzed by blot hybridization (Figure 4). The region of the inserted globin gene was first analyzed by secondary digestion with Barn HI, which yields a 2.1 kb fragment. Nuclear digestion with Msp I produces a clear band of 1.0 kb, which represents cutting in the nuclease-hypersensitive region of the promoter, and Hinf I treatment shows a barely detectable band also resulting from specific digestion within this region. This gene, however, is not cut by other restriction enzymes that do not have sites within the nucleosome gap region. Although this experimental approach was useful for demonstrating restriction enzyme sensitivity in nuclei containing the unmethylated gene, the heavily methylated construct could not be analyzed by Barn HI digestion, since methylation at the CpG residues flanking this restriction site prevented cleavage by this enzyme. As an alternative, we examined the small 585 bp Pvu II fragment for specific cleavage. While nuclei containing the unmethylated constructs are sensitive to specific Msp I and Hinf I digestion, nuclei harboring the methylated DNA are totally unaffected by restriction enzyme treatment in this region, even at very high enzyme concentrations. A similar restriction enzyme-sensitive region has been demonstrated in the promoter region of the HSV fk gene inserted into mouse L cells (Sweet et al., 1982). To test the effect of DNA methylation on the nuclease sensitivity of rk, this gene was cloned into Ml3 and inserted into L cells following total CpG methylation. Although we did not observe a specific region of restriction enzyme sensitivity in these constructs, DNA methylation completely inhibited enzyme sensitivity at all sites (data not shown). Previous studies of the unmodified tk gene inserted into L cells

DNA Methylation 539

A

564-

and Active Chromatin

Formation

B

1234

I)

Figure 4. Detection of Restriction

Enzyme-Sensitive

Sites

Nuclei (107/100 ~1) containing methylated or unmethylated copies of the chicken 6-globin gene were digested with 90 units of Msp I or Hinf I at 37V for 1 hr. The reaction was terminated, and samples were deproteinized and digested with appropriate restriction enzymes, run on agarose gels, blotted, and hybridized to the Pvu II fragment of the globin gene. (A) Nuclei containing unmethylated chicken globin gene were digested with Msp I (lane 3) Hinf I (lane 2) or no enzyme (lane 4). Extracted DNA (20 pg) was digested with Barn HI and run on a 1% agarose gel. Lane 1 shows a A Hind Ill digest. (B) Nuclei containing the unmethylated (lanes 5-7) and methylated (lanes 2-4) chicken globin insert were digested with Msp I (lanes 3 and 6) Hinf I (lanes 2 and 5) or no enzyme (lanes 4 and 7). Extracted DNA (20 pg) was digested with Pvu II and run on 1.5% agarose gel. Lane 1 contains a Pvu II digest of the chicken globin plasmid. Blot hybridization was carried out using a gene screen plus transfer membrane, rather than nitrocellulose. The map represents the Ml3 construct containing the chicken globin gene showing the sites of the restriction enzymes Barn HI (B), Pvu II (P), Hinf I (H), and Msp I (M). The arrow shows the start site and direction of transcription.

showed that counterselection for the r/r-- phenotype is associated with the appearance of clones in which the t/r gene has undergone de novo methylation and an alteration in its DNAase I sensitivity (Sweet et al., 1982). Our experiment indirectly suggests that DNA methylation may in fact be one of the causal factors that directs the formation of the new structure typical of the inactive state. The above data all indicate that DNA methylation has an effect on the local chromatin structure in the immediate vicinity of the modified DNA. It has recently been shown that the active domain has a distinctive higher order structure that stretches over large regions encompassing at least 40-50 kb (Weintraub, 1984). When nuclei are digested with low concentrations of micrococcal nuclease and electrophoresed as nuclear protein complexes, two types of discrete supranucleosomal particles are observed. The “a” particles, which run relatively slowly, represent inactive domains, while the “b” particles contain active genes and run in an accelerated fashion on the gel. In both cases these particles are held together by higher order protein-DNA interactions despite the abundant micrococcal nuclease cleavage points within the DNA. It

