Cell- and Sequence-specific Binding of Nuclear Proteins to 5’-Flanking DNA of the Rat Growth Hormone Gene*

the 236 base pairs of 5’-flanking DNA of the rat hormone found between -104 and -49 base pairs, relative to the transcription (cap) to nuclear protein(s) which to be cell type and generate a DNase I-resistant footprint on both strands between -95 and -68. A distinct protein component(s) selectively binds to DNA between -236 and -146 but is not cell type specific. These regions correspond to those in transfer important in expression (-104/+7) expression (-2361 of

Thyroid hormone stimulates growth hormone gene expression in somatotrophic cells of the rat anterior pituitary and in several rat pituitary cell lines (1)(2)(3)(4)(5)(6)(7). This effect is mediated by a chromatin-associated receptor which is a DNA binding protein (1,7,8). We have presented evidence that thyroid hormone rapidly stimulates the transcription rate of the rat growth hormone gene in cultured GC cells which is proportional to the level of thyroid hormone-receptor complexes (7). Recent gene transfer studies have indicated that 5"flanking DNA of the rat growth hormone gene contains cis-acting element(s) which mediate regulated expression by thyroid hormone (9)(10)(11). These sequences appear to be contained within the first 236 base pairs of 5'-flanking DNA of the gene Although the rat growth hormone gene has been widely used to study regulated gene expression by thyroid hormone (1-lo), no studies have been reported which have examined the cell-and sequence-specific binding of putative transacting factors with the gene. Nondenaturing polyacrylamide gel electrophoresis (12-17) was used to identify whether nuclear proteins specifically interact with 5"flanking DNA of the gene. This procedure is based on the observation that DNA.protein complexes have a decreased electrophoretic mobility compared with the unbound DNA fragment resulting * This research was supported in part by Grants AM 16636 and AM 21566 from the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ Supported in part by the Sackler Institute of Graduate Biomedical Sciences at New York University. (10,11). from alterations in charge density and/or DNA conformation (15). This approach has been effectively used to study the interaction of purified procaryotic regulatory proteins with specific DNA sequences (12-14). More recently this technique has been applied to impure protein preparations where a synthetic duplex alternating DNA polymer, poly(d1-dC),' is added to inhibit nonspecific DNA-protein interactions (17).
In this study using gel electrophoresis we provide evidence that a protein(s) found in growth hormone-producing cells selectively binds to DNA between -104 and +7 relative to the cap site of the gene. DNase I footprinting further localized the binding to sequences between -95 and -68 base pairs which contains the "CAAT" homology (18). A separate protein(s) of lower abundance, which is not cell type specific, binds to 5'-flanking DNA between -236 and -146. These sequences are located in the regions found by gene transfer studies to be involved in mediating basal (-104/+7) and regulated expression of the gene by thyroid hormone (-236/ -146) (11).

EXPERIMENTAL PROCEDURES
Materials-[~~-~~P]dCTP, [LP~'P]~ATP, or [Y-~'P)ATP (each at 3000 Ci/mmol) were obtained from Du Pont-New England Nuclear. DNase I (ribonuclease free) was from Worthington and calf intestinal alkaline phosphatase (molecular biology grade) was from Boehringer Mannheim. All other enzymes were obtained from either New England Biolabs or Boehringer Mannheim and, unless indicated otherwise, were used under the conditions recommended by the suppliers. Duplex poly(d1-dC) was obtained from Pharmacia P-L Biochemicals. Reagents used for gel electrophoresis were obtained from Bio-Rad or Eastman.
Cell Culture Conditions-The growth hormone-producing rat pituitary cell lines GC and GH&, were cultured with Dulbecco's modified Eagle's medium containing 10% (v/v) fetal calf serum (GIBCO) (7,9) while GH, cells, a related cell line, were cultured with Ham's F-10 medium containing 12.5% horse serum and 2.5% fetal calf serum (v/v) (GIBCO) (3). H4TG rat hepatoma cells (American Type Culture Collection) and Rat2 cells, a fibroblastic like line (19), were cultured using the same media as GC cells.
