Proc. Natl. Acad. Sci. USAVol. 82, pp. 2560-2564, May 1985Biochemistry

Effect of regional DNA methylation on gene expression(herpes tk gene/hamster aprt gene/M13)

1. KESHET*, J. YISRAELI*, AND H. CEDAR*tDepartments of tMolecular Biology and *Cellular Biochemistry, The Hebrew University-Hadassah Medical School, Jerusalem, Israel

Communicated by Richard Axel, December 10, 1984

ABSTRACT The effect of DNA methylation on the tran-scriptional activity of the hamster adenine phosphoribosyl-transferase (aprt) and the herpes thymidine kinase (tk) geneshas been investigated. By using M13 constructs containingthese gene sequences, specific segments of each gene weremethylated in vitro by restriction fragment primer-directedsecond-strand synthesis using the substrate 2'-deoxy-5-methyl-cytidine triphosphate (dmCTP). These hybrid-methylatedmolecules were inserted into mouse Ltk- cells by DNA-medi-ated cotransfer. In all cases, the integrated sequences retainedthe in vitro-directed methylation pattern. The aprt gene wasinhibited by CpG methylation in the 5' region but was unaf-fected by methylation at the 3' end or in adjacent M13 se-quences. In contrast to this, DNA methylation in both the 5'promoter region and the 3' structural region of the tk gene hada strong inhibitory effect. This suggests that this modificationmay affect transcription by mechanisms that do not involvethe direct alteration of recognition sequences for RNA poly-merase.

Several lines of evidence suggest that DNA methylation mayplay a role in gene regulation in eukaryotes. Studies of nu-merous tissue-specific genes using methyl-specific restric-tion enzymes have shown a clear correlation between themethylation level of active and inactive genes (1). Thus,most genes are undermethylated in the tissues in which theyare expressed, while they are heavily methylated in nonex-pressing tissues and in the germ line. Furthermore, in vitroDNA methylation of a few specific gene sequences inhibitedthe activity of these genes when inserted into animal cells invivo (2-6). These data suggest that changes in the methyl-ation pattern during differentiation may modulate gene activ-ity, but they do not provide any information on the mecha-nism of action of DNA methylation. Both the hamster ade-nine phosphoribosyltransferase (aprt) gene and the herpesthymidine kinase (tk) gene have been shown to be inhibitedby DNA methylation after their insertion into cells in cultureby DNA-mediated gene transfer (2, 7, 8). In this paper, weprobe the mechanism of this process by asking which regionsof the gene domain are influenced by DNA modification. Tothis end, we have prepared hybrid molecules, methylated ei-ther at the 5' or 3' end of the gene, and have assayed theirtranscriptional activity. The results indicate that methylationof the 5' end of these genes is sufficient to inhibit transcrip-tion. The structural nonregulatory portion of the tk gene isalso strongly influenced by DNA modification.

MATERIALS AND METHODS

M13 Constructs and Synthesis. The M13 clones containingthe 2.0-kilobase (kb) Pvu II herpes tk fragment (9) were ob-tained from S. McKnight and those containing the 3.8-kb

BamHI hamster aprt fragment (10) were made by D. Littmanand kindly provided by him for these experiments. Single-stranded DNA was prepared (11) from these recombinantphages by polyethylene glycol precipitation followed bystandard DNA extraction techniques. The plasmid pTM(12), containing the bacterial agpt gene with the tk promoterregion, was obtained from R. Flavell and was prepared bythe method of Clewell (13).

Second-strand synthesis of single-stranded M13 moleculeswas performed under conditions described (14) for 1 hr at15°C using a DNA concentration of 2-3 ,g per 100 ,ul and apolymerase concentration of 75 units/ml. The 15-mer uni-versal M13 primer was obtained from New England Biolabsand was used at a level of 25 ng per pg of M13 DNA. Underthese conditions, the synthesis reaction went to 90%-95%completion, as judged by gel electrophoresis analysis, but itdid not produce strand displacement. These reactions werecarried out using either dCTP or 2'-deoxy-5-methylcytidinetriphosphate (dmCTP) (P-L Biochemicals) to obtain methyl-ated molecules and their unmethylated controls. Restrictionfragment primers were isolated from low-melting agarosegels and were hybridized to M13 DNA in 0.1 M NaCl at 60°Cafter denaturation for 3 min at 100°C in 1 mM EDTA. Theprimer was used at a 5-fold molar excess to get optimum useof the template. Double-stranded molecules were gel-puri-fied before their introduction into L cells by DNA-mediatedgene transfer. The synthesis reaction was sometimes per-formed using a small amount of radioactive nucleotides totest these molecules for strand displacement by specific re-striction enzyme analysis (Fig. 1A).

