T w o CGTCA Motifs and a GHFlPitl Binding Site Mediate CAMP-dependent Protein Kinase A Regulation of Human Growth Hormone Gene Expression in Rat Anterior Pituitary GC Cells*

We established the cis-acting elements which mediate cAMP responsiveness of the human growth hormone (hGH) gene in transiently transfected rat anterior pituitary tumor GC cells. Analysis of the intact hGH gene or hGH S’-flanking DNA (W-FR) coupled to the hGH cDNA or chloramphenicol acetyltransferase or luciferase genes, indicated that cAMP primarily stimulated hGH promoter activity. Cotransfection of a protein kinase A inhibitory protein cDNA demonstrated that the cAMP response was mediated by protein kinase k Mutational analysis of the hGH promoter identified two core cAMP response element motifs (CGTCA) located at nucleotides -187/-183 (distal cAMP response element; dCRE) and -W/-95 (proximal cAMP response element; pCRE) and a pituitary-specific transcription factor (GHFl/Pitl) binding site at nucleotides -123/-112 (dGHF1) which were required for CAMP responsiveness. GHFl was not a limiting factor, since overexpression of GHFl in cotransfec-tions increased basal but not forskolin induction levels. Gel shift analyses indicated that similar, ubiquitous, thermostable protein(s) specifically bound the pCRE and dCRE motifs. The CGTCA motif-binding factors were C A M P response element binding protein (CREB)/ activating transcription factor-1

We established the cis-acting elements which mediate cAMP responsiveness of the human growth hormone (hGH) gene in transiently transfected rat anterior pituitary tumor GC cells. Analysis of the intact hGH gene or hGH S'-flanking DNA (W-FR) coupled to the hGH cDNA or chloramphenicol acetyltransferase or luciferase genes, indicated that cAMP primarily stimulated hGH promoter activity. Cotransfection of a protein kinase A inhibitory protein cDNA demonstrated that the cAMP response was mediated by protein kinase k Mutational analysis of the hGH promoter identified two core cAMP response element motifs (CGTCA) located at nucleotides -187/-183 (distal cAMP response element; dCRE) and -W/-95 (proximal cAMP response element; pCRE) and a pituitary-specific transcription factor (GHFl/Pitl) binding site at nucleotides -123/-112 (dGHF1) which were required for CAMP responsiveness. GHFl was not a limiting factor, since overexpression of GHFl in cotransfections increased basal but not forskolin induction levels. Gel shift analyses indicated that similar, ubiquitous, thermostable protein(s) specifically bound the pCRE and dCRE motifs. The CGTCA motif-binding factors were C A M P response element binding protein (CREB)/ activating transcription factor-1 (ATF-1)-related, since the DNA-protein complex was competed by unlabeled CREB consensus oligonucleotide, specifically supershifted by antisera to CREB and ATF-1 but not ATF-2, and was bound by purified CREB with the same relative binding affinity (pCRE e dCRE e CREB) and mobility as the GC nuclear extract. UV cross-linking and Southwestern blot analyses revealed multiple DNA-protein interactions of which 4 0 0 -and 45-kDa proteins were predominant; the 45-kDa protein may represent CREB. These results indicate that CREB/ATF-1-related factors act coordinately with the cell-specific factor GHFl to mediate CAMP-dependent regulation of hGH-1 gene transcription in anterior pituitary somatotrophs.
Human growth hormone (hGH)l belongs to a family of hormones which includes chorionic somatomammotropin (hCS) * This work was supported by National Institutes of Health Grant DK41206 (to N. L. E.). 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 U.S.C. Section 1734 solely to indicate this fact.
Doctor of Philosophy, Mayo Graduate School.
Regulation of the hCS-1 gene by CAMP has not been established. Because GHRH acts via CAMP to increase rat (r) GH gene transcription (Barinaga et al., 1983,19851, it was assumed that the -75% identical hGH gene would also be transcriptionally controlled by CAMP. The regulation of a large number of genes by CAMP is mediated by certain members of either the activator protein-2 (AP-2; Imagawa et al., 1987) or CAMP response element binding proteidactivating transcription factor (CREB/ATF) family (reviewed in Meyer and Habener (19931, Montminy et al. (1990), Roesler et al. (1988)). The CREB/ATF family belongs to the bZIP (basic regiodeucine zipper) class of DNA binding proteins (Vinson et al., 1989) and is involved in mediating the actions of ElA, Ca2+, and CAMP (Liu et al., 1993;Hai et al., 1989). To date, CREB (Gonzalez et a l . , 1989;Montminy et al., 1990), CREM (Foulkes et al., 19921, JunD (Kobierski et al., 1991), and ATF-1 (Rehfuss et al., 1991) are the best characterized CAMP-responsive bZIP transcription factors. Both CREB and ATF-1 mediate CAMP responsiveness on a number of genes by binding to a CAMP response element (CRE) consisting of CGTCA motifs of the palindromic consensus sequence TGAC-GTCA (reviewed in Meyer and Habener (1993), Montminy et al. by the hGH-1 gene); rGH, rat growth hormone; GHRH, hypothalamic growth hormone releasing hormone; GHFVPitl, pituitary-specific transcription factor; pGHFl and dGHF1, proximal and distal GHFl binding coded by the hCS-1 gene); PRL, prolactin; CAT, chloramphenicol acet-sites, respectively; hCS, human chorionic somatomammotropin (enyltransferase; LUC, luciferase; 5'-FR, 5"flanking region; CRE, CAMP response element; pCRE and dCRE, proximal and distal CREs, respectively; CRU, CAMP response unit; CREB, CAMP response element binding protein; ATF, activating transcription factor; PKI, protein kinase A inhibitory peptide; hCG, human chorionic gonadotropin; PKImut, protein kinase A mutant inhibitory peptide; bp, base pair; nt; nucleotide. (1990), andRoesler et al. (1988)). These genes contain common CGTCA motifs yet their flanking sequences differ considerably which may be important in dictating binding specificity (Deutsch et al., 1988). CREB/ATF-related factors may mediate the actions of cAMP by binding as homodimers or heterodimers to these various CGTCA motifs (reviewed in ). Alternatively, these sites may bind heterodimers, consisting of members of the CREB/ATF family and unrelated factors, as shown by Maguire et al., (1991) for the hepatitis B virus X protein.
