INTRODUCTION

The rat growth hormone (rGH)¹ gene is expressed selectively in a subpopulation of anterior pituitary cells. rGH promoter sequences between 237 and +8 (where +1 is the transcription start site) drive the maximal expression of a linked reporter gene transiently transfected into cultured pituitary cells (reviewed in Refs. 1, 2, 3). Promoter mutagenesis and transient transfection demonstrated that several different sequence elements are critical for rGH promoter activity. The nonadditive effects of those promoter mutants suggested that cooperation between those factors was crucial to the high level and pituitary-specific regulation of the rGH promoter (3). The importance of these cooperative activities was underscored by studies showing that the rGH promoter could be strongly activated in nonpituitary cells (4, 5, 6) only by the co-expression of the pituitary-specific factor (7, 8, 9, 10, 11, 12, 13) Pit-1, and either the thyroid hormone receptor (T₃R) or ZN15. The transcriptional consequences of repositioning binding sites within the rGH promoter (7, 14, 15) further suggested that the integration of binding factor activities is crucial to promoter function. Thus, a detailed knowledge of the factors that bind to the rGH promoter and their cooperative and/or mutually disruptive activities will be required to understand the physiologic and ontologic regulation of GH transcription.

Previously, we identified a factor termed GHF3 that binds to rGH promoter sequences between 239 and 219 and is one of the strongest determinants of rGH promoter activity; promoter mutants that disrupted the GHF3 binding site reduced the activity of the rGH promoter transfected into GH-secreting rat pituitary GC cells by 75% (16). Gel mobility shift experiments showed that complexes of differing electrophoretic mobility bound to the GHF3 binding site and that the identical site was footprinted in each of those complexes (16). Two-dimensional gel mobility shift assays showed that each complex shared a DNA-binding subunit of similar electrophoretic mobility (16). Consistent with the presence of a common DNA-binding subunit, uv cross-linking of the different complexes to the GHF3 binding site demonstrated that each complex contained a DNA cross-linked factor of identical size, approximately 46 kDa (16). This agreed well with the predominant 43-kDa GHF3 DNA-binding subunit detected by a Southwestern blot using a radiolabeled GHF3 DNA binding site to probe crude GC cell extracts separated electrophoretically according to size (16). Thyrotroph embryonic factor (TEF), which binds to multiple sites within the promoter for the gene encoding the -subunit of the thyroid-stimulating hormone, also bound to the rGH promoter over the GHF3 site (17). However, the 29.3-kDa size of TEF (17) does not agree well with the ~43-kDa size of the predominant GHF3 DNA-binding subunit present in GC cells or even with other minor 35-, 58-, 72-, and 140-kDa GHF3-binding proteins detected by Southwestern blot (16).

Given the importance of the GHF3 site, it was necessary to identify the factor that binds to it to understand the control of rGH expression. This factor was of additional interest because it is the DNA-binding subunit common to the array of alternate multisubunit GHF3 complexes. We demonstrate here, by antibody interference with gel mobility shift complexes and by Western blots of GHF3 factors purified by DNA affinity chromatography, that the GHF3 DNA-binding subunit contains C/EBP. A role for C/EBP in the pituitary-specific regulation of the GH gene was further supported by a number of observations: 1) C/EBP and Pit-1 synergistically activated the rGH promoter in human monocyte U937 cells; 2) full-length C/EBP was detected in extracts of GH-secreting GC cells and prolactin-secreting 235-1 cells but not in Pit-1-containing, GHFT1-5 pituitary progenitor cells transformed at a developmental stage prior to the expression of any mature pituitary hormones (18); and 3) transient expression of C/EBP in GHFT1-5 cells strongly activated a co-transfected rGH promoter, implying that C/EBP may be limiting for rGH expression in these pituitary progenitor cells. 60% of the C/EBP activation in GHFT1-5 cells required the GHF3 binding site, whereas the remaining activation required rGH promoter sequences between 106 and 33. rGH promoter activation by C/EBP was synergistic with cultivating the GHFT1-5 cells with PMA and forskolin, activators of protein kinases C and A, respectively (19, 20) further demonstrating the importance of integrative functions in the regulation of high level rGH promoter activity. Thus, C/EBP is an important factor regulating high level transcription of rGH promoter activity and appears to function in synergy with Pit-1, activators of protein kinases A and C, and possibly other factors through the GHF3 and secondary activation sites.

