Enhancement of Reporter Gene de Novo Methylation by DNA ...

THEJOURNAL OF BIOLOGICAL CHEMISTRY 0 1994 by The American Society for Biochemistry and Molecular Biolow, h e .

Vol. 269, No. 3, Issue of January 21, p p . 1821-1826, 1994 Printed in U.S.A.

Enhancement of Reporter Gene de Novo Methylation by DNA Fragments &om the a-Fetoprotein Control Region* (Received for publication, July 8, 1993)

Andreas Hasse and WolfgangA. SchulzS From the Znstitut fur Physiologische Chemie I, Heinrich-Heine-Universitat,0-4000 Dusseldorf; Germany

The 5’-upstream region of the rat a-fetoprotein (AFP) whereasothers are intrinsicallyprotected against de nouo gene strongly increased de novo methylation of an adja- methylation (13-15). These findings support a model for the cent chloramphenicol acetyltransferase (CAT) gene generation of DNA methylation patterns stating that methylupon transfection intoF9 mouse embryonal carcinoma ation starts at certain sequences termed “centers of methylcells. The same effect was exerted by a distal 776-base ation’’ (16). Such sequences should become evident ina transpair (bp) fragment and by 300- and 1-kb fragments pre- fectionexperimentby their influenceon the methylation ceding the transcriptional start site, but not by other pattern of adjacent DNA sequences. partsofthecontrolregion.Furtherdivisionofthe Using the mouse embryonal carcinoma cell lineF9 we have larger, strongly active fragments resulted in a gradual previously shown that someeukaryoticpromotersequences decrease ofmethylationandclonalvariationinthe indeed affect the methylation pattern of an adjacent CAT’ remethylation patterns. The effect of the 775-bp fragment porter gene (11). When the CAT gene was introduced next to did not dependon its orientation. It was ablated by insertion of themousemetallothionein I promoterbe- the mouse MT-I promoter, its methylation was significantly CAT gene, but not reduced as compared to a promoterless plasmid. In contrast, tween theAFP gene fragment and the by its insertion upstreamof the AFP gene fragment.Two 6.7 k b of t h e rat a-fetoprotein (AFP) control region caused a of the CAT gene and even of fragments from the AF’P control region increasing meth-strong increase in the methylation ylation contained B1 and B2 small interspersed repeti- vector sequences in the integrate (11) that are otherwise only weakly methylated in F9 cells (15). Surprisingly, integration tiveelements,respectively. B1 and B2 sequencesof of the transfected plasmids did different origin also acted strongly to increase methyl- site and transcriptional activity not significantly affectthe resulting methylation patterns(11). ation. Here, we report an analysis of t h e effects of different parts of Thesefindingssupporttheideathatmammalian the AF’Pupstream region onthe methylation of a CAT reporter genes contain specific sequences involved in regulating their methylation. The effects of these sequences appear gene. to be exerted in cis, to be dependent on proximity, but MATERIALSANDMETHODS not on orientation, and to require an optimal size of 500-700 bp.Smallretrotransposonsequenceswithin Cell Culture and Dansfection”F9 murine embryoniccarcinoma such elements may be particularly effective in attract- stem cells were maintained as described previously (17) in Dulbecco’s modified Eagle’s mediumsupplemented with L-glutamine,gentamycin, ing de novo methylation. and 10%heat-inactivated fetal calf serum. They wereassayed regularly for the absence of differentiation markers. Transfections were performed by the calcium phosphate method as “he methylation patterns of vertebrate genes undergo com- described elsewhere (11). 10 pg pertissue culture dish of the respective plex changes during embryonic development (fora review, see CAT plasmid and pRSVneo at a 2:l ratio were incorporatedinto a 500-pl Ref. 1) whichare essential for normal differentiation.Homozy- precipitate. The precipitate was left on the cultures for 8 h which were gous disruption of t h e DNA methyltransferase locus in trans- then treated for 90 s with a HEPES-buffered salt solution containing 15% glycerol. Selection for expression of the resistance marker was genic mice caused embryonic lethality resulting from severely started 40 h later in Dulbecco’s modified Eagle’smedium containing 500 abnormal morphology (2). Both de novo methylation and de- pg/ml geneticin (Life Technologies,Inc.). Resistant cell clones were isomethylation participate in the generation of methylation pat- lated after 14 days of continuous selection. terns (3-5). De nouo methylation and maintenance methylation Plasmids-Several AFP deletion plasmids were kindly provided by are both catalyzed by DNA-methyltransferase (61, b u t i tis not Dr. J. Locker, University of Pittsburgh, and have beendescribed known how the enzyme distinguishes sequences to be methyl- (PAFP~OOOCAT,pAFPAlCAT, and pAFPA2CAT (18); pAFPlOOOCAT and SpeAXCAT (19)). pGCATC and pGCATA are promoterless CAT ated de novo. vectors (20) differing in the orientation of the multiple cloning site. Certain embryonic cells in culture possess a particularly “pXCAT” constructs were generated by inserting the following XbaI high d e novo methylation activity toward foreign, e.g. retrovi- fragments from pAFPA1CAT into the XbaI site of pGCATC (numbered ral, DNAintegrated into their genomes (7,8). DNAmethylation as in Ref. 21): pX75OCAT, -7041- to -6265-bp fragment; pX13OOCAT, -6266 to -4931 bp; PX~OOCAT, -4932to -4490 bp; PX~OOOCAT, -4491 is dependent on propertiesof the inserted DNA sequences (9to -3365 bp, -2527 to -2339 bp, and -2017 to -1294 bp; pX3OOCAT, 11). Somesequences are efficientlymethylated(7,11,12), -1295 to -977 bp. pA21OCAT contains the XbaIIAuaII fragment from -7041 to -6833 bp, pA3OOCAT the AuaII fragment from -6834 to -6528 * This work was supported by the Deutsche Forschungsgemeinschaft bp,pA26OCAT the AuaIUXbaI fragment from-6529to-6265bp, Grant Schu60U3-2.The costs of publication of this article were defrayed pS34OCAT the XbaI/StuI fragment from -7041 to -6700 bp, in part by the payment of page charges. This article must therefore be pXA35OCAT an XbaUAuaI fragment carrying -978 to -643 bp in pGhereby marked “aduertisement”in accordance with 18 U.S.C. Section CATA and pA65OCAT the AuaUHpaII fragment from -642 to +7 bp. 1734 solely toindicate this fact. $ To whom correspondence should be addressed: Institut fur PhysiologischeChemieI, Heinrich-Heine-Universitat, Postfach 101007, The abbreviations used are: CAT, chloramphenicol acetyltransferD-40001 Dusseldorf, Germany.Tel.: 49-211-311-2715; Fax: 49-211-311- ase; AFP, a-fetoprotein; MT-I, metallothionein I; bp, base pair; kb, kilo3029. base.

