Chromatin structure, transcription, and methylation of the prolactin gene domain in pituitary tumors of Fischer 344 rats.

In order to study prolactin gene transcription and chromatin structure in a relatively homogeneous population of prolactin-producing cells, we induced lactotroph proliferation (pituitary tumor formation) in Fischer 344 rats by chronic diethylstibestrol treatment. We show that the prolactin gene is highly sensitive to digestion by DNase I in the chromatin of these pituitary tumors but not in liver. Furthermore, sequences immediately flanking the prolactin gene exhibit a similar high DNase I sensitivity. This contrasts with previous reports that the coding region of some genes is highly sensitive to DNase I digestion and that adjacent noncoding regions exhibit an intermediate sensitivity. This high level of DNase I sensitivity in the flanking regions was not due to transcription; we did not detect RNA transcripts homologous to any unique DNA sequence within 12 kilobase pairs upstream or downstream of the prolactin gene. We have identified two DNase I-hypersensitive sites 5' to the prolactin gene in pituitary tumors but none in liver. Also, the coding region, but not adjacent noncoding regions, of the prolactin gene domain is hypomethylated at MspI/HpaII and HhaI restriction enzyme sites in pituitary tumors. The prolactin gene domain is methylated at these sites in liver.

In order to study prolactin gene transcription and chromatin structure in a relatively homogeneous population of prolactin-producing cells, we induced lactotroph proliferation (pituitary tumor formation) in Fischer 344 rats by chronic diethylstilbestrol treatment. We show that the prolactin gene is highly sensitive to digestion by DNase I in the chromatin of these pituitary tumors but not in liver. Furthermore, sequences immediately flanking the prolactin gene exhibit a similar high DNase I sensitivity. This contrasts with previous reports that the coding region of some genes is highly sensitive to DNase I digestion and that adjacent noncoding regions exhibit an intermediate sensitivity. This high level of DNase I sensitivity in the flanking regions was not due to transcription; we did not detect RNA transcripts homologous to any unique DNA sequence within 12 kilobase pairs upstream or downstream of the prolactin gene. We have identified two DNase I-hypersensitive sites 5' to the prolactin gene in pituitary tumors but none in liver. Also, the coding region, but not adjacent noncoding regions, of the prolactin gene domain is hypomethylated at MspIIHpaII and HhaI restriction enzyme sites in pituitary tumors. The prolactin gene domain is methylated at these sites in liver.
Estrogen increases prolactin mRNA levels (1, 2) and the rate of transcription of the prolactin gene (3,4) in rat anterior pituitaries. It has been hypothesized that estrogen modulates gene activity through a specific nuclear localization of the estrogen receptor. Interaction of the estrogen-receptor complex with a specific DNA sequence has been suggested as a possible mechanism; but as most hormone-inducible genes are transcribed only in one or a few tissues out of the many with functional estrogen receptors, other factors also must be involved. It has been suggested that certain "developmentally significant" events must occur for hormonal controls to function ( 5 ) .
Several structural features of chromatin have been associated with transcriptionally active genes. These include a general DNase I sensitivity of the chromatin gene domain (6,7), DNase I-hypersensitive sites located near regulatory regions (8), hypomethylation of cytosine residues (9, lo), and * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  binding of high mobility group proteins 14 and 17 to transcribed gene regions (11). Reports from two research groups have indicated that coding regions of the a-and @-globin genes (12,131 and the ovalbumin gene (14) of the chicken are highly DNase I-sensitive. Adjacent noncoding regions (or previously transcribed embryonic genes) were reported to display an intermediate sensitivity to DNase I digestion. More recently, it has been reported that coding and immediately adjacent nontranscribed DNA sequences exhibit the same high level of sensitivity to DNase I digestion (15-17).