EXPERIMENTAL PROCEDURES

Affinity-purified Pit-1 and 20 µg of nuclear extracts, prepared as described previously (21), were separated by discontinuous SDS-polyacrylamide gel electrophoresis and electroblotted to nitrocellulose. GHF3 purified 600-fold by one round of DNA binding site affinity chromatography (16) was subjected to a second round of affinity purification and simultaneously blotted. Blots were blocked with 5% nonfat dry milk powder and probed with the indicated primary antibodies. C/EBP, C/EBP, and C/EBP polyclonal antibodies were a gift from W.-C. Yeh and S. L. McKnight (Tularik, South San Francisco). Anti-TEF antibodies were obtained from D. Drolet and M. G. Rosenfeld (University of California, San Diego), and anti-DBP antibodies were received from D. Lavery and U. Schibler (University of Geneva). Pit-1 polyclonal antibodies were purchased from Berkeley Antibody Company (BAbCo). The blots were then probed with anti-rabbit horseradish peroxidase-linked secondary antibodies (Amersham Corp. and Promega Corp., respectively) and developed using the ECL chemiluminescence solutions (Amersham).

Gel mobility shift assays were conducted essentially as described previously (21) except that affinity-purified GHF3 (16) was first incubated for 15 min at room temperature with the indicated antibodies or nonimmune serum. Following preincubation, a radiolabeled 87-base pair fragment containing rGH promoter sequences between 285 and 198 that span the GHF3 site (239 to 219), was added with or without 50 ng of a cold competitor oligonucleotide (240 to 216) (16). pGHF1 and GHF2 control competitor oligonucleotides were previously described (11). Because of nonspecific DNA binding activities in the nonimmune and C/EBP-primed sera, 4 µg of poly(dI-dC) (Pharmacia Biotech Inc.) had to be added to each incubation. Poly(dI-dC) was added at the time of probe and oligonucleotide addition.

The 237/+8 promoter and 230/226 mutation thereof (see Fig. 6, mut/237) (16), the 106/+8 (8), the 33/+8, and the 33/+8 promoter to which six copies of the GHF3 binding site were appended (16) were previously described. +1 is the transcription start site. All rGH promoters were cloned upstream of the coding sequences of the bacterial chloramphenicol acetyltransferase (CAT) gene carried in a pUC vector deleted of sequences between AatII and EcoRI (4, 22). Use of the modified pUC vector was critical, since the 33/+8 rGH promoter carried in the standard pUC was strongly activated by C/EBP expression (data not shown), whereas parallel transfected 33/+8 promoters carried in the modified pUC vector were not (see Figs. 6 and 7). Thus, there appears to be a C/EBP activation site present in the standard pUC vector, but this site was eliminated in the experiments described here.

PMA and forskolin-dependent and -independent activation of the wild type 237/+8 rGH promoter (

) by C/EBP in GHFT1-5 cells is reduced by the 230/226 promoter mutant (

GHF3 binding sites confer C/EBP activation to the nonresponsive, 33/+8 rGH promoter.

8

5

U937 or GHFT1-5 cells were grown in RPMI 1640 or DME-H21, respectively, containing 10% fetal calf serum. Transfection was by electroporation using the Bio-Rad electroporation apparatus set at 960 microfarads and 0.3 V. The electroporation buffer consisted of phosphate-buffered saline containing 0.1% glucose and 10 µg/ml of Biobrene Plus (Applied Biosystems Inc.). 5 µg of each promoter was co-transfected with 5 µg (see Fig. 6) or 10 µg (see Figs. 3, 4, 5) of a C/EBP cDNA expression vector, 5 µg (see Fig. 4) or 10 µg (see Fig. 3) of a Pit-1 cDNA expression vector, or compensatory amounts of a control, empty (no cDNA inserted) expression vector. The amounts of C/EBP and control expression vectors were varied in Fig. 7 as indicated. 1 µg of a firefly luciferase cDNA expressed under the control of the Rous sarcoma virus promoter was co-transfected with each point. Transfected cells were induced the following day with 10⁵ forskolin (Sigma) and 10⁸ PMA or control delivery vehicles and collected 1 day after PMA and forskolin induction. CAT activity, luciferase activity, and protein amounts were determined for each extract. Luciferase activity normalized to the amount of extract protein was marginally increased by C/EBP expression when GHFT1-5 cells were incubated with PMA and forskolin (2.0 ± 0.7-fold over 10 experiments) but not when incubated with control delivery vehicles (1.0 ± 0.4). CAT activities, derived from the rGH237 promoter and normalized to the amount of extract protein, were much more dramatically affected by C/EBP expression (see Figs. 3, 4, 5, 6, 7). A minimum of three independent experiments (see figure legends for n) were normalized to the expression level of the rGH237 promoter activated by C/EBP expression and PMA/forskolin incubation (100%) and presented as the percentage mean ± S.D.