Assay of DNA-Protein Binding by Gel Electrophoresis-Cell nuclei were isolated (20) and extracted with buffer containing 0.4 M KCI, 20 mM Tris-HC1 (pH 7.85 at 25 "c), 0.25 M sucrose, 1.1 mM MgC12, and 10 mM 2-mercaptoethanol (20). These conditions give the most efficient extraction of thyroid hormone nuclear receptor and the salt and buffer conditions are similar to those used to prepare nuclear extracts for in uitro transcription assays (21). The nuclear extracts were stored in small aliquots at -80 "C in 33% glycerol, and freshly thawed material was used in each experiment. A PstI digestion fragment of the cloned rat growth hormone gene (-530 to +69) (18) was cleaved with XhoI, and the fragment extending from -530 to +7 was subcloned and amplified (22). A restriction map of the fragment is shown in Fig. 2a. After cleavage by the appropriate restriction enzymes, The abbreviations used are: poly(d1-dC), alternating duplex polydeoxyinosinic acid-polydeoxycytidylic acid Hepes, N-2-hydroxyethylpiperazine-N'-2-ethanesulfonic acid L -T~, 3,5,3'-triiodo-~-thyronine. of the Rat Growth Hormone Gene various fragments were isolated by gel electrophoresis and labeled by filling in the cohesive ends with the large (Klenow) fragment of DNA polymerase I using [cx-~'P]~CTP or [LU-~'P]~ATP (22). The DNA fragment(s) (6000 dpm; about 0.06 ng or 10-12 fmol) were incubated with 2 pg of nuclear extract protein at 25 "C for 30 min in 20 pl of buffer composed of 15 mM Hepes (pH 7.5 at 25 "C), 60 mM KCl, 5 mM MgCl', 1 mM EDTA, 5 mM dithiothreitol, 12% glycerol, and varying amounts of duplex poly(d1-dC). The samples were chilled to 0 "C and loaded onto a 4% polyacrylamide gel (acrylamide:bisacrylamide, 291) which had been pre-electrophoresed for 120 min at 4 "C using buffer consisting of 6.7 mM Tris-HC1 (pH 7.5 at 25 "C), 1 mM EDTA, and 3.3 mM sodium acetate (16). The electrophoresis was performed at 4 "C at a voltage gradient of about 10 V/ cm with rapid buffer circulation and the electrophoresis was stopped when the bromphenol blue marker migrated to the bottom of the gel. The gel was soaked in 5% glycerol for 20 min with gentle shaking and, after air drying, was autoradiographed at -80 "C for 5-20 h using Kodak X-Omat AR film and a Du Pont Lightning Plus intensifying screen.
DNase I Footprinting Studies-The FnuDII site at -145 ( Fig. 2a) was converted to a BglII site using linkers and the resultant BglII/ XhoI (-145/+7) fragment was subcloned. The DNA was first cleaved at the BglII site to label the end at -145 with either the Klenow enzyme and [cx-~'P]~ATP (noncoding strand) or T4 polynucleotide kinase and [y3'P]ATP (coding strand). The DNA was then cleaved with XhoI, and the labeled -145/+7 fragment was isolated by polyacrylamide gel electrophoresis. DNase I footprinting experiments (23) were performed using the same conditions as for gel electrophoresis except that the reaction was scaled up 10-50-fold. After 30 min at 25 "C the reactions were adjusted to 10 mM Mg2f and 2.5 mM Ca2+. DNase I was added to final concentrations of 0.1-0.5 pg/ml and the samples were incubated for 1 min at 25 "C. EDTA was added to 25 mM and after phenol/chloroform extraction and ethanol precipitation, the DNA was electrophoresed in 8% sequencing gels (22). Maxam-Gilbert chemical sequencing ladders (24) were used as standards.