Knowledge of the chromatin structure of the prolactin gene will be important in understanding how this gene acquires the capability to respond to estrogen and the specific mechanisms of this response. In this report, we examine the tissue specificity of the DNase I sensitivity of the prolactin gene chromatin in diethylstilbestrol-treated rats of the Fischer 344 strain (18). We use limited nicking with DNase I coupled with filter blot hybridization to examine the relative DNase I sensitivity of the coding and noncoding regions of 35 kilobase pairs of the prolactin gene domain in pituitary tumors and liver. We also identify DNase I-hypersensitive sites near the prolactin gene, quantitate transcription from the prolactin gene domain, and examine the tissue-specific methylation pattern of the prolactin gene.

MATERIALS AND METHODS
Anim~ls"F344~ male rats were obtained from Harlan-Sprague Dawley, Madison, WI. These rats were chronically treated with diethylstilbestrol by implanting 1-cm sections of silastic tubing containing 5 mg of DES under the skin of 3-week-old rats (19). The rats were then maintained on a 12-h light/l2-h dark cycle with continuous access to rat chow and water.
Source of Cloned DNA-Approximately 37 kb of cloned prolactin genomic DNA was provided by Dr. Brad Thompson, National Institutes of Health (20). Serum albumin cDNA was provided by Dr. Richard Hanson, Case Western University (21).
Preparation and Digestion of Nuclei"F344 rats, treated with DES for 3 months (unless specified differently in the text), were anesthetized and decapitated. Pituitary tumors were removed and immediately placed in ice-cold homogenization buffer (50 mM Tris-HC1, pH 7.5, 20 mM KC1, 5 mM MgClz, 0.5 M sucrose, 0.15 mM spermine, and 0.5 mM spermidine). All subsequent steps were done at 4 "C. The pituitaries were homogenized by 10 strokes, at medium speed, of a motor-driven Teflon-glass homogenizer. The homogenate was layered onto a 0.88 M sucrose cushion in homogenization buffer and centrifuged 5 min at 5000 rpm in an HB-4 rotor (Sorvall). The pelleted nuclei were resuspended in homogenization buffer and again pelleted through a 0.88 M sucrose cushion as above. Liver nuclei were prepared as above except they were homogenized again before pelleting through the second sucrose cushion.
Nuclei were resuspended in DNase I digestion buffer (6)  ton, DPFF) were added to identical aliquots of nuclei, and the mixture was incubated at 37 "C for 10 min. Digestion was stopped by adding EDTA to a final concentration of 10 mM and mixing vigorously.
Restrictwn Endonuclease Cleauage of DNA, Transfer to Nitrocellulose, and Hybridization-DNA was purified from the nuclei as previously described (22). Aliquots of purified DNA (30 pg) were digested to completion with an excess of restriction endonuclease according to the procedures recommended by the manufacturers (Bethesda Research Laboratories or New England Biolabs). After digestion, the DNA was fractionated by agarose gel electrophoresis in TEA buffer (40 mM Tris-HC1, 55 mM sodium acetate, 20 mM EDTA, pH 8.3). DNA was transferred from the gel to nitrocellulose filters (Schleicher & Schuell, BA85) as described (23). After overnight transfer, the filters were blotted dry and baked 2 h at 80 "C. All hybridization and wash steps were done at 68 "C (unless specified) as described previously (24). Briefly, the filters were washed 30 min with 3 X SSC (1 X SSC is 0.15 M NaCl and 0.015 M trisodium citrate, pH 7.0). Next the filters were treated 3 h with 3 X SSC and 0.1% polyvinylpyrrolidone, 0.1% Ficoll, and 0.1% bovine serum albumin (modification of the procedure of Denhardt (25)). This solution was replaced with a similar solution that contained, in addition, 20 mM NaH2P04, 0.1% SDS, and 50 pg/ml of denatured herring sperm DNA. The incubation was continued for 2 h. This solution was in turn replaced with a minimal volume of identical solution plus 1 X lo6 cpm/ml of 3ZP-labeled DNA.
Hybridization was continued 20-36 h. The nitrocellulose filters were washed three times with hybridization solution minus the labeled probe at room temperature and 30 min at 68 "C. A final 30-min wash was with 0.1 X SSC, 0.1% SDS at 68 "C. Filters were rinsed at room temperature with 3 X SSC and blotted dry. The dried filters were autoradiographed by exposing them to Kodak XAR-5 film for approximately 20 h at -80 "C with a DuPont Cronex Lightning Plus intensifying screen.