C/EBP synergizes with Pit-1 to activate the rGH promoter in U937 cells.

8

5

The rGH promoter is activated by C/EBP expression in mouse pituitary progenitor cells.

8

5

Maximal synergistic activation of the 237/+8 rGH promoter is achieved when GHFT1-5 cells transfected with C/EBP are incubated with both PMA and forskolin.

8

5

RESULTS

C/EBP Co-purifies with GHF3 Binding Activity

Although TEF transcripts were detected in GH-secreting GC cells (17) and TEF protein was capable of binding to the GHF3 site in the rGH promoter (17), the size of TEF is different from the size of the major GHF3-binding protein found in GC cells (16). We also did not detect TEF in Western blots (data not shown) of GHF3 DNA-binding subunit purified from GC cell extracts by GHF3 binding site affinity chromatography (16), although the lack of a Western signal with extracts from a variety of pituitary and nonpituitary cells limited the conclusion that no anti-TEF cross-reacting material is present in affinity-purified GHF3.

TEF is a member of a large family of transcription factors, the bZIP proteins, that bind to similar DNA sequences and share homology in their DNA binding and dimerization domains (23, 24). Since only the portions of TEF that were conserved with other bZIP proteins were necessary for binding to the GHF3 site (17), it was possible that the authentic GHF3 DNA-binding subunit was another bZIP factor. Western blots using polyclonal antibodies directed against a 42-kDa bZIP protein, C/EBP (25, 26), detected an appropriately sized factor present in the GHF3-binding fractions enriched from GC cell extracts by DNA affinity chromatography (Fig. 1A, purified GHF3). 34- and 30-kDa proteins that cross-reacted with the C/EBP antibody were also detected in the affinity-purified material. In contrast, C/EBP antibodies did not cross-react with Pit-1 prepared from the same extracts by DNA affinity chromatography (purified Pit-1), although Pit-1 antibodies were cross-reactive with affinity-purified Pit-1 (Fig. 1B).

C/EBP is enriched in affinity-purified GHF3 DNA binding subunit.

8

5

C/EBP Antibodies Inhibit GHF3 Binding Activity

Thus, C/EBP co-purifies with GHF3 binding activity and is of a size consistent with the GHF3 DNA-binding factor detected by uv cross-linking and Southwestern analysis (16). To determine if C/EBP merely co-purified with the binding activity or whether C/EBP is an integral component of the GHF3 complex, we examined the effect of C/EBP antibodies on GHF3 binding. Anti-C/EBP antibodies inhibited the formation of the gel shift GHF3 complex formed with affinity-purified GHF3-binding factor (Fig. 2), whereas they did not affect the ability of affinity-purified Pit-1 to bind to DNA (data not shown). Nonimmune sera or polyclonal antibodies against other bZIP transcription factors including C/EBP (26), TEF (17), and DBP (27) did not affect the gel shifts.² C/EBP therefore co-purifies with the DNA-binding subunit of the multisubunit GHF3 complexes and is required for GHF3 factor binding to the GHF3 site.

Antibodies against C/EBP disrupt gel mobility shift complexes formed between the affinity-purified GHF3 DNA-binding subunit (

) and a radiolabeled rGH promoter fragment spanning the GHF3 binding site (285/198).

Synergistic Activation of the rGH Promoter by C/EBP and Pit-1

C/EBP is present in a number of cell types in which the GH gene is not transcribed (28, 29, 30), yet the activity of the rGH promoter is strongly dependent upon the GHF3 binding site (16). The rGH promoter is inactive in human monocyte U937 cells but can be activated by Pit-1 and TR co-expression (4). The rGH promoter was also synergistically activated by the co-expression of C/EBP and Pit-1 to a level that was 3.6 ± 0.3-fold greater than would be expected from the summation of the increases in activity from independent Pit-1 and C/EBP expression (Fig. 3). Mutation of the GHF3 binding site abolished synergy (data not shown).