is not known, however, what property of this active domain is responsible for the altered gel migration. To determine whether DNA methylation plays some role in the formation of active domains, methylated and unmethylated M13chicken /3-globin were introduced into mouse L cells by DNA-mediated gene transfer in the absence of any additional carrier to ensure that many methylated copies would integrate in tandem and thus produce an optimal effect on higher order structure. Nuclei from pooled clones containing these methylated and unmethylated constructs were treated with micrococcal nuclease, and the nucleoprotein particles were analyzed by gel electrophoresis followed by blot hybridization (Figure 5). While the unmethylated DNA was contained in the activity-associated “b’ particles, the methylated DNA-containing complexes were retarded on the gel and migrated like inactive particles. The endogenous inactive mouse fi-globin gene was found exclusively in the “a” particle fraction in both types of nuclei. It should be noted that the DNA methylation had no apparent effect on the overall nucleosome structure and spacing, since modified DNA yielded a normal nucleosome ladder when nuclei were treated with mi-

Cell 540

B

a-

-

b-

-b

2

a x4.3

< 2.3 < 2.0

Figure 5. Detection of Supranucleosome

Particles

(A) Nuclei containing methylated or unmethylated MIS-chicken globin were digested with micrococcal nuclease (0.3 &ml), and the EDTAreleased nucleoprotein was electrophoresed directly on a 1% agarose minigel (20 m M Tris, pH 7.4; 50 m M sodium acetate; and 2 m M EDNA). The photograph shows the same blot hybridized to an Ml3 probe (lanes 1 and 2) and to a mouse P-globin probe (lanes 4 and 5). lanes 1 and 4 show nucleoprotein particles released from L cells transfected with the ummethylated Ml3 chicken &globin. Lanes 2 and 5 show nucleoprotein particles released from L cells transfected with the methylated construct. For both cell types, digestions were carried out with several concentrations of micrococcal nuclease, but only the concentrations showing the limit particle are included in the figure. Lanes 3 and 6 show end-labeled Hind Ill-digested DNA. (B) Nuclei containing methylated Ml3tk sequences were digested for 20 min at 22oC with micrococcal nuclease (lane 1, 0.4 &ml; lane 2, 0.6 pg/ml). The reactions were terminated, and the samples were deproteinized, run on 1.5% agarose gels, blotted using a gene screen plus transfer membrane, and hybridized to an Ml3 nick translated probe. An end-labeled Hind Ill digest of A DNA is shown in lane 3.

nuclease, and the resulting purified DNA was analyzed by gel electrophoresis and blot hybridization (Figure 5).

crococcal

Discussion Although there is a strong correlation between gene activity, DNA undermethylation, and DNAase I sensitivity, it has been hard to unravel the causal relationships among all of these factors. It has already been demonstrated that for several genes that were artificially inserted into specific cell types, DNA methylation acts in a causal manner to inhibit gene expression (Cedar, 1984). These experiments do not, however, explain the mechanism of this methyl-directed repression. The data in this paper suggest that one of the modes of action of DNA modification may involve an alteration in DNA-protein interactions over gene-sized domains, leading to the inability to set up the conformational elements characteristic of active chromatin. Consistent with previous data, all unmethylated DNA inserted into mouse L cells was found to be in a DNAase l-sensitive conformation, regardless of the transcriptional potential of the segment. Total CpG DNA methylation of any DNA fragment acted in a dominant manner to prevent the formation of this structure. Other more specific structural features of the active state, including hypersensitive

sites for DNAase I and restriction enzymes, were also affected by DNA modification. Any proposed mechanism to explain the inhibitory effect of DNA methylation on gene transcription must invoke alterations in protein-DNA interactions. Thus, methylation may directly modify the interaction of RNA polymerase or other transcription factors with their DNA binding sites, or could influence transcription by altering the placement of ubiquitous structural nuclear proteins, such as histones, HMG, or matrix proteins, which may be responsible for organizing the conformation unique to the active state. In some cases methylation-induced repression probably involves interference with the binding of specific proteins on DNA sites known to be involved in gene regulation. Thus, for Adeno type 2 and type 12 genes (Kruczek and Doerfler, 1983; Langner et al., 1984) human r-globin (Busslinger et al., 1983), and hamster aprt (Keshet et al., 1985) only the Yupstream sequences are sensitive to DNA modification when tested in an expression assay, suggesting that in these cases the methyl cytosines may act as a type of point mutation that alters a specific recognition site. In none of these cases, however, has it been directly shown that DNA methylation alters the binding of specific proteins. The fact that the methyl group can indeed influence the affinity of specific proteins at their recognition site has been demonstrated for the E. coli lac repressor (Fisher