Previous efforts to identify the CAMP-responsive elements in the rat and human GH genes have focused on the 5'-flanking region (5'-FR) (Copp and Samuels, 1989;Brent et al., 1988;Dana and Karin, 1989). These studies showed that the rGH and hGH 5"FR mediate cAMP responsiveness; however, precise identification of the CREs was not achieved. Brent et al., (1988) narrowed the cAMP responsive region of the hGH 5'-FR to within nt -212". In contrast, Dana and Karin (1989) reported the hGH promoter CRE to lie within 82 bp upstream of the transcriptional start site. Both studies suggested GHFl involvement in the cAMP response. GHFl's role in mediating the cAMP response was supported by the identification of two CREB binding sites in the 5'-FR of the GHFl gene (McCormick et al., 1990;Chen et al., 1990) and by the regulation of GHFl gene transcription by forskolin (McCormick et al., 1990). GHFl may also be phosphorylated in vitro by protein kinase A (Kapiloff et al., 1991). Accordingly, intranuclear increases in GHFl levels or posttranslational modification of GHFl might account in part for the cAMP-mediated increase in hGH gene transcription.
We determined the cAMP regulation of hGH gene expression by analyzing the intact hGH gene or hGH 5'-FR coupled to the hGH cDNA or chloramphenicol acetyltransferase (CAT) or luciferase (LUC) genes in transiently transfected rat anterior pituitary GC cells exposed to the cAMP-elevating agent forskolin (Seamon and Daly, 1986). Our results demonstrate that cAMP regulation of the hGH promoter is protein kinase Amediated at the level of transcription. The elements controlling CAMP responsiveness of the hGH gene are localized to the promoter region and require two CGTCA motifs and a GHFl binding site. The two CGTCAmotifs are located distally (dCRE; nt -187/-183) and proximally (pCRE; nt -99/-95) to the distalmost GHFl binding site (dGHF1; nt -123/-112). DNA binding studies and immunocharacterization revealed that the protein(s) binding to the CGTCA motifs were CREB/ATF-1-related. UV cross-linking and Southwestern blot analyses revealed multiple DNA-protein interactions of which -100-and -45-kDa proteins were predominant; the -45-kDa protein may represent CREB. Thus CREB/ATF-1-related factors may act coordinately with the tissue-specific factor GHFl to mediate cAMP/protein kinase A regulation of hGH-1 gene transcription in anterior pituitary somatotrophs.  (Sambrook et al., 1989). In this study hGH and hCS refer to the hGH-1 and hCS-1 genes exclusively. All nucleotide numbering is relative to the transcription start site. The hGH promoter (nt -49UEcoRI to +G/BamHI) is designated GHp. The hCS promoter (nt -493/EcoRI to +6/BamHI) is designated CSp. The human p-actin promoter (4.3-kilobase EcoRI to AZuI fragment) is designated ACTp. The hGH cDNA (nt +l/BamHI to +689/SmaI) is designated GHc. The hGH structural gene (nt +l/BamHI to +2651/EcoRI) is designated GHs. The hGH (nt +2029 to +2651) and SV40 (HindIIIINdeI) 3'-untranslated/3'-FR are designated GH3' and SV3', respectively. Con-struction of the GHp.GHs.GH3', GHp.GHc.GH3', CSp.GHc.GH3', AC-Tp.GHc.GH3', GHp.GHc.SV3', and CSp.GHc.SV3' plasmids was described in detail by Zhang et al. (1992). The GH and CS hybrid promoters in plasmids CSp/GHp.GHc.SV3' and GHp/CSp.GHc.SV3' were fused at their common NsiI site (nt -83) as described by Nachtigal et al. (1989). Deletions of the hGH 5"FR were constructed by Bal31 digestion as described by Cattini et al. (1986). The -109/+51 thymidine kinase promoter/CAT construct (TKp.CAT) was described by Cattini and Eberhardt (1987). The Rous sarcoma virus promoter (RSVp) fused to the rat GHFl cDNA(RSVp.rGHF1) was a giR from Dr. Michael Karin (University of California, San Diego). RSVp fused to the p-galactosidase gene (RSVp.p-Gal) was described by Walker et al. (1983). Plasmids expressing wild-type (RSVp.PKI) and mutant (RSVp.PKImut) protein kinase A inhibitory protein were described by Grove et al. (1987).
Site-specific mutagenesis of the hCS and hGH 5"FR was done with minor modifications of the protocol of Hemsley et al., (1989). The hGH 5"FR (nt -49WEcoRI to +6/BamHI) in pUC8 (hGHp.pUC8) was modified by insertion of a Sal1 linker into the EcoRI site. Five ngs of the hGHp.pUC8 plasmid was subjected to inverse polymerase chain reaction with 1Ck100 pmol of each primer (listed below) in a 100-pl reaction consisting of 20 m~ Tris-HC1 (pH 8.2), 10 m~ KCI, 1.5-2.0 m~ MgCl,, 6 m~ (NH&S04, 0.1% Triton X-100,200 p each dNTP, and 1.25 units of Pfu DNA polymerase (Stratagene). The reaction underwent 25 cycles of 1 min at 94 "C, 1 min at 50 "C, and 8 min at 75 "C with a final cycle 10 min at 72 "C extension. The following oligonucleotides were used for the polymerase chain reactions (all sequences are written 5' to 3'; + = sense strand, -= antisense strand): GH(dCRE)p.LUC, CCGCTGGAAGG-GAAAGATG (+); CCGCTAGTGTTGTGAGGGTT (-); GH(pCRE)p.LUC, ACCGGGCCCCATGCATAAAT (+); ACCGGCTTGTGCTAATGGAT (-); GH(dGHFl)p.LUC, TTATCCACTAGCACAAGCCCGTCAGTGGCC (+I; TCGCGAAGCTCCTCCCACA (-1. Mutant polymerase chain reaction products were isolated on a n 1% agarose gel, purified (Geneclean, BiolOl, Inc.) and 5'-end phosphorylated with T4 polynucleotide kinase. The products were then re-isolated on an 1% agarose gel, purified (Geneclean), ligated, and transformed into Escherichia coli HB101. Individual colonies were screened for correct mutations and nonspecific Pfu polymerase errors by dideoxy chain termination sequencing (U. S. Biochemical Corp.). The mutated clones were digested with BamHI, blunt-ended with Klenow, and re-digested with SalI. These fragments were subcloned into a pA3.LUC (Maxwell et al., 1989) vector that was previously modified by introduction of a Sal1 linker into the SmaI sites. The modified pA3.LUC vector was first digested with HindIII, bluntended with Klenow and then digested with Sal1 to accommodate the DNA fragments, containing the mutated promoter sequences, upstream of the LUC gene. Plasmid DNA was purified by two sequential CsCl gradients and the DNAconcentration measured spectrophotometrically. Plasmids were also checked for purity, concentration, supercoiling and restriction digestion pattern by agarose gel electrophoresis. In most cases, at least two separate plasmid preparations were tested in the transfection experiments.