In U937 cells, the rGH promoter activated by both Pit-1 and C/EBP was 37.9 ± 12.9% as active as the Pit-1- and TR-activated promoter in parallel experiments, whereas the much weaker activations by independent TR or C/EBP expression were equivalent (TR was 1.2-fold more active on average). The 72-125 mutation in Pit-1 that selectively reduces Pit-1 synergy with TR without affecting the intrinsic activation ability of Pit-1 (6) marginally reduced Pit-1 synergy with C/EBP from 3.6- to 2.0-fold (Fig. 3). In contrast, the same mutation dramatically reduced Pit-1 synergy with TR from 10.2- to 3.0-fold in parallel experiments. Thus, there are some differences in the sequence requirements for Pit-1 synergy with C/EBP or with TR.

C/EBP Activation of the rGH Promoter in GHFT1-5 Pituitary Progenitor Cells

GHFT1-5 cells are an excellent cell line for studying the regulation of rGH gene expression, since they represent an embryonic pituitary cell type transformed during the developmental window in which Pit-1 is expressed prior to commitment to the GH-, TSH-, or PRL-secreting cell lineages (18) (Fig. 1B).² Therefore, factors other than Pit-1 expression must limit rGH expression in these cells. The 43-kDa C/EBP species was not detectable in Western blots of GHFT1-5 extracts, although C/EBP was observed in GC cells, prolactin-secreting 235-1 cells, and human cervical carcinoma HeLa cells (Fig. 1A). A 34-kDa factor that cross-reacted with the anti-C/EBP antibody was detected in extracts from all three pituitary cell types (Fig. 1A), and a 30-kDa cross-reacting factor was detected in the GC and 235-1 cell extracts. Neither the 30- nor the 34-kDa species were detected in HeLa cell extracts (Fig. 1A).

The wild type 237/+8 rGH promoter was weakly active when transfected into GHFT1-5 cells (Fig. 4). Co-expression of C/EBP resulted in a 4.5-fold, on average (n = 9), enhancement of CAT activity expressed from the rGH promoter. In contrast, Pit-1 co-expression did not activate the rGH promoter and did not further enhance the C/EBP-activated rGH promoter (Fig. 4). The Pit-1 expression vector is active in GHFT1-5 cells under these conditions, since its transfection strongly activates the prolactin promoter,³ minimal promoters to which rGH Pit-1 binding sites are multimerized and appended (6), and even the 237/+8 rGH promoter, but only if TR is co-expressed and the transfected GHFT1-5 cells are incubated with PMA and forskolin (6).

Synergistic Activation by C/EBP, PMA, and Forskolin

Since Pit-1 and TR activate the rGH promoter only in PMA- and forskolin-induced GHFT1-5 cells, the effect of PMA and forskolin on the activation of the rGH promoter by C/EBP and/or Pit-1 was investigated. Incubation of C/EBP-transfected GHFT1-5 cells with PMA and forskolin caused a 15.4-fold, on average (n = 9), enhancement of the rGH promoter, much greater than the sum of the 1.8- and 4.5-fold effects caused by independent PMA/forskolin incubation and C/EBP activation, respectively (Fig. 4). Pit-1 did not enhance C/EBP activation of the rGH promoter even when the GHFT1-5 cells were incubated with PMA and forskolin.

Synergistic activation by C/EBP and PMA/forskolin incubation was maximal when both protein kinase inducers were present (Fig. 5). C/EBP activation after incubation with both PMA and forskolin was 1.8-fold higher than the sum of the increases by independent PMA and forskolin incubation over C/EBP activation in the absence of PMA and forskolin. PMA was independently able to synergize with C/EBP, albeit more poorly than when both PMA and forskolin were present.