1821

Enhancement of DNA de Novo Methylation

1822

pBlCAT was constructed by inserting a 750-bp PstI fragment from p64.7T (-0.2)into the PstI siteof pGCATC, and pB2CAT by inserting a 0.48-kb PuuII fragment from the same plasmid into the SmaI site of pGCATC. p64.7T (-0.2) was kindly provided by Dr. R. Cox, London.For the construction of pL1-54OCAT and pL1-39OOCAT, EamHI fragments 540 and 3900 bp long were excised from pMRB1.1 and pMRB5, respectively (22) and cloned into the EamHIsite of pGCATC. pMT750CATwas constructed by inserting a 670-bp KpnIIBglIIfragment from the mouse metallothionein I promoter (23) into the SmaI site of pX75OCAT (5' to the X750 fragment); p750MTCAT contains the same fragment inserted into the HincII site 3' to X750. Isolation and Analysis of genomic DNA-High molecular weight DNAfrom stably transfected F9 clones was isolated. The DNA was analyzed for the presence ofCAT genes by polymerase chain reaction using the primers CAT5 (GATITTCAGGAGCTAAGGAAGCTAA) and CAT3 (TAACACGCCACATCTTGCGAATATA)under standard polymerase chain reaction conditions which result in the amplification of a 406-bp fragment from the CAT gene. Subsequently,positive clones were analyzed for methylation of the integrated plasmids as described (11) using restriction endonucleases at a 5-fold excess.DNA was separated on 1.3%agarose gels and was transfered to Hybond N membranes in 10 x SSC. Hybridization was performed at 65 "C in 7% sodium dodecyl sulfate, high phosphate buffer, with high stringency washes as described (17). Probes were labeled with L3*P]dCTP using the random oligonucleotide-priming method. The SacI-Sau3A fragment from pGCATC indicated in Fig. 1was used as a probe for CAT sequences.

