CpG methylation represses the activity of the rat prolactin promoter in rat GH3 pituitary cell lines

ELSEVIER

Molecular and Cellular Endocrinology 108 (1995) 9.5-105

CpG methylation represses the activity of the rat prolactin promoter in rat GH, pituitary cell lines V.M. Ng8*, J.-N. Laverrike,

D. Gourdji

Groupe de Biologie de ia Cellule Neuroendocrine. CNRS URA 111.5, CollPge de France, 1 I place Marcelin Berthelot, 75231 Paris cedex 05, France

Received 7 October 1994; accepted 7 December 1994

Abstract In the present report, we have investigated the role of DNA methylation on the binding and trarzs-acting properties of transcription factors involved in the regulation of the rat prolactin (rPRL) gene, specifically Pit- 1. To this aim we took advantage of a model system composed of three GH3 rat pituitary tumor cell lines that greatly differed in the extent of rPRL gene methylation and in the level of rPRL gene expression. Northern blot analyses indicated that identical species of Pit-l mRNA were present to similar extent in the three GH3 cell lines. Electrophoretic mobility shift assays further demonstrated that Pit-l was present in nuclear extracts and displayed equal affinities to bind the IP responsive element encompassing the -65 to -38 region of the rPRL promoter, whatever the GH3 cell line tested. These data suggested that differential expression of the rPRL gene among cell lines did not result from variable amounts of Pit1. By combining in vitro methylation and transient transfection experiments with a rPRL promoter-driven CAT construct, we showed that extensive methylation at CpG sites abolished the expression of the reporter gene. Furthermore, in vivo competition assays demonstrated that CpG methylation inhibited gene expression by preventing the binding of transcription factors We propose that related mechanisms linked to DNA methylation might alter the activity of the endogenous PRL gene in the low expressing cell line. Keywords:

DNA methylation; rPRL gene; Pit-l; GH3 cell lines

1. Introduction In the mammalian genome, methylation of cytosine residues engaged in CpG dinucleotides constitutes an epigenetic and inheritable modification of DNA (see Holliday, 1993). Once established, the methylation status is believed to play an important role in the maintenance of patterns of gene expression (Razin and Cedar, 1991; see Yeivin and Razin, 1993). In particular, tissue-specific genes undergo a specific demethylation concomitant to their activation process (Waalwijk and Flavell, 1978; Bird, 1986; Shemer et al., 1990). Conversely, de novo methylation is often associated with the silencing of genes (Doerfler, 1983; Cedar, 1988). The current hypothesis for this inhibitory effect is that DNA methylation could modify the availability of regulatory sequences to interact with the transcription apparatus. This hindrance is thought

*Corresponding 84.

author, Tel.: +33 1 44 27 15 88; Fax: +33 1 44 2710

to be induced by the combination of three mechanisms: (i) direct inhibition of DNA binding by methyl groups; (ii) indirect inhibition following the occupancy of methylated DNA by methylcytosine binding proteins; and (iii) methylation-induced formation of an inactive chromatin structure (see Levine et al., 1991). The expression of the rat prolactin (rPRL) gene is primarily restrained to lactotrope and lactosomatotrope cell types of the pituitary gland. Consistently, the frequency of CpG dinucleotides in the rPRL promoter, i.e. 15 sites dispersed along 2 kb, is close to that observed in tissuespecific genes known to be regulated by methylation (Yeivin and Razin, 1993). In line, the rPRL gene exhibits coincident demethylated and transcriptionally active states in rat pituitary gland during pregnancy, lactation and also in estrogen-induced rat prolactinomas (Durrin et al., 1984; Kumar and Biswas, 1988; Zhang et al., 1989). Furthermore, similar correlations were shown in our laboratory using different tumor-derived rat cellular models (Laverribre et al., 1986). In GH,B6 cells, which secrete large amounts of rPRL and contain high levels of rPRL

0 1995 Elsevier Science Ireland Ltd. All rights reserved 0303-7207/95/$09.50 SSDI 0303-7207(94)03462-3

96

V.M. NgB et al. I Molecular and Cellular Endocrinology 108 (1995) 95-105

mRNA, the rPRL gene is hypomethylated. In contrast, in GH$DL cells selected by long-term culture of GH, cells in a medium supplemented with charcoal dextran-stripped serum, minute amounts of rPRL mRNA are found and the rPRL gene is hypermethylated. Moreover, in GH,CDL cells treated with 5-azacytidine (5azaC), a potent inhibitor of DNA methyltransferase activity, the rPRL gene was specifically and partially demethylated, and rPRL mRNA levels were strongly increased. Thus, evidence existed for a link between methylation and rPRL gene expression, but the mechanisms involved were still ignored. In contrast, a number of data have demonstrated that the rPRL gene is the ultimate target for tissue-specific and hormone-dependent transcription factors (see Gourdji and Laverribre, 1994). Among these factors, Pit- 1 or GHF- 1 is essential for the tissue-specific activation and expression of the rPRL gene (see Bodner et al., 1988; Mangalam et al., 1989; Rosenfeld, 1991). This transcription factor binds four related c&elements present in both the distal enhancer (sites 1D to 4D) and the proximal promoter regions of the rPRL gene (sites 1P to 4P) (Nelson et al., 1988). Additional factors have been demonstrated to act in synergy with Pit-l to ensure pituitary-specific expression of the rPRL gene, namely the estrogen receptor and the octamer binding protein Ott-1 (Day et al., 1990; Simmons et al., 1990; Voss et al., 1991). These proteins are co-expressed with Pit-l in lactotrope cells. Sitespecific deletions have also revealed the roles of two additional transcription factors, the lactotrope-specific factor LSF-1 and the basal transcription factor BTF (GutierrezHartmann et al., 1987; Harvey et al., 1991; Jackson et al., 1992). The responsive elements corresponding to these factors either overlap or are contiguous to Pit-l proximal sites. Concerning the regulated rPRL gene expression, Pit-l also appears to mediate the calcium-dependent pathway of thyrotropin-releasing hormone (TRH), a peptide stimulating rPRL gene expression (Laverrikre et al., 1983; Murdoch et al., 1983; Yan and Bancroft, 1991; Yan et al., 1991). In addition, the activation of the rPRL gene can be pharmacologically induced by activating the PKCand CAMP-dependent pathways. Linker-scanning analyses reveal that TRH, PKC- and CAMP-responsive elements closely overlap Pit-l proximal sites (Iverson et al., 1990). All these data thus point out the central role of PitI in pituitary-specific and in regulated rPRL gene expression (Gourdji and Laverri&re, 1994). In the present report we address the question of the respective roles of transcription factors and DNA methylation in the regulation of the rPRL gene. Differential expression of the rPRL gene could result from two nonexclusive mechanisms. The methylation status within the rPRL gene domain may modulate the binding or trunsactivating capacities of transcription factors. Alternatively, level of rPRL gene expression could be linked to modified levels of tissue-specific transcription factors, namely those of Pit- 1.