Cell Culture and DNA Dansfections-GC cells were grown in monolayers at 37 "C, 5% CO, and 100% humidity in Dulbecco's modified Eagle's medium (DMEM, high glucose, Celox) supplemented with 10% fetal bovine serum (FBS; Whittaker), 100 units/ml penicillin, 100 pg/ml streptomycin (Life Technologies, Inc.), and 1 m~ L-glutamine (Celox). One to two days prior to transfection, GC cells were rinsed with Dulbecco's phosphate-buffered saline containing MgCI, and CaCl, (PBS+MC, Celox) and deinduced with DMEM containing 4% hormonedepleted FBS (STR-FBS) (Samuels et al., 1979). The CAT and hGH reporter genes were transfected essentially as described by Zhang et al. (1992). Briefly, GC cells were harvested and transfected (4 x lo7 cellsf 0.4 ml) at 350 V a t 960 microfarads (Gene Pulser, Bio-Rad) with either 15 pg of CAT plasmid or 15 pg of hGH reporter plasmid and 3 pg of a human actin promoter driven human chorionic gonadotropin (hCG) cDNA expression plasmid to control for transfection efficiency. Transfected cells were diluted in 4% STR-FBSDMEM and plated onto 6 x 6-cm2 dishes. Forskolin was added a t 2.5 p final concentration for the entire 20-24-h transfection period. For the LUC reporter gene transfections, deinduced cells were harvested in 0.5 x Trypsin-EDTA (Life Biotechnologies Inc.), rinsed twice with 4% STR-FBS/DMEM, remspended a t 5 x lo7 celldml in a 0.1% glucose/PBS+MC solution, and aliquoted a t 0.2 d 4 -m r n gap electroporation cuvette (Bio-Rad). Plasmid DNA (3.5 pmol, -15 pg) was added, and the cell suspension was electroporated at 350 V a t 500 microfarads at room temperature. The electroporated cells were placed at room temperature for 7-10 min, resuspended in 22 ml of 4% STR-FBS/DMEM, and divided to 4 x 6-cm2 dishes (Falcon). After a 16-20-h incubation, two of the four dishes were by guest, on November 26, 2009 www.jbc.org Downloaded from adjusted to 10 p~ forskolin or 0.2% ethanol vehicle for a final 6-h exposure.
Reporter Assays-CAT activity was measured by a modified twophase liquid scintillation assay (Neumann et aE., 1987;Eastman, 1987;Zhang et al., 1992) and normalized per mg of protein (countdmin x min-l x mg'). &Galactosidase assays were performed essentially as previously described (Sambrook et al., 1989). hGH and hCG protein assays were performed with commercial solid-phase radioimmunoassay kits (Hybritech) from media collected 20-24 h post-transfection. The hGH radioimmunoassay is specific for human but not rat GH. The final results are expressed as nanograms of hGH normalized/mIU of hCG expression. LUC protein from transfected cells was harvested by rinsing cells with 5 ml of Dulbecco's phosphate-buffered saline without MgCl, and CaClz (Celox), scraping the cells with a rubber policeman into 1 ml of phosphate-buffered saline, spinning the suspension 5 min at 4 "C, resuspending in 100 plO.1 M potassium phosphate (pH 7.8), 1 IIIM dithiothreitol, freezing, and thawing a total of three times, spinning 5 min at 4 "C, and transferring the supernatant to a new 1.5-ml tube. Immediately, or after overnight storage at -85 "C, 10 pl of cell extract were assayed for LUC activity (deWet et al., 1987) by measuring light emission in a luminometer (model 2010, Analytical Luminescence Laboratories). Protein concentration was measured by Coomassie binding (Pierce) uersus bovine serum albumin standard. LUC activity is expressed as light unitdpg of protein 2 S.E.