AFP control region

-5

~5

-*

1

-3::(,..*:2I I

... .: :...

I

0

-1

I

.... ....

I

*:/

.' II

A pAFPAlCAT

L 7

pAFPA2CAT

-

.

jY / $

123 L CAT

P

....

....

[ PWAXCAT x

pGCATC

IEZa::: P

b::::

pX75OCAT

=

.... ....

m::::

n

pX13OOCAT

::1'pXL00CAT '

0:

RESULTS

First, several CAT plasmids containing parts of the ratAFP upstream regions weretested that had been used for the analysis of transcriptional regulation of the AFP gene (18, 19). The plasmid pAlCAT contains 6160 bp of the original control sequence, pA2CAT contains its distal-most 3000bp plus the proximal 1600 bp, and pSpeAXCAT only the 65-bp minimal promoter (Fig. lA). Fig. 1also shows the locations of the HpaII sites in the CAT gene and the CAT probe used for Southern hybridization. Following stable transfection of these plasmids into F9cells, the resulting methylation patterns of the CAT reporter gene were analyzed by Southern hybridization following digestion with HpaII and MspI. As a standard for methylation, the promoterless CAT vector pGCATC was used which becomes only moderately methylated in F9 cells (11).Although our previous study had not provided evidence for clonalvariation in methylation patterns, several individual cell clones were investigated for each construct. For quantitative evaluation of methylation, the intensities of the individual bands obtained by hybridization with the CAT probe were determined by laser densitometry of appropriately exposed autoradiographs. Methylation of the fragments Mp-M1 and M3-M4 was calculated as follows. The fraction of the signal intensity at the correct size on all signal intensities of bands of equal or larger size in the HpaII lane was determined and divided by the same quotient obtained from the MspI lane. Methylation of the M,-Ma band was not evaluated, because the Mz site was almost completely methylated in all constructs including pGCATC, with the exceptions described later. In addition it could not be quantitated in all clones because of its small size. The increase in methylation toward pGCATC was calculated from the average of methylation of the fragments Mp-M1 and MrM4 using a value of 42% average methylation for pGCATC. The quantitative data for methylation are summarized in Table I. In the plasmids pAlCAT and PASCAT, the increase in the methylation of the CAT sequence was as strong asinthe construct pAFP7OOOCAT investigated previously (Fig. 2 A ) . pSpeAXCAT became methylated to a similar overall extent as pGCATC (Fig. 2 4 ) . Since the M,-M, and M3-M4 bands in thisconstruct possess similar lengths (350 and 408 bp, respectively), the methylation of individual sites couldnotbe evaluated. Since pAlCAT and pA2CAT are shorter than pAFP7OOOCAT by 838

pAFprOOOCAT *

r:::

pAZ1OCAT

l-l pA300CAT

._..