To answer this question, we first established a variant of GH$DL cell line selected after multiple 5-azaC treatments and containing stable rPRL mRNA levels, intermediate between GH&DL and GH3B6 levels. Taking advantage of these three cell lines, we have then compared the overall activity of truns-acting factors using transient expression assay with PRL-CAT constructs, and determined Pit-l levels by Northern blot and electrophoretie mobility shift assay. We did not find any appreciable difference in Pit-l levels that could be correlated to the dramatic variation in the endogenous rPRL gene expression. In contrast, in vitro CpG methylation of PRL-CAT constructs inhibited both the cell-specific and the ligandinduced activity of the rPRL promoter. Finally, in vivo competition assays demonstrated that such inhibition of the reporter gene resulted from the alterations in trunsactivating factors binding. 2. Materials and methods and oligonucleotides pPRL-CAT (Lufkin et al., 1989) was kindly provided by Dr. C. Bancroft (New York, NY). A mutated vector derived from pPRL-CAT, totally devoid of CAT activity, was prepared as detailed by Ng8 et al. (1993). In brief, the CAT coding sequence was altered by an insertional mutation performed by digestion of pPRL-CAT construct at the unique BspEI site. The resulting S-protruding ends were filled in with Klenow fragment of DNA polymerase I and the plasmid was recircularized using T4 DNA ligase. This vector designed for competition experiments was referred to as pPRL-(CAT*) competitor vector. The oligonucleotides containing the human 1P sequence (hlP) and the APl site of the human collagenase promoter as well as the expression vector pCMV$gal were kindly provided by Dr J.A. Martial (Peers et al., 1990). Pit-l partial cDNA probe and pSPT18 plasmid were kindly given by Dr M. Delhase (Brussels, Belgium) and Dr. J.A. Martial (Libge, Belgium), respectively. pCAT-Basic expression vector was purchased from Promega (France). All plasmids were prepared by alkaline lysis, purified by centrifugation in CsCYethidium bromide gradients, treated with RNase, centrifuged through a cushion of 1 M NaCl, extracted with phenol/chloroform and finally precipitated by ethanol (Sambrook et al., 1989). The concentrations of all plasmid preparations were determined by measuring the absorbance at 260 nm.

2.1. Plasmids

2.2. Cell lines and culture conditions The different cell lines used, i.e. GH3B6, GHFDL and GH3AZA3 were cloned or selected in our laboratory from the GH3 rat pituitary lactosomatotrope cell line. The GH3B6 cells (Gourdji et al., 1982) were routinely grown in Ham’s F-12 medium (Grand Island Biological CO., Grand Island, NY) supplemented with 15% heatinactivated horse serum (ATGC Biotechnologie, France)

V.M. NgGet al. I Molecularand Cellular Endocrinology108 (1995) 95-105

and 2.6% fetal calf serum (Grand Jsland Biological Co., Grand Island, NY). This medium was referred to as N medium. The GH&DL cell strain was selected from the GHs cell line (Brunet et al., 1977; Laverriere et al., 1986) and was routinely grown in Ham’s F-12 medium supplemented with 15% horse serum and 2.5% fetal calf serum, both of them twice extracted with charcoal-dextran, i.e. 75-90% depleted in total estrogens, and referred to as CD medium. For the experiments reported below, GH$DL cells were cultured in N medium. GH,AZA, cells were derived from the GHsCDL cell line submitted to three successive treatments with 5 PM 5-azaC (see Section 3); they were routinely maintained in N medium. All the GHs cell lines were used between their seventh and eighteenth subcultures and grown in monolayers at 37°C in a humidified atmosphere (95% air, 5% CO*). The AtT,s mouse corticotrope and C6 rat glial cell lines, which do not express the rPRL gene and are devoid of Pit-l transcription factor, were used as controls when indicated. For determination of the steady-state level of rPRL mRNA, 5 x lo5 cells were plated per 60-mm plastic tissue culture dish (Falcon Division, Benton and Dickinson and Co., Oxnard, CA) in 5 ml culture medium and were grown for 7 days, the medium being renewed on days 2 and 5. For transient expression assay, 2 million cells were plated per 75 cm2 Falcon flask in 10 ml culture medium and grown for 7 days, the medium being renewed on days 2 and 5. 2.3. 5AzaC treatments GHsCDL cells in the exponential growth phase were exposed to 5 PM 5-azaC (Sigma Chemical Co., St Louis, MO) for 48-60 h, a duration longer than one population doubling time (35-40 h). After the incubation period, the medium was replaced by fresh N or CD medium, renewed after 24 h and every 2 days thereafter. Treated and control cells were then trypsinized weekly and subcultured. The 5-azaC treatment was repeated up to sixfold, after -20 days delay. Control cultures were carried out in parallel throughout the experiments. Cells were rinsed twice with PBS, and total RNA was prepared for determination of rPRL and Pit-l mRNA levels after the durations indicated. 2.4. Dot and Northern blot assay of rPRL and Pit-l mRNA.

Total RNA was extracted essentially as described by Cathala et al. (1983) with the modifications previously reported (Laverribe et al., 1989). The concentration of total RNA was measured in triplicate by ethidium bromide fluorescence. For dot blot assay of rPRL mRNA, duplicate samples of RNA (1.5-5 pg) were denatured and immobilized on nitrocellulose filters as previously reported (Laverriere et al., 1989). Standard curves were obtained by including in each filter several dilutions (from -7 to 700 megacopies) of a unique solution of

91

rPRL RNA in vitro-transcribed from an 800-bp rPRL cDNA inserted in pSPT18 plasmid. For Northern blot assay of rPRL and Pit-l mRNA, IO-2Opg glyoxalated total RNA were separated by agarose gel electrophoresis and transferred onto Gene Screen membranes (NEN Research Products, Boston MA) using a vacuum blotting unit (Vacugene 2016, Pharmacia LKB Biotechnology, France). Hybridization was performed as described by Thomas (1980) with the 800 bp rPRL cDNA insert or with the 800 bp Pit-l cDNA insert labeled by nick translation with [a-32P]dATP and [a-32P]d’ITP (800 Wmmol, Amersham France) to a specific activity of 600-800 millions cprn&g (Rigby et al., 1977). After washing, the hybridized filters were submitted to autoradiographic analyses using Kodak X-Omat films with intensifying screens, and exposed for 4-24 h at -80°C. To determine the concentration of rPRL mRNA, dots were subsequently punched out and the bound radioactivity was measured by liquid scintillation counting. Results were then expressed in megacopies (1 Mcopy = lo6 copies) per microgram of total RNA. 2.5. Electroporation and analysis of gene reporter activities

