A Cis-acting Element Accounts for a Conserved Methylation Pattern ...

THEJOURNAL OF BIOLOGICAL CHEMIsTRY 0 1993 by The American Society for Biochemistry and Molecular Biology,Inc.

Vol. 268, No. 1, Issue of January 5, ~ p 552-558 : 1993 nnted an &LA.

A Cis-acting Element Accounts for a Conserved Methylation Pattern Upstream ofthe Mouse Adenine Phosphoribosyltransferase Gene* (Received for publication, June 29, 1992)

Padmaja MummaneniSO, Peggy L. Bishop$, and Mitchell S . TurkerSllIl** From the $Department of Pathology and the qDepartment of Microbiology and Immunology, University of Kentucky College of Medicine and the 11Markey Cancer Center, Lexington, Kentucky 40536

A 2.1-kilobase pair region located just upstream of the mouse aprt (adenine phosphoribosyltransferase) gene has a methylation pattern that is conserved in mouse tissues and culture cell lines. This upstream region includes four HpaIIIMspI sites. Two of these sites are fully methylated, one is partially methylated, and one is unmethylated. Transfection experiments have demonstratedthat the conserved methylation pattern can be reproduced in a mouse embryonal carcinoma stem cell line via de novo methylation (Turker, M. S., Mummaneni, P., and Bishop, P. L. (1991)Somat. Cell Mol. Genet. 17, 151-157). To examine the molecular basis of the conserved methylation pattern, a plasmid-based deletion analysis was conducted by removing and rearranging specific portions of the upstream region. Unmethylated versions of these plasmid constructs werethen transfected into the mouse stem cell line and the methylation status of the remaining HpaIIl MepI sites determined with a Southern blot analysis. By using this approach, a cis-acting sequence within the upstream region of approximately 0.8 kilobase pairs wasidentified which appears responsible for the conserved methylation pattern. We use the term “de novo methylation center” to denote this sequence. Based on the results obtained, a model is offered to explain the formation of the conserved methylation pattern in the upstream region.

A role for DNA methylation in the control of gene expression has been proposed for mammalian cells. A correlation between increased levels of cytosine methylation, particularly in the dinucleotide pair CpG, and decreased levels of transcription has been well documented (Cedar, 1988; Razin and Cedar, 1990). CpG dinucleotide pairs are often clustered at the 5’ ends of housekeeping genes (Bird, 1986; GardinerGarden, 1987). De m u 0 methylation of such clusters, termed CpG islands, is one of the hallmarks of X-chromosome inactivation (Monk, 1990a). Moreover, de novo methylation associated with expansion of a repeat sequence in a recently isolated gene, termed fmr-1, is believed to result in mental retardation that is associated with the fragile X syndrome

* This work wassupported in part by National Institutes of Health Grant AG08199. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertkement” in accordance with 18 U.S.C. Section 1734 solelyto indicate this fact. 5 Supported in part by a PSP grant from the University of Kentucky. ** To whom all correspondence should be addressed Dept. of Pathology, Markey Cancer Center, University of Kentucky College of Medicine, 800 Rose St., Lexington, KY 40536. Tel.: 606-257-4951; Fax: 606-258-2094.

(Oberle et al., 1991). Recent work has indicated that de m u 0 methylation of autosomal loci occurs early in development therebyestablishing the embryonic methylation pattern (Monk et al., 1987; Shemer et al., 1991a, 1991b; Kafri et al., 1992). De mu0 DNA methylation of autosomal loci is also associated with genomic imprinting, a phenomenon in which allele-specific expression is a function of the gender of the contributingparent (Monk, 1990b; Solter, 1988). Inappropriate imprinting of tumor suppressor genes has been implicated in anumber of childhood cancers including retinoblastoma, Wilm’s tumor,and rhabdomyosarcoma (Sapienza, 1991). De novo methylation of regions on human chromosomes 3, 11, and 17 has also been correlated with malignant tumors (deBustros et al., 1988; Makos et al., 1992). Despite a potential role for de m u 0 DNA methylation in human disease, the signals that specify this heritable epigenetic modification in mammalian cells are essentially unknown (Doerfler et al., 1990; Selker, 1990). In vitro experiments with purified methyltransferase and synthesized or natural templates have suggested that the spacing between CpG clusters may provide a signal for de novo methylation (Bolden et al., 1986; Ward et al., 1987) and that themammalian methyltransferase can recognize unusual DNA structures (Smith et al., 1991). However, these possibilities have not been tested with cellular experiments. Unfortunately, a general rule for such experiments is that de novo methylation of unmethylated transfected DNA does not occur in cultured mammalian cells (Selker, 1990; Razin and Cedar, 1991). A well studied exception is the spreading of methylation from in vitro methylated sites in the adenovirus E2A promoter to distal sites in the viral genome subsequent to transfection into hamster cells (Toth et al., 1989, 1991). Additional exceptions have been reported when embryonal stem cells are used as recipient cells (Stewart et ul., 1982; Szyf et al., 1990) and when an adrenal expressed genewas transfectedinto an adrenocortical cell line (Szyf et al., 1989). However, in none of these cases was it possible to reproduce an endogenous methylation pattern via de novo methylation. We have recently described the development of a model cellular system allowing a molecular analysis of de mu0DNA methylation in mammalian embryonal carcinoma stem cells (Turker et al., 1991). This system examines de mu0 methylation of a 2.1-kbp’ region, which begins approximately 0.4 kbp upstream of the mouse aprt (adenine phosphoribosyltransferase) gene. Unique components of this system are that the upstream region has a conserved methylation pattern in mouse tissues and cultured cells (Turker et al., 1989a) and that both aprt alleles and their respective upstream regions have been deleted from the recipient stem cells (Turker et al., 1989b). Using the aprt gene as the selectable marker, we

552

The abbreviations used are: kbp, kilobase pair(s); bp, base paids).