P

....

pA260CAT

n

pSWCAT

*

n: pXA350CAT

.n::::

pA650CAT '

I :

FIG.1. AFP-CAT constructs used for methylation analysis.All plasmids have in common the CAT sequence (hatched boxes) and are flanked by SV40 polyadenylation signals on the right and pBR322 sequences on the left, respectively (not shown). The four MspI sites inthe CAT sequence ( I - f ) , the proximal one in the AFP part ( P ) , and the probe used for hybridization (bar) are indicated. An X in front of the constructs indicates that the P MspI site lies within the pBR322 sequence. A, AFP deletion constructs; E , constructs with XbaI fragments from pAFPA1CAT; C, constructs with subfragments from pX75OCAT or pAFP1OOOCAT.

and 2458 bp, respectively,the level of CAT methylation did not drop gradually with the size of the upstream region, but appeared to depend on the presence of specific sequences in the control region. The major part of the AFP upstream sequence present in pAlCAT which was sufficient to mimic the effect of the entire AFP control region on methylation, was subdivided arbitrarily into XbaI fragments as depicted in Fig. lB.All fragments were then inserted into pGCATC 5' to the CAT gene. The resulting plasmids were designated pX75OCAT, pX13OOCAT, pX4OOCAT, pX2OOOCAT, and pX3OOCAT in 5' to 3' order. The remaining 1000-bp proximal promoter fragment is contained in

Enhancement of DNA de Novo Methylation TABLE I Methylation of AFP-CAT constructs transfected into F9 cells Values were determined for three to six individual clones per construct asdescribed in the text. Methylation of fragment

Construct MS",

MAf1 96

A 7 a T

% HM

b~

Increase in methylation

I

96

pAFPA1CAT pAFPA2CAT

85 2 5 92 f 1

87 9 94 2 2

*

105 121

pX75OCAT pX13OOCAT pX4OOCAT pX2OOOCAT pX3OOCAT pAFPlOOOCAT 9 0 2 2

92 2 4 52 2 14 58 2 21 57 f 14 84 11 8929

82 1 6829 61 2 21 6628 89 * 9

*

107 43 42 46 106 113

pA21OCAT pA3OOCAT pA26OCAT pS34OCAT pXA35OCAT pA650CAT

68 2 12 75 2 25 73 2 18 41 2 34 66 2 13 64 f 13

56 17 66 2 34 74 2 17 63 25 81 2 17 93 2 2

48 68 75 24 75 87

pAFFX750-CAT p75OMTCAT pMT750CAT

78 2 7 33 2 18 77 f 13

86 2 12 45 2 17 58 13

95 -7 68

*

1823

pAFPlOOOCAT (19).These constructs were transfected into F9 cells, and the methylation pattern of the CAT sequence was investigated in at least three independent cell clones for each plasmid. The resulting patterns are shown in Fig. 2 B , and a quantitation of the resultsis presented in Table I. As compared to the promoterless control plasmid pGCATC, each fragment caused an increase in thelevel of methylation, which, however, was only moderate with the fragments X1300 (43%),X400 (42%),and X2000 (46%) derived from the central part of the AFP upstream region (Fig. 1B ).In contrast, the most distal partof the control region contained in pX75OCAT drastically elevated the methylation of the adjacent reporter gene (107%). In clones harboring this construct, CAT methylation was as high as inpAFP7OOOCAT (110%).A second region causing elevated methylation was contained within the 1300 bp preceding the transcription initiation site as indicated by increases in methylation of 106% and 121%,respectively, in pX3OOCAT and pAFP1OOOCAT. The fragments contained in pX75OCAT and pAFPlOOOCAT were divided further and cloned into pGCATC. X750 was cut with AvaII yielding the fragments A210, A300, and A260 (Fig. X ) . These retained a detectable, albeit reduced effect on CAT gene methylation with increasesof 44%for pA21OCAT, 68% for pA3OOCAT, and 75% for pA26OCAT. The significantly lower value for the 5'-most fragment, A210, was not only due to its small size, since the methylation level of the CAT sequence in pS34OCAT, which contains the region present in pA21OCAT plus the adjacent 130 bp was even lower (24% increase) than that of pA21OCAT. The proximal AFP upstream region was analyzed further using two constructs, pXA35OCAT and pAX65OCAT, obtained by cutting theproximal 1000 bp of the AFP upstream region with AvaI and inserting therespective 350- and 650-bp fragments upstream of the CAT coding sequence (Fig. 1C). The 350-bp XbaI-AvaI fragment caused a significant increase in CAT gene methylation (75%), and the effect of the 650-bp sequence containing the actual promoter was even stronger (87%).As in thecase of pX75OCAT, the individual parts caused less methylation than the entirefragment. Since two of the regions in theAFP control region increasing methylation contained small interspersed repetitive elements belonging to the B1 and B2 class, respectively (21), the effect of

B

HM

HM

lp17-

HM

HM

-0

.