For transfection assays, cells grown under routine conditions were harvested with trypsin/EDTA and resuspended in Ham’s F-12 medium at a concentration of 12 million cells per ml and mixed with purified plasmid DNA. Two to thirty micrograms of pPRL-CAT and 10 pg of pCMV-/?gal were used in all experiments, plus 2016Opg pPRL-(CAT*) competitor vector or pCAT-Basic vector where indicated. This suspension was exposed to a single pulse of 250 V/4-mm and 18OOpF capacitance with an Electropore 2000 apparatus (Eurogentec, Liege, Belgium). After electroporation, one volume of Ham’s F12 medium twofold concentrated in N or CD serum was added and transfected cells were transferred into 36-mm culture dishes in 3 ml N or CD medium. When exposed to regulatory factors, the transfected cells were transferred in a minimum serum-free medium (Ham’s F-12 medium containing 0.5 ng/ml parathyroid hormone, 30 nM sodium selenate, 5 pg/ml transferrin, lOpg/ml insulin, 0.1% bovine serum albumin) supplemented either with 60 nM TRH (Calbiochem, San Diego, CA) or 16 nM TPA (Sigma Chemical Co., St Louis, MO) plus 1 ,uM forskolin (Calbiochem, San Diego, CA) or with vehicle. Drugs were added as 1000X stocks in water or in dimethylsulfoxide. The medium was renewed 24 h later. Forty-eight hours after electroporation, the culture medium was removed, the cells were washed twice with phosphatebuffered saline and harvested by scrapping in 500~1 of phosphate-buffered saline containing 0.25 U/ml aprotinin, centrifuged at low speed and resuspended in 50 or 100~1 of 0.25 M Tris-HCl (pH 7.6) plus 0.25 U/ml aprotinin. The cells were then disrupted by three cycles of rapid freezing and thawing, centrifuged at 12 000 X g for

98

V.M. NgB et al. 1 Molecular and Cellular Endocrinology IO8 (1995) 95-105

10 min at 4°C and the levels of CAT and #I-galactosidase (B-gal) activities a nd protein contents were measured in the supernatants. CAT assays were performed in duplicate or in triplicate as described in Sambrook et al. (1989). The reaction mixture containing 3Op1 cell extract, unlabeled chloramphenicol and [3H]acetyl coenzyme A (3 Ci/mmol, Simersham France) was overlaid with waterimmiscible scintillation liquid (Econofluor-2, NEN Research Products, Boston MA). The tritiated acetylated forms of chloramphenicol produced by CAT activity, which were diffusing in the organic phase, were monitored in each sample by liquid scintillation at selected time intervals (every -30 min) for a total duration of 16 h. Data were automatically registered on a Macintosh microcomputer (Apple Computer, Inc., Cupertino, CA) using personal software, and treated by linear regression analysis. The kinetic data corresponding to the initial velocity only were taken into account. /I-Gal activity was determined in 2-~1 samples by measuring the optical density, at 420 nm, of the o-nitrophenyl-@-D-galactopyranoside hydrolysed after 20 min incubation at 37°C (Sambrook et al., 1989). CAT activities in untreated and treated cells originating from a single pool of transfected cells were always referred to p-gal activity measured in untreated cells. This overcomes the increase in b-gal activity elicited by TPA plus forskolin, likely conferred by the 5’ TGACGTCA 3’ palindromic sequence present in the cytomegalovirus enhancer. Protein content was measured using the Bio-Rad microassay procedure as recommended by the supplier (Bio-Rad, France). 2.6. In vitro DNA methylation of plasmids Methylation of pPRL-CAT construct or pPRL-(CAT*) competitor vector was carried out at 37°C for 2 h with 1 unit of M. SssI methylase per pg DNA under conditions recommended by the supplier (New England Biolabs, Ozyme, France). Mock-methylated DNA was prepared in parallel under the same conditions but with omission of methylase. The completeness of methylation was verified by digestion with HpaII and MspI (Boehringer Mannheim, France). DNA was purified by phenol/chloroform extractions and ethanol precipitation. 2.7. Electrophoretic mobility shif assays Nuclear extracts were prepared as described by Dignam et al. (1983). Double-strand oligonucleotides were end-labeled with [a-32P]dCTP and [a-32P]dGTP (800 Ci/ mmol, Amersham France) in the presence of the two other unlabeled deoxyribonucleotides by filling in Sprotruding ends with Klenow fragment of DNA polymerase I (Boehringer Mannheim, France). Variable amounts of nuclear extracts were incubated with 0.1 ng (3 fmol) of specific end-labeled oligonucleotide duplexes (20 000 cpm/O.l ng), 2pg bulk carrier DNA [poly(dIdC)] in a total volume of 17~1 containing 3OOpg/ml bovine serum albumin, 12% glycerol, 12 mM HEPES-

NaOH (pH 7.9), 4 mM Tris-HCl (pH 7.9), 60 mM KCl, 0.8 mM EDTA and 1 mM DTT. Binding reactions were performed for 4 h on ice. Electrophoresis through a nondenaturing gel (5% acrylamide/0.125% bisacryl&de) was performed in TBE buffer at 80 V during 1 h (Chodosh, 1992). The gel was then fixed in 20% acetic acid, 10% ethanol, dried at 80°C during 2 h and submitted to autoradiographic analysis using Kodak X-Omat films without screen at room temperature for 2 h. 2.8. Statistical analysis Each experiment was performed using duplicate or triplicate culture dishes and experiments were repeated two to four times. 3. Results We have previously shown that GH3CDL cells treated by the demethylating agent 5-azaC underwent an increase in hormone secretion (Laverriere et al., 1986). However, this increase was transient, and we first attempted to select a cell model displaying noticeable and sustained levels of rPRL gene expression. For this purpose, we tested two different culture media combined with increasing numbers of 5-azaC treatments. 3.1. 5-AzaC treatments of GH3CDL cells lead to a stable stimulation of rPRL gene expression in N medium GH3CDL cells were cultured in medium supplemented either with serum (N medium) or with charcoal dextranstripped serum (CD medium), and exposed to 5pM 5azaC once or several times at -20-day intervals, as described in Section 2. Total RNA was isolated at various times and the level of rPRL mRNA was monitored by dot blot analysis. Three successive treatments resulted in cumulative stimulations of rPRL mRNA levels (Fig. lA,B). When cells were grown in N medium, the maximum absolute increases in rPRL mRNA levels induced above control values were 25 + 2 Mcopies@g total RNA for one treatment, 40 + 1 Mcopies/pg total RNA for two treatments and 55 2 4 Mcopies/pg total RNA for three treatments (Fig. 1A). When cells were grown in CD medium, the maximum absolute increases in rPRL mRNA levels were not significantly different from those observed with N medium (Fig. 1B). However, changes in rPRL mRNA levels plateaued following the first, second or third treatment when cells were grown in N medium (Fig. lA), whereas rPRL mRNA levels decreased when cells were cultured in CD medium (Fig. 1B). Additional treatments, up to 6, did not induce further augmentation in rPRL mRNA levels (not illustrated). These results indicated that three 5-azaC treatments of GH,CDL cells grown in N medium were required to achieve maximum and SUStained increase in rPRL mRNA accumulation. The cell line obtained following this experimental schedule was referred to as GH3AZA3. These cells were thus grown in

V.M. NgB et al. I Molecular and Cellular Endocrinology 108 (1995) 95-105

A

.

steady-state level of rPRL mRNA (300-800 Mcopieslpg of total RNA), while GH3CDL cells contained minute

Normal medium

60

amounts of rPRL mRNA, i.e. 0.5-l -4% of that observed Aza2 “/’

I

I

I

.Azal -1/ 0

I 0

20

40

. 60

I

. 80

B 60

99

CD medium

I

.