Gel Shifi Analysis-The following hGH 5"FR (pCRE or dCRE) or synthetic (CREB, Romega) oligonucleotides were used for the gel shift assays (all sequences are written 5' to 3'; only the sense strand is shown; bold, uppercase nucleotides represent CREs; bold, lowercase GTCAGTGGCCCC(-88); pCREm, (-109)GCACAAGCCgflaccgGGC-nucletotides represent mutations (m)): pCRE, (-1oS)GCACAAGCCC-dCREm, ( -1 9 7 ) C A A C A C T a G c G g C c G -l 7 5 ) ; CREB, AGA-GATTGCC!I'GACGTCAGAGAGCTAG, CREBm, AGAGA'ITGCCTag-atctAGAGAGCTAG. Ten picomoles of sense strand wild-type or mutated oligonucleotide (pCRE, dCRE, or CREB) were phosphorylated with T4 polynucleotide kinase and 5 pmol of [y-32PlATP at 37 "C for 30 min and the kinase inactivated at 65 "C for 5 min. An excess (13.2 pmol) of antisense strand oligonucleotide was added to the labeling reaction mix, incubated at 70 "C for 10 min, and cooled to mom temperature. The annealed, phosphorylated oligonucleotides were separated on a Bio-Gel P60 column (Bio-Rad), and -175-225 pl were recovered from the two or three initial elution peaks containing the highest amount of radioactivity. Unlabeled, double-stranded competitor oligonucleotides were prepared by diluting single-stranded oligonucleotides to 5 pmoVpl in 10 m~ Tris.HC1 (pH 7.5),10 m~ NaCl, and 3 m~ MgCl,, mixing equal volumes of sense and antisense oligonucleotides together, heating to 70 "C for 10 mi n, and cooling to room temperature. The binding reactions were performed by diluting GC, HeLa, JEG3, or BeWo nuclear extract or purified, bacterially expressed rat CREB327 (generous gift from Dr. Richard A. Maurer, University of Iowa) (Sun et al., 1992), at the amounts indicated in the figure legends, to 15 pl with Dignam Buffer D (Dignam et al., 1983), supplemented with 6.25 m~ MgCl,, 0.5 m~ Dm, and 150 ng of sonicated fish sperm DNA in a total volume of 23 pl. Fish sperm DNA was used as a nonspecific competitor because it allowed the best detection of DNA-protein interactions (Lee and Schwartz, 1992) and provided an unbiased representation of competitor DNA sequences. At this point, the mixture was incubated for 10 min at 24 "C under the conditions indicated in the figure legends. Polyclonal rabbit antisera to glutathione S-transferase or glutathione S-transferase fusion proteins of ATF-1 (amino acids 57-2711, ATF-2 (amino acids 1-505), or CREB were generous giRs from Drs. Susanne Wagner and Michael R. Green (glutathione S-transferase, ATF-1, ATF-2; University of Massachusetts) and Dr. Richard A. Maurer (CREB; University of Iowa). End-labeled, double-stranded oligonucleotide (10,000 cpm, 1 fmol) was added for a final 10 min, 24 "C incubation. Samples were loaded on a prerun nondenaturing 4% polyacrylamide, 0.33 x Tris borate-EDTA gel, and electrophoresed ( P r o t e a n IIxi, Bio-Rad) for 2.5 h at 160 V and 4 "C. The gel was then dried onto Whatman No. 3 " chromatography paper at 80 "C and exposed to Kodak XAR2 film with two intensifymg screens at 24 "C for 16 h. W Cross-linking-A standard gel shift reaction with GC or HeLa CCC(-88); dCRE,

( -1 9 7 ) C A A C A C T G G T G A C G T~~~-l 7 5 ) ;
nuclear extract and end-labeled pCRE or dCRE probe (described above) was scaled-up 5-fold and run for 2 h. The gel was left on one glass plate, covered with Saran Wrap, placed on ice in a glass dish, and exposed to W light for 1 h in a Strataliier (Stratagene). The wet gel was exposed to Kodak XAR2 film overnight at 4 "C. Shifted bands were visualized and excised from the gel. The gel slice was equilibrated with 1 x SDS sample buffer (2.5 m~ TriwHCl (pH 6.81, 2.5% SDS, 100 m~ dithiothreitol, 10% glycerol, 0.05% bromphenol blue) for 20 min. The gel slice was then stuffed into the well of a 4% stacking, 10% resolving polyacrylamide, SDS-gel. Prestained broad-range protein markers (Bio-Rad) were run in parallel. The gel was run at 4 "C for 10 min at 25 mA then 5.5 h at 35 mA. The gel was then dried onto Whatman No. 3 " chromatography paper and exposed to Kodak XAR2 film with two intensifying screens at -85 "C for 72 h. Southwestern Blotting-Protein blots for Southwestern analysis were generated by separating 100 pg of GC or HeLa nuclear extract by 10% SDS-polyacrylamide gel electrophoresis, electroblotting onto nitrocellulose (BioTrace NT, Gelman), and denaturinghnaturing the protein with guanidine hydrochloride as described by Vinson et al. (1988). I4C-Labeled protein molecular weight markers (Amersham Corp.) were run in parallel. The filters were incubated with 1 x lo6 cpdml (specific activity, -1 x lo6 cpdpmol) of end-labeled, monomeric oligonucleotide for 2 h at 4 "C. Blots were exposed to Kodak XAR2 film with two intensifying screens at -85 "C for 72 h.

CAMP Stimulates hGH-1 Gene Expression by a lhnscrip-
tioml Mechanism-Cyclic AMP-mediated regulation of hGH gene expression was analyzed through a series of plasmids containing various segments of the hGH gene (Zhang et al., 1992). These plasmids were transiently transfected into GC cells and exposed to forskolin. The first series of plasmid constructions utilized radioimmunological detection of secreted hGH protein as a measure of forskolin stimulation of hGH gene expression. Initially, -500 bp of the hGH 5"FR was coupled to the hGH structural gene containing introns and exons (designated GHs) and -500 bp of the hGH 3"FR (designated GH3') to form GHp.GHs.GH3' (Fig. 1). GHp.GHs.GH3' expression was stimulated -5-fold in forskolin uersus ethanol treated cells (Fig. 1). The hGH structural gene was replaced with the hGH cDNA (designated GHc) to form GHp.GHc.GH3'. Forskolin regulation of GHp.GHc.GH3' was not significantly different from GHp.GHs.GH3' regulation ( Fig. 1). Thus, intronic sequences did not significantly contribute to CAMP regulation of the hGH gene. Next we determined if the -93% identical hCS promoter could mediate the CAMP effect by replacing the hGH 5"FR in GHp.GHc.GH3' with the hCS 5"FR (designated CSp) to form CSp.GHc.GH3'. This switch resulted in significantly reduced (p < 0.0001) forskolin regulation ( Fig. 1) which suggested that the hGH promoter contained CAMP responsive elements not present in the hCS promoter. As a control the human actin promoter (designated ACTp) was inserted upstream of GHc.GH3' to form ACTp.GHc.GH3'. The ACTp.GHc.GH3' construct, like CSp.GHc.GH3', was minimally regulated by forskolin (Fig. 11, further supporting the concept that the hGH promoter mediated forskolin responsiveness. The possibility that the hGH 5"FR required the hGH 3"FR to mediate forskolin responsiveness was explored by replacing the hGH 3"FR in GHp.GHc.GH3' with the heterologous SV40 3"FR (designated SV3') to form GHp.GHc.SV3'. The GHp.GHc.GH3' and GHp.GHc.SV3' plasmids were regulated similarly by forskolin (Fig. 1). Thus the hGH 3"FR did not contribute significantly to CAMP responsiveness of the hGH gene. Conversely, since the hGH 3"FR may have dampened any hCS 5'-FR mediated forskolin responsiveness, the hGH 5"FR in GHp.GHc.SV3' was replaced with the hCS 5'-FR (CSp.GHc.SV3'). CSp.GHc.SV3' forskolin regulation was significantly lower than GHp.-GHc.SV3' (p < 0.001; Fig. l), reconfirming the hGH promoter CAMP response specificity.