De Novo CenterMethylation demonstrated that the conserved methylation pattern in the upstream region could be accurately reproduced via de novo methylation. The capacity for de novo methylation was lost upon differentiation of the stem cells (Turker et al., 1991). We report here the results of a plasmid-based deletion analysis in which portions of the upstream region weredeleted and/or rearranged and the resultant constructs assayed for de nouo methylation upon transfection. Significantly, a cis acting sequence of approximately 0.8 kbp was identified that provides the signals for de nouo methylation. We use the term “de novo methylation center” to denote this sequence. Based on our results, we present a model to explain how the de novo methylation center is responsible for the conserved methylation pattern in the upstream region. EXPERIMENTAL PROCEDURES

Cell Culture and Transfection-The cell line used for the transfection experiments in thisstudy is termed DELTG3. It is a thioguanineresistant subclone of the H4D2 stem cell line which lacks both aprt alleles and their respective upstream regions (Turker et al., 1989b). The methods used to transfect the DELTG3 cells via electroporation and to select independent transfectants have been described previously (Turker et al., 1991). Two modifications were made. The first was to use 60 /IM adenine and 50 /IMazaserine for selection of APRTpositive transfectants. The second modification was to linearize plasmid constructs with ScaI instead of EcoRI prior to electroporation. A single ScaI site is present in the vector sequence. Analysis of Methylation in Transfectants-All plasmids used were unmethylated prior to electroporation via growth in Escherichia coli strain JM109. Each independent transfectant clone was expanded until approximately 5-10 X lo6 cells were obtained. Genomic DNA was then isolated from these cells (Miller et al., 1988) and analyzed with a Southern blot analysis to determine the number of integrated plasmids. No more than four integrated copies were noted per transfectant. To examine the methylation profile in the integrated plasmids, the genomic DNA preparations were digested separately with the restriction enzymes MspI and HpaII and thenused for a Southern blot analysis. Both enzymes recognize the sequence CCGG, but HpaII function is inhibited when the internalC is methylated. Other restriction enzymes, used to fix one or both end pointsof the digested DNA fragments prior to hybridization, are noted under “Results.” The conditions for enzyme digestion have been reported elsewhere (Turker et al., 1989a). After Southern blotting onto nylon membranes, the digested DNA preparations were hybridized with one of the three probes shown in Fig. 1. The probes were chosen near the 3’ end of the upstream region as the presence of repetitive DNA elements at the 5’ end made the use of probes from this region impractical. The methylation patterns of the HpaII/MspI sites studied were deduced by considering the molecular sizes of the hybridization bands and the probe or probes that they hybridized with. Construct Preparation-The starting plasmid construct used for this study is termed pSam 6.3 (Fig. 1).This plasmid contains the 2.3kbp mouse aprt gene and the HO through H2 HpaII/MspI sites that comprise the upstream region. The specific schemes used to create each construct inthis study are presented inthe appropriate subsection under “Results.” In each case, DNA sequencing was used to confirm that the appropriate insert was present and that the insert was in the 5’ to 3’ orientation relative to its usual position in the upstream region. When DNA fragments containing different restriction enzyme site ends were to be ligated, both ends were filled in to create bluntends and the DNA fragments were then ligated as described by Perbal (1988). To create deletions in the HS fragment (see Fig. 5), the Erase-a-base kit from Promega (Madison, WI) was used according to the manufacturer’s instructions. Densitometry-Scanning densitometry for quantitation of methylation profiles was performed using a Quik Scan densitometer (Helena, Beaumont, TX). RESULTS

The Conserved Methylation Pattern in the Region Upstream under study is included within the pSam 6.3 plasmid construct (Fig. 1).This region contains four HpaIIIMspI sites termed HO, Hla, H1,