1

20

0

.

I

40 Time (days)

.

I

60

.

80

c

(>

M

=’

+-

Prolactin

mRNA

Fig. 1. Analysis of rPRL mRNA in control and 5-azaC-treated cells. (A,B) Kinetic study of the effect of 5-azaC on rPRL mRNA levels in GH3CDL cells. Cells were treated one, two or three times with 5 PM 5azaC for 60 h (arrows) as detailed in Section 2, and subcultured in normal medium (N medium, A) or in charcoal dextran-stripped medium (CD medium, B) for the indicated durations. Levels of rPRL mRNA were quantified by dot blot analysis. Results are expressed as Mcopies of rPRL mRNA per pg total RNA and are the mean & SD of three independent determinations. Error bars, when not visible, are included in the symbols. (C) Northern blot analysis of rPRL mRNA in the three GH3 cell models. Ten microgram samples of total RNA isolated from GH$DL, GH3AZA3 or GH3B6 cells (lanes marked CDL, AZA3 and B6, respectively) were electrophoresed through 1% agarose gel, transferred onto Gene Screen membranes and hybridised with 32P-labelled rPRL cDNA probe as described in Section 2. The arrow denotes the position of the 1 kb rPRL mRNA band.

large amounts and frozen in aliquots for use later in the experiments reported here. Results of Northern blot analysis performed on RNA preparations isolated from GH,CDL, GH,AZA, and GH3B6 cells illustrate the drastic differences in rPRL mRNA levels (Fig. 1C). As quantified by dot blot analysis (not illustrated), the GH3B6 cell line exhibited the highest

in GH3B6 cells. GH3AZA3 cells displayed intermediate levels corresponding to 4-9.5% of the GH3B6 levels. These levels remained unchanged after several (up to 20) subcultures in N medium. This observation was consistent with a stable demethylation of the rPRL gene in GH3AZA3 cells. However, taking into account the overall effect of 6-azaC on gene demethylation, the increased expression of the rPRL gene might also result from an enhanced activity of genes required for tissue-specific expression, particularly those encoding transcription factors. 3.2. Transient expression of the PRL promoter-CAT construct In order to evaluate the possible consequences of 5azaC treatments on transcription factors, we performed transient expression assays using a rPRL promoter-driven CAT construct, bearing 2.0 kb upstream sequences of the rPRL gene and thus including the distal enhancer and the proximal promoter (pPRL-CAT). The PRL-CAT gene being intrinsically unmethylated, CAT activity reflected the overall efficiency of tram-acting factors regulating the rPRL promoter in the three cellular models. Cells were co-transfected with pPRL-CAT construct together with pCMV$gal plasmid as a control for transfection efficiency. After electroporation, cells were cultured in N or CD medium and transient expression of the reporter genes was estimated 48 h later. The PI&CAT gene was expressed to significant levels in the three GH3 cell lines and with the same pattern whatever the culture medium (Fig. 2). However, when cells were cultured in CD medium, CAT activity was twofold lower than that observed in N medium. Compared with rPRL mRNA levels, the I

0

N.D. AtT20

-I

CDL

AZA3

06

Fig. 2. Differential expression of the transfected CAT gene in the three GH3 cell lines. Twelve million cells (AtT20, GH$DL, GH3AZA3 or GH3B6) were co-transfected with 3Opg CAT reporter gene containing 2 kb upstream sequences of the rPRL gene (pPRL-CAT), together with 1Opg pCMV-pgal. After electroporation, cells were plated in triplicate and cultured either in CD medium (open boxes) or in N medium (black boxes) for 48 h. Cell extracts were then prepared and CAT and B-gal activities were assayed as described in Section 2. CAT activity was expressed relative to /Igal activity. Results are the mean f SD of three independent determinations. N.D.: not detectable.

V.M. NgB et al. I Molecularand Ceffular Endocrinology108 (1995) 95-l 05

100

CD

A

N

I-

4.2 kb --w 2.5 kb e

2.32 L

1.9 kb --_) 1.5 kb --_)

lane

123456789

Fig. 3. Analysis of Pit-l mRNA in GH3 cell lines. (A) Northern blot analysis of Pit-l mRNA in SazaC-treated GH3CDL cells. Total RNA originating from cells treated one, two or three times with 5 yM 5-aza-C and isolated at day 66 in the experiment illustrated in Fig. 1 were hybridised with 32Plabelled Pit-l cDNA. The cellular origin of RNA is indicated on the top, and brackets refer to the nature of the culture medium used. Lane M indicates migration of 1 DNA digested by BsrEII and end-labelled by exchange reaction using T4 polynucleotide kinase (Samhrook et al., 1989). The corresponding _ _ molecular weights are indicated in kb. (B) Nortbem blot analysis of Pit-I mRNA in AtT20, GH3CDL and GH3B6 cells grown in N medium. Arrows denote the positions of the four Pit-l mRNA species. ’

difference in CAT gene expression among the cell lines was noticeably reduced: in four independent experiments performed in N medium, relative CAT activity in GH3CDL cells reached 41 + 12%, while GH3AZA3 cells displayed 87 + 19% of that observed in control G&B6 cells. As expected, AtT,, mouse corticotrope cells transfected with pPRL-CAT displayed no CAT activity. Treatments with 5-azaC thus restored the activity of rPRL gene trans-acting factors to levels almost equivalent to that observed in GH3B6 cells. In view of such results, we asked if the expression of the gene encoding Pit-l was modified by 5-azaC treatments. 3.3. Analysis of Pit-l levels by Northern blot and electrophoretic

mobility shif assay

In a first step, we assessed the abundance of Pit-l mRNA in the three cell lines. This was conducted by Northern blot analysis performed using total RNA previously submitted to rPRL mRNA analysis illustrated in Fig. 1. In all the GH3 cell lines and whatever the number of 5-azaC treatments or the culture medium used, the Pit1 cDNA probe hybridized to similar extent with four mRNA species (4.2 kb, 2.5 kb, 1.9 kb and 1.5 kb), the 2.5 kb species being the most abundant (Fig. 3A). Consequently Pit-l gene expression appeared insensitive to 5azaC treatments. Moreover, the amount of Pit-l mRNA was rather lower in cells cultured in N medium, compared

to cells cultured in CD medium. Equivalent levels of Pit- 1 mRNA were detected in GH3B6 and GH3CDL cells. Thus, no change in Pit-l mRNA species could be correlated with the modulations in rPRL mRNA level illustrated in Fig. 1. As expected, no Pit-l mRNA was detected in ATu, cells (Fig. 3B). In order to define whether the level of Pit-l protein was correlated with the concentration of Pit-l mRNA, we performed electrophoretic mobility shift assay using a 36 bp probe containing the most proximal Pit-l response element from the human PRL gene (hlP, Fig. 4A). This sequence is able to bind the Pit-l protein with high specificity and strong affinity, and was thus used to monitor the concentration of the transcription factor (Peers et al., 1990). In presence of GH3CDL nuclear extract, several bands were visualised. The fastest band detected corresponded to the free probe, as illustrated in the absence of nuclear extract (Fig. 4B, lane 1). The more discrete and less retarded species marked Cl was identified as a complex formed between the labeled oligonucleotide and a monomer of Pit-l described by Voss et al. (1991) (lane 2). The second retarded species, marked C2 and migrating slower than Cl, was resolved into several discrete species on high-resolution gel (not illustrated). These bands likely corresponded to complexes between the hlP probe and heterodimers including Pit-l and Ott-1 (Voss et al., 1991). Increasing concentrations of GH3CDL nuclear