The approximate location of the CREs in the hGH promoter was determined by hybrid promoter plasmid constructions. Replacement of the -493" hCS 5"FR in CSp.GHc.SV3' with the -492/-83 hGH 5'-FR to form GHp/-83CSp.GHc.SV3' yielded forskolin regulation which was not significantly different from GHp.GHc.SV3' (Fig. 11, suggesting that the forskolin responsiveness was localized upstream of n t -83. The reciprocal experiment replacing the -492" hGH 5"FR in GHp.GHc.SV3' with the -4931-83 hCS 5"FR to form CSp/-83 GHp.GHc.SV3' yielded forskolin regulation which was not significantly different from CSp.GHc.SV3', thus supporting the upstream localization of the hGH 5'-FR CAMP responsive region. CAMP Stimulation of the hGH Promoter Is Mediated by Protein Kinase A-Forskolin is known to stimulate adenylate cyclase, elevate CAMP levels, and activate protein kinase A (Seamon and Daly, 1986) but reports suggesting that forskolin has other specific effects (Hoshi et al., 1988;Wagoner and Pallotta, 1988) prompted us to verify forskolin specificity in GC cells. The hGH or hCS promoter was cloned upstream of the CAT gene (GHp.CAT or CSp.CAT) and cotransfected into GC cells with either wild-type or nonfunctional mutant protein kinase A inhibitory protein cDNA driven by the Rous sarcoma virus long terminal repeat (RSVp.PKI or RSVp.PKImut) and assayed for forskolin inducibility. Heterodimers of the protein kinase A catalytic subunit and protein kinase A inhibitory protein can effectively block CAMP stimulation of protein kinase A activity (Grove et al., 1987). Cotransfection of RSVp.PKI led to complete loss of CAMP stimulation but did not affect basal GHp.CAT and CSp.CAT activity in transfected GC cells ( Table   I). Cotransfection of RSVp.PKImut was without effect ( Table I).
This suggested that the forskolin regulation of the hGH pro- GHp.CAT and CSp.CAT plasmids (15 pg) were co-transfected with 15 pg wild-type (RSVp.PKI) or mutant (RSVp.PKImut) PKA inhibitory protein cDNA expression vectors (Grove et al., 1987) and 5 pg of RSVp.P-Gal as a transfection efficiency control.

Protein kinase inhibitor effects on hGH-1 and hCS-1 promoter activity
Basal CAT activity was determined as countdmin x min" and normalized to p-galactosidase activity (Zhang et al., 1992). E Forskolin induction ratios were determined as in the legend to Fig. 1.
Number of independent transfections. moter, and even the marginally regulated hCS promoter, were mediated by the CAMP-dependent protein kinase A pathway.
CAMP Responsive Elements in the hGH Promoter Localize to a t Least Two Distinct Regions-To localize the hGH 5"FR CREs, the hGH promoter in GHp.CAT was progressively deleted from its 5' end ( Fig. 2). Deletion of the hGH 5"FR from nt -190 to -163 led to a small, but significant ( p < 0.05) drop in forskolin inducibility (Fig. 2). A second, more pronounced drop ( p < 0.001) in forskolin inducibility was evident upon deletion from n t -135 to -87 (Fig. 2). This suggested that CREs were located in two regions, nt -1901-163 and -1351-87, of the hGH promoter. These results confirm the hybrid promoter results obtained in Fig. 1 which suggested localization of CREs upstream of nt -83. Deletion of the hGH 5"FR to nt -45 resulted in no further loss of forskolin stimulation (Fig. 2). However, complete removal of the hGH 5'-FR further reduced forskolin responsiveness, suggesting that the hGH 5"FR nt -45/+6 may contain a partial CRE. Transfection of the heterologous thymidine kinase (TK) promoter (nt -109/+51)/CAT plasmid (TKp.CAT) (Cattini and Eberhardt, 1987) resulted in an -2fold forskolin induction. This is essentially the same stimulation as seen with -45 GHp.CAT and suggests that this regulation may be mediated by common elements.