Analysis of Transfected Cells. DNA-mediated gene transferinto mouse Ltk- cells was performed as described (15) using100-200 ng of the selective plasmid vector and 1-2 ,ug of co-transfected DNA construct per plate (15-17). We obtained100-200 colonies per plate, and these were pooled and grownto mass culture for DNA and RNA analysis. In some cases,these clones were grown in [3H]thymidine for several hoursand then subjected to autoradiography (18) to determine thetk enzymatic activity of these cells. Positive colonies werethose in which 509-60% of the cells were labeled. Quantita-tive differences in the amount of label per cell were also ob-served.DNA was purified from mouse L cells (14) and subjected

to Southern blot analysis (19) using gel-purified restrictionfragment probes labeled by nick-translation (20) to a specificactivity of 2-5 x 108 cpm/hg. RNA was isolated by guani-dine thiocyanate extraction followed by ultracentrifugationthrough a CsCl cushion (21). RNA dot blots were preparedby formaldehyde denaturation (22) and were hybridized tonick-translated probes. S1 nuclease analysis was done as de-scribed using a single-stranded tk probe prepared by second-strand synthesis of an M13 construct containing a 230-base-pair (bp).region of the 5' sequences (23).

Abbreviations: aprt, adenine phosphoribosyltransferase; tk, thymi-dine kinase; kb, kilobase(s); bp, base pair(s); dmCTP, 2'-deoxy-5-methylcytidine triphosphate.

2560

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Proc. Natl. Acad. Sci. USA 82 (1985) 2561

B

1 2 3 4

11 -C

7 | _

2.8 -

B P Hm .

MA

5 6 7 8 1 2 3 4 5 6a b a b a b

_

a b

p 11

7

2.3

- 0.5

HB3. .

I -2.35

-1.7

-1.2

MB MC MD

m-

0.4 kb

FIG. 1. Methylation pattern of aprt. (A) Single-stranded M13 DNA containing the aprt gene was used as a template for second-strandsynthesis using [a-32P]dCTP. DNA was isolated and subjected to restriction enzyme analysis and gel electrophoresis. Samples 1 and 2 were

synthesized using the 3.8-kb BamHI fragment as primer, and samples 3 and 4 were synthesized with the 15-mer universal primer. These DNAswere electrophoresed either uncut (lanes 1 and 3) or after HincIl digestion (lanes 2 and 4). Note that the large 2.8-kb HincII fragment remainscompletely unlabeled when the BamHI fragment is used as primer. DNA synthesized using the 2.6-kb BamHI/EcoRI 3' fragment as primer isshown uncut (lane 5) and cut with HincII/EcoRI (lane 6), and DNA synthesized with the 1.2-kb BamHI/EcoRI 5' fragment as primer is shownuncut (lane 7) and cut with HincII/EcoRI (lane 8). Note that the HincII/EcoRI 0.5-kb fragment within the 5' region is labeled in lane 6 butremains unlabeled when protected as in lane 8. The 1-kb fragment observed in lane 6 is the result of digestion at the HincII site within the 5'region and the EcoRI site in the M13 polylinker. The small 0.3-kb fragment results from digestion at the 3' HincII site and an HincII site withinthe M13 polylinker. The 2.3-kb fragment, which is completely included within the 2.6-kb 3' EcoRI/BamHI fragment, is labeled only in lane 8.The complete M13 construct migrates at -1 kb and the M13 DNA without the insert is -7 kb. The map shows the following restriction sites: B,BamHI; E, EcoRI; P, Pvu II; and H, HincII. MB, MC, and MD represent individual Msp I sites; MA represents a region containing four MsP Isites. (B) DNA (20 pg) from transfected L cells was restricted with Pvu II (lanes a) or Pvu II/Hpa II (lanes b), subjected to gel electrophoresison 1% agarose gels, blotted, and hybridized to the 1.8-kb Pvu II fragment (see map). Sample 1 was obtained from cells in which the 3' end of thegene is methylated, but the 5' end is unmethylated. Thus, all of the Msp I sites are methylated except for the group of sites near the 5' Pvu II