of the Mouse aprt Gene-The 2.1-kbp region

553

and H2. The HO and H2 sites represent the 5‘ and 3’ ends, respectively, of the sequence referred to as the “upstream region.” For thepurpose of this paper we will refer to theHO through H2 HpaII/MspI sites asmethylation sites. The location of these methylation sites relative to the aprt gene and their usual methylation pattern in both the endogenous genomic location (Turker etal., 1989a) and the transfected pSam 6.3 construct in cultured mouse stem cells (Turker et al., 1991) is shown in Fig. 1 and summarized in Table I . The HO site is normally unmethylated, the Hla and H1 sites are fully methylated, and the H2 site is partially methylated. An additional HpaII/MspI site located approximately 0.25-kbp upstream of the HO site was recently identified with DNA sequencing and is termed HOa (Fig. 1).Its methylation status has not been determined. Position Effect for de Novo Methylation of the HO Site and the aprt Gene CpG Island-To determine if the normally unmethylated HO site could be methylated when relocated, it was moved to the position normally occupied by the H1site. This was accomplished by using a construct, termed pBL, from which a 282-bp BglII fragment containing the H1 site (Fig. 1) had been deleted (Turker et al., 1991) and ligating into the remaining BglII site of pBL a 163-bp PstI fragment containing the HO site. The PstI fragment was termed PS, and therefore the resultantconstruct was termed pBLPS(Fig. 1).The pBLPS construct contains two HO sites, one at its normal position at the5‘ end of the upstream region and the second at the position normally occupied by the fully methylated H1 site. Thislatter HO site is referred toasthe relocated HO site. Similarly, to determine if the normally unmethylated H3-H8 sites, which comprise the aprt gene CpG island, could be methylated when relocated to the upstream region, a 384-bp TaqI fragment containing these sites was ligated into theBglII site of pBL. The resultant construct was termed pBLTS (Fig. 1). A Southern blot analysis with high molecular weight DNA preparations from DELTG3 cells stably transfected with the pBLPS and pBLTS constructs demonstrated that the relocated HO and H3-H8 sites were completely methylated, with a single exception (Fig. 2). To specifically examine de novo methylation of the relocated HO and H3-H8 sites, genomic DNA preparations from pBLPS and pBLTS transfectants were digested separately with XmnI, XmnI and MspI or with XmnI and HpaII and hybridized with the pXSprobe (see Fig. 1for location of pXS probe). XmnI sites bracket the relocated HO site in pBLPS and therelocated H3-H8 site in pBLTS as well as the H l a site present in both constructs (Fig. 1).The H l a site maintains it status of full de novo methylation in pBL-derived constructs (Turker et al.,1991). Complete methylation of the relocated HO site in pBLPS and the relocated H3-H8 sites in pBLTS was demonstrated by hybridization bands inthe respective XmnI-HpaII digestion lanes that were identical in size to the hybridization bands observed in the XmnI digestion lane. Using HpaII and MspI digests and the pN1probe, we also determined that the HO site in its normal location in the pBLPS construct andthe H3-H8 sites within the CpG island in the pBLTS construct were unmethylated. Using the p400 probe with Southern blotted DNA digested with EcoRV and TaqI (see Fig. 1for location of restriction sites)we found that the H2 site was partially methylated in both constructs (data not shown) as previously shown for the pSam6.3 and pBL constructs (see Table I ) . Deletion of a 1.5-kbp HpaI Fragment Results in Loss of the de Novo Methylation Signal-The above results demonstrated that the HO siteand the H3-H8 sites couldbe de novo

De Novo CenterMethylation

554 Hla

HOa HO

pSam 6.3

p

+I?,

0.5

?

H9 0.8

HI 0.8

?

H2 H8

7

H3 .

E

/

I

,

T

,,

I

H I O A l HI2 H13H14

7

0.4

&+

',T

0.5Kbp

pNI

S

lpxsl

I

J I I

H la

HO I

P

HI

0.83

II I HPX

I 8

H2 I 8

0.79

I H

E

I X

I

V

\'\H3 I . T

I .76

' HO'

H lo

HO

p8LPS

Hli"

H3

... X

H la

HO

p8LTS

X

1.66Kbp 0.67Kbp

H3 H4567 H8

H2

H3

...

I

X

X

I.a8

IPXSI

0.70

FIG. 1. The plasmid constructs pSam 6.3, pBLPS, and pBLTS. The pSam 6.3 plasmid contains the 2.3-kbp mouse aprt gene and an additional 4.0 kbp of upstream sequence. The five exons of the aprt gene are represented by closed boxes above the line for pSam6.3. The closed boxes below the lines for the pBLPS and pBLTSconstructs represent the PstI fragment containing the HO site and theTag1 fragment containing the H3-H8 sites, respectively. Each bubble represents a HpaIIIMspI methylation site numbered sequentially from the 5' end of pSam 6.3 to the 3' end. Closed bubbles indicate the site is completely methylated in transfectants, partially closed bubbles indicate the site is partially methylated, and open bubbles indicate the site is unmethylated. The relative locations of the three probes used in this study, pN1, pXS, and p400, are also shown. The lines beneath each plasmid map refer to specific hybridization bands observed in the Southern blot analysis shown in Fig. 2. The restriction site symbols follow: B, BglII; E , EcoRI; EV, EcoRV; H , HpaI; P, PstI; S , SphI; T,T q I ; XXmnI. ,

TABLE I Summary of de novo methylation profiles incultured stem cells for HpaII sites located within the region upstream of the mouse aprt gene No' analyzed

HO

Hla

%

%

%

H1

H2

%

%

Endogenous region" 100 NPb 100 19-27 5 0 0 100 NP 100 25-61 6 pSam6.3' 1 0 100NPNP 46 pBL' 0 100 100 NP 34-49 5 pBLPS 0 100 72-100d NP NQ' 5 pBLTS NP0 NP NP 0 3 PHL 5 0 NP 0 NP 0 pHLPS 4 0 NP NP 80-100 16-45 pHLBS 0 NP 0 NP 0-20 7 pHLHS1217PS 3 83-100 100 100 NP 46-72 pHLHS154,357PS 3 42-50 NP 100 NP pHLHS511,357PS NQ Turker et al. (1989a). * NP, site is not present in construct. Values calculated from data presented by Turker et al. (1991). this construct the value of 100% methylation refers to the relocated H3-H8 methylation sites of the CpG island (see Fig. 1). e NQ, data not quantitated.