Two rGH Promoter Sites for C/EBP Activation

Point mutations disrupting the GHF3 binding site (16) in the 237/+8 promoter showed that 60% of the C/EBP activation in GHFT1-5 cells, in either the presence or absence of Pit-1 or PMA/forskolin, required the GHF3 binding site (Fig. 6, compare wild type (wt/237) promoter with mutant (mut/237) promoter). Whereas the rGH promoter truncated to 106 was activated by C/EBP expression to the same extent as the 237/+8 promoter mutated in the GHF3 binding site, the rGH promoter truncated to 33 was not activated by C/EBP expression, even at high amounts of transfected C/EBP expression vector (Fig. 7A). The 33/+8 promoter could be made strongly responsive (43.2-fold activation on average when co-transfected with 10 µg of c/EBP expression vector) to C/EBP by inserting six copies of the GHF3 binding site immediately upstream of it (Fig. 7B), demonstrating that the 33/+8 promoter was competent for C/EBP activation.

DISCUSSION

The GHF3 site is a major locus controlling the transcription of the rGH promoter and is bound by a series of different multisubunit complexes via a single, common DNA-binding subunit (16). As a first step toward elucidating the structures and activities of the different GHF3 complexes, we have determined in the following ways that the common DNA-binding subunit contains C/EBP: C/EBP co-fractionated with factors enriched greater than 600-fold in the minimal DNA-binding subunit of GHF3 (Fig. 1A); antibodies directed against C/EBP disrupted GHF3 gel mobility shift complexes (Fig. 2) formed with affinity-purified GHF3; Southwestern blots previously identified the predominant GHF3-binding factor as having a molecular mass of 43 kDa (16), similar to the 42-kDa size of C/EBP (25, 26) (Fig. 1A); and the GHF3 binding site contains strong similarity (Fig. 8) to an optimal, palindromic C/EBP binding site (31, 32).

C/EBP is a member of the bZIP family of transcription factors, and the known ability of C/EBP to heterodimerize with some other bZIP factors (26, 33, 34, 35, 36, 37) might implicate other bZIP proteins in the GHF3 complexes. Current evidence suggests that C/EBP is likely to be the only DNA-binding factor in the minimal DNA-binding subunit, since protein-DNA uv cross-linking of the minimal GHF3 DNA-binding subunit detected exclusively a 46-kDa cross-linked species (16). Similarly, Western blots of affinity-purified, GHF3 DNA-binding subunit probed with antibodies against the bZIP proteins DBP (27), TEF (17), and C/EBP (26) (data not shown) did not indicate that these factors co-purified with the GHF3 DNA-binding subunit. These antibodies also did not disrupt the gel shift complexes formed between affinity-enriched GHF3 and the GHF3 binding site (data not shown). It is possible that these or other bZIP proteins may oligomerize with C/EBP to participate as other subunits of the higher order GHF3 complexes that did not co-purify with the DNA-binding subunit. Although no such C/EBP oligomers have been reported to date, oligomerization through leucine zippers has been demonstrated (37, 38). Further studies will be required to define the other subunits of the multisubunit GHF3 complexes.

The relative lack of full-length (42-kDa) C/EBP in pituitary progenitor GHFT1-5 cells compared with adult-derived GC or 235-1 pituitary tumor cells (Fig. 1), combined with the strong activation of the rGH promoter in GHFT1-5 cells by C/EBP expression (Fig. 4), provided additional evidence for an important role of C/EBP in the pituitary-specific expression of the rGH promoter. Previously reported distributions of C/EBP RNA or protein in rats (28), mice (28, 29), or humans (30) did not specifically include the pituitary gland. C/EBP expression is associated with the terminally differentiated state in other tissues and has antiproliferative actions in some cells (39, 40, 41, 42, 43), making it possible that C/EBP expression promotes both rGH expression (Fig. 4) and concomitant differentiation into a nonproliferative cell type. Pituitary-specific activation of the rGH promoter by C/EBP is likely to be dependent upon Pit-1 (Fig. 3), although the presence of both Pit-1 and C/EBP in lactotroph 235-1 cells (Fig. 1) that do not express GH indicates a role for other factors.