T w o CGTCA Motifs Are Required for CAMP Regulation of the hGH Promoter-Sequence inspection of the two CAMP-responsive regions of the hGH promoter (nt -190/-163 and -1351-87) revealed a CGTCA motif at nt -188/-184 (antisense strand; designated dCRE) and nt -loo/-96 (sense strand; designated pCRE) (Fig. 3A). To determine if the CGTCA motifs represented the CREs, these sequences were mutated within the 500-bp hGH promoter and cloned upstream of the more sensitive firefly luciferase reporter gene (designated GHp.LUC) (deWet et al., 1987). The dCRE sequence was substituted with a Not1 restriction enzyme site to form GH(dCRE)p.LUC whereas the pCRE sequence was substituted with a KpnI restriction enzyme site to form GH(pCRE)p.LUC (Fig. 3A). A combination of the above pCRE and dCRE substitution mutations was also constructed to form GH(pdCRE)p.LUC. Elimination of one or both CGTCA motifls) resulted in a significant drop in GH promoter forskolin inducibility relative to GHp.LUC ( p < 0.0005, Fig. 3B). The pCRE is apparently more critical than the dCRE since the loss in forskolin induction is more pronounced with the mutated pCRE than the mutated dCRE (Fig. 3B). This is consistent with the 5"deletion data (Fig. 2) which showed a smaller drop in forskolin induction from -190 to -163 than from -135 to -87. The promoterless (pA3.LUC) and RSV proximal promoter (RSV13,p.LUC) control plasmids were slightly negatively regulated by forskolin, which suggested that there was specificity to the residual forskolin inducibility seen with the hGH promoter mutants. Nonetheless, these data indicate that maximal CAMP stimulation of the hGH promoter requires at least two CGTCA elements.

Distal GHFl Binding Site Is also Required for CAMP-regu-
luted hGH Gene Dunscription-The hGH promoter 5' deletion data suggested that the -135/-87 region was necessary for forskolin responsiveness. In addition to the pCRE site, this region contains the distal-most GHFl binding site (designated dGHF1) of the two known hGH 5"FR GHFl binding sites. To test GHFl involvement in CAMP regulation, we mutated the dGHFl binding site in GHp.LUC to form GH(dGHF)p.LUC and assessed its forskolin responsiveness in transiently transfected GC cells. The dGHFl site was mutated by substitution in the core GHFl binding site from TAAA'M'ATCCAT to gcgA'M'ATC-CAcTA (Fig. 3A 1. Mutation of the TAAAT motif has been shown to reduce GHFl binding affinity about 4-fold in a linker-scanning mutation (Lefevre et al., 1987;Bodner et al., 1988) and the T to C transition at nt -112, which is a natural difference between hGH and hCS, resulted in reduced GHFl binding (data not shown). Forskolin induction of GH(dGHF)p.LUC was significantly reduced ( p < 0.0005) relative to GHp.LUC (Fig.   3B 1, indicating that the dGHFl site is critical for CAMP responsiveness. The incomplete loss in forskolin response seen with GH(dGHF)p.LUC, GH(pCRE)p.LUC, and GH(pdCRE)p.LUC, may be due to the presence of response elements downstream of nt -45 as seen in Figs. 1 and 2. Based on the above deletion and mutation results it appears that at least three distinct cis elements are necessary for maximal forskolin-regulated hGH transcription.
To further assess GHFl involvement in CAMP regulation of the hGH promoter, we overexpressed GHFl in GHp.LUC transfected GC cells. GHp.LUC basal activity was elevated -10-fold but forskolin inducibility was unaffected (Fig. 3C). The elevated GHp.LUC basal activity in GHFl uersus control p-galactosidase cotransfected cells suggested that GHFl was indeed being overexpressed and that GHFl levels were limiting GHp.LUC activity. The unaffected forskolin response, however, suggests that a mechanism other than modulation of GHFl protein levels or posttranslational modification status is responsible for CAMP regulation of the hGH promoter.
Specific Binding of CREB IATF-related Proteins to the Proximal and Distal CGTCA Motifs-% support the functional role of the CGTCA motifs in the hGH promoter CAMP response, we attempted to show specific binding of protein to these sites. Numerous DNase I footprinting analyses of the hGH 5"FR with crude and partially purified GC nuclear extracts by us and others have failed to show protection over either CGTCA motif (data not shown) (Lefevre et al., 1987;Bodner and Karin, 1987;Nickel et al., 1991). However, given the relative insensitivity of DNase I footprinting we performed the more sensitive gel shift analysis using end-labeled oligonucleotides spanning the hGH dCRE and pCRE CGTCA motifs. As a positive control we included a synthetic, palindromic TGACGTCA-containing oligonucleotide (CREB consensus, Promega). Increasing concentrations of crude GC nuclear extract incubated with end-labeled pCRE, dCRE, or CREB oligonucleotides, generated specific DNA-protein complexes (Fig. 4A). GC nuclear extract bound the radiolabeled probes with the relative binding affinity order pCRE < dCRE < CREB (Fig. 4A, compare lanes 3 , 8, and 13). The pCRE and dCRE DNA-protein complexes migrated at the same relative mobility as the CREB DNA-protein complex suggesting that similar CREB/ATF-related factors may bind these sites.

FIG. 4. Gel shift analyses of pCRE and dCRE oligonucleotides.
Gel shift analyses were performed as described under "Experimental Procedures." Essentially, each gel shift was incubated for 10 min a t 24 "C with protein and competitor DNA followed by a 10-min, 24 "C incubation with 10,000 cpm (1 fmol) end-labeled probe. A, end-labeled pCRE, dCRE, or CREB oligonucleotide was incubated with increasing concentrations by guest, on November 26, 2009 www.jbc.org

Downloaded from
To determine the tissue specificity of the CGTCA-binding proteins we tested nuclear extracts prepared from human placental choriocarcinoma (BeWo and JEG-3) and human cervical carcinoma (HeLa) cell lines in the gel shift assay. JEG-3 and HeLa nuclear extracts formed a complex with pCRE (Fig. 4F,  lanes 3 and 4, respectively), dCRE (Fig. 4F, lanes 7 and 8, respectively), and CREB (Fig. 4F, lanes 11 and 12, respectively) oligonucleotide which migrated to the same relative position as that with GC nuclear extract (Fig. 4F, lanes 1,5, and 9). This suggested that the CGTCA-binding proteins were present in a non-tissue-specific fashion. BeWo nuclear extract was unable to form this specific complex with the end-labeled DNA (Fig. 4F,  lanes 2,6, and 10) which may indicate that the CGTCA-binding proteins are not present in these cells or that the nuclear extract was degraded upon preparation. The significance of the other DNA-protein complexes formed with JEG-3, HeLa, and BeWo nuclear extracts is unclear but may represent specific or nonspecific complexes or proteolytic degradation products.