site. Digestion with Pvu II/Hpa II yields a band '100 nucleotides shorter than the Pvu II fragment. Sample 2 was from cells in which the entireaprt gene was unmethylated and the Pvu II/Hpa II digest reveals the 1.2-kb band bordered by MA and MB. The other bands are too small to bevisualized on this gel. The DNA in sample 3 is from cells in which the aprt gene is totally methylated; sample 4 was taken from cells in whichonly the 5' region is methylated. Digestion with Pvu II/Hpa II yields a 1.3-kb band that is formed by cleavage at MB and the 5' Pvu II site. Twomarkers, the 1.7-kb Pvu II fragment (lane 5) and the 1.2-kb and 2.4-kb EcoRI/BamHI fragments (lane 6), are also included.

RESULTS

To study the effect of DNA methylation on eukaryoticgenes, we previously developed a convenient method formethylating all CpG residues in vitro (8, 14). To this end,gene sequences are cloned into the single-stranded phagevector M13, and the complementary strand is synthesized inthe presence of dmCTP. These molecules are thus hemi-methylated at every cytosine residue. When these moleculesare introduced into animal cells in culture by DNA-mediatedgene transfer, the CpG-specific maintenance methylase per-

petuates all of the CpG methylations, while cytosine modifi-cation at other sites is diluted out by growth over many gen-

erations. This system is also very convenient for preparing

DNA molecules that are methylated in one region and un-

methylated in other regions. To this end, restriction frag-ments are used as primers in the M13 second-strand synthe-sis reaction. After synthesis, the whole molecule will containmethylated cytosine except for the region corresponding tothe primer. Using constructs of this nature, one can test theeffect of DNA methylation on specific regions of a gene.

The 3.8-kb BamHI fragment containing the hamster aprtgene (10) was cloned into M13 mp7, and clones containingboth orientations were isolated. Fully methylated moleculeswere obtained by using the 15-mer M13 primer, which is ad-jacent to the aprt gene insert. Molecules methylated only inthe 5' region were synthesized using the 2.6-kb EcoRI/BamHI fragment as primer. Specific methylation of the 3'end of the gene was accomplished by the use of the 1.2-kbBamHI/EcoRI restriction fragment as primer. Unmethylat-ed controls were synthesized with the same primers, using

dCTP as the substrate in the DNA polymerase reaction.One of the problems associated with this technique is that

the DNA polymerase Klenow fragment may continue syn-thesizing DNA into the primer region of these circular tem-plates either by strand displacement or by residual activity ofthe 5'-3' exonuclease. To prevent these unwanted reactions,polymerization was carried out for 1 hr at 150C using an en-zyme excess. Under these conditions, the polymerase can-not synthesize the entire molecule. Radioactively labeledmolecules made in this way were tested by restriction map-ping to ensure that the primer region was not replaced in thereaction (i.e., remained unlabeled) (Fig. 1A). In the case ofpartially methylated constructs, it was essential to use theaprt gene in the proper orientation to ensure the continuityof the entire gene sequence.These methylated gene constructs were inserted into

mouse L cells by cotransfection using the tk gene as the se-lective marker, and all of these hypoxanthine/aminopter-in/thymidine (HAT)-resistant colonies were grown togetherto mass culture. The method of cotransfection was chosenover that of direct selection of the aprt phenotype to preventartifacts of activity and methylation that might be due to theselection process. Transfected cultures made from variousdifferent constructs all contained about the same amount ofintegrated aprt. Using the restriction enzymes Msp I andHpa II, it could be shown that these integrated copies of theaprt gene retained the original methylation patterns made invitro (Fig. 1B). Thus, assuming that the C-C-G-G sites are

representative of all CpG residues, unmethylated regions re-

mained unmethylated while methylated regions retainedtheir methyl moieties in vivo.