V. M. Nsi, et al. / Molecular

A

and Cellular Endocrinology

Human 1 P sequence

-60

-35

I

I

5’ CTAGAATGCCTGAATCATTATATTCATGAAGATATC

B

CDL

cell line

nuclear extract (pg) competitor

0

2

AZAB

6 16

2

_ -__-

3

B6

6 16 16

2

6 16 16

-_+---+

c2Cl -

free probe -----t lane

I

2

3

4

5

6

7

8

9 10 11 12

Fig. 4. Analysis of hlP binding activity among GH3 cell lines. (A) Sequence of the 36-bp probe corresponding to the 1P Pit-l site of the human prolactin gene. Only the sense strand is represented. (B) Electrophoretic mobility shift assay of the hlP sequence binding activity resent in nuclear extracts from GH$DL, GH3AZA3 and GH3B6 cells. R P-Labelled double-strand hlP probe (0.1 ng) was incubated with O16pg nuclear extracts isolated from GH$DL (lanes 24), GH3AZA3 (lanes 5-8) or GH3B6 cells (lanes 9-12). A XXI-fold excess of unlabelled hlP probe was added as competitor in lanes 8 and 12. The samples were submitted to polyacrylamide gel electrophoresis and proteinDNA complexes visualised by autoradiography as described in Section 2, Arrows indicate the free probe and Cl and C2 complexes described in Section 3. extract resulted in a decrease in the abundance of the Cl complex, with a parallel increase in the concentration of the C2 species (lanes 2-4), in accordance with data previously reported (Voss et al., 1991). Assays with nuclear extracts from GH3AZA3 and GH3B6 cells led to identical mobility shift patterns (Fig. 4B, lanes 5-7 and 9-11). Competition with a 500-fold excess of the unlabeled oligonucleotide resulted in a strong attenuation of the signal corresponding to the above described complexes. On the contrary, none of these complexes was detected when using nuclear extract from C6 rat glial cells, attesting to their specificity (not illustrated). The three GH, cell lines thus contained similar amounts of transcription factors displaying an equal efficiency to bind the Pit-l response element in vitro. Consequently Pit-l level unlikely accounted for the differential expression of the rPRL gene. We then used in vitro-metbylated DNA sequences to examine the possibility that DNA methylation could prevent binding or trans-activating properties of transcription factors, including Pit-l.

3.4. Expression of methylated reporter gene The first in vivo approach was to transfect cells with a pPRL-CAT construct fully methylated. Methylation was performed in vitro using M. SssI methylase. This enzyme methylates cytosines engaged in CpG dinucleotides, similar to eukaryotic DNA methyltransferases. As a control, cells were transfected with mock-methylated con-

101

108 (I 995) 95-105

struct. Like the native plasmid, mock-methylated PRLCAT gene was expressed to equivalent levels in GH3AZA3 and GH3B6 cells, but threefold less in GH3CDL cells (Fig. 5A). In contrast, CAT activity was close to background level when the PRL-CAT construct was fully methylated, whatever the cell line transfected. The effect of methylation on regulated rPRL gene expression was then tested. As shown in Fig. 5B, the expression of mock-methylated vector was increased by 1.2and 2.3-fold in GH3AZA3 cells exposed to TRH or TPA plus forskolin. This indicated that trans-activating factors mediating TRH, PKC- and CAMP-dependent regulations of the rPRL gene were present and active. However, when the methylated plasmid was used, these regulations were fully abolished (Fig. 5B). Similar results were obtained when using mock-methylated versus methylated PRL-CAT construct which contains the proximal rPRL promoter only, from positions -425 to +38 (pO.4kbPRLCAT, described by Ng8 et al,, 1993). In two independent experiments, each performed using the three cell lines, the expression levels of the methylated vectors were inhibited by 96.7 2 2.2% and 98.4 rt 0.8% (means f SD of six independent determinations) in control and TRHtreated cells, respectively. In vitro DNA methylation by M. &I methylase thus inhibited not only the tissuespecific but also the regulated expression of the transfected gene. 3.5. Does methylation alter in vivo binding capacity or cis-acting eficiency of rPRL gene regulatory sequences? To answer this question, we have applied an in vivo

A

B _ mock

methyl.

mock

methyl.

Fig. 5. Effect of CpG methylation on the transient expression of pPRLCAT construct. Twelve million cells were co-transfected with 3Opg mock-methylated (mock) or M. &I-methylated (methyl.) pPRL-CAT construct together with 1Opg of pCMV-j3gal. (A) Transient expression of mock-methylated and methylated pPRL-CAT construct in GHJCDL, GH,AZA, and GH3B,j cells cultured in N medium. CAT activity was expressed relative to p-gal activity. Results are the mean of triplicate f SD. Note that right side scale is lOO-fold lower than the left side scale. (B) Transient expression of mock-methylated and metbylated pPRL-CAT constructs in GH3AZA3 cells cultured in serum-free medium and untreated (Un.) or treated with TRH or TPA plus forskolin (T/F) (see Section 2). CAT activity was expressed relatively to @-gal activity obtained from transfected untreated cells. Note that the right side scale is 20-fold lower than the left side scale. Error bars, when not visible, arc included in the symbols.

V.M. NgB et al. I Molecular and Cellular Endocrinology 108 (1995) 95-IO5

102

B

Untreated 250 r

B .I 200 ‘: 1s

200

E

8 si 150 g 2 1 100

501

I

0

1

1

10

1

1

20

f

1

1

30

1

40

’

’