1.7.

d 0.7n n m r r m . .

..

PXS

FIG.2. Southern blot analysis of cells transfected with the pBLPS and pBLTS constructs. Genomic DNA preparations from pBLPS and pBLTS transfectants and from the wild type P19 cell line were digested with XmnI and then redigested with either MspI (M) or HpaII ( H ) . Digestion lanes that arenot indicated with either an M or H were treated with XmnI only. The digested DNA preparations were then separated by agarose gel electrophoresis, Southernblotted onto nylon membranes, and hybridized with 32P-labeledpN1 probe. Approximate molecular sizes of the hybridization bands are indicated.

kbp hybridization bands inboth the HpaII and MspI digestion lanes after hybridization of the integrated pHL DNA with the pN1 probe (Fig. 4). The 0.39-kbp hybridization band represents the fragment between the H2 and H3 sites. The 0.62methylated if relocated to the position normally occupied by kbp hybridization band represents the fragment between the the H1 site. Therefore, we assumed that the signals for de H2 and HO methylation sites, indicatingthat theHO site was nouo methylation were present within the upstream region unmethylated as well (see Fig. 3). When the nylon membrane and hypothesized that these signalsmight also accountfor de used in this experiment was rehybridized with the p400 probe, nouo methylation of the H2 site. To testthese possibilities, a only the 0.39-kbp hybridization band was observed, confirm1.5-kbp HpaI fragment (Fig. l),termed HS, was deleted from ing that theH2 site was unmethylated. To further demonstrate that the pHL construct lacked de the upstream region. The resultantplasmid construct, termed pHL (Fig. 3), lacked both the Hla and H1 methylation sites, m u 0 methylation signals, the 163-bp PstI (PS) fragment but retained the HO and H2sites. The pHL constructallowed containing the HO site was ligated into the remaining HpaI us to determine that the deleted 1.5-kbp HS fragment was site of pHL to createa construct termed pHLPS (Fig. 3). As responsible for de m u 0 methylation of the H2 site. In trans- with the pBLPS construct (Fig. l), the pHLPS construct fectants containing this construct, the H2 site was unmeth- contains two HO sites, one being relocated to a second locaylated as shown by the presence of identical 0.62- and 0.39- tion. A Southern blot analysis demonstrated that the relo-