A

a a a I K A aN 9 I I

Biochemistry: Keshet et aL

low,

.K.:'.

2562 Biochemistry: Keshet et al.

To determine the effect of DNA methylation on geneexpression, RNA was isolated from individual cultures andsubjected to dot blot analysis using a purified aprt nick-translated probe (Fig. 2). Methylation of the entire aprt genecaused a 15-fold inhibition of aprt transcription. This sameeffect could be obtained by modification of the 5' region ofthe gene alone. In contrast to this, methylation of the 3' re-gion had very little effect on expression. Constructs pre-pared using the entire BamHI fragment as primer anddmCTP in the synthesis reaction also showed normal geneactivity, indicating that methylation of the M13 sequences

12 3 4 5 6 7 8 9

A

* B

*..* *.* C

.*.*- * * *' D

surrounding the gene have no effect on transcription. Thisconstruct was made in such a way that the M13 sequencesadjacent to the 5' end of the gene were methylated. Thismethylation had no apparent effect, despite the fact that thisregion is only -50 nucleotides from the promoter sequences(I. Lowy, personal communication).The experiments using the aprt gene gave straightforward

results, showing that the 5' region of the gene is influencedby methylation, while the 3' region is not, and these resultsare in agreement with studies on other genes (5, 6). To testwhether this is a general phenomenon, the herpes simplex tkgene was subjected to the same analysis. Using M13 mp2clones containing the 2.0-kb Pvu II tk fragment in both orien-tations (9), hybrid rnethylated molecules were synthesized.The 5' end of the gene was represented by the 250-bp PvuII/Bgl II fragment, while the 3' end was defined by the 1.8-kb BgI II/Pvu II fragment, which includes almost the entirecoding sequence of the gene. These constructs were insertedinto L cells by cotransfection, using as the selective markerthe bacterial gene agpt, which provides resistance to the

1 2 3 4 5 630pg lOpg

CpG mapM13* MJ13

- 3'

4-

Activity %c 5mc100 91

100 12

100 83

100 7

FIG. 2. Transcription of methylated aprt genes in L cells. RNAwas isolated from mass cultures obtained from transfection with var-ious hybrid methylated molecules and subjected to dot blot analysisusing 1.5 ,g (A), 3 Iug (B), 6 ,Ig (C), or 15 ,ug (D) of RNA. The RNAfrom lane 1 was obtained from cells containing no aprt. All con-structs were made using a specified primer and synthesis with eitherdCTP (even numbered lanes) or dmCTP (odd numbered lanes).RNA in lanes 2 and 3 were from cells containing the construct madewith the 3.8-kb BamHI fragment (i.e., the entire gene is unmethylat-ed). The 2.6-kb 3' EcoRI/BamHI primer construct (i.e., 5' regionmethylated) RNA is shown in lanes 4 and 5, and the 1.2-kb 5'EcoRI/BamHI primer construct (i.e., 3' region methylated) RNA isshown in lanes 6 and 7. The RNA in lanes 8 and 9 was made fromcells containing the M13 aprt construct synthesized using the 15-meruniversal primer. These cells appear to have a greater overall levelof aprt RNA, because they contain 2-3 times as many copies of theDNA sequences. Although equal amounts of RNA were used foreach cell type, each RNA sample was also counterhybridized with aprobe for dihydrofolate reductase (DHFR) and this experimentshowed that all samples contained about the same amount ofDHFR-specific sequences (±30%). The chart shows the results of this ex-periment after scanning and quantitation of the dot blot. The map inthis diagram shows the BamHI (0) sites, the location of integrationof the insert in M13, and the EcoRI site (A), which was used to makethe primer restriction fragments. The line immediately below the re-striction map indicates the RNA transcript including the exons(heavy blocks). Each hybrid-synthesized molecule is shown sche-matically below. Heavy line indicates the restriction fragment prim-er used for synthesis. Arrow indicates direction of this synthesis.Small primer in the bottom line is the M13 15-mer universal primer.Each synthesis was carried out using either dCTP or dmCTP. Thetranscription activity of each construct was measured by dot blotanalysis, and results are expressed as percentage activity of themethylated molecules as compared to the same molecules synthe-sized with dCTP. The line above the aprt restriction map shows thelocations of all of the CpG residues (sequence obtained from unpub-lished data of I. Lowy and R. Axel). Regions below the dotted lineshave not been sequenced.