50

Fold excess competitor

$j

TRH 250 i-

TPA + forskolin 600 r 500 400

150 100

300

50

50

200 100

0

0

0

abed

-abed

abed

Fig. 6. Titration of trans-activating factors regulating rPRL promoter activity. (A) Dose-dependent effect of pPRL-(CAT*) competitor vector on transient expression of pPRL-CAT construct in GH3AZAy cells. Twelve million GH3AZA3 cells were transfected with 2pg pPRL-CAT expression vector together with increasing amounts pPRL-(CAT*) competitor vector (20-1OOpg). Total concentration of transfected DNA was kept constant by adding the required amount of pCAT-Basic expression vector. After electroporation, cells were plated in duplicate, either untreated (black squares) or treated with 50 nM TRH (open squares) or with 16 nM TPA plus 1yM forskolin (open circles) and cultured in serum-free medium for 48 h. Cells were then processed as described in Section 2. Data are the mean f range of two independent experiments performed in duplicate. In the main graph, the results are expressed as percent of relative CAT activity obtained from untreated cells transfected without pPRL-(CAT*) competitor vector (18.5 X lo3 + 0.2 x low3 for one experiment and 15.8 x low2 r 0.8 X 10e2 for the other). In the inset, results are expressed as percent of relative CAT activity obtained from the corresponding untreated and TRH- or TPA plus forskolin-treated cells transfected without pPRL-(CAT*) competitor vector. As a control, j3gal activities from the same experiments are illustrated in the lower part of the graph, showing the absence of inhibitory effect of the pPRL(CAT*) competitor vector. Results are expressed as percent of control values measured in the absence of pPRL-(CAT*) competitor vector (2.8 * 0.3 and 0.76 f 0.01 U/well respectively). Error bars, when not visible, are included in the symbols. (B) Titration of trans-activating factors by methylated pPRL-(CAT*) competitor vector in GH3AZA3 cells. GH3AZA3 cells were co-transfected with 4yg of pPRL-CAT expression vector plus 16Opg of the following constructs per 12 million cells: (a) pCAT-Basic expression vector; (b) native pPRL-(CAT*) competitor vector; (c) M. &I-methylated pPRL-(CAT*) competitor vector; (d) mock-methylated pPRL-(CAT*) competitor vector. After electroporation, cells were. plated in duplicate, either untreated (black boxes) or treated with 50 nM TRH (open boxes) or with 16 nM TPA plus 1yM forskolin (grey boxes) and cultured in serum-free medium for 48 h. Cells were then processed as described in Section 2. Results are expressed as percent of relative CAT activity obtained from untreated cells transfected with pCAT-Basic (0.358 k 0.012). Results are the mean of duplicate * range. Error bars, when not visible, are included in the symbols.

competition assay (Ng6 et al., 1993): GHD3AZA3 cells were first co-transfected with a fixed amount of native pPRL-CAT construct and increasing quantities of pPRL(CAT*), a mutated vector virtually identical to the expression vector but totally devoid of CAT activity (see Section 2). The competitor DNA sequences of this vector possess a supercoiled structure, and were thus equivalent to those present in the expression vector and fully competent to titrate transcription factors. The total concentration of transfected DNA was kept constant by adding the required amount of pCAT-Basic expression vector. As in the preceding experiments, pCMV$gal was included in each transfection assay and results were referred to Pgal activity. Following electroporation, cells were either untreated or treated with TRH or TPA plus forskolin in order to compare basal and regulated expression. As shown in Fig. 6A (upper part), increasing excess pPRL-(CAT*) competitor vector led to a dose-dependent decrease in the

relative CAT activity in untreated cells. A lo-fold excess was sufficient to exert a significant inhibition, while a 50fold excess (the maxima1 concentration tested) resulted in a 61 + 16% decrease. When results were expressed relatively to non-competed values for untreated cells, TRH and TPA plus forskolin treatments yielded 131 f 4% and 264 f 32% stimulation, respectively, in the absence of competitor DNA. Increasing concentrations of pPRL(CAT*) competitor vector resulted in a progressive inhibition of relative CAT activity, reaching 36 4 12% and 32 + 8% of the reference level in TRH- and TPA plus forskolin-treated cells, respectively. In order to compare the absolute efficiency of the pPRL-(CAT*) competitor vector in untreated and treated groups, the data were also expressed relative to the non-competed value measured for each treatment. As shown in the inset of Fig. 6A, the resulting competition curves were superimposed. This provided clear-cut evidence that the competitor vector

V.U. Ngd et al. I Molecular and Cellular Endocrinology 108 (1995) 95-105

bound the trns-acting factors involved in constitutive as well as in regulated rPRL gene expression with equal efficiency. In contrast, the expression levels of pgal reporter gene were not decreased (Fig. 6A, lower part) indicating that general transcription machinery was not affected, even by the highest concentration of pPRL(CAT*) competitor vector. This ruled out the hypothesis that repression of CAT reporter gene could result from titration of general transcription factors. We then used this assay to determine the consequence of DNA methylation on the binding of truns-acting factors involved in both types of regulation. The pPRL(CAT*) competitor vector was thus methylated using M. SssI methylase and its efficiency as a competitor was compared with that of its mock-methylated counterpart. The assay was performed in GHsAZAs cells untreated or treated with TRH or TPA plus forskolin. In untreated cells, the highest relative CAT activity was observed in the presence of a 40-fold excess methylated pPRL(CAT*) competitor vector (Fig. 6B, group c). In the presence of the same excess of mock-methylated mutated vector (group d), this level was inhibited by -50%. In cells treated either with TRH or with TPA plus forskolin, the mock-methylated pPRL-(CAT*) competitor vector induced a similar range of inhibition. As a control, experiments were performed either with the same excess pCAT-Basic vector, i.e. devoid of rPRL promoter (group a), or with the native mutated vector (group b). Strikingly, the values were consistently lower than in presence of the methylated pPRL-(CAT*) competitor vector. Nevertheless and in agreement with data illustrated in Fig. 6A, relative CAT activity was significantly decreased by 33% in the presence of excess native competitor vector in untreated cells, and by 43% and 38% in TRH- and TPA plus forskolin-treated cells, respectively (group a versus b). Similar results were obtained in additional independent experiments using the pOAkbPRL-CAT construct: in the presence of a 40-fold excess competitor vector, the relative CAT activities were inhibited by 73 + 3%, 70 rt 6% and 68 + 10% in control, TRH- and TPA plus forskolintreated cells, respectively (mean + SD of three independent experiments). Transfection assays with the p0.4kbPRL-CAT expression vector and mock-methylated or methylated counterparts confirmed that methylation prevented the decrease in CAT expression level induced by the competitor vector. In the presence of methylated competitor vector, the levels of relative CAT activity were 2.5 +0.4-fold, 3.3 &0.6-fold and 3.7 +0.9-fold higher than in the presence of mock-methylated competitor vector in control, TRH- and TPA plus forskolintreated cells, respectively (mean + range of two independent experiments). Taken together, these results showed that methylation of DNA sequences drastically altered the ability of the entire rPRL promoter, as well as that of the proximal promoter, to bind and titrate transcription factors in vivo.