Preparation and Labeling of Hybridization Probes-Fragments of DNA spanning the prolactin gene domain have been subcloned into the plasmid pBR322.' Plasmid DNA containing different prolactin gene fragments was isolated (26), and the prolactin DNA was excised from the plasmid by restriction endonuclease cleavage. Prolactin DNA fragments were purified from plasmid DNA by preparative agarose gel electrophoresis (27). Isolated DNA fragments were labeled to specific activities of 1-6 X 10' cpmlpg of DNA by nick translation (28) using [a-3ZP]dTTP and [(u-~'P]~CTP (New England Nuclear, 600 Ci/mmol).
Isolation of RNA-Nuclear and cytoplasmic RNA was prepared as described (29) from pituitary tumors and liver. First, nuclei were prepared as described above and the cytoplasmic fraction was saved. Guanidine thiocyanate (Kodak, 4 M in 50 mM Tris-HC1, pH 7.5, 10 mM EDTA, 2% Sarkosyl, 1% 2-mercaptoethanol) was added to the nuclei, and DNA was sheared by forcing the solution through a small gauge needle four to five times. One gram of CsCl was added per 2.5 ml of suspension. This solution was layered onto a cushion of 5.7 M CsCl in 100 mM EDTA in an SW 41 (Beckman) centrifuge tube.
After 14 h of centrifugation at 32,000 rpm at 20 "C, the supernatant was discarded, walls of the tubes were dried, and the RNA pellet was resuspended in 10 mM Tris-HC1, pH 7.5, 5 mM EDTA, 1% SDS. Resuspended RNA was extracted once with a mixture of chloroform:butanol (4:l). The organic phase was re-extracted once with the resuspension buffer, and the combined aqueous phases were precipitated overnight at -20 "C after addition of 0.1 volume of 2.5 M sodium acetate and 2 volumes of absolute ethanol.
Cytoplasmic RNA was prepared similarly. Two volumes of a 6 M stock of guanidine thiocyanate in 75 mM Tris-HCI, pH 7.5, 15 mM EDTA, 3% Sarkosyl, 1.5% 2-mercaptoethanol were added to 1 volume of the cytoplasmic fraction, and the solution was treated as above.
Preparation of RNA Dot Blots and Hybridization-Dot blots were prepared by modification of a method described previously (30). RNA was diluted to various extents with TE buffer (10 mM Tris-HC1, pH 7.5,l mM EDTA) and mixed with Escherichia coli tRNA to a uniform final concentration of nucleic acid. The RNA was denatured by heating to 65 "C for 5 min and immediately placed on ice for 5 min. Two microliters of each dilution were spotted next to each other onto nitrocellulose filters that had previously been wetted with water, equilibrated in 20 X SSC, and air-dried. A DNA standard curve was prepared by mixing known amounts of unlabeled cloned DNA, homologous to the labeled hybridization probe, with E. coli tRNA to a to Mol. Cell. Endow.
Weber, J. L., Durrin, L. K., and Gorski, J., manuscript submitted final concentration of nucleic acid identical to the amount in the RNA samples. The DNAtRNA mixture was denatured at 100 "C for 10 min and immediately placed on ice for 5 min. DNA dilutions were spotted next to each other onto the same nitrocellulose filters containing the liver and pituitary tumor RNAs. The filters were air-dried and baked 2 h at 80 "C. Filters were prehybridized and hybridized according to a modification of a previously published procedure (31). Briefly, the filters were treated 2 h at 42 "C with 50% formamide (deionized), 5 X SSC, 0.1% SDS, 20 mM NaHzP04, pH 7.0, 250 pg/ ml of denatured herring sperm DNA, and 0.1% polyvinylpyrrolidone, 0.1% Ficoll, and 0.1% BSA. Hybridizations were done in fresh prehybridization buffer plus 1 X lo6 cpm/ml of nick-translated DNA at 42 "C for 36 h. The filters were washed with 3 X SSC, 0.1% SDS three times at room temperature and once for 30 min at 50 "C. A final stringent wash was with 0.1 X SSC, 0.1% SDS at 50 "C for 30 min. The filters were rinsed with 5 X SSC at room temperature, airdried, and exposed to Kodak XAR-5 film for 3-24 h at -80 "C with a DuPont Cronex Lightning Plus intensifying screen.