C/EBP antibodies also cross-reacted with 30- and 34-kDa proteins present both in affinity-purified GHF3 and in extracts from the GC and 235-1 cells; only the 34-kDa species seemed to be present in GHFT1-5 cells (Fig. 1A). A 30-kDa C/EBP variant originating from an internal translation initiation site within the C/EBP mRNA has been described previously in mice, rats, and chickens (43, 44, 45). This amino-terminal truncated, 30-kDa C/EBP lacks the transcriptional activation functions (45) and antiproliferative activity (43) of C/EBP but retains its DNA binding and dimerization domains, consistent with its purification by DNA affinity chromatography (Fig. 1A). The previously reported extinction of GH gene expression in fusions of GH-secreting GC cells and L-cells may involve factors binding to the GHF3 site (46), and it is possible that an altered balance of activating 42-kDa C/EBP and inactive 30-kDa C/EBP could contribute to this suppression. The origins and activity of the 34-kDa form have not been reported, and it may represent a pituitary-specific isoform of C/EBP. Basal rGH promoter activity in GHFT1-5 cells was inhibited by a point mutation (Fig. 6, mut/237) that inhibits GHF3 factor binding, suggesting that GHFT1-5 cells contain an activating factor, possibly the 34-kDa C/EBP variant, that binds to the GHF3 site.

The same promoter mutation reduced, but did not eliminate, activation of the rGH promoter by ectopic C/EBP expression in GHFT1-5 cells (Fig. 6). Because the promoter containing the GHF3 binding site mutation was still activated by C/EBP, it was impossible to differentiate whether C/EBP activation through the GHF3 binding site was a result of direct binding of C/EBP to the GHF3 site or whether C/EBP was binding to a separate site and merely synergizing with other GHF3-binding factors present in GHFT1-5 cells. C/EBP activation of the 33/+8 rGH promoter truncated to contain little more than the TATA box was wholly GHF3 binding site-dependent (Fig. 7B), suggesting that C/EBP activation can occur directly through the GHF3 binding site. Thus, it is likely that C/EBP directly binds to the GHF3 site in vivo (Fig. 7) as well as in vitro (Figs. 1 and 2) to cause the activation of the rGH promoter.

The current data also suggest that C/EBP participates, directly or indirectly, in the cAMP- and PMA-dependent effects on the rGH promoter. rGH promoter activation by C/EBP was synergistic with incubating GHFT1-5 cells with PMA and forskolin, activators of protein kinases C and A, respectively (Figs. 4, 5, 6). GH expression is tightly regulated by the hypothalamic factor, growth hormone-releasing factor, through a cAMP intermediary (47, 48). As forskolin increases the intracellular concentration of cAMP (20), it is possible that the effects of forskolin on rGH promoter activity may be mimicking the effects of growth hormone-releasing factor. Pit-1 is another target for PMA and forskolin action and is phosphorylated by both protein kinase A and C in vitro and in vivo (49). Forskolin activation of the rGH promoter has also been suggested to be Pit-1 binding site-dependent (50, 51), although this is difficult to ascertain given the dependence of the rGH promoter on the Pit-1 binding sites. Pit-1 mutated in its protein kinase A and C phosphorylation sites was 65% as effective as wild type Pit-1 at activating the rGH promoter (52), demonstrating that other rGH promoter-binding factors, possibly C/EBP, contribute significantly to both protein kinase A and C activation.

PMA and forskolin synergy with C/EBP displayed components dependent upon both the GHF3 binding site and a second C/EBP activation site that maps to between 106 and 33 (Fig. 6). The molecular nature of this secondary C/EBP activation site is currently unknown, but since purified GHF3 did not bind to a rGH promoter fragment spanning sequences 106 to 15 (Fig. 2), it is not likely to be mediated directly through GHF3 factor binding at that site. We have also observed that C/EBP expression activates the human GH promoter in GHFT1-5 cells and is synergistic with PMA and forskolin.⁴ We currently do not know if C/EBP activation of the human promoter requires the homologous 106 to 33 sequences conserved between the rat and human promoters (53).

Thus, the GHF3 binding site strongly contributes to the control of the rGH promoter in GH-secreting pituitary cells derived from adult pituitary tumors (16) and probably in transgenic mice (54). The GHF3-binding factor, identified here as C/EBP, synergizes with PMA and forskolin or with Pit-1 and strongly activates the rGH promoter in C/EBP-deficient pituitary progenitor cells. The role of C/EBP as the central, DNA-binding subunit of the multisubunit GHF3 complexes is also interesting given the unique architecture of the GHF3 complexes: a number of physiologic control mechanisms may impinge on this single factor and be integrated at this regulatory bottleneck. Determining the molecular nature of C/EBP-interacting proteins in those higher order complexes will also be essential to understanding the central role of the GHF3 binding site and C/EBP in the regulation of rGH promoter activity.