103

4. Discussion In an attempt to understand the interplay between DNA methylation and trans-acting factors in rPRL gene expression, we have investigated the level of the pituitary-specific transcription factor Pit-l and the role of CpG methylation on rPRL promoter activity. The results presented here show that the drastic difference in rPRL gene expression observed in three pituitary cell lines do not correlate with concomitant modifications in Pit-l gene expression. Identical species of Pit-l mRNA are indeed detected at similar significant levels whatever the cell line studied. The presence of the transcription factor as well as its binding ability are further attested by gel shift analysis using the most proximal Pit-l response element. The mobility shift patterns observed are consistent with previous reports (Voss et al., 1991) and strongly suggest that the ubiquitous factor Ott-1 is functional and that the capacity of Pit-l and Ott-1 to form homodimeric and heterodimeric complexes is preserved in GH&DL cells that express the rPRL gene to very low levels. Moreover, GHsCDL cells are able to direct efficient expression of a transfected PRL-CAT construct, thus demonstrating that tissue-specific transcription factors are functional in these cells. Our results agree with those previously reported by Arnold et al. (1991) with PRL-deficient GHsB, cells, but they differ from those observed in pituitary x fibroblast hybrids in which rPRL gene extinction results from the repression of Pit-l gene (Supowit et al., 1992). They also differ from those recently reported using in vivo transplanted GH3 cells in which the specific repression of the rPRL gene was correlated with a decrease in in vitro binding activity of Pit-l (Day and Day, 1994). Taking into account the primary role of Pit-l in pituitary development, the noticeable level of expression measured in GHsCDL cells strongly suggests that these cells remain committed to the lactosomatotrope lineage. In keeping with this hypothesis, 5-azaC treatments, when restoring a high level of rPRL expression, unravel a preexisting pattern of differentiation. The stimulatory effect of the demethylating agent is nevertheless transient when cells are grown in CD medium, despite the presence of Pit-l. This indicates that additional factors present in N medium, are required for stabilizing the 5-azaC-induced increase in rPRL gene expression. Estradiol may be one of these factors, given the synergistic roles of the estrogen receptor together with Pit-l in the establishment of the lactotrope phenotype during anterior pituitary development (Day et al., 1990; Simmons et al., 1990). These results are also consistent with the procedure by which GHsCDL cells were selected (see Section 1). Furthermore, the involvement of estradiol in the establishment of a demethylated and active state of the chicken vitellogenin gene reinforces this hypothesis (Jost et al., 1991; see Jost and Saluz, 1993). However, GHsAZAs cells display an increased ability to direct the expression of trans-

104

KM Ngd et al. I Molecular and Cellular Endocrinology IO8 (1995) 95405

fected PRL-CAT construct compared to GHsCDL cells. In view of the equivalent levels of Pit-l gene expression, additional tram-acting factor(s) are likely reactivated by 5-azaC treatments. LSF-1 and BTF, which are thought to be involved in the rPRL gene expression, constitute potential candidates that could be activated either directly or indirectly (Gutierrez-Hartmann et al., 1987; Harvey et al., 1991; Jackson et al, 1992). The stimulatory effect of 5 azaC may thus result from several mechanisms, including the demethylation of genes encoding these factors. It is significant that change in transcription factor activity would occur together with the 5-azaC-induced demethylation of the rPRL gene established in previous experiments (Laverriere et al., 1986). Consequently, in addition to the increase in transcriptional activity revealed by transfection assay, demethylation might further enhance the level of rPRL gene expression by extending the accessibility of the promoter to transcription factors and specifically to Pit-l. This attractive hypothesis is supported by the fact that the level of rPRL mRNA is 6-10fold lower in GH,CDL cells than in GHsAZAs cells whereas the exogenous PRL-CAT gene is expressed at a 2-fold lower level only. Furthermore, in vitro methylation of the PRL-CAT construct yields full abolition of CAT expression and impairs TRH, PKC- and CAMP-dependent regulation of the rPRL gene. This suggests that DNA methylation also precludes interaction of nuclear factors mediating these regulations, including Pit-l (see Section 1). Clear-cut evidence for such a mechanism is provided by in vivo competition experiments: when in vitromethylated, the pPRL-(CAT*) competitor vector does not inhibit either constitutive or regulated PRL-CAT gene expression, indicating that full CpG methylation prevents the binding of transcription factors to their &elements. Whether the effect of DNA methylation on transcription factor binding was direct or not remained to be elucidated. According to the published sequence of the rPRL promoter (see references in Gourdji and Laverriere, 1994), none of the Pit-l cis-elements contains a CpG dinucleotide in the core of the binding site. Moreover, we have recently observed by genomic sequencing that the CpG located at position -38, adjacent to the 1P site, was replaced by a CpT dinucleotide in the three GHs cell lines used in this study. This suggested that repression of Pit-l binding occurred indirectly and involved methyl-CpG binding proteins, namely MeCP-1 and MeCP-2 (Meehan et al., 1989; 1992; Lewis et al., 1992). It is tempting to propose that related mechanisms are involved in the repression of the resident rPRL gene in GHsCDL cells. However, the expression of the methylated rPRL gene in GHsCDL cells differs from that of the in vitro methylated PRL-CAT gene in two aspects. First, the endogenous gene is expressed at low but significant levels whereas the methylated exogenous gene is totally silent. In addition, the rPRL gene in GH$DL cells is strongly regulated by PKC- and PKA-dependent path-

ways (results not shown) whereas these regulations are fully abolished by in vitro methylation of the PRL-CAT construct. These differences may be attributed to a lower number of methyl-CpG sites in the endogenous gene in comparison to the ectopic gene. Kass et al. (1993) have indeed recently shown that inhibition of expression vector by in vitro methylation drastically depends upon the number of methylated CpG sites rather than upon their location inside or outside the promoter region. The high density of CpG dinucleotides in the vector sequence of the PRL-CAT construct (8-fold the rPRL promoter content) may account for the strong inhibition induced by in vitro methylation. Alternatively, the methylation which happens spontaneously within the endogenous rPRL gene namely in GHsCDL cells may concern a limited set of CpG sites. Depending on the extent of methylation, the gene would be accurately modulated from null expression level to full transcriptional activity. This could provide a way to regulate rPRL gene expression independently of the differentiation pattern specified by the presence of the tissue-specific transcription factor Pit- 1. Acknowledgements The authors wish to thank Dr. C. Bancroft (New York) and Dr. M. Delhase (Brussels, Belgium) for their generous gifts of the pPRL-CAT expression vector and the Pit1 cDNA containing plasmid, respectively. The authors are greatly indebted to Dr. J.A. Martial (Liege, Belgium) and to Dr. A. Belayew (Leuven, Belgium) for helpful discussions and for providing numerous tools used in this study. The authors would like to acknowledge the advice of Dr. A. Morin (GifTYvette, France) for electroporating the cells and Dr. E. Vila-Porcile for her critical reading of the manuscript. The authors also thank Mr. C. Pennarun for his skilful assistance in the preparation of the illustrations, and Dr. G. Vassent for designing the hardware connection and computer software used in the CAT assay. This work was supported by the Centre National de la Recherche Scientifique (URA 1115). V.N. was supported by a fellowship from the Minis&e de la Recherche et de 1’Enseignement Superieur. References Arnold, T.E., Farrance, I.K., Morris, J. and lvarie, R. (1991) DNA Cell Biol. 10, 105-112. Bird, A.P. (1986) Nature 321,209-213. Bodner, M., Castrillo, J.L., Theill, L.E., Deerinck, T., Ellisman, M. and Karin, M. (1988) Cell 55.505-5 18. Brunet, N., Gourdji, D., Moreau, M. F.. Grouselle, D., Boumaud,F. and Tixier-Vidal, A. (1977) Ann. Biol. Anim. Biochim. Biophys. 17.413-424. Cathala, G., Sabouret, J.F., Mendez, B., West, B.L., Karin, M., Martial, J.A. and Baxter, J.D. (1983) DNA 2,327-333. Cedar, H. (1988). Cell 53.3-t. Chodosh, L.A. (1992) in Current Protocols in Molecular Biology (A~subd, F.M., Brent, R., Kingston, R.E., Moore, D.D., Seidman,