RESULTS
Gel Electrophoretic Identification of Protein-DNA Inteructions Using Poly(dI-dC) as u Competitor- Fig. 1 shows the effect of GC cell nuclear extracts on the electrophoretic migration of a "P-end-labeled BglII/XhoI restriction fragment which extends from -236 to +7 relative to the cap site of the rat growth hormone gene (18). When nonspecific DNA-protein interactions were inhibited by increasing amounts of poly(d1-dC) (lunes 2-7), two ["PIDNA species were detected (lune 7) with mobilities less than the free [32P]DNA fragment (lune 1). Proteinase K digestion (lune 8) indicates that the decrease in electrophoretic mobility reflects the interaction of protein with DNA. The more abundant (lower) protein. DNA complex (Fig. 1, lane 7) was always observed while the amount of the other (upper) complex varied among different experiments. When Escherichia coli DNA was used instead of poly(d1-dC), similar protein -DNA complexes were formed but the effective concentration range was very narrow. With 2 pg of nuclear extract protein, 1 pg of E. coli DNA gave similar protein-DNA complexes as 2-5 pg of poly(d1-dC), while 2 pg of E. coli DNA significantly decreased the amount of protein.
DNA complex formed. This has been observed in studies with other genes (17) and presumably results from the fact that natural DNA contains short sequences which have sufficient homology to compete with specific DNA sequences at high competitor concentrations.
Interaction of Nuclear Proteins of GC Cells with Different 5'-Flunking Regions of the Rat Growth Hormone Gene-To further define the binding regions in the 5"flanking DNA, we examined a series of DNA restriction fragments ranging from -530 to +7 (Fig. 2u). In these studies we identified two regions which selectively bind to distinct protein(s) from GC cell nuclei. For the purpose of discussion we refer to these two binding domains as regions I and 11. Region I binding occurs within the 5'-flanking DNA between the AluI (-104) and the XhoI restriction site (which is cleaved between +7 and +8) to form an abundant complex which electrophoreses as a broad radiolabeled band of lower mobility than the free DNA fragment (Fig. 2b, lune 2). The TuqIIXhoI fragment (-48/+7), containing the TATA homology, showed no interaction with nuclear extract protein under the conditions used (Fig. 2b, lane I) suggesting that sequences in the -104/+7 fragment which form the protein.DNA complex lie between -104 and -49. The FnuDII/XhoI fragment (-145/+7) (Fig. 2b, lune 3) generated the same abundant complex seen with the -104/ +7 fragment and an additional species of lower mobility to give a similar pattern as the -236/+7 fragment (BglII/XhoI) (Fig. 1, lune 7).
The different electrophoretic pattern formed with the -104/+7 and -145/+7 fragments suggests that additional proteins bind to the longer fragment. However, the amount of this additional complex varied among different experiments. DNase I footprinting studies (Fig. 2c) were carried out in an experiment in which the two complexes could be identified by gel electrophoresis. Only one DNase I protected region was detected which was more predominant on the noncoding strand (-68 to -95) (lunes 3 and 4 ) than the coding strand (-68 to -90) (lunes 5 and 6). No DNase I protected regions were detected between -104 and -145 under the conditions used to identify the DNase I footprint between -68 and -95. When the BglII/FnuDII fragment extending from -236 to -146 was incubated with GC cell nuclear extract, it formed a complex which migrated as a narrow rather than a broad labeled band (Fig. 2b, lune 4). We refer to DNA in the -236/ -146 fragment as binding region 11. An identical gel electrophoretic complex was formed using a labeled -236/-130 fragment (not illutrated). In contrast, sequences from -312 to -237 and from -530 to -313 did not generate any high affinity protein.DNA complexes (Fig. 2b, lunes 5 and 6). In uitro addition of L-T3 did not alter the results of Fig. 2b. Although we have been able to use DNase I to footprint the binding region in the -104/+7 fragment (Fig. 2c), no DNase I protected regions were identified using a -236/-130 fragment even when the complex was isolated from the gel. Gel electrophoresis has been reported to detect protein .DNA complexes which are not identified by DNase I protection experiments (25). Therefore, other techniques will be required to precisely locate the sequences in the -236/-146 region which are involved in the protein-DNA interaction (25).
