Pit-1 and GATA-2 Interact and Functionally Cooperate to Activate the Thyrotropin β-Subunit Promoter*

The molecular determinants governing cell-specific expression of the thyrotropin (TSH) β-subunit gene in pituitary thyrotropes are not well understood. The P1 region of the mouse TSHβ promoter (−133 to −88) region interacts with Pit-1 and an additional 50-kDa factor at an adjacent site that resembles a consensus GATA binding site. Northern and Western blot assays demonstrated the presence of GATA-2 transcripts and protein in TtT-97 thyrotropic tumors. In electrophoretic mobility shift assays, a comigrating complex was observed with both TtT-97 nuclear extracts and GATA-2 expressed in COS cells. The complex demonstrated binding specificity to the P1 region DNA probe and could be disrupted by a GATA-2 antibody. When both Pit-1 and GATA-2 were combined, a slower migrating complex, indicative of a ternary protein-DNA interaction was observed. Cotransfection of both Pit-1 and GATA-2 into CV-1 cells synergistically stimulated mouse TSHβ promoter activity 8.5-fold, while each factor alone had a minimal effect. Mutations that abrogated this functional stimulatory effect mapped to the P1 region. Finally, we show that GATA-2 directly interacts with Pit-1 in solution. In summary, these data demonstrate functional synergy and physical interaction between homeobox and zinc finger factors and provide insights into the transcriptional mechanisms of thyrotrope-specific gene expression.

␤-subunit gene is restricted to the thyrotrope cells of the anterior pituitary gland, where it combines with the ␣-subunit gene product to produce intact TSH (3). Little information is available to explain the highly restricted expression of the TSH␤ subunit gene to this small population of thyrotrope cells. To study these thyrotrope-specific factors, we have utilized the TtT-97 tumor model, which consists of a homogeneous thyrotrope cell population that expresses the TSH␤ gene without the confounding influence of other pituitary cell types (4). Characterization of the transcription factors present in thyrotropes that functionally interact with this promoter will enhance our understanding of the pivotal role they may play in pituitary development and in the maintenance of the differentiated phenotype.
Cell-specific activity of the mouse TSH␤ promoter in thyrotrope cells requires sequences located between Ϫ270 and Ϫ80 of the 5Ј-flanking region (5,6). Within this broad area, four DNase I protected regions have been identified using nuclear extracts from TSH␤ expressing mouse TtT-97 thyrotropic tumor cells (7)(8)(9). These have been termed D1 (Ϫ253 to Ϫ222), D2 (Ϫ196 to Ϫ176), P1 (Ϫ133 to Ϫ88), and P2 (Ϫ86 to Ϫ64). Recently, using scanning mutagenesis, DNase I footprinting, and gene transfer studies, we have focused on the contribution of the functionally important proximal P1 region to TSH␤ gene expression in thyrotropes (10). Selected mutations in the P1 region resulted in loss of DNA binding of at least two different thyrotropic nuclear proteins in TtT-97 thyrotrope cells. We previously identified the more distal of these two binding proteins as the POU homeodomain Pit-1 transcription factor (9,10). By Southwestern blot analysis, we determined that the more proximal factor is a 50-kDa protein, which binds adjacent to the Pit-1 consensus sequence where its binding is independent of Pit-1. Targeted mutagenesis showed that when the binding of either Pit-1 or the 50-kDa protein was disrupted, functional promoter activity was reduced by 60 -80% to a level exhibited by a minimal Ϫ80/ϩ43 promoter fragment that lacks the P1 region (10). Thus, full TSH␤ promoter activity required the functional participation of both Pit-1 and an additional protein at the P1 promoter region.
Pit-1 is a transcription factor present in somatotropes, lactotropes, and thyrotropes of the anterior pituitary (11,12), where it plays a critical role in the maturation of these cell types during pituitary development (13,14). Its role in somatotrope cell growth and its ability to transactivate the growth hormone and prolactin gene promoters are well documented (12,15,16). Based on binding studies with the growth hormone, prolactin, and Pit-1 promoters, a consensus DNA recognition element ((T/A)(T/A)TATNCAT) has been derived (12,17). In the P1 region of the mTSH␤, gene an imperfect consensus sequence occurs twice on the antisense strand at positions Ϫ122 to Ϫ114 and Ϫ107 to Ϫ100. Mutations that disrupt either of these sites abrogated Pit-1 binding and dramatically lowered functional promoter activity (10). Interestingly, a mutation that drastically altered the downstream site was without detrimental effect on binding or function. Other studies have shown that regulation of TSH␤ promoter activity by TRH or cyclic AMP colocalized with Pit-1 binding sites and that mutations of these consensus sequences blunted the stimulatory effect (18 -20). Given this information, it is reasonable to speculate that Pit-1 may be important for both basal and hormonestimulated activity of the TSH␤ promoter. However, additional factors most likely are involved in regulating basal activity, since the endogenous TSH␤ gene is not expressed in Pit-1 containing lactotrope and somatotrope cells. Furthermore, the addition of Pit-1 alone was not sufficient to stimulate TSH␤ promoter activity in ␣TSH cells that lack endogenous Pit-1 (9). Finally, we have recently shown that a mutation that disrupts binding of a 50-kDa thyrotropic protein without disrupting Pit-1 binding to the adjacent site diminished promoter activity in transfected TtT-97 thyrotrope cells (10).