HM

RG.2. Methylation patterns of the CAT sequence in AFP-CAT plasmid Constructs transfected into F9 cells. 10 pg of genomic DNA from each cell clonewas digested withHpaII (H) or MspI ( M ) and hybridized with a CAT probe followingSouthern blotting. See Fig.1 for plasmids and the extension of the probe. A typical pattern isshown for each construct. The respective sizes of the M3-M4 fragment (408 bp), of the M2-M31 fragment (128 bp), and of the M,-Ml fragment (variable in each construct, cf. Fig. 1) are indicated, as well as the sizes of some bands resulting from methylation of certain MspI sites (e.g. 536 = 408 + 128 resulting from methylation of site M3, but not Mz and M4).A42-bp band corresponding to the Ml-M2 fragment cannot be detected by Southern analysis. B1 and B2 sequences derived from a different source on de novo methylation of the CAT gene was also tested by using the constructs pBlCAT and pB2CAT (Fig. 3A). For comparison, two constructs containing parts of L1 long interspersed repetitive elements, pL1-54OCAT and pL1-39OOCAT (Fig. 3A),were also investigated. The larger L1 construct contains 3.9 kb from the central region of L1, whereas the smaller fragment in pL154OCAT corresponds to the BAM5 region of L1 elements (see Ref. 24 for a review). m i c a 1 methylation patterns of the CAT sequences in these constructs are shown in Fig. 3B. The strongest effect was seen with the B2 element, but methylation was also strongly enhanced by the B1 element. Of the sequences from the L1 element, the 3.9-kb fragment also had a strong effect on CAT methylation, whereas the effect of the 540-bp fragment was minor (Fig. 3B). Densitometric analysis yielded values of 129, 108, 104, and 43% average increase in methylation as compared to pGCATC for pB2CAT.pBlCAT,pL139OOCAT and pL1-54OCAT, respectively. pB2CAT thus attained the strongest methylation of all constructs investigated. To test whether its effect on CAT methylation was dependent on orientation, the AFP fragment in pX75OCAT was inverted, yielding pX750-CAT. Methylation of the reporter sequence in this plasmid was about as strong as inpX75OCAT (95%versus 107%increase in methylation, Fig. 4).


1824

pX75O'CAT

A

5 6 7 8 9

...

pL1-SLOCAT

...

H M H M H M H M H M

... .._

pL:-39%CAT

pB1CAT

"-

I,.

"

pB2CAT ' CAT-Sonde

'

lkb

pL1B bp

pL15COCAT 39OOCAT HM

i a

pB1 CAT HM

HM

1. .

&

1

pB2CAT HM

I I c

9Lo-

RG.3. Methylation of CAT constn~ctec o n t a i i g repetitive DNAsequences.A, structure of the constructs, the first MspI 5' of the CAT gene is indicated as Mp.B , methylation analysis as described in Fig. 2. The sizes of the M,-M, fragment are 710, 650,940, and 450 bp for pL1-54OCAT, pL1-39OOCAT, pBlCAT, and pBBCAT, respectively.

To test the interaction between the sequences contained in the AFP control region that increase methylation of the CAT gene and the670-bp mouse MT-I promoter sequence which had been found to reduce it (ll),two constructs were investigated that contained both the distal 775-bp fragment from the AFP control region and theMT-I promoter. In pMT750CAT the AFP fragment wascloned next to CAT gene and theMT-I promoter further upstream, whereas in p750MTCAT the order of the two sequences was reversed. In pMT750CAT, the CAT sequence retained significantly increased methylation (74%),whereas in p750MTCAT its methylation was lower than inpGCATC (Fig. 5). In particular, site M2in theCAT gene, which becamemethylated in all AFP-CAT constructs as well as in pGCATC, remained completely unmethylated in these clones. Variations of the methylation pattern in individual cell clones with the same construct were small in constructs containing large parts from the AFP control region, but became significant in some constructs harboring short fragments. This is indicated by the standarddeviations of the quantitative values in Table I, and is directly evident from the respective autoradiographs. An example for divergent methylation patterns of thq same construct is given in Fig. 6 showing CAT gene methylation in four cell clones harboring pA3OOCAT. Clonal variations of the methylation patterns are infrequent in our system (cfi Fig. 4) and had not been observed in a previous study (11).A survey of the data inTable I indicate that clonal variations became prominent with some fragments shorter than about 400 bp.