Proteins of -100 a n d -45 kDa Bind Predominately to the Proximal a n d Distal CGTCA Motifs-The molecular weight of the proteins bound to the pCRE and dCRE sites was determined by two methods. We first attempted in situ UV crosslinking of DNA-protein complexes by UV irradiating a gel shift gel containing pCRE or dCRE probe shifted with GC or HeLa nuclear extract, isolating the specifically shiRed bands, and separating the cross-linked from uncross-linked products by SDS-PAGE. This yielded many cross-linked species of which an -130-kDa complex was predominate for both the pCRE and dCRE oligonucleotides (Fig. 5 A ) . The specificity of the crosslinked products is implicit to in situ UV cross-linking since the isolated gel shift bands were shown previously (Fig. 4B) to be specific. Due to the low efficiency of UV cross-linking, the -130-kDa band probably consists of a 1:l DNAprotein complex. Subtraction of the oligonucleotide molecular weight (-15 kDa) from the cross-linked complex yields a protein of -115 kDa.
To more accurately estimate the molecular weight of the proteins binding the pCRE and dCRE oligonucleotides we performed Southwestern blotting (Viison et al., 1988). Filter strips were probed with end-labeled, monomeric wild-type or mutated pCRE or dCRE oligonucleotides. We also probed with wild-type CREB oligonucleotide as a positive control. The pCRE (Fig. 5B,  lanes 1 and 3) and dCRE (Fig. 5B, lanes 5 and 7) oligonucleotides predominately bound -100-and -45-kDa proteins in both GC and HeLa nuclear extracts. These interactions were specific for the CGTCAmotifs since the mutated pCRE (Fig. 5B, lanes 2 and 4) and dCRE (Fig. 5B, lanes 6 and 8) oligonucleotides did not bind these proteins or bound them with reduced affinity. The other bands present on the blots were probably nonspecific since they bound both wild-type and mutated DNA. The positive control CREB probe (Fig. 5B, lanes 9 and 10) also bound a n -100-kDa protein and a protein in the -45-kDa range.

DISCUSSION
We have demonstrated that two CGTCA motifs, designated pCRE and dCRE, and the distal GHFl binding site comprise a functional CAMP responsive unit (CRU) on the hGH promoter. The pCRE, dCRE, and dGHFl sites are located between nt -190 and -85 and are required for most of the forskolin responsiveness of the hGH promoter (Figs. 1-31. The remaining forskolin responsiveness is located downstream of these elements and may be part of the general transcriptional machinery. Our results support and extend those of Brent et al. (1988) who localized the cAMP-responsive region of the hGH promoter to within n t -2121-83. The slight forskolin responsiveness seen downstream of hGH 5"FR nt -85 also partially supports the conclusions of Dana and Karin (1989) who localized the CRE downstream of nt -85. However, the bulk of CAMP regulated activity arises from elements upstream of nt -85. The reason for this discrepancy is unclear, but may have to do with the less sensitive CAT constructs used by Dana and Karin (1989). Both reports suggested GHFl involvement in the CAMP response which is consistent with our finding with the dGHFl site. The CAMP regulation of the hGH promoter is fully dependent on protein kinase A, since it is completely blocked by coexpression of a protein kinase A inhibitory protein ( Table I).
The hGH promoter contains high affinity (nt -285I-278) and low affinity (nt -163I-156) AP-2 binding sites that are potentially capable of mediating CAMP responsiveness . However, our deletion data (Fig. 2 A ) indicate that in GC cells these sites are not fbnctional in the hGH promoter, confirming the findings of Dana and Karin (1989). Copp and Samuels (1989) reported similar results with the homologous rGH promoter. Thus AP-2 probably does not contribute to CAMP stimulation of the hGH promoter in GC cells, possibly due to the relatively low abundance ofAP-2 in these cells (Dana and Karin, 1989).
The exact identity of the hGH promoter CGTCA-binding proteins is unknown. We have shown that the CGTCA-binding proteins are CREBIATF-1-related by thermostability, competition with consensus CREB oligonucleotide, and immunoreactivity with CREB and ATF-1 antisera (Fig. 4). Our UV crosslinking and Southwestern blot data (Fig. 5) suggest that -45and -100-kDa proteins may be confemng CAMP responsiveness on the hGH promoter. The -100-kDa CGTCA-binding factor may be related to a previously identified -120-kDa TGACGTCA-binding factor purified from HeLa cells (Andrisani and Dixon, 1990). The -100-kDa CGTCA-binding factor may also be related to the hPRL promoter CGTCA-binding factor (Peers et al., 1992) especially given the homology between the GH and PRL genes. Cloning of the hGH and hPRL promoter -100-kDa CGTCA-binding proteins and HeLa cell -120-kDa TGACGTCA-binding factors will be necessary to determine their relatedness. The -45-kDa CGTCA-binding protein may correspond to the CREB factor of -43-kDa (Gonzalez et al., 1989) but is unlikely to represent ATF-1 since its molecular mass is only -38-kDa (Liu et al., 1993). The similarity between the pCRE, dCRE, and CREB oligonucleotide-bound proteins (Figs. 4 and 5) supports this possibility. The ability of polyclonal antisera to both ATF-1 and CREB to react with the pCRE, dCRE, and CREB DNA-protein complexes is not surprising given the immunological-relatedness of the CREBIATF family (Hai et al., 1988). By the same reasoning it is perhaps of GC nuclear extract (0-38 pg). B, end-labeled wild-type pCRE, dCRE, or CREB oligonucleotides were incubated with GC nuclear extract (15 pg) in the absence (lanes 1,6,11) or presence (lanes 2-5,7-10, and 12-15) of unlabeled wild-type or mutated oligonucleotides (25 pmol). C, GC nuclear extract (15 pg) was incubated at 24,37,55,70, or 100 "C for 10 min prior to incubation with end-labeled pCRE, dCRE, or CREB oligonucleotides. (lanes 1, 6 , l l )  using GC and HeLa nuclear extracts and "P-labeled pCRE or dCRE were scaled up 5-fold. After separating the free from the bound probe on a native polyacrylamide gel, the gel was exposed to UV light for 1 h. DNA-protein complexes were visualized, excised, equilibrated in SDS sample buffer, and resolved on an SDSpolyacrylamide gel. Molecular mass marker locations are indicated between gel lanes and sizes (kDa) are indicated in the left margins. Specific DNA-protein complexes are indicated by an arrow in the right margins. Uncross-linked probe migrated off the gel. B , Southwestern blot analysis was performed as described under "Experimental Procedures." GC or HeLa nuclear extracts (100 pg) were resolved on an SDS-polyacrylamide gel and electrotransferred to nitrocellulose. The nitrocellulose was cut into strips, denaturedhenatured with guanidine hydrochloride, and probed with end-labeled pCRE, pCREm, dCRE, dCREm, or CREB oligonucleotides. Molecular mass marker sizes ( m a ) and locations are indicated in the left margin. Specific DNA-protein complexes are indicated by arrows in the right margins. surprising that ATF-2 antisera is unable to react with the CGTCA-binding factors, but underscore the specificity of the antisera.