Examination of the promoter sequence in the proximal P1 region reveals two areas of sequence homology to a GATA consensus binding site (21,22): AGATGC, from Ϫ110 to Ϫ105, and AGATAA, from Ϫ98 to Ϫ93. The GATA transcription factor family consists of six members, GATA-1 through -6. In general, they are approximately 50 kDa in size (hGATA-2) and contain a highly conserved DNA-binding domain consisting of two zinc fingers, and although originally found in cells of hematopoietic lineage (23), they are expressed in other cell types (24). GATA factors have been shown to functionally interact with other classes of transcription factors such as AP1 (25), Sp1 (26 -28), and the estrogen receptor (29). The AGATAA sequence is also present in the human, equine, and mouse ␣-subunit promoters, and Steger et al. (30) have shown that GATA family members are involved in both pituitary and placental expression of the ␣-glycoprotein hormone gene that encodes the common subunit of TSH, follicle-stimulating hormone, luteinizing hormone, and chorionic gonadotropin. In the current study, we present data showing that the 50-kDa protein in thyrotropes that binds to the proximal P1 region of the mTSH␤ promoter and functionally synergizes with Pit-1 is likely to be GATA-2.

EXPERIMENTAL PROCEDURES
Plasmid and Probe Constructions-Construction of the Ϫ392 to ϩ40 mTSH␤ luciferase vector and those containing the P1 region mutations (P1-M3, P1-M7) have been described previously (6,10). The P1 probes containing sequences from Ϫ144 to Ϫ74 were produced by a PCR strategy using as templates the wild type or mutant Ϫ392 to ϩ40 luciferase constructs. The sense strand primer was 5Ј-CGAGTCGA-CAAGTTTTATTTG-3Ј, and the antisense strand primer was 5Ј-GAT-GTCGACATTCGAATTGCT-3Ј. This strategy introduced a SalI site (underlined) at each terminus. We performed 30 cycles of denaturation at 94°C for 30 s, annealing at 58°C for 30 s, and extension at 72°C for 30 s with 2.5 units of Taq DNA polymerase (Perkin-Elmer) under buffer conditions suggested by the manufacturer. Products of the appropriate size were digested with SalI, subcloned into pGEM5Zfϩ (Promega Corp., Madison, WI), and sequenced. P1 probes were purified after digestion with SalI and isolation on a 2% agarose gel using Qiaex II resin (Qiagen Inc., Chatsworth, CA).
Oligonucleotides encompassing the Ϫ117 to Ϫ88 wild type and selected mutated sequences were synthesized with SalI overhangs and annealed using standard techniques (31). For the wild type sequence the sense strand was 5Ј-TCGACTTTTCAATAGATGCTTTTCAGATA-AGAAAG-3Ј, and the antisense strand was 5Ј-TCGACTTTCTTATCT-GAAAAGCATCTATTGAAAAG-3Ј. For the P1-M7 probe, the sense strand was 5Ј-TCGACTTTTCAATAGATGCTTTTCAGATTCTAAAC-3Ј, and the antisense strand was 5Ј-TCGAGTTTAGAATCTGAAAAGCATC-TATTGAAAAG-3Ј, with the underlined bases representing the mutation.
A 700-bp mouse GATA-2-specific sequence, generated by PCR from a genomic DNA clone, was isolated from a pUC19 vector by digestion with EcoRI. This plasmid was a gift from Drs. F-Y. Tsai (33) to a specific activity of 4 -8 ϫ 10 8 cpm/g using a commercial kit (Life Technologies, Inc.). The conserved GATA DNAbinding domain (DBD) containing codon sequences from 259 -344 (34) was amplified from pRSV-mGATA-3 by a PCR strategy using the following oligonucleotides. The sense strand was 5Ј-CAGGATCCGAAG-GCAGGGAGTGTGTGAAC-3Ј, and the antisense strand was 5Ј-CA-GAATTCGTACAGCCCACAGGCATTG-3Ј. We performed 30 cycles of denaturation at 94°C for 1 min, annealing at 57°C for 1 min, and extension at 72°C for 1 min with 2.5 units of Taq DNA polymerase, 250 ng of each primer, and 1 ng of the GATA-3 plasmid. The resultant product was subcloned into pCRII (Invitrogen, San Diego, CA) and sequenced.