D e Novo Methylation Center H4

HO H 3 0.39 0.62

HZ

555

(BS)that contains theH1 methylation site. This site is completely methylated in its normal position (Turker et al., H 0.5 Kbp 1989a, 1991). The BSfragment was ligated into theHpaI site of the pHL construct to create pHLBS. Results from the Southern blot analysis with DNA preparations isolated from HO 'HO' HZH4 H3 0 . 6 4 pHLBS transfectants demonstrated that the BS fragment ? 0 . 3 9 ? ? pHLPS .... containedsignals specifying de nouo DNA methylation of both the H1 and H2 sites. Complete methylation of the H1 HO H4 HI H3 H2 site was observed in three of four transfectants and near complete methylation (80%) in the other transfectant (Fig. 0.9 Kbp 4). Specifically, a 0.56-kbp hybridization band detected by the 1.3 Kbp pN1 probe in the MspI digestion lanes was absent or signifiI PN I I cantly decreased in the HpaII digestion lanes. This hybridiI p400 I zation band represents the fragment between the H2 and H1 FIG.3. The plasmid constructs pHL, pHLPS, and pHLBS. methylation sites in the pHLBS construct (Fig. 3). In all four The closed boxes below the lines for pHLPS and pHLBS represent the PstI fragment containing the HO site and the BglII fragment transfectants analyzed, the 0.56-kbp hybridization band was replaced by two new hybridization bands of 0.9 and 1.3 kbp containing the H1 site, respectively. Each bubble represents a HpaII/ MspI methylation site with the relative level of methylation in trans- in the HpaIIdigestion lanes. The 0.9-kbp hybridization band fected cells indicated by the degree to which the bubble is filled in. indicated a methylatedH1 site, and the1.3-kbp hybridization Shaded regions for the pHLPS construct indicate variation in the band indicated both a methylated H1 site and a methylated methylation pattern for different transfectants tested. The lines shown beneath the pHLBS construct represent specific hybridization H2 site. Specific hybridization of the 1.3-kbp hybridization bands observed in cells transfected with this construct (see Fig. 4). band with the p400 probe (Fig. 4) confirmed methylation of the H2 site at levels ranging from 16 to 45% in theintegrated The H beneath the pHL construct refers to a HpaI restriction site. pHLBS construct. The HO site in thepHLBS construct remained unmethylated. Therefore,the BSfragment could de nouo methylate, a t least partially,the H2 site located approximately 0.6 kbp downstream, but could not de M U O methylate the HO site located approximately 0.3 kbp upstream. This -1.3 result suggested that thede mu0 methylation signal from the -0.9 BS fragment was unidirectional. -0.4 Approximate Localization of Additional Sequences in the HS Frugment That Promote de Novo Methylation-The above experiment demonstratedthat the282-bp BS fragment,which is included within the larger HS fragment (see Fig. l),could de nouo methylate both the H1 and H2 sites. However, the BS fragment was missing from the pBLPS and pBLTSconstructs (see Fig. l), indicating that the remaining region of the HS fragment contained additional de nouo methylation activity. To localize the additional sequences within the HS FIG. 4. Southern blot analysis of cells transfected with the fragment that promoted de nouo methylation, three truncated pHL, pHLPS, and pHLBS constructs. Genomic DNA prepara- pieces of HS were tested for cis-acting de mu0 methylation tions from pHL, pHLPS, and pHLBS transfectants were digested signals when ligated into the pHLconstruct. In each case the with either MspI ( M ) or HpaII ( H ) ,separated by agarose gel electro- deleted region of the HSfragment included the H1 methylaphoresis, Southern-blotted onto a nylon membrane, and hybridized with the 32P-labeled pN1 probe. The nylon membranewas then tion site andone BglII site (Fig. 5A). The first truncated HS fragment used, termed HS1217, stripped and rehybridized with 32P-labeledp400 probe. Approximate contained a deletion for all but 265 bp of the 3' end of the molecular sizes of the hybridization bands are indicated. HS fragment (Fig. 5A). To determine if the region remaining in cated HO site remained unmethylated in pHLPS transfectants HS1217 contained de nouo methylation activity, the PS (Fig. 4). This conclusion was based on the presence of iden- fragment containing the HO site was ligated into the BglII tical 0.64- and 0.39-kbp hybridization bands in the HpaII and site of HS1217 and the resultant fragment was then ligated the HpaI site of the pHL construct to create MspI digestion lanes after hybridization with the pN1probe. into In this case, the 0.64-kbp hybridization band represents the pHLHS1217PS (see Fig. 5B). Southernblot analysisof seven fragment between the H2 site and relocated HO methylation pHLHS1217PS transfectants failed to demonstrate methylasite (see Fig. 3). Only the 0.39-kbp hybridization band was tion of the relocated HO site (Fig. 6). Moreover, the H2 site recognized by the p400 probe (Fig. 4), demonstrating that the was methylated in only three of the seven pHLHS1217PS H2 site was also unmethylated in the pHLPS construct. A transfectants and occurred at relatively low levels ranging lack of methylation for all sites was also noted when a 160- from 7 to 20%. Methylation of the H2 site was indicated by bp fragment that contained the H3 site was ligated into the the presence of a 1.25-kbp hybridization band in the HpaII HpaI site of pHL and transfected into the DELTG3 cells digestion lanes (Fig. 6) that specifically hybridized with the p400 probe (data not shown). (data not shown). The second truncated HS fragment used, termed A 282-bp BglIIfragment Restores de Novo Methylation Activity to the pHL Construct-Since the pHLplasmid lacked HS154,357, contains a 154-bp deletion at its 5' end and a 357the signals for de nouo methylation of the H2site, we decided bp deletion at its 3' end (Fig. 5A). The PS fragment containing the HO site was ligated into the BglII site of HS154,357, t o use it as a test system to assay for cis-acting signals in various portionsof the 1.5-kbp HS fragment. The first portion and the resultant fragment was then ligated into the HpaI of the HS fragment tested was the 282-bp BglII fragment site of the pHL construct to create pHLHS154,357PS (Fig. pHC

I

?

De Center Nouo Methylation

556 A. Hla

H

HLHS154,357PS

HS

HI

A

90

0

A

H

HS1217

HS 154.357

3.4-

3.0-

a

*

.$

2.11.7-

0.5 Kbp

B

0.8-1 pHLHS1217PS HO'HO'HOa

H2

9

I

0.86

L

H3 H4

0.4 -

__

0.39

I

1

PN1

E

p400

FIG.7. Southern blot analysis of cells transfected with the pHLHS154,357PS construct. Genomic DNA preparations from

pHLHS 154,357 PS HO HOa Hla r//

1.06 p0.25? 030

E

+'

3.0 Kbp 3.4 Kbp+/

7

'HO' 0.63

7

H3H4 H2

...

0.79 0.39

1.7Kbp 2.1 Kbp

I

pN I

I

W

I

cells transfected with the pHLHS154,357PS construct were digested or HpaII ( H ) . with EcoRI and then redigested with either MspI (M) The DNA preparations were then separated by agarose gel electrophoresis, Southern-blotted onto a nylon membrane, and hybridized with 32P-labeledpN1 probe. The nylon membrane was then stripped and rehybridized with 3ZP-labeledp400 probe. Approximate molecular sizes of hybridization bands are indicated.