Competition Studies Using Homologous and Heterologous DNA Fragments-Competition studies (Fig. 3) indicate that proteins which bind to region I1 (-236/-146) are distinct from those which interact with sequences in region I (-145/ +7). When 50 ng of the unlabeled -145/+7 or -104/+7 fragment was incubated with 0.05 ng of 3ZP-labeled DNA from -145/+7, the formation of labeled protein.DNA complexes was completely inhibited (Fig. 3u, lunes 2 and 3). In contrast, 50 ng of the -236/-146 fragment did not inhibit protein. DNA complex formation of the labeled -145/+7 fragment (Fig. 3u, lune 4). Complex formation of labeled DNA from -104/+7 was also inhibited by the unlabeled homologous fragment but not heterologous DNA from -236/-146 (Fig.  3u, lunes 6 and 7). In like fashion, complex formation of labeled -236/-146 DNA was inhibited by the homologous fragment but not by DNA from -145/+7 (Fig. 3b). When GC cell nuclear extracts were incubated with 0.05 ng of labeled DNA from -236/+7 and 30 ng of unlabeled DNA from -145/ +7, the protein-DNA complex formed migrated as a narrow labeled species characteristic of the complex formed by the a.

b.
C. The radiolabeled fragment -236/-146 (about 0.05 ng) was incubated with 3 pg of GC cell nuclear extract protein without unlabeled DNA ( l a n e l ) , with 30 ng of the same unlabeled fragment ( l a n e 2), or 30 ng of the -145/+7 unlabeled fragment ( l a n e 3). c, a [32P]DNA fragment (about 0.05 ng) containing binding regions I and I1 (-236/+7) was incubated with 2 pg of GC nuclear extract protein in the absence of any unlabeled DNA fragment ( l a n e I ) or with 30 ng of the unlabeled fragment from -145/+7 (lane 2) which contains only binding region I.

Protein-DNA Interactions Using Nuclear Proteins from Rat Growth Hormone-producing Cells and Rat Cells Which Do Not Express the Gene-The
protein.DNA complexes formed by nuclear extracts from rat cells (H4TG and Rat2) which do not express the growth hormone gene are shown for fragments -236/+7, -145/+7, and -236/-146 (Fig. 4, a-c). Nuclear extracts from GC and GH, cells formed the same protein.
DNA complexes with the -236/+7 (Fig. 4a) and -145/+7 (Fig. 4b) fragments. GH& cells, another growth hormoneproducing rat pituitary cell line, generated a pattern similar to GC and GH, cells (not shown). In contrast, extracts from the rat hepatoma H4TG and the Rat2 fibroblastic cell lines did not give the same pattern of binding. Instead, nuclear proteins from the H4TG and the Rat2 lines generated a complex having a very low electrophoretic mobility seen near the top of the gel. Gel electrophoretic shift studies were also carried out with nuclear extracts from rat liver and testes (not shown). These extracts, when incubated with the -145/+7 fragment, gave results identical to that observed with nuclear proteins from Rat2 and H4TG cells (Fig. 4b). In contrast with the -145/+7 fragment, nuclear proteins from each of the rat cell lines formed the same high affinity protein. DNA complex with the -236/-146 fragment (Fig. 4c). More of the complex appeared to be formed by nuclear extracts of GC and GH, than Rat2 and H4TG cells. Nuclear extracts from rat liver and testes also generated the same gel shift band with the -236/-146 fragment which was about 50% less than that observed with extracts from GC cells (not shown).