280- -t

122'- O

50- 1

hi1 ii1M13-109

M130 50

122 nt50 nt

FIG. 3. S1 nuclease analysis of tk RNA. Total RNA was isolatedfrom mass cultures of pooled L-cell clones containing different hy-brid-methylated M13 tk molecules. Analysis was performed by hy-bridization of this RNA (20 pAg) to a single-stranded probe synthe-sized from an M13 clone (mP8TP) containing a portion of the 5' re-gion stretching from -109 to +122 (see map). After S1 nucleaseanalysis and gel electrophoresis, the normal tk transcript protects alabeled fragment of 122 nucleotides (nt) (this is not a unique bandbecause the tk gene contains several transcription start sites in thesame region). The RNA made from unmethylated tk is shown in lane4, while that made from totally methylated tk (using the 15-nier prim-er and dmCTP) is shown in lane 3. RNA from the tk gene methylatedonly in the 5' region (using the. 1.8-kb Pvu II/Bgl II primer anddmCTP) is shown in lane 5, and that made from the gene methylatedonly in the 3' region (using the 0.25-kb Pvu II/Bgl II primer anddmCTP) is shown in lane 6. The protected fragments in the region of50 nucleotides is from the agpt RNA, which contains the tk promoterregion up to the Bgl II site (+50). This gene was used as the selectionvector for introducing the tk constructs into the cells by DNA-medi-ated gene transfer. All cell lines make about equal quantities of thisRNA. Lane 1 contains RNA from untransfected mouse L cells andlane 2 contains RNA from L cells transfected with the normal tkgene and a shorter pseudogene and induced by herpesvirus infection(23).

Proc. Natl. Acad Sci. USA 82 (1985)

4

Proc. NatL. Acad. Sci. USA 82 (1985) 2563

drug G-418. Mass cultures were grown from 100-200 trans-formed clones and were then tested for the amount andmethylation status of the methylated molecules. All culturescontained approximately equal amounts of the tk gene se-quence, and the methylation patterns of the integrated geneswere tested with the methyl-sensitive restriction enzymeHpa II. By this criterion, all of the C-C-G-G sites within the2.0-kb region maintained the same methylation status as theinput constructs (data not shown).

Transcription of the tk gene was studied by an S1 nucleaseassay using a single-stranded highly labeled probe specificfor the 5' end of the tk transcript (23). This assay was essen-tial because other assays were found to be less sensitive andbecause the assay must be specific enough to distinguish be-tween real tk RNA and the tk sequences present at the 5' endof agpt transcripts (see legend to Fig. 3). In this assay, the tkRNA reveals a protected fragment of 122 bp, while the agpttranscript yields a series of bands at -50 bp (Fig. 3).While the completely unmethylated constructs show a

fairly high level of tk RNA synthesis, the totally methylatedDNA showed no detectable RNA product (even when 75 jugof total RNA was used in the assay). This large differencewas confirmed by an autoradiographic assay of tk activity inthese cells. While all cells containing the unmethylated tkgene incorporated thymidine, those cells containing themethylated construct showed no measurable incorporationof thymidine as judged by autoradiography of fixed cells.When hybrid methylated tk genes were tested for their abili-ty to produce message, it was found that methylation at ei-ther the 5' or 3' end is sufficient to cause inhibition of tktranscription. For both constructs, the transcription of the tkgene was at least 1/20th to 1/50th that of control levels (Fig.4). Complete methylation of the adjacent M13 sequences,however, had no effect on the tk transcription activity (datanot shown). It should be noted that the level of transcriptionof the agpt gene is equal in all samples, serving as a controlfor the quality and uniformity of the RNA. Cells containingthese hybrid genes were capable of incorporating [3H]thymi-dine but at 1/3rd to 1/5th the level of that in cells harboringunmethylated genes.