V.M. Ng6 et al. I Molecular and Cellular Endacrinalogy 108 (I9951 95-105

J-G., Smith, J.A. and Struhl, K., eds.), GreenlWiley-Interscience, New York. Day, R.N. and Day, K.H. (1994) Mol. Cell. Endocrinol. 8, 12-20. Day, R.N., Koike, S., Sakai, M., Muramatsu, M. and Maurer, R.A. (1990) Mol. Cell. Endocrinol. 4, 1964-1971. Dignam, J.D., Martin, P.L., Shastry, B.S. and Roeder, R.G. (1983) Methods Enzymol. 101,582-598. Doertler, W. (1983) Annu. Rev. Biochem. 52.93-124. Durrin. L.K., Weber, J.L. and Gorski, J. (1984) J. Biol. Chem. 259, 7086-7093. Gourdji, D. and Laverribre, J.N. (1994) Mol. Cell. Endocrinol. 100, 133-142. Gourdji, D.. Tougard, C. and Tixier-Vidal. A. (1982) in Frontiers in Neuroendocrinology, Vol. 7 (W.F. Ganong and L Martini, eds.), pp. 317-357, Raven Press, New York. Gutierrez-Hartmann. A., Siddiqui, S. and Loukin, S. (1987) Proc. Natl. Acad. Sci. USA 84,5211-5215. Harvey, C., Jackson, S., Siddiqui, SK. and Gutierrez-Hartmann, A. (1991) Mol. Cell. Endocrinol. 5,836843. Holliday, R. (1993) in: DNA Methylation: Molecular Biology and Biological Significance (J.P. Jost and H.P. Saluz, eds.), pp. 452468, Birkhauser Verlag, Basel. Iverson, R.A., Day, K., d’Emden, M., Day, R.N. and Maurer, R.A. (1990) Mol. Cell. Endocrinol. 4, 1564- 1571. Jackson,S.M., Keech, C.A., Williamson, D.J. and Gutierrez-Hartmann, A. (1992) Mol. Cell. Biol. 12, 27023-27 19. lost. J.P. and Saluz, H.P. (1993) in DNA Methylation: Molecular Biology and Biological Significance (J.P. Jest and H.P. Saluz, eds.), pp. 425451, Birkhiiuser Verlag, Basel. Jost, J.P., Saluz, H.P. and Paw& A. (1991) Nucleic Acids Res. 19, 5111-5775. Kass, S.U., Goddard, J.P. and Adams, R.L.P. (1993) Mol. Cell. Biol. 13.7372-7379. Kumar, V. and Biswas, D.K. (1988) J. Biol. Chem. 263,12645-12652. Laverriere, J.N., Morin, A., Tixier-Vidal, A., Truong, A.T., Gourdji, D. and Martial, J.A. (1983) EMBO J. 2.149331499. Laverribre, J.N., Mullen M., Buisson, N., Tougard. C., Tixier-Vidal, A., Martial, J.A. and Gourdji, D. (1986) Endocrinology 118,198-206. I-aver&e, J.N., Richard, J.L., Buisson, N., Martial, J-A., Tixier-Vidal, A. and Gourdji, D. (1989) Neuroendocrinology 50.693-701. Levine, A., Cantoni, G.L. and Razin, A. (1991) Proc. Natl. Acad. Sci. USA 88.6515-6518. Lewis, J.D., Meenan, R.R., Henzel, W.I., Maurer-Fogy, I., Jeppesen, P., Klein, F. and Bird, A. (1992) Cell 69,905-914. Lufkin, T., Jackson, A.E., Pan Wayne, T. and Bancroft, C. (1989) Mol. Cell. Endocrinol. 3, 559-566.

105

Mangalam, H.J., Albert, V.R., Ingraham, H.A., Kapiloff, M., Wilson, L., Nelson, C., Elsholtz, H.P. and Rosenfeld, M.G. (1989) Genes Dev. 3.946-958. Meehan, R.R., Lewis, J.D. and Bird, A.P. (1992) Nucleic Acids Res. 20.5085-5092. Meehan, R.R., Lewis, J.D., McKay, S., Kleiner, E.L. and Bird, A.P. (1989) Cell 58,499-507. Murdoch, G.H., France, R., Evans, R.M. and Rosenfeld, M.G. (1983) J. Biol. Chem. 258.15329-15335. Nelson, C.. Albert, V.R., Elsholtz, H.P., Lu Li, W. and Rosenfeld, M.G. (1988) Science 239, 1400-1405. Ngo, V., Laverriere, J.N. and Gourdji, D. (1993) Nucleic Acids Res. 21, 5795-5796. Peers, B.. Voz, M.L., Monget, P., Mathy-Hartert, M., Bewaer, M., Belayew, A. and Martial, J.A. (1990) Mol. Cell. Biol. 10, 46904700. Razin, A. and Cedar, H. (1991) Microbial. Rev. 55,451458. Rigby, P.W., Dieckmann, M., Rhodes, C. and Berg, P. (1977) J. Mol. Biol. 113,237-261. Rosenfeld. M.G. (1991) Genes Dev. 5,897-907. Sambrook, J., Fritsch, E.F. and Maniatis, T. (1989) Molecular Cloning, A Laboratory Manual. Cold Spring Harbor Laboratory, Cold Spring Harbor, NY. Shemer, R., Walsh, A.M., Eisenberg, S., Breslow, J.L. and Razin, A. (1990) 1. Biol. Chem. 265, lOlO-1015. Simmons, D.M., Voss, J.W.. Ingraham, H.A., Holloway, J.M., Broide, R.S.. Rosenfeld, M.G. and Swanson, L.W. (1990) Genes Dev. 4, 695-7 11, Supowit, SC., Ramsey, T. and Thompson, E.B. (1992) Mol. Cell. Endocrinol. 6,786792. Thomas, P.S. (1980) Proc. Natl. Acad. Sci. USA 77,5201-5205. VOSS, J.W., Wilson, L. and Rosenfeld, R.M.G. (1991) Genes Dev. 5, 1309-1320. Waalwijk, C. and Flavell, R.A. (1978) Nucleic Acids Res. 5, 32313236. Yan. G.-Z. and Bancroft, C. (1991) Mol. Cell. Endocrinol. 5, 14881497. Yan. G-Z., Pan, W.T. and Bancroft, C. (1991) Mol. Cell. Endocrinol. 5, 535-541. Yieivin. A. and Razin, A. (1993) in DNA Methylation: Molecular Biology and Biological Significance (J.P. Jost and H.P. Saluz, eds.), pp. 523-568, Birkhiauser Verlag, Basel. Zhang, Z.X., Kumar, V., Riveral, R.T., Pasion, S.C., Chilshom, J. and Biswas, D.K. (1989) DNA 8.605-613.