Quantification of RNA-The amount of prolactin-specific RNA hydridizing to radioactive cloned DNA probe was estimated by first comparing the intensity of the autoradiographic signal from the RNA dots to the DNA dots in the standard curve. Then, the nitrocellulose filter dots containing RNA:DNA or DNA:DNA hybrids were excised and counted in a scintillation counter. The counts obtained from the RNA dots were compared to the counts obtained for the known concentration of DNA in the standard curve. If several concentrations of RNA were spotted onto the filters, the count from each concentration was normalized to the amount of specific RNA/pg of total RNA and the different values were averaged to determine the amount of specific RNA/pg of total RNA.
Methylation-Aliquots of DNA (20 pg) prepared from pituitary tumors, nonestrogen-treated pituitaries, or DES-treated liver were digested to completion with MspI, HpaII, or HhuI. MspI and HpaII are isoshizomers that recognize the nucleotide sequence CCGG. HpaII cuts at this sequence only if the internal cytosine is unmethylated. MspI cuts at this sequence regardless of the methylation status of the internal cytosine. HhuI cuts at the sequence GCGC only if the internal cytosine is unmethylated. Restriction endonuclease-cleaved DNA was fractionated by agarose gel electrophoresis, transferred to nitrocellulose, and hybridized to specific fragments of DNA from the prolactin gene domain as described above.

Tissue-specific DNase I Sensitivity of the Prolactin Gene-
The prolactin gene is transcribed only in lactotrophs, a subpopulation of the five major cell types that comprise the anterior pituitary gland. The other major cell types (somatotrophs, corticotrophs, thyrotrophs, and gonadotrophs) normally represent 90-95% of the cells in intact anterior pituitaries of adult male rats (32). F344 rats, when chronically treated with the synthetic estrogen DES, develop pituitary tumors (18,19) due to the predominant proliferation of lactotrophs (19). After 3 months of DES treatment, we observed tumors approximately 30 times the size of normal pituitaries (data not shown), confirming previous publications (19, 33).
In Fig. la, a partial restriction map is shown of the recombinant clones containing the prolactin gene (34).' Five exons and four introns, extending over approximately 10 kb of DNA, comprise the prolactin gene. Twelve kb of DNA 5' and 15 kb of DNA 3' to the prolactin gene also are included in these recombinant clones. Fig. l b shows the location and size of specific restriction fragments referred to in the text and figures. Fig. IC shows restriction fragments of prolactin DNA that were excised from the plasmid pBR322 by restriction enzyme cleavage and isolated by preparative agarose gel electrophoresis. These fragments were labeled to high specific activity by nick translation and used as probes in hybridization assays.  The DNA was fractionated by electrophoresis on 1% agarose gels and blotted onto nitrocellulose filters (Southern). The blots shown in a and c were hybridized to probes El and E3, respectively (see 37 "C for 10 min. DNA was purified, cut with restriction endonucleases, separated on 1% agarose gels, transferred to nitrocellulose (Southern), and hybridized to labeled prolactin restriction fragments or serum albumin cDNA. digested a t approximately the same rate. As a control, the serum albumin gene in the same DNA preparation was digested only slightly, even at the highest concentration of DNase I (Fig. 26).