DISCUSSIONThe use of M13 containing various gene sequence insertsprovides the technology for region-directed DNA modifica-tion. We have assayed the effect of such hybrid DNA meth-ylation on both the tk and aprt genes in vivo by using DNA-mediated gene transfer. All of the studies presented in thispaper were performed on pools of transfected clones andrepresent the average effect of DNA methylation on many

M13 Pvu II Bgl II=,% 5S"~~'

0.2 kb

individual gene copies.Our results show that the aprt gene is influenced by meth-

ylation at the 5', but not at the 3', end. This observation isconsistent with results from other laboratories, which haveshown, using transfection assays, that the human -y-globingene (5) and the adenovirus Ela gene (6) are affected byDNA methylation in the same manner. This result is also inkeeping with the in vivo status of the aprt gene in hamstercells, because in all tissues, including the germ line, this geneis highly methylated in the entire region 3' to the EcoRI site,while all of the sites 5' to EcoRI are completely unmethylat-ed (24). Although this methylation mapping was limited to afew representative restriction enzyme sites, it is reasonableto assume that other CpG residues in the 3' region are alsohighly methylated. Our results show that this gene can betranscribed at its full level even when every CpG 3' to theEcoRI site is methylated. This pattern of 5' undermethyla-tion and 3' methylation is typical for all housekeeping genesso far analyzed and may be related to the fact that thesegenes are expressed constitutively in all tissues and do notundergo demethylation during development (1).

Tissue-specific genes have been found to be highly meth-ylated in germ line DNA and undermethylated in the tissuein which the gene is expressed, suggesting that the differenti-ation of the tissue is accompanied by demethylation of thegene. In some cases, this undermethylation probably occursafter the actual activation of the gene (25, 26), and it has beensuggested that transcription itself may be the factor causingdemethylation (27, 28). The results obtained on the methylat-ed aprt gene argue against this possibility, because contin-ued transcription of the gene was not accompanied by de-methylation. In a more direct experiment, two constructs ofthe agpt gene containing or lacking the tk promoter weremethylated in vitro with the Hpa II methylase and were in-serted into L cells by DNA-mediated gene transfer. Al-though only the promoter-containing gene was transcribedand biologically active, both genes remained fully methylat-ed after >150 generations in culture (unpublished observa-tions). Thus, available data suggest that in these systemstranscription per se is not a factor in causing demethylation.

Results obtained by methylating various regions of the tkgene clearly indicate that its activity can be affected by DNAmodification of the structural part of the gene alone. Itshould be noted that herpes tk has a very high concentrationof CpG residues over the entire gene, both the promoter andthe structural regions (Fig. 4). On the other hand, both thehamster aprt and the human -globin gene have a relativelyhigh density of CpG moieties at their 5' end, but very few atthe 3' end of the molecule. If the degree of inhibition of tran-scription is dependent on the amount of methylation, it is not

CpG mapPvu II M13

3'

Activity

4-

4-- P

FIG. 4. Effect of DNA methylation on tk transcription. Map shows restriction sites used to make the primer fragments for M13 synthesisreactions. The tk RNA transcript is shown below the map. Mode of synthesis of the hybrid-methylated molecules is shown schematically.Blocked regions represent the different primers used, the small one being the M13 15-mer universal primer. The first construct was made usingdCTP (solid line), and the others were synthesized with dmCTP (broken line). Fully methylated molecules showed no transcription activity, thepartially methylated molecules had -2% (*) the activity of the control. Top line shows locations of all CpG moieties in the tk gene region.

Biochemistry: Keshet et aL

*

_ *

2564 Biochemistry: Keshet etaLP

surprising that these other genes are more sensitive to theeffect of modification at their 5' end. Another factor thatshould be taken into consideration in comparing the effect ofmethylation of tk to other genes is that, unlike other se-quences, this viral gene is always found in the unmethylatedstate in the virus (29) and probably in infected cells.