For cotransfection experiments, a 2.4-kb end-filled EcoRI fragment containing the hGATA-2 coding region (35) was cloned downstream of the immediate early cytomegalovirus (CMV) enhancer/promoter at a unique NotI site within the expression vector pCMV-␤gal (CLONTECH Laboratories, Palo Alto, CA), in which the ␤-galactosidase region had been removed. Both cDNA and vector fragments were blunt-ended by treatment with avian myeloblastosis reverse transcriptase (Promega Corp., Madison, WI) and all four dNTPs and ligated with T4 DNA ligase (Boehringer Mannheim). Constructs with the correct orientation were determined by DNA sequencing using primers adjacent to the cloning site and Sequenase (U.S. Biochemical Corp.). Mouse Pit-1 coding sequences were blunt end-cloned into the same vector as described (9).
RNA Blot Analysis-Total RNA was isolated from TtT-97 thyrotropic tumors, mouse ␣TSH cells, or rat GH 3 cells by the guanidinium isothiocyanate-CsCl method (36). Poly(A) ϩ RNA was isolated by affinity chromatography over an oligo(dT) cellulose column (type 7, Pharmacia Biotech, Inc.). The RNA was size-fractionated through a 1% agarose gel containing 6% formaldehyde as described (37) and transferred to a nylon membrane (Nytran, 0.2 m, Schleicher and Schull) using 20 ϫ SSC buffer (3 M sodium chloride, 0.3 M sodium citrate), and fragments were fixed by ultraviolet light cross-linking (model FB-UVXL-1000, Fisher) and hybridized to a cDNA probe for the DBD, which is highly conserved in all GATA family members, or with cDNA probes specific for mouse GATA-2, GATA-3, or GATA-4. Prehybridization, hybridization, and wash conditions have been described previously (37). The membranes were stripped by incubating them in boiling 0.1 ϫ SSC, 0.1% sodium lauryl sulfate, three times for 15 min each and then reprobed with a radiolabeled mouse ␤-actin cDNA probe.
Western Blot Analysis-CV-1 cells transiently transfected with hemagglutinin (HA) epitope-tagged mPit-1 and/or hGATA-2 were harvested with phosphate-buffered saline containing 3 mM EDTA, pelleted, and resuspended in 100 l of TEA-SDS solubilization buffer (55 mM triethanolamine, 111 mM NaCl, 2.2 mM EDTA, and 0.44% SDS) as described (38). Lysed extracts were passed through a 25-gauge needle seven times to shear genomic DNA. Protein concentration was determined by the method of Bradford (39) using a commercial kit (Bio-Rad). Equal amounts (50 g) of protein were separated on a 12% polyacrylamide-SDS gel (40) and transferred to an Immobilon-P (polyvinylidene difluoride) membrane (Millipore, Bedford, MA) by electroblotting. The membrane was dried by immersion in methanol for 10 s followed by air drying for 15 min. Nonspecific binding was blocked with 7.5% nonfat milk in TBST (20 mM Tris-Cl, pH 7.5, 137 mM NaCl, 0.2% Tween 20) for 1 h. Immobilized proteins were incubated for 1 h at room temperature with a mouse monoclonal anti-HA-peroxidase antibody (0.1 g/ml; Boehringer Mannheim) at a dilution of 1:1000 in TBST supplemented with 1% milk. After three 10-min washes with TBST, HA-tagged proteins were detected using an ECL chemiluminescent kit (Amersham Life Science).
Nuclear extracts from TtT-97 thyrotropic tumors were prepared as described (41), and hGATA-2 transfected COS cell extracts were electrophoresed, transferred, and blocked with 7.5% milk as described above. GATA-2 protein was visualized with a mouse monoclonal anti-GATA-2 antibody (Santa Cruz Biotechnology, Inc.) at a dilution of 1:5000 for 1 h in TBST containing 1% milk. After washing in TBST, the membrane was incubated with horseradish peroxidase-conjugated goat antimouse immunoglobulin G (Life Technologies, Inc.) (1:10,000 dilution) for 1 h in TBST containing 1% milk and washed three times in TBST, and proteins were detected using a chemiluminescent assay.
Library Screening for Mouse GATA Recombinants-A once amplified expression library containing TtT-97 cDNA fragments cloned into the EcoRI site of EX-lox (42) was screened for GATA sequences as described (43) using a conserved DNA-binding domain probe containing the zinc finger domains corresponding to codons 259 -344 of mouse GATA-3 (34). Approximately 750,000 recombinant phage were initially screened with the probe. Following four rounds of screening, five recombinant phage were were purified, and recombinant plasmid was produced by autoexcision using the Cre