removal of the 154-bp fragment at the 5' end of the HS fragment allowed methylation of the HO site in its normal FIG.5. A , HS fragment and deletion fragments HS154,357 and location. HS1217 derived from it. Bubbles refer to HpaII/MspI methylation The determination of a bidirectional signal from the sites. The restriction sites shown are: A , ApaI; B, BglII; H , HpaI. B, HS154,357 fragment was based on the results shown in Fig. the plasmid constructs pHLHS1217PS and pHLHS154,357PS. The open boxes below the lines for these constructs represent the HS1217 7, in which EcoRI digested DNA preparations from and HS154.357 inserts, respectively. Within these inserts, the closed pHLHS154,357PS transfectants were further digested with boxes represent the PstI fragment containing the HO site. Each bubble MspI or HpaII andhybridized with the pN1 andp400 probes refers to a HpaII/MspI restriction site. The degree of filling of the after Southern blotting. Complete methylation of the relobubble indicates the relative degree of methylation of the site in transfected cells. The lines shown beneath the pHLHS154,357PS cated HO site was demonstrated by the absence of a 0.79-kbp construct refer to specific hybridization bands observed in the South- hybridization band in the HpaII digestion lanes that was ern blot analysis with cell lines transfected with this construct (see present in the MspI digestion lanes. This hybridization band Fig. 7). represents the DNA fragment from the H2 site to the relocated HO site. IntheHpaII digestion lane, the 0.79-kbp c9 fragment was replaced by two strong hybridization bands of HLHS1217PS (3 3.0- and 3.4-kbp and two weak hybridization bands of1.73 and 2.1-kbp. The predominant 3.0- and 3.4-kbp hybridization bands indicated DNA fragments containing methylated HOa, HO, Hla, and relocated HO sites, with the 3.4-kbp hybridiza1.3tion band also containing a methylated H2 site. To account 0.9for the weak 1.7- and 2.1-kbp hybridization bands, the HO site in its normal location would be unmethylated. Methyla0.4tion of the H2 site in the 3.4- and 2.1-kbp hybridization bands was confirmed with the p400 probe (Fig. 7). Finally, a third construct was created by removing a 436PN' bp XhoI/PstI fragment containing the Hla site (restriction FIG.6. Southern blot analysis of cells transfected with the enzyme sites not shown) from the pHLHS154,357PS conpHLHS1217PS construct. Genomic DNA preparations from cells transfected with the pHLHS1217PS construct were digested with struct. The resultant construct, termed pHLHS511,357PS, or HpaII ( H ) ,separated by agarose gel electropho- contained a total of 551-bp from the region between the Hla either MspI ( M ) resis, Southern-blotted onto a nylon membrane, and hybridized with and H1sites and exhibited both an upstream and downstream "P-labeled pN1 probe. Also shown is a genomic DNA preparation de m u 0 methylation signal. The strength of the downstream from the DELTG3 cell line that was used as the recipient cell line signal was not altered, as evidenced by complete methylation for all transfection experiments in this study. Approximate molecular of the relocated HO site. However, the strength of the upsizes of hybridization bands are indicated. stream signal was reduced by approximately 50%, as evidenced by reduced methylation of the HO site in its normal 5 B ) . TheSouthern blotanalysis with pHLHS154,357PS transfectants revealed that this region of the HS fragment location (data not shown). Unfortunately, three attempts at contains a bidirectional de m u 0 methylation signal. Methyl- cloning the 436-bp fragment containingthe Hlasite into pHL ation in the downstream direction was indicated by complete were unsuccessful. methylation of the relocated HO site and partialmethylation DISCUSSION of the H2site. A signal in the upstream directionwas indicated by methylation of the HO site in itsnormal location a t levels We have used a deletion analysis to investigate de mu0 ranging from 83 to 100%. This result also suggested that methylation of four HpaII/MspI sites ina region upstream of 0.5 Kbp

De Center Novo Methylation the mouse aprt gene. These sites, termed HO, Hla, H1, and H2, have a conserved methylation pattern for the endogenous upstream region in all mouse tissues and cultured cell lines examined (see Turker et al. (1989a)). Based on our analysis, a model is presented in Fig. 8 to explain how the conserved methylation pattern is formed. The salient features of this model are as follows. 1) A cis-acting element, which is approximately 0.8-kbp in size, is responsible for de novo methylation of the Hla,H1, and H2 sites. This element, termed a de novo methylation center, is divided into two primary domains; each domain imparts a unidirectional de nouo methylation signal. 2) The signal traveling downstream from the de nouo methylation center is responsible for complete methylation of the H1site and partialmethylation of the H2 site. 3) Partial methylation of the H2 site is the result of the downstream signal being of variable strength. 4) The signal traveling in the upstream direction is responsible for complete methylation of the Hla site. 5) Sequences exist that “stop” the upstream signal from methylating the HO site and the downstream signal from methylating the H3 site. The experimental basis for this model is discussed below. Four lines of evidence supportour conclusion that the region upstream of the aprt gene contains acis-acting element that signals de nouo methylation within this region. The first was the previous demonstration (Turker et al., 1991) that the transfected pSam 6.3 plasmid would obtain the conserved methylation pattern characteristic of the upstream region. The second line of evidence was that the normally unmethylated HO site and H3-H8 sites of the CpG island were de nouo methylated when moved to a location within the upstream region that is normally occupied by the H1 site. The third line of evidence was that methylation of the H2 site and a relocated HO site was abolished by removing a 1.5-kbp HpaI fragment from the upstream region. Finally, de nouo methylation of both these sites was restored by reintroducing portions of the HpaI fragment. Evidence for a de nouo methylation signal traveling in the downstream direction was obtained by demonstrating methylation of the H2 site and the relocated HO site in constructs suchaspBLPS (Fig. 1) and pHLHS154,357PS (Fig. 5 B ) . Both sites remained unmethylated in the pHLPS construct (Fig. 3) and nearly so in the pHLHS1217PS construct (Fig. 5 B ) , thereby localizing the 3’ end of the downstream signal approximately to the H1site. The presence of de nouo methylation signals in both the pBL construct, which lacked the 282-bp fragment (BS) containingthe H1site, and the pHLBS construct, which contains only the BS fragment, indicates that the downstream signal can be broken into at least two parts. However, we still believe theyare part of a single functional domain. Evidence for a de nouo methylation signal traveling in the upstream direction was obtained with the pHLHS154,357PS construct (Fig. 5 B ) , in which methylation of both the HO and HOa sites was noted at very high levels (83-100%). The data also suggested the presence of a stop signal that prevented HO