210-

a

128-

FIG.4. Methylation patterns of pX76O-CAT in four F 9 cell clones. DNA samples from four distinct F9 cell clones(5-9) containing pX750-CAT were analyzed for methylation of the CAT sequence as described in the legend to Fig. 2. DISCUSSION

In this study, we provide evidence that a cellular control region contains DNA elements influencing the extent of de nouo methylation of an adjacent reporter gene. Previously,the intact rat AFP upstream sequence had been shown to elevate the methylation of a transfected CAT gene in cis. Each part of the control region possessed an intrinsic ability to modify the methylation pattern of the CAT gene. Somedid not drastically affect it, but two regions strongly elevating methylation were found in thedistal-most 775 bp of the control region and in the1300 bp preceding the transcription initiation site. Shorter subfragments from both regions retained a significant effect, but a gradual decline in methylation was observed. Moreover,strong deviations in the methylation patterns of independent cell clones emerged with some fragments shorter than about 500 bp. Both observations are best explained by a minimal size requirement for DNAfragments in order to modulate the methylation of adjacent sequences. Some short fragments may only be able to create an effect on methylation when supported by other sequences at the integration site. From our data, the minimal length for a consistently active sequence can be estimated as between 500 and 700 bp. However, some constructs with small DNA fragments such as pX3OOCAT did not show variations. Possibly, they contain stronger recognition sites for methylation that areless easily influenced by sequences at the integration site. So far, few other studieshave provided evidence for sequence elements directing methylation. In integrated adenovirus DNA, methylation starts within a defined region of about 1kb and spreads gradually throughout the viral genome and into flanking host sequences (9). In the human APRT upstream region, an 800-bp fragment was found necessary for the generation of a specific methylation pattern followingtransfection into P19 embryonal carcinoma cells(12).Conversely, a sequence in the human Thy-I gene has been reported to inhibit the methylation of surrounding DNA (13), thus behaving similar to the MT-I promoter (11).In toto, these findings support the idea that centers of methylation in theregulatory sequences of cellular genes (16) are involved in regulating de novo methylation and in controlling the changes of DNA methylation dur-


pMT750 p750MT CAT CAT

** 1

bp

2

1

1825 1

2

4

5

I--ld-l”

H MH MH MH M

bp

2

3

H M H M H M H M

2323

-

1929

-

1371 1264

--

702

-

t 578536 -

408350FIG.5. Methylation patterns of constructs containing both the AFPX7M) fragment and the MT-I promoter. Autoradiograph of a Southern blot with HpaII (HI and MspI ( M ) digests of genomic DNA from two pMT750CATand two p750MTCAT cell clones hybridized with the CAT probe as in Fig. 2.

406

-

.

FIG.6. Methylation patterns of pASOOCAT in four FYI cell clones. DNA from four cell clones ( I , 2,4, and 5) containing pA3OOCAT was analyzed for methylation of the CAT sequence as described in the legend to Fig. 2; H , HpaII; M, MspI.