DISCUSSION
Sequences which form protein-DNA complexes in the -104/+7 fragment appear to be localized between -104 and -49 (Fig. 2b) and generate a DNase I footprint between -68 and -95 which spans the CAAT homology found at about -80 (18). This gel electrophoretic complex was only identified in cells which express the rat growth hormone gene (Fig. 4, a   a. b.  H4TG ( l a n e 3), and the Rat2 fibroblastic cell line ( l a n e 4 ) were prepared as described under "Experimental Procedures." The binding reactions were carried out using 2 pg of nuclear extract protein and 4000 ng of poly(d1-dC). and b), suggesting that sequences in this region recognize cellspecific proteins which mediate basal and/or cell-specific expression of the gene. This notion is supported by transient expression experiments which indicate that basal expression in GC and GH, cells is dependent on sequences located between -104 and +7 and that no expression occurred in Rat2 or H4TG cells (11). When sequences between -236 and -146 were deleted, regulated expression by thyroid hormone was eliminated (11). This suggests that proteins which bind to sequences from -236 to -146 may play a role in mediating regulated expression by hormone. Protein(s) which bind to this region were found in each of the rat cell lines examined but were less abundant in the Rat2 and H4TG cells (Fig. 4c) which have about 60-75% less thyroid hormone receptor than the somatotrophic cell lines.'

C.
The amount of this protein in GC cells is much less abundant than the protein(s) which bind to DNA from -104 to +7 and is in the same range as the thyroid hormone nuclear receptor. The amount of the protein which binds to the -236/ -146 fragment can be estimated from the gel electrophoretic assays. Using 10 fmol of labeled fragment, about 20% of the DNA forms a protein. DNA complex (Fig. 2b, lane 4). Increasing the amount of the fragment does not substantially increase the amount of complex, suggesting that there is about 2 fmol of binding protein/2 pg of GC cell nuclear extract protein. This amount is in the same range as the amount of receptor extracted from GC and GH, nuclei by 0.4 M KC1 (100 fmol/ 100 pg nuclear extract protein) (7,20). The observation that rat liver nuclear extracts generate about 50% of the protein.
DNA complex with the -236/-146 fragment is in keeping with the relative difference in abundance of receptor in GC cells and rat liver (7,9,20). However, nuclear extracts from rat testes also generated the same gel shift pattern with the -236/-146 fragment which was about 50% of that found with GC cell extracts. Studies of receptor abundance in various rat tissues by in vivo injection of L -[ ' *~I ] T~ suggest that nuclei from testes contain only 0.4% of the receptor of rat liver (26). I n vitro L-['*~I]T~ binding studies using testes nuclear extracts, however, indicated a receptor abundance which was about 10% of that found in GC cell extracts.' The differences noted in the in vitro' and the in vivo abundance of receptor (26) in the testes may be secondary to low levels of perfusion of the testicular cells which contain receptor or to a decrease in cell entry of L-["~I]T~.
The discrepancy between receptor abundance and the amount of protein. DNA complex formed with testicular nuclear extract and the -2361-146 fragment does not permit us to conclude that the protein(s) which binds to the fragment is receptor. Our results indicate, however, that a wide variety of rat cells contain a nuclear protein(s) which binds in a sequence-specific manner to the region of the rat growth hormone gene which is involved in mediating regulated expression by thyroid hormone (11). Recent studies with chick and human cells indicate that one of the cellular homologues of the avian erythroblastosis virus v-erb-A gene encodes a thyroid hormone receptor (27,28). Several other distinct cellular homologues of the v-erb-A gene have been cloned (29)(30)(31) which, like the 75-kDa viral gag-erb-A gene product (27), may retain the DNA binding domain but may not bind thyroid hormone. Whether the protein(s) which can interact with the -2361-146 fragment represents receptor as well as related erb-A protein homologues which do not bind hormone is under investigation.
Our results suggest, however, that the unique expression of