?

r-7 I

\

L-J

4

methylation of the HO site and localized this stop signal to a 154-bp sequence at the 5’ end of the HS fragment that was absent in the HS154,357 fragment (Fig. 5A). Methylation of the HO site at a level of approximately 50% still occurred when the Hla site and surrounding sequences were removed from the HS154,357PS fragment to create a construct termed pHLHS511,357PS (Table I). This result demonstrates that a significant portion of the de nouo methylation signal traveling in theupstream direction was present in the region 3’ of the H l a site. Therefore, we speculate at this time that the upstream signal accounts for complete methylation of the Hla site and, at thistime, propose this site as the approximate 5’ end of the de nouo methylation center. In a previous study, we reported that the partial methylation profile for the H2 site is stably maintained in mouse tissues and cultured cell lines. This observation led to the prediction that a mechanism exists to account for the partial methylation profile (Turker et al., 1989a). We note that in the work presented in this study and in a previous publication (Turker et al., 1991), de nouo methylation of the H2 site in transfected plasmids, when it occurred, was always partial. We propose here that the strength of the downstream de nouo methylation signal is variable and reaches the H2 site only a fraction of the time. Consistent with this hypothesis is the observation that placing the de nouo methylation center closer to the H2 site in the pHLHS157PS construct resulted in greater degree of methylation of the H2 site (see Table I). From recent data (Cooper et al., 1992), we know that methylation of the H3 site is not compatible with transcription of the aprt gene. For this reason we have added to our model a stop sequence to prevent methylation of the H3 site and therest of the CpG island of the aprt gene. The presence of sequences that prevent de novo methylation of CpG islands has been proposed by others (Shemer et al., 1991b).Moreover, the isolation of a DNA fragment capable of preventing de nouo methylation of the thy-1 gene has been demonstrated by Szyf et al. (1990). The removal of the CpG island from the influence of a putative stop sequence could explain its complete methylation when moved within the methylation center. The importance of de nouo methylation of autosomal genes for both physiological and pathological processes is becoming increasingly apparent (see Introduction). Interestingly, de novo methylation of autosomal gene regions in developing embryos follows a genomewide demethylation event that occurs even earlier in embryonic development, several cell divisions after fertilization. It is believed that the very early demethylation event removes the methylation pattern contributed by the sperm and egg genomes (Cedar and Razin, 1990). Although a single mammalian methyltransferase enzyme has been described that is considered responsible for both maintenance andde nouo methylation reactions (Adams, 1990), the basic molecular mechanisms involved in the de novo methylation reaction within cells are currently unknown (Doerfler et al., 1990; Selker, 1990). Our data demonstrated that one mechanism to de nouo methylate DNA sequences is

H la

HI

H2

T

T

P

Methylation Center

(STOP?

557

)

H3 r -

r

-

9

\

(yoP?

b

L-J

FIG. 8. Model to explain conserved methylation pattern for region upstream of the mouse aprt gene. A de m u 0 methylation center, approximately localized to theregion between the Hla andH1 methylation sites, imparts a methylation signal to upstream (Hla) and downstream sites (H1 and H2). Partial methylation of the H2 site is dependent upon the strength of the signal traveling downstream. Sequences that stop the spreading of de nouo methylation in both the upstream and downstream direction are postulatedto punctuate the de novo methylation signals.

558

De Center Novo Methylation

via the action of a cis-acting element, which we have termed a de nouo methylation center. A similar proposal has been made by Syzf (1991). Others have postulated that de m u 0 methylation, when it occurs, is relatively nonspecific and that the CpG islands are protected from this event (Cedar and Razin, 1990). Another general hypothesis for de nouo methylation is that it isa relatively nonspecific event that follows chromatin changes associated with gene inactivation. This hypothesis works well to explain methylation of genes on the inactive X chromosome because this de novo methylation event occurs after gene inactivation (Lock et al., 1987). A hypothesis to explain increased methylation of specific chromosomal regions in malignant cells is that such methylation is due to increased levels of the methyltransferase enzyme. This increase then results in de nouo methylation of specific “hot spots” that are preferentially recognized by the enzyme (El-Deiry et al., 1991). Elements of this hypothesis are consistent with our observation of a de nouo methylation center. Given the different roles that methylation may play in mammalian cells, it is likely a variety of mechanisms exist to promote de m u o methylation in these cells. Therefore, aspects of the all of the above hypotheses may be correct and perhaps will ultimately prove to be interrelated as well. In conclusion, we have demonstrated for the first time that a cis-acting sequence can account for the formation of an endogenous methylation pattern. Although we do not believe that our model is applicable to all de nouo methylation events, i t is unlikely that the presence of such a sequence near the mouse a p t gene is a unique occurrence. Therefore, the presence of cis-acting elements may ultimately explain de nouo methylation events affecting a subset of autosomal genes. Acknowledgments-We thank M. Pieretti, G . Cooper, and N. Khattar for a critical reading of this manuscript. REFERENCES Adams, R. L. P. (1990)Biochem. J. 266,309-320 Bird, A. P. (1986)Nature 321,209-213