mal region, the 300- and 350-bp fragments increasing methyling development. However, the dataobtained with the AFP control region are ation are probably also devoidof transcriptional regulatory difficult to reconcile with a narrow definition of centers of meth- elements (191,but share between them a B2 element (21).We ylation. The CAT gene became methylated to a moderate de- therefore tested whether repetitive elements from different gree and with a reproducible pattern even in the absence of sources might also enhance de novo methylation of an adjacent sequences from the AFP control region. Ifde novo methylation CAT sequence. B1 and B2 elements, classified as small interwere strictly to depend on the presence of a center of methyl- spersed repetitive elements (24),were found to be most ation, the bacterial CAT gene would have to contain one. In strongly active, although a systematic study of further eleaddition, most of the sequences from the control region caused ments has yet to be undertaken. This finding is intriguing in some increase in its methylation. Thus, methylation was the light of the hypothesis that methylation in eukaryotes may strongly influenced by, but was not completely dependent on, have evolved froman ancient host defense mechanism directed specific sequences in theAFP control region. primarily against foreign DNA (26).If B1 and B2 elements are In theAFP control region, two regions increased methylation indeed recognized by the de nouo methylation system, they to a particular strongextent. For a 775-bp fragment, this effect could function as starting points for methylation in many was shown to be independent of its orientation toward the CAT genes, since they are widely distributed in the genome. The third active fragment comprises within its 650 bp the gene. This is reminiscent of a property of transcriptional enhancers. However, the increase of methylation in theCAT gene promoter and, most probably overlapping with it, a silencer was lost when the mouse metallothionein I promoter was in- region, whose boundaries have not yet been exactly defined (19, serted between the AFP gene fragment and its target. Con- 25). Therefore, we cannot distinguish whether the promoter or versely, insertion of the 775-bp fragment between MT-I pro- the silencer sequences or both are responsible for the increase moter and theCAT gene interfered with the effect of the MT-I in methylation. sequence on DNA methylation. Therefore, effects on methylWith regard to their CpG density, none of these sequences is ation seem to depend not only on a cis position, but also on expected to be a particularly good substrate for DNA methylproximity. data (27-30). There transferase as judged by available in vitro The entireAFP upstream region has been sequenced, and the are only 5 CpG sites in the775-bp fragment, 3 sites in theX300 localization of transcriptional regulatory elements, potential fragment, and 14 in theproximal 1000 bp. In comparison, the methylation sites, and repetitive elements is known (19, 21,CpG frequency in the CAT gene (31) approximates the optimum 25). The 775-bp fragment is located next to the most upstream value of 1 CpG per 25 nucleotides determined in vitro.Thereenhancer, but does not contain transcriptional regulatory ele- fore, the sequences increasing methylation do not necessarily ments itself (19, 21). Instead, it consists of an assortment of correspond to transcriptional regulatory elements or represent repetitive elements, starting with the 5’-end of an L1 element, optimal substrates for DNAmethyltransferase. They may, howfollowed bya sequence with homology to Alu elements and a B1 ever, possess a DNA conformation that is optimal for the enelement. Incidentally, these elementsare essentially separated zyme, as suggested by the preference of DNA methyltransferby the AvaII digestion used for subcloning, and the subfrag- ase for unusual structures ofDNA in vitro(29, 30, 32). This ments of the 775-bp sequence eliciting increased methylation could also explain the finding that larger regions are more therefore correspond to the Alu and B1 element. In the proxi- effective in increasing methylation.