Alternatively, Fig. 2, c and d, shows EcoRI-cut DNA from DNase I-digested liver nuclei. The DNA in Fig. 2c was hybridized to probe E3, and the DNA in Fig. 2d was hybridized to the serum albumin cDNA. In the liver, the prolactin gene is not highly sensitive to DNase I, but the serum albumin gene, in the same DNA preparation, is highly sensitive. Identical results as those shown in Fig. 2, a-d, were obtained when a nitrocellulose filter blot that had first been hybridized to either the serum albumin cDNA or probe El, was stripped of the first probe by boiling 2 min in water, and rehybridized to the other hybridization probe (data not shown). The persistence of faint bands in the DNase I-sensitive genes, a t all concentrations of DNase I used to digest chromatin, is probably due to the heterogeneous nature of the pituitary tumor or liver cell populations with respect to prolactin or serum albumin production, respectively.
Relative hybridized to probe AB and DNA in Fig. 36 to probe C. Under these conditions, DNA regions 5' to the prolactin gene are detected. The 4.5-kb restriction fragment in Fig. 3a is situated greater than 7 kb upstream from the first exon of the prolactin gene (see Fig. Ib for location of fragments). The 6.9-kb restriction fragment in Fig. 36 extends immediately upstream from the first exon. The chromatin associated with these DNA fragments exhibits the same relative sensitivity to DNase I digestion; DNA is digested extensively a t a DNase I concentration greater than 3.0 pg/ml. This is similar to the DNase I sensitivity of the chromatin associated with the coding region of the prolactin gene, shown in Fig. 3, c and d. DNA in Fig. 3c was cut with MspI and hybridized to probe E,; DNA in Fig. 3d was cut with Hind111 and hybridized to probe G. In both cases, the chromatin was digested a t a DNase I concentration greater than 3.0 pg/ml. The restriction fragments in Fig. 3, As a control for these experiments, DNA from these same DNase I digests was cleaved with EcoRI and hybridized to the serum albumin cDNA. The serum albumin gene was insensitive to DNase I digestion even a t a nuclease concentration as high as 20 pg/ml (data not shown).
Localization of Hypersensitiue Sites near the Prolactin Gene-The sub-band observed in Fig. 3b (arrow) is a DNase I-hypersensitive site that results when DNase I cuts both strands of DNA a t a specific site in chromatin. This hypersensitive site is situated approximately 1.8 kb upstream from the transcription start site in a region that contains repetitive sequences (see Fig. 7). We examined this site further by cleaving genomic DNA from DNase I-digested chromatin to completion with XbaI, and then hybridizing to probe E2 which represents DNA situated 3' to the hypersensitive site. Under these conditions, we observed the 3.8-kb XbaI restriction fragment and, in addition, two sub-bands (Fig. 4, third lane) which are observed only in samples treated with DNase I (Fig.   4, compare second and third lanes). The presence of two subbands indicates that two hypersensitive sites are located near the prolactin gene. The second hypersensitive site maps about 150 base pairs upstream from the transcription start site.
These DNase I-hypersensitive sites are not observed in liver chromatin (Fig. 4, first lane). We are currently characterizing these sites more thoroughly. No other hypersensitive sites were detected in the 35 kb of prolactin chromatin examined (Fig. 3, a-f, and results confirmed by mapping experiments using different restriction endonucleases and the hybridization probes used here; data not shown).