Bolden, A. H., Nalin, C. M., Ward, C. A., Poonian, M. S., and Weissbach, A. (1986)Mol Cell. Biol. 64.915-918 Ce&, H . (1988)Cell 63,3:4 Cedar, H., and b i n , A. (1990)Biochirn. Biophys. Acta 1049, 1-8 Cooper, G. E., Khattar, N. J., Bishop, P. L., and Turker, M. S. (1992)Somat. CeU Mol Genet. 18, 215-225 deBustros A. Nelkin B. D. Silverman A. Ehrlich, G., Poiesz, B., and Baylin, S. B. (lb88i Proc. Natl. A’cad. Sci. U.’S. k.86,5693-5697 Doerfler, W., Toth, M., Hermann, R., Lichtenberg, U.,and Hoeveler, A. (1990) in Nucleic Acid Methylation (Clawson, G. A., Willis, D. B., Weissbach, A., and Jones, P.A., e&) pp. 329-349, Wiley-Liss, New York El-Deiry, W. S., Nelkin, B. D., Celano, P., Yen, R.-W. C., Falco, J. P., Hamilton, S. P., and Baylin, S. B. (1991)Proc. Nutl. Acad. Sa. U.S. A. 88,3370-3374 Gardiner-Garden, M., and Frommer, M. (1987)J. Mol. Biol. 196,261-282 Kafri, T., Ariel. M.. Brandeis. M., Shemer. R.. Urven, L., McCarrev. .. J., Cedar, H.,.and Razin, A: (1992)Genes.& Deu. 6,705-714 . . Makos, M., Nelkim, B. D., Lerman, M. I., Latif, F., Zbar, B., and Baylin, S. B. (1992)Proc. Nutl. Acud. Sci. U.S. A. 89, 1929-1933 Miller, S. A., Dykes, D. D., and Polesky, H. F. (1988)Nucleic Acids Res. 16, 1215 Monk, M., Boubelik, M., and Lehnert, S. (1987)Deuehpment 99,371-382 Monk. M. (199Oa)Phil. Truns. R. SOC.London 326.299-312 Monk; M. (199Ob)Trendp Genet. 6,110-114 Lock, L. F., Takagi, N., and Martin, G. R. (1987)Cell 48,39-46 Oberle, I., Rousseau, F., Heitz, D., Kretz, C., Devys, D., Hanauer, A., Boue, J., Bertheas, M. F., and Mandel, J. L. (1991)Science 262,1097-1102 Perbal. B. (1988)A Proctieal Guide to Molecular CloninP. -. 2nd Ed.. John Wilev & Sons, New York Razjn, A., and Cedar, H . (1991),Microbiol.Rev. 66,451-458 Sa lenza C (1991)Bcochem.Bcophys. Acta 1072,51-61 SePker, E: (i990)Trends Biochem. 16, 103-107 Shemer. R.. Eisenbern, S.. Breslow. J. L., and Razin, A. (1991a)J. Biol. Chem. 266,‘23676-23681-. . Shemer R., Kafr, T., O’Connell, A., Eisenberg, S., Breslow, J. L., and Razin, A. (1691b)Proc. Nutl. Acud. Sci. U.S. A. 88,11300-11304 Smith. S. S.. Kan. J. L. C.. Baker.. D. J.. Kadan. B. E., and Dembek, P. (1991) J. M O L . s Wr 2i7,39-5i Solter, D. (1988) , Annu. Rev. Genet. It2,127-146 Stewart, C:L., Stehlman, D., Jahner, D., and Jaenisch, R. (1982)Proc. Natl. Acud. Scz. lI S. A. 79,4098-4102 Szyf, M.,Schinner, B. P., and Seidman, J. G. (1989)Proc. Nutl A d . Sci. U. A. -S. . . . 86.6853-6857 _ ~ . ~ , ~. ~ . SZ:yf, M., Tanigawa, G., and McCarthy, P. L.,Jr. (1990)Mol. Cell. Biol. ~~

4396-4400 - - - - - -. -

Szyf, M. (1991)Biochem. Cell Biol. 69,764-767 Toth, M., Lichtenberg, U.,and Doerfler, W. (1989)Proc. Natl. Acad. Sci. U.S. A. 86,3728-3732 Toth M Muller, U.,and Doerfler, W. (1990)J. Mol. Biol. 214,673-683 Turier. 3.S.. Mummaneni., P.., and BishoD. - . P. L. (1991)Somat. Cell Mol. Genei.-17,-151-157 Turker. K.. Smith., A. C.. , and Martin. G . M.(1989a)J. BioL .”, M. S.. ... Swisshelm. . Chem. 264, 11632-11636 Turker, M. S., Stambrook, P. J., Tischfield, J. A,, Smith, A. C., and Martin, G . M. (1989b)Somat. Cell Mol. Genet. 16,159-166 Ward, C., Bolden, A., Nalin, C. M., and Weissbach, A. (1987)J. Biol. Chem. 262,11057-11063 ~

~

”

_

~~~~

~

~~~

~~~~

I~

I

~