D, GC nuclear extract (15 pg) and end-labeled pCRE, dCRE, or CREB oligonucleotides were incubated in the absence
Based on oligonucleotide competition gel shift experiments, Peers et al. (1991) concluded that CREB does not bind the CGTCA motif in the hPRL gene. Liang et al. (1992) concluded that CREB is not responsible for the CAMP regulation of the homologous rPRL promoter, since CREB antibodies did not react with rPRL gel-shifted proteins and co-transfection of a dominant negative inhibitor of CREB had little effect on CAMPmediated rPRL promoter activity. Using DNase I protection analysis, Keech et al. (1992) demonstrated that affinity purified CREB from GC cells did not interact with the proximal rPRL promoter. Nevertheless, our results suggest that a CREB-related factor might be involved in regulating the hGH promoter CAMP response. Our finding that the hGH gene is transcriptionally up-regulated by CAMP parallels findings with the rGH gene (Barinaga et al., 1983(Barinaga et al., , 1985. This raises the possibility that the rat and human GH genes could be similarly regulated by CAMP. However, the rGH gene does not have CGTCA motifs in its 5'-FR. Interestingly, it does contain two tandemly repeated CGTCA motifs at nt 2255 and 2276 on the antisense strand of the 3"FR downstream of the polyadenylation signal but these have not been characterized. Based on these dissimilarities it is likely that the rGH and hGH genes may be transcriptionally regulated by CAMP but by different mechanisms. The homologous hCS 5"FR does contain a CGTCA motif, analogous to the hGH pCRE, which represses basal activity but not CAMP-responsiveness on this marginally CAMP responsive promoter (data not shown).
Multiple CGTCA motifs are a common feature of many CAMP-responsive promoters. CGTCA motifs spaced by 10 bp are present in the human vasoactive intestinal polypeptide gene (Fink et al., 1988). Tandem repeats of the palindromic TGACGTCA sequence with 1-bp intervening DNA are found in the human glycoprotein hormone a gene (Drust et al., 1991). Three CGTCA repeats are present in the HTLV-1 long terminal repeat (Beimling and Moelling, 1992;Zhao and Giam, 1992). Multimers of the rat PRL promoter CGTCA motif support CAMP regulation (Liang et al., 1992). Palindromic TGACGTCA and non-palindromic GTACGTCA sequences with 35-bp intervening DNA are present in the GHFl promoter (McCormick et al., 1990;Chen et al., 1990). The human CREB gene 5"FR contains three CGTCA motifs with one located upstream of the transcriptional start site and two located 21 bp apart between the transcriptional and translational start sites . Multiple CGTCA sites may allow cooperative interactions between related or unrelated proteins, thereby stabilizing individual CGTCA-binding protein interactions and providing more efficient CAMP responsiveness. The hGH promoter CGTCA motifs are unusual given their relatively long separation. The center-to-center spacing of the CGTCA motifs corresponds to 8.5 turns (88 bp) of B form DNA helix, which places these motifs out of phase with one another. However, the proximal CGTCA motif is on the sense strand and the distal CGTCA motif is on the antisense strand which might consequently place their cognate factors in phase. This would allow interaction of their respective binding factofis) with each other or the basic transcriptional apparatus, provided that the DNA could bend to allow protein-protein contacts.
Our results overexpressing GHFl indicate that GHFl levels affect the basal GHp.LUC activity but not its CAMP responsiveness (Fig. 3C). Perhaps GHFl plays a structural role on the hGH promoter and needs merely be present for CAMP inducibility. In fact, Verrijzer et al. (1991) showed that GHFl induces DNA bending when bound to its cognate sequence in vitro, suggesting that GHFl binding might enhance interaction be-tween the distant CGTCA-binding proteins. Further studies will be required to test this model.
The hGH CRU appears to be similar to the GHFl CRU (McCormick et al., 1990), hPRL CRU (Peers et al., 1991) and rPRL CRU (Liang et al., 1992;Keech et al., 1992) which consists of one or more CGTCA motifs and a GHFl binding site. The hGH and hPRL genes are structurally related members of the same gene family (Miller and Eberhardt, 1983), are expressed in pituitary cells, and are dependent on GHFl for their cell-specific expression. The GHFl gene is positively autoregulated by GHFl and requires both CREB and GHFl for CAMP responsiveness (McCormick et al., 1990). Coordinated control by GHRH of GHFl and GH gene expression is therefore possible. Cooperative interactions between GHFl and the factors that recognize the hGH or PRL CGTCA motifs may account for the cell-specific, CAMP regulation of these promoters. Similar cooperative mechanisms have been proposed for the repression of tyrosine aminotransferase gene expression in non-liver cells by the tissue-specific extinguisher Tse-1 which has an absolute dependence on a CRE for its function (Boshart et al., 1990).