Correlation of DNase I Sensitivity with Transcriptionul Ac- tiuity-The 35 kb of highly DNase I-sensitive chromatin we observed may be due to the presence of transcribed regions 5' or 3' to the prolactin gene. Most of the DNA sequences greater than 1.5 kb 3' to the prolactin gene are repeated throughout the genome. There are 7 kb of unique sequence DNA 5' to the prolactin gene (Fig. 7). In order to address the possibility that another gene might be located here and to quantitate the transcripts homologous to discrete regions of the prolactin gene, we used a dot blot hybridization assay (30). Different dilutions of RNA prepared from the nuclear or cytoplasmic fractions of pituitary tumor or liver of DES-treated rats were spotted next to each other on a single nitrocellulose filter in dots of uniform size. After hybridization of the filters to the labeled DNA fragments shown in Fig. IC, the quantity of RNA transcripts homologous to prolactin restriction fragments was determined by comparison to a DNA standard curve. Fig. 5 and Table I show  or N) in either pituitary tumors or liver (faint dots are seen for 5 pg of liver nuclear RNA, but the radioactivity in these dots, determined by scintillation counting, was no higher than counts in the control dots). The dot blot hybridization assay used for this work can detect less than 1 pg of homologous RNA in 5 pg of total RNA (31). However, we cannot exclude the possibility that these DNA sequences are transcribed but the transcripts are rapidly degraded. We examined this by fractionating cytoplasmic and nuclear RNAs on formaldehyde gels. After the RNA was transferred from the gel to a nitrocellulose filter and hybridized to probe El, we were able to detect nuclear precursors of the prolactin mRNA (which are rapidly processed). Alternatively, if the blot was hybridized instead to probe C, we could not detect transcripts homologous to unique DNA sequences outside the prolactin gene (data not shown). We conclude that transcripts homologous to unique DNA sequences 5' or 3' to the prolactin gene must account for less than of the total RNA. Table I shows an estimate of the picograms of RNA/pg of total RNA, homologous to specific prolactin restriction frag-0 e . 0 0 . P ments. The data in this table were obtained by excising the spots from the nitrocellulose filters, whose autoradiograms are shown in Fig. 5, and quantitating the radioactivity as described under "Materials and Methods." These estimates of the number of homologous RNA transcripts are only semiquantitative due to the limited accuracy of the dot blot hybridization assay. However, this dot blot hybridization assay is valuable for estimating the relative abundance of one transcript as compared to another. As expected, a high concentration of prolactin transcripts is present in both the cytoplasm and nucleus of pituitary tumors but not in liver (E, and G).
Transcripts homologous to the fourth intron of the prolactin gene (F), which contains Alu-like repeats (34), are not detected in the cytoplasm and are present in low concentration in the nucleus. This is expected since introns are usually processed and degraded rapidly. This repetitive sequence does not appear to be transcribed abundantly elsewhere in the genome. The repetitive sequences A,, I, J, and K hybridize to homologous nuclear RNA transcripts. Unexpectedly, probe A, also hybridizes to transcripts present in the cytoplasm. A similar result was obtained for liver RNA. Transcripts ho- Fig. IC) SA is the serum albumin cDNA. Quantity of RNA transcripts was determined as described under "Materials and Methods." The values given reflect the average of three determinations. Tissue-specific Hypomethylation of the Prolactin Gene-Transcribed DNA sequences are generally unmethylated, while adjacent nontranscribed DNA sequences are methylated. Hypomethylation of a gene is tissue-specific. We used the restriction endonucleases MspI, HpaII, and HhaI to define the methylation pattern of the prolactin gene in DES-induced pituitary tumors, control pituitaries, and liver. Fig. la and 6e show the MspIIHpaII and HhaI restriction sites in the prolactin gene domain.

Prolactin DNA fragment
Genomic DNA was digested with MspI, HpaII, or HhaI, fractionated in adjacent lanes by agarose gel electrophoresis, transferred to nitrocellulose, and hybridized to one of the labeled unique sequence restriction fragments shown in Fig. IC. When MspI-cleaved DNA is hybridized to probe El, 4.6and 1.8-kb bands are observed (Fig. 66, first, fourth, and seuenth lanes). A 6.4-kb band is never seen, indicating that whenever the two outer sites are unmethylated, the inner site is also unmethylated. In liver, no discrete bands are observed for HpaII-cleaved DNA (Fig. 6b, eighth lane), suggesting the three MspI/HpaII sites within the coding region of the prolactin gene are methylated (this experiment does not prove conclusively, though, that all three sites are methylated). In control pituitaries and pituitary tumors, however, both bands are observed (Fig. 6b, second and fifth lanes). A 10.8-kb band is observed when pituitary tumor, but not control pituitary or liver DNA, is digested with HhaI and hybridized to probe El (Fig. 6b, sixth, third, and ninth lanes, respectively). The autoradiographic intensity of the HhaI band is less than for the HpaII band. These results indicate that the HhaI sites located immediately 5' and in the last exon of the prolactin gene are a t least partially unmethylated in lactotrophs; the lesser intensity of the HhaI band as compared to the HpaII band could result if one or both of these HhaI sites are methylated in a subset of the DNA molecules examined but the HpaII sites in the same DNA molecules are unmethylated. The 10.8-kb HhaI restriction fragment in control pituitary would not have been observed if only a few per cent of the DNA molecules were unmethylated at this site (see below). Alternatively, this site may be methylated in control pituitaries, and estrogen induces demethylation of this site.