Enhancement de of DNA

1826

Acknowledgments-We are grateful to Dr. H. Sies. Dusseldorf. for helpful discussions and encouragement,to Dr. J. Locker, Pittsburgh, for providing plasmids and sequence data, and to U. Adamek for technical assistance. REFERENCES 1. Razin, A,, and Cedar, A. (1993)in DNA Methylation (Jost, J. P., and Saluz, H. P., eds), pp. 343357,Birkhauser, Base1 2. Li, E., Bestor, T. H., and Jaenisch, R. (1992)Cell 89,91&926 3. Razin, A,, Webb,C., Szyf, M., Yisraeli, J., Rosenthal, A., Naveh-Many,T., Sciaky-Gallili, N., and Cedar, H. (1984)Proc. Natl. Acad. Sci. U.S. A. 81, 2275-2279 4. Monk, M., Boubelik, M., and Lehnert, S . (1987)Deuelopment 99, 371-382 5. Shemer, R., Kafri, T., O'Connell, O., Eisenberg, S., Breslow, J. L., and Razin, A. (1991)Proc. Natl. Acad. Sci. U. S. A. 88, 1130CL11304 6. Bestor, T., Laudano,A., Mattaliano, R., and Ingram, V. (1988)J. Mol. Biol. 203, 971-983 7. Niwa, 0.(1985)Mol. Cell. Biol. 6, 232&2331 8. Jahner, D., and Jaenisch, R. (1985)in DNA Methylation: Biochemistry and Biological Significance (Razin, A,, Cedar, H., and Riggs, A. D., eds), pp. 189-219,Springer, New York 9. Orend, G., Kuhlmann, I., and Doerfler, W. (1991)J. Virol. 65,4301-4308 10. lhrker, M. S., Mummaneni, P., and Bishop, P. L. (1991)Somatic Cell Mol. Genet. 17, 151-157 11. Hasse, A,, Schulz, W. A,, and Sies, H. (1992)Biochim. Biophys. Acta 1131, 16-22 12. Mummaneni, P., Bishop, P., and Turker, M. (1993)J. Biol. Chem. 268,552458 13. Szyf,M., Tanigawa, G., and McCarthy, P. (1990)Mol. Cell. Biol. 10,4396-t400 14. Shemer, R.,Eisenberg, S., Breslow, J. L., and Razin, A. (1991b)J. B i d . Chem. 2sS, 23676-23681

Novo Methylation 15. Frank, D., Keshet, I., Shani, M., Levine, A., Razin, A,, and Cedar, H. (1991) Nature 361,239-241 16. Szyf,M. (1991)Biochem. Cell Biol. 89,764-767 17. Schulz, W. A,, and Gais, G . (1989)Biochim. Biophys. Acta 1013,125-132 18. Muglia, L., and Rothman-Denes, L. B. (1986)Proc. Natl. Acad. Sei.U. S. A. 89, 7653-7657 19. Wen, P., Groupp, E. R., Buzard, G . , Crawford, N., and Locker, J. (1991)DNA Cell Biol. 10, 525-536 20. Frebourg, T.,and Brison, 0.(1988)Gene 6 6 , 3 1 5 3 1 8 21. Buzard, G., and Locker, J. (1990)DNA Sequence 1,33-48 22. Meunier-Rotival,M., Soriano, P., Cuny, G., Strauss, F., and Bernardi, G . (1982) hoc.Natl. A c a d . Sci. U. S. A. 79,355359 23. Stuart, G . W., Searle, P. E, Chen, H. Y.,Brinster, R. L., and Palmiter, R.D. (1984)Proc.Natl. Acad. Sci. U. S. A. 81,731S7322 24. Hastie, N. D. (1989)in Genetic Variants and Strainsof the Laboratory Mouse (Lyons, M. F., and Searle, A. G., eds) 2nd Ed., pp. 559-573,Fischer, Stuttgad

25. Guertin, M., LaRue, H., Bernier, D., Wrange, 0.. Chevrette, M., Gingras, M. C., and Belanger, L. (1988)Mol. Cell. Biol. 8, 1398-1407 26. Doerfler, W. (1991)Biol. Chem. Hoppe-Seyler 372,557464 27. Bolden, A. H., Ward, C. A,, Nalin, C. M.,and Weissbach, A. (1986)Prug. Nucleic Acid Res. Mol. Biol. 33, 231-250 28. Carotti, D., Palitti, F., Mastrantonio, S., Rispoli, M., Strom, R., Amato, A, Campagnari, F., and Whitehead, E. P. (1986)Biochim. Biophys. Acta 866, 135-143 29. Bestor, T. (1987)Nucleic Acids Res. 15,38353843 30. Bestor, T., Gundersen, G., Kolsto, A,, and F'rydz, H. (1992)GATA 9,48-53 31. Alton, N. K., and Vapnek, D. (1979)Nature 282,864-869 32. Smith, S . S., Kan, J. L. C., Baker, D. J., Kaplan, B. E., and Dembek, P. (1991) J. Mol. Biol. 217, 39-51