It should be noted that the region of influence of methyl-ation on the aprt gene is hard to define. The fragment (1.2-kbBamHI/EcoRI) that was used to represent the 5' region ofthe gene includes the promoter region as well as a large partof the transcribed sequences. It is therefore impossible fromthis experiment to resolve the exact boundaries of methyl-ation inhibition. It is certainly possible that DNA modifica-tion of sequences within the upstream exons and introns, aswell as those in the promoter region, affects expression. Inretrospect, it might have been more instructive to use the 5'Pvu II site as the boundary between the 5' and 3' regions,because this site is close to the start of transcription. Thiswas not done because at the time this experiment was initiat-ed it was thought that this site was within the promoter se-quence. By using plasmid constructs containing adenovirusgene promoters linked to the CAT gene, it was shown thatmethylation of these upstream regions is sufficient to inhibitgene activity. Because of the nature of this experiment, theeffect of modification on the 3' downstream sequences wasnot assayed, and a role for DNA methylation in this regioncannot be ruled out.The fact that methylation of the structural portion of the tk

gene has a strong inhibitory effect on its transcription shouldshed light on the mechanism of action of this modification.The exact regions that are necessary for accurate transcrip-tion initiation of the tk gene have been thoroughly character-ized and are concentrated in two hexanucleotide sequenceslocated at sites 5' to the start of transcription (30). Theseelements are located 100-150 bp 5' to the Bgl II site. Thus, inhybrid methylated molecules in which the entire structuralgene region, beginning with the Bgl II site, is methylated, theidentified regulatory region for tk is completely unmodified.A similar conclusion can be drawn from previous experi-ments showing that in vitro Hpa II methylation inhibits boththe activity and SV40 late transcription since there are no C-C-G-G sites located in the promoters of these genes (3, 7).One model to explain the inhibition of transcription causedby methylation is that this modification alters the recognitionsignals for RNA polymerase on the gene, thus acting as apotentially reversible mutation. The fact that methylation in-hibits RNA transcription of the tk gene without altering theserecognition sites argues that this modification may act byalternative mechanisms. One obvious suggestion is thatDNA methylation in the structural portion of the gene actsby inhibiting the elongation step of the RNA polymerase re-action. If shortened RNA transcripts had been produced inthese cells, they would probably have been degraded andwould therefore go undetected in the S1 nuclease assay. An-other possibility is that DNA modification affects local chro-matin structure and may drive modified DNA sequences intoan inactive conformation. Recent experiments in our labora-tory suggest that methylated DNA inserted into animal cellsin culture by DNA-mediated gene transfer acquires a DNaseI resistant conformation despite the fact that surrounding un-methylated sequences are DNase I sensitive.

Methylation restriction mapping of many tissue-specificgenes reveals that the undermethylation observed in activegenes almost always involves all parts of the gene, includingboth the promoter and structural gene regions, and evensome sites downstream from the gene itself (1). This also

suggests that methylation may play a regulatory role over theentire gene domain.

We would like to thank S. McKnight and D. Littman for providingM13 clones containing the Herpes tk and hamster aprt gene se-quences, and I. Lowy and R. Axel for providing the aprt sequenceprior to publication. We are indebted to S. Silverstein, who kindlyprovided the M13 construct and the expertise for the tk S1 nucleaseassay and who personally helped carry out the analysis. This projectwas supported by a grant from the National Institutes of Health (GM20483).

1. Yisraeli, J. & Szyf, M. (1984) in DNA Methylation: Biochemis-try and Biological Significance, eds. Razin, A., Cedar, H. &Riggs, A. D. (Springer, New York), pp. 353-376.