When MspI-cleaved DNA is hybridized to probe C, a 10.9kb band is observed (Fig. 6a, first, fourth, and seventh lanes). When pituitary tumor, control pituitary, or liver DNA is digested with HpaII, no discrete bands are found (Fig. 6 a , second, fifth, or eighth lane). Thus, both MspIIHpaII sites 5' to the prolactin gene are methylated in all tissues studied. When Hhd-cleaved DNA from either of the three tissues is hybridized to probe C, no discrete bands are observed (Fig.  6a, third, sixth, or ninth lane). HhaI sites further 5' than the one nearest the first exon must be methylated.
Hybridization of MspI-cleaved DNA to probe N results in a complex pattern of bands (Fig. 6c, first, fourth, and seuenth lanes) due to the repetitive nature of this sequence in the genome. The restriction map of cloned prolactin DNA indicates that the prolactin bands are 12.5, 3.5, and 0.6 kb in length. Only the larger two bands are actually observed, however, possibly due to the poor binding of smaller DNA fragments to nitrocellulose filters after transfer from agarose gels. When probe N is hybridized to HpaIIor HhaI-cleaved DNA from pituitary or liver, no discrete bands are seen (Fig.  6c, second, third, fifth, sixth, eighth, and ninth lanes). The 3' region of the prolactin gene domain is, like the 5' region, methylated.
The method we employed above can detect the presence of methylation when only 10-20% of the DNA sequences in a DNA sample are methylated (14). To determine the extent of methylation of the transcribed region of the prolactin gene in DES-induced pituitary tumors, DNA was prepared from rats treated with DES for various times from 0 to 15 weeks. Aliquots of these purified DNAs were digested with HpaII and SstI. SstI cleaves DNA regardless of the methylation pattern. The DNA was then prepared for hybridization as described under "Materials and Methods." When DNA is digested with SstI only and hybridized to probe El, a 7.8-kb band is detected (see Fig. 6e for restriction map). HpaII cleaves prolactin DNA three times within the SstI restriction fragment, resulting in the appearance of three bands (Fig. 6d, second to fifth lanes). The 7.8-kb band would be observed after HpaIIISstI digestion only if some of the prolactin genes in a population of DNA remain methylated. The autoradiographic intensity of the three HpaII bands, as compared to the SstI band, indicates the relative extent of methylation of the prolactin gene. In Fig. 6d, the DNA in the first lane was digested with MspI. Two major bands are observed at 4.6 and 1.8 kb (as well as fainter bands resulting from partial digestion). The other lanes contain HpaII/SstI-cleaved prolactin DNA from F344 rats treated with DES for the number of weeks shown. With increasing time of DES treatment, there is an increase in the proportion of prolactin DNA that is unmethylated.

DISCUSSION
We have summarized in Fig. 7 the methylation pattern, location of repetitive DNA sequences,' and DNase I-hypersensitive sites associated with the prolactin gene domain in anterior pituitary tumors of F344 rats. In addition, this entire chromatin domain (35 kb) is sensitive (as compared to liver chromatin) to DNase I digestion.
In several systems, acquisition of a DNase I-sensitive chromatin conformation has been demonstrated to occur during differentiation (12,35,36). Only in the Xenopus vitellogenin gene chromatin system has a steroid (estrogen) been shown to directly induce a more DNase I-sensitive conformation (37). Demethylation of transcribed DNA sequences likewise has been shown to occur during differentiation (12,13); and, in a few cases, a specific MspI/HpaII site becomes hypomethylated upon administration of a steroid (38, 39).