DNA Methylation 541

and Active Chromatin

Formation

and Caruthers, lS79), and for bacterial restriction enzymes. Although micrococcal nuclease treatment revealed a normal nucleosome ladder on methylated DNA, we still consider it likely that DNA methylation may also have subtle effects on nucleosome placement along the DNA. In several instances, including SV40, HSV-t/r, and the chicken adult globin gene, the presence of an extensive region of nuclease sensitivity has suggested the existence of a nucleosome gap (Weisbrod, 1982), and in the case of the SV40 origin region, this can indeed be observed by electron microscopy in a portion of the minichromosomes (Saragosti et al., 1980). Such a restriction enzyme-sensitive region was also observed for the chicken globin promoter, linked to the t/r gene and inserted into L cells (G. Felsenfeld, personal communication, and this paper), and the formation of this presumed nucleosomefree segment was inhibited by DNA methylation, suggesting that this modification can influence nucleosome placement either directly or through protein factors. Previous studies have indeed shown that DNA within the nucleosome core is unexpectedly rich in methyl moieties, suggesting that nucleosomes may, indeed, have some preference for the way in which they distribute themselves with regard to methyl groups on the DNA (Razin and Cedar, 1977; Solage and Cedar, 1978). The idea that DNA methylation plays a role in the organization of chromosome conformation may be helpful for understanding several biological processes. Although the active state of tissue-specific genes is probably determined by protein factors that appear at critical stages of development, this state is maintained over many generations and in some cases despite the absence of the original determination factors (Burch and Weintraub, 1983). Our data suggest that the DNA methylation pattern could provide a template for directing the inheritance of chromatin structure following DNA replication and cell division. The sequence of events occurring at the replication fork is consistent with this proposal, since maintenance methylation takes place less than 1 min following replication (Kappler, 1970; Gruenbaum et al., 1983), while the establishment of the DNAase l-sensitive state takes 3-5 min (Weintraub, 1979) and the final deposition of all histone proteins is only complete after 10 min (Worcel et al., 1978). Methylation patterns may also be responsible for the establishment of chromatin structure during early stages of development. We have previously shown that several constitutive genes have a fixed methylation pattern in all somatic and germ-line DNA, usually characterized by a methylated 3’end and an undermethylated 5’region (Stein et al., 1983). This pattern carried in sperm DNA could serve as a map for directing the organization of the active chromatin structure allowing these genes to be expressed immediately even in cells at early stages of development, while tissue-specific genes, which are fully methylated in sperm, would remain in an inactive conformation. Consistent with this model is the fact that the regions that are unmethylated in housekeeping genes are also sites of DNAase I hypersensitivity in somatic cells (Groudine and Conklin, 1985). Since most of the genetic material is prob-

ably inactive in any given cell type, the dominant nature of the effect of DNA methylation could provide a simple and efficient mechanism for ensuring that these genes remain in an inactive chromosomal conformation. The idea that DNA methylation affects the overall chromatin structure in addition to specific binding sites is consistent with other observations on the distribution of DNA methyl groups in the regions of tissue-specific genes. These gene sequences are generally fully methylated at every local restriction site tested in DNA from sperm or other nonactive tissue specimens (Yisraeli and Szyf, 1984). Assuming that these sites are representative, this suggests that every CpG in these regions is methylated. This level of methylation is equivalent to that obtained with the Ml3 constructs used in this study, and one would expect this level of DNA modification to have a profound effect on chromatin structure in vivo. In the active state these genes are almost always undermethylated at numerous sites distributed over the entire gene domain and flanking sequences, and in some cases every testable restriction site has been found to have undergone demethylation (Yisraeli and Szyf, 1984). This regional effect suggests that the influence of DNA methylation is on the general structure of the gene as opposed to specific interactions in unique regulatory sequences. While these data suggest that DNA methylation influences chromatin structure, it is clear that chromosomal conformation is also determined by other factors not associated with the DNA template itself. Hormone induction of chicken vittelogenin, for example, causes the appearance of a sharp hypersensitive site in the promoter region of the gene, and this is followed by the demethylation of a Hpa II site within the same domain a few generations later (Burch and Weintraub, 1983). It should be noted, however, that while protein factors may be responsible for the induction of the active state, changes in DNA methylation could subsequently be required to ensure the maintenance of this structure. It has been noted that some eukaryotic organisms do not contain detectable levels of DNA methylation (UrieliShoval et al., 1982), which some have interpreted as an indication that DNA methylation may not be important even in those organisms in which it exists. We suggest that the universal mechanism for regulating gene expression involves a complicated network of protein-DNA interactions. These structural features are probably set up and maintained by specific protein-DNA recognitions. In organisms that have methylated cytosine, this modification is one of the factors playing a role in this recognition process and provides a novel way of altering DNA sequences in a reversible manner. Experimental