2. Stein, R., Razin, A. & Cedar, H. (1982) Proc. Nail. Acad. Sci.USA 79, 3418-3422.

3. Fradin, A., Manley, J. L. & Prives, C. L. (1982) Proc. Nail.Acad. Sci. USA 79, 5142-5146.

4. Vardimon, L., Kressmann, A., Cedar, H., Maechler, M. &Doerfler, W. (1982) Proc. Natl. Acad. Sci. USA 79, 1073-1077.

5. Busslinger, M., Hurst, J. & Flavell, R. (1983) Cell 34, 197-206.6. Kruczek, I. & Doerfler, W. (1983) Proc. Nail. Acad. Sci. USA

80, 7586-7590.7. Wigler, M., Levy, D. & Perucho, M. (1981) Cell 24, 33-40.8. Cedar, H. Stein, R., Gruenbaum, Y., Naveh-Many, T.,

Sciaky-Gallili, N. & Razin, A. (1983) Cold Spring HarborSymp. Quant. Biol. 47, 605-609.

9. McKnight, S. L. (1980) Nucleic Acids Res. 8, 5949-5%3.10. Lowy, I., Pellicer, A., Jackson, J. F., Sim, G.-K., Silverstein,

S. & Axel, R. (1980) Cell 22, 817-823.11. Messing, J. (1979) in Recombinant DNA Technical Bulletin,

NIH Publication No. 79-99 (National Institutes of Health, Be-thesda, MD), Vol. 2, pp. 43-48.

12. Grosveld, F. G., Lund, T., Murray, E. J., Mellor, A. L.,Dahl, H. H. M. & Flavell, R. A. (1982) Nucleic Acids Res. 10,6715-6732.

13. Clewell, D. B. (1972) J. Bacteriol. 110, 667-676.14. Stein, R., Gruenbaum, Y., Pollack, Y., Razin, A. & Cedar, H.

(1982) Proc. Natl. Acad. Sci. USA 79, 61-65.15. Wigler, M., Sweet, R., Sim, G.-K., Wold, B., Pellicer, A.,

Lacy, E., Maniatis, T., Silverstein, S. & Axel, R. (1979) Cell16, 777-785.

16. Wigler, M., Pellicer, A., Silverstein, S., Axel, R., Urlaub, G.& Chasin, L. (1979) Proc. Natl. Acad. Sci. USA 76, 1373-1376.

17. Jimenez, A. & Davies, J. (1980) Nature (London) 287, 869-871.

18. Capecchi, M. R. (1980) Cell 20, 479-488.19. Southern, E. M. (1975) J. Mol. Biol. 98, 503-517.20. Rigby, P. J. W., Dieckmann, M., Rhodes, C. & Berg, P. (1977)

J. Mol. Biol. 113, 237-251.21. Chirgwin, J. M., Przybyla, A. E., MacDonald, R. J. & Rutter,

W. J. (1979) Biochemistry 18, 5294-5299.22. Mencher, D., Cohen, H., Benvenisty, N., Meyuhas, 0. & Re-

shef, L. (1984) Eur. J. Biochem. 141, 199-203.23. El Kareh, A., Murphy, A. J. M., Fichter, T., Efstratiadis, A.

& Silverstein, S. (1985) Proc. Natl. Acad. Sci. USA 82, 1002-1006.

24. Stein, R., Sciaky-Gallili, N., Razin, A. & Cedar, H. (1983)Proc. Natl. Acad. Sci. USA 80, 2422-2426.

25. Burch, J. B. E. & Weintraub, H. (1983) Cell 33, 65-76.26. Wilks, A. F., Cozens, P. J., Mattaj, I. W. & Jost, J. P. (1982)

Proc. Natl. Acad. Sci. USA 79, 4252-4255.27. Ott, M. O., Sperling, L., Cassio, D., Levilliers, T., Sala-Tre-

pat, J. & Weiss, M. C. (1982) Cell 30, 825-833.28. Benvenisty, N., Mencher, D., Meyuhas, O., Razin, A. & Re-

shef, L. (1984) Proc. Natl. Acad. Sci. USA 82, 267-271.29. Desrosiers, R. C., Mulder, C. & Fleckenstein, B. (1979) Proc.

Natl. Acad. Sci. USA 76, 3839-3843.30. McKnight, S. L., Kingsbury, R. C., Spencer, A. & Smith, M.

(1984) Cell 37, 253-262.

Proc. NatL Acad Sci. USA 82 (1985)