Estrogen induces an increase in prolactin gene transcription (3,4), and estrogen may also be involved in differentiation of lactotrophs (40). The F344 pituitary tumors we used to study the prolactin gene chromatin are induced by chronic estrogen treatment. Therefore, we were not able to determine whether differences we observed in chromatin conformation and DNA structure of the prolactin gene result from a direct effect of estrogen or whether they result from other processes occurring during differentiation. Very likely, in the F344 pituitary system we employed, estrogen simply causes proliferation of fully differentiated lactotrophs. The prolactin genes in the tumor cells then retain the chromatin and DNA structures associated with their progenitor cells. Some support of this idea is given by the methylation data we presented, especially treatment, a fraction of the prolactin genes in the anterior pituitary and all those in liver are methylated. Estrogen may induce demethylation of one or both of the HhaI sites within the first exon or immediately 5' to the prolactin gene in lactotrophs, although we have not shown this conclusively.
The coding region of the prolactin gene exhibits the same degree of sensitivity to DNase I as nontranscribed regions up to 12 kb upstream or downstream. In contrast, early reports in the literature showed that the coding regions of the 8globin genes (13), a-globin genes (12), and ovalbumin gene (14) of the chicken, as well as integrated adenovirus genes in hamster cells (41), are highly sensitive to digestion by DNase I, whereas noncoding regions several kb upstream or downstream of these genes display an intermediate sensitivity.
Subsequently, Wood and Felsenfeld (15) and, more recently, Nicolas et al. (16) found no differential DNase I sensitivity between coding and noncoding regions of the @-globin gene. Lawson et al. (17) studied the chromatin structure of the ovalbumin and X and Y genes of the chicken and found approximately 100 kb of DNA within and around these genes in a uniformly DNase I-sensitive conformation. The latter group suggested that the high level of DNase I digestion they employed might have obliterated more labile DNase I-sensitive structures detected previously by limited DNase I nicking and filter blot hybridization (12)(13)(14). Stalder et al. (13) presented evidence that supports this claim. They detected a differential sensitivity of the P-globin gene and surrounding sequences, after limited DNase I nicking of chromatin and blot hybridization, but not by solution hybridization after extensive DNase I digestion. Flint and Weintraub (41), however, extensively digested chromatin with DNase I and showed by solution hybridization that the highly DNase I-sensitive adenovirus genes, integrated into hamster cell DNA, are bordered by relatively DNase I-insensitive regions a few nucleosomes upstream or downstream.
Differential sensitivity to DNase I digestion between the coding and noncoding regions of a gene does not appear to be a universal phenomenon and is not consistently observed in chromatin of a single gene ( i e . ovalbumin or &globin genes of the chicken). If such differences in DNase I sensitivity of coding and noncoding regions of a gene domain do exist, our inability to detect this difference in the prolactin gene domain may be due to the presence of other transcription units around this gene. However, we were not able to detect transcripts homologous to any of the unique sequence regions.
We did not detect 5' or 3' boundaries to the DNase Isensitive chromatin region, as have been previously reported for the ovalbumin gene family (17) and for the CY-and 8globin genes (12,13). Possibly such DNase I-insensitive regions do exist farther 5' or 3' from the prolactin gene than examined here. It has been suggested that large DNase Isensitive regions found in chromatin may be analogous to the 30-90-kb DNA loops observed after deproteinization of metaphase chromosomes (16) or of lampbrush chromosomes (13).
Chromatin hypersensitive sites have been localized to regions 5' to the CY-and &globin genes (12,13), Drosophila heat shock genes (8), vitellogenin gene (38), as well as many other genes (see Ref. 42 for review). These sites have often been localized to specific DNA sequences near the transcription start site of genes; additional hypersensitive sites have been observed further upstream from the vitellogenin gene of the chicken (38) and the glue protein genes of Drosophila (43). Hypersensitive sites have occasionally been observed 3' to genes or in regions not known to be regulatory regions (42). The precise role of hypersensitive sites is not known, but they