Procedures

Ml3 Conetructe and Synthesis p o-CAT, which contains the rat a-actin promoter region linked to the bacterial CAT (chloramphenicol acetyl transferase) gene (Melloul et al., 1984) was obtained from D. Yaffe and U. Nudel, and the 4.2 kb Pst I fragment containing the &AT was recloned into M13mp8. M13mp8 containing a 2.1 kb Sac I partial digest of the chicken adult globin gene inserted into the Hint II site of the Ml3 polylinker (BDH-10) was obtained from G. Felsenfeld and D. Jackson. S. McKnight kindly provided

the Ml3 clone containing the 2.0 kb Pvu II Herpes simplex virus thymidine kinase gene (HSV-tk) fragment (f&Knight, 1980). M13mp8 containing the 4.4 kb Pst I fragment of the human 8globin gene was obtained from Ft. Flavell. Single-stranded DNA was prepared (Stein et al., 1982) from these recombinant phages by polyethylene glycol precipitation followed by standard DNA extraction techniques. Second-strand synthesis from single-stranded Ml3 molecules was performed under conditions previously described for 1 hr at 15OC using a DNA concentration of 2-3 rg per 100 PI and a polymerase concentration of 75 units/ml. The 1Bmer universal Ml3 primer was obtained from New England Biolabs and was used at a level of 25 ng per N of Ml3 DNA. These reactions were carried out using either dCTP or dmCTP (2’-deoxy-5-methylcytidine triphosphate; P-L Eiochemicals) to obtain methylated molecules and their unmethylated controls. Double-stranded molecules were purified by phenol-chloroform extractions and ethanol precipitation before their introduction into L cells by DNA-mediated gene transfer (Wigler et al., 1979). Analysis of Transfected Cells DNA-mediated gene transfer into mouse Ltk- cells was performed as described (Wigler et al., 1979) using 100-200 ng of the selective plasmid vector and l-2 rg of cotransfected DNA construct per plate. pBR322 containing the 3.4 kb HSV-Barn HI thymidine kinase gene frag ment (pTK) was used for most transfections, but the M13-tk construct was introduced using the pHaprt vector followed by growth in the appropriate selection medium (Lowy et al., 1980). We obtained 100-200 colonies per plate, which were pooled and grown to mass culture. Nuclei were prepared (Sweet et al., 1982) by washing the cells in 1 x SSC and 10 m M Tris (pH 7.4) and resuspending them in RSB (10 m M Tris, pH 7.4; 10 m M NaCI; and 3 m M MgCld containing 1% NP40 and 1 m M PMSF. A nuclear pellet was precipitated by centrifugation (2500 xg for 5 min), washed twice in RSB without detergent and PMSF, and resuspended in the appropriate digestion buffer. DNAase I and micrococcal digestions were carried out in RSB containing 10m4 M CaClz. Restriction endonuclease digestions were carried out in 20 m M Tris (pH 79) 50 m M NaCI, 3 m M MgClz, 0.1 m M EGTA, and 1 m M f3-mercaptoethanol. Digestions were usually performed at a nuclear DNA concentration of 1 .O mg/ml, and the reaction was terminated with an equal volume of buffer containing 1% SDS, 0.8 M NaCI. 20 m M Tris (pH 7.9), 10 m M EDTA, and Proteinase K (400~glml) and was incubated for 2 hr at 3pc. The DNA was finally purified by phenol-chloroformisoamylalcohol25:24:1 extractions and ethanol precipitation. Purified DNA was subjected to restriction digestion, gel electrophoresis on agarose gels, and blot hybridization using nick translated restriction fragment probes at a specific activity of 2-5 x 10s cpmbg (Stein et al., 1982). Supranucleosome particles were prepared (Weintraub, 1984) from nuclei suspended in RSB containing 10m4 M CaCl*. Mild micrococcal nuclease digestion was performed at 22OC for 15 min in a 400 PI reaction volume with a final enzyme concentration of 0.3 rglml. Treated nuclei were precipitated by centrifugation for 1 min at 4OC in a microfuge, and the pellets were washed once in cold RSB and then resuspended in cold TE (10 m M Tris-HCI, pH 7.4; 1 m M EDTA). Before electrophoresis, nuclear debris was removed by low-speed centrifugation for 1 min at 4OC. The samples were run directly on a 1% agarose minigel in 20 m M Tris-HCI (pH 7.4) 50 m M sodium acetate, and 2 m M EDTA. Gels were stained with ethidium bromide prior to blot hybridization analysis (Stein et al., 1982). Acknowledgments We would like to thank R. Adelstein, J. Yisraeli, G. Felsenfeld, D. Yaffe, U. Nude& R. Flavell, and S. McKnight for providing the Ml3 constructs used in this study, and J. Yisraeli, S. Handeli, and A. Razin for critical reading of the manuscript. This research was supported by grant GM20463 from the N.I.H. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “‘sdverrisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Received September

16, 1965; revised December

6, 1985

Burch, J. B. E., and Weintraub, H. (1983). Temporal order of chromatin structural changes associated with activation of the major chicken vitellogenin gene. Cell 33, 65-76. Busslinger, M., Hurst, J., and Flavell, R. A. (1983). DNA methylation and the regulation of globin gene expression. Cell 34, 197-206. Carmon, Y., Czosnek, H., Nudel, U., Shani, M., and Yaffe, D. (1982). DNasel sensitivity of genes expressed during myogenesis. Nut. Acids Res. 10, 3085-3098. Cedar, H. (1984). DNA methylation and gene expression. In DNA Methylation: Biochemistry and Biological Significance, A. Razin, H. Cedar, and A. D. Riggs, eds. (New York: Springer Verlag), pp. 147-164. Davies, R. L., Fuhrer-Krusi, S., and Kucherlapati, R. S. (1982). Modulation of transfected gene expression mediated by changes in chromatin structure. Cell 31, 521-529. Felber, B. K., Gerber-Huber, S., Meier, C., May, F. E. B., Westley, B., Weber, R., and Ryffel, G. U. (1981). Quantitation of DNasel sensitivity in Xenopus chromatin containing active and inactive globin, albumin and vitellogenin genes. Nucl. Acids Res. 9, 2455-2474. Fisher, E. F., and Caruthers, M. H. (1979). Studies on gene control regions. XII. The functional significance of a Lac operator constitutive mutations. Nucl. Acids Res. 7, 401-416. Groudine, M., and Conklin, K. F (1985). Chromatin structure and de novo methylation of sperm DNA: implications for activation of the paternal genome. Science 228, 1061-1068. Gruenbaum, Y., Szyf, M., Cedar, H., and Razin, A. (1983). Methylation of replicating and postreplicated mouse L-cell DNA. Proc. Natl. Acad. Sci. USA 80, 4919-4921. Hofer, E., and Darnell, J. E., Jr. (1981). The primary transcription of the mouse Smajor globin gene. Cell 23, 585693.

unit

Jones, P (1964). Gene activation by 5azacytidine. In DNA Methylation: Biochemistry and Biological Significance, A. Razin, H. Cedar, and A. D. Riggs, eds. (New York: Springer Verlag), pp. 165187. Kappler, J. W. (1970). The kinetics of DNA methylation mouse adrenal cell line. J. Cell Physiol. 75, 21-32.

in cultures of a

Keshet, I., Yisraeli, J., and Cedar, H. (1985). Effect of regional DNA methylation on gene expression. Proc. Natl. Acad. Sci. USA 82, 2560-2564. Kruczek, I., and Doerfler, W. (1983). Expression of the chloramphenicol acetyl transferase gene in mammalian cells under the control of adenovirus type 12 promoters: effect of promoter methylation on gene expression. Proc. Natl. Acad. Sci. USA 80, 7586-7590. Langner, K. D., Vardimon, L., Renz, D., and Doerfler, W. (1984). DNA methylation of three 5’CCGG 3’sites in the promoter and 5’ region inactivate the E2a gene of adenovirus type 2. Proc. Natl. Acad. Sci. USA 81, 2950-2954. Lowy, I., Pellicer, A., Jackson, J. F., Sim, G.-K., Silverstein, S., and Axel, R. (1980). Isolation of transforming DNA: cloning the hamster aprf gene. Cell 22, 617-623. McGhee, J. D., Wood, W. I:, Dolan, M., Engel, J. D., and Felsenfeld, G. (1981). A 200 base pair region at the 5’ end of the chicken adult fl-globin gene is accessible to nuclease digestion. Cell 27, 45-55. McKnight, S. L. (1980). The nucleotide sequence and transcript map of the herpes simplex virus thymidine kinase gene. Nucl. Acids Res. 8, 59494964. Melloul, D., Aloni, B., Calvo, J., Yaffe, D., and Nudel, U. (1964). Developmentally regulated expression of chimeric genes containing muscle actin DNA sequences in transfected myogenic cells. EMBO J. 3, 983-990. Razin, A., and Cedar, H. (1977). Distribution of 5-methylcytosine chromatin. Proc. Natl. Acad. Sci. USA 74, 2725-2728.

in

Saragosti, S., Moine, G., and Yaniv, M. (1980). Absence of nucleosomes in a fraction of SV40 chromatin between the origin of replication and the region coding for the late leader RNA. Cell 20, 85-73. Sheffery, M., Rifkind. R. A., and Marks, P A. (1982). Murine erythroleukemia cell differentiation: DNasel hypersensitivity and DNA methylation near the globin genes. Proc. Natl. Acad. Sci. USA 79, 1180-1184.

DNA Methylation 543

and Active Chromatin

Formation

Solage, A., and Cedar, H. (1978). Organization of 5-methylcytosine chromosomal DNA. Biochemistry 77, 2934-2938.

in

Stein, Ft., Gruenbaum, Y., F’ollack, Y., Razin, A., and Cedar, H. (1982). Clonal inheritance of the pattern of DNA methylation in mouse cells. Proc. Natl. Acad. Sci. USA 79, 81-85. Stein, R., Sciaky-Gallili, N., Razin. A., and Cedar, H. (1983). Pattern of methylation of two genes coding for housekeeping functions. Proc. Natl. Acad. Sci. USA 80, 2422-2426. Sweet, R. W., Chao, M. V., and Axel, R. (1982). The structure of thymidine kinase gene promoter: nuclease hypersensitivity correlates with expression. Cell 31, 347-353. Urieli-Shoval, S., Gruenbaum, Y., Sedat, J., and Razin, A. (1982). The absence of detectable methylated bases in Drosophila melanogaster DNA. FEES Lett. 146, 148-152. Weintraub, H. (1979). Assembly of an active chromatin structure during replication. Nucl. Acids Res. 7, 781-792. Weintraub, H. (1983). A dominant role for DNA secondary structure in forming hypersensitive structures in chromatin. Cell 32, 1191-1203. Weintraub, H. (1984). Histone-Hl-dependent chromatin superstructures and the suppression of gene activity. Cell 38, 17-27. Weisbrod,

S. (1982). Active chromatin.

Nature 297, 289-295.

Wigler, M.. Sweet, R., Sim, G.-K., Wold, B., Pellicer, A., Lacy, E., Maniatis, T., Silverstein, S., and Axel, R. (1979). Transformation of mammalian cells with genes from procaryotes and eucaryotes. Cell 76, 777-785. Worcel, A., Han, S., and Wong, M. L. (1978). Assembly cated chromatin. Cell 15, 989-977.

of newly repli-

Yisraeli, J., and Szyf, M. (1984). Gene methylation patterns and expression. In DNA Methylation: Biochemistry and Biological Significance, A. Razin, H. Cedar, and A. D. Riggs, eds. (New York: Springer Verlag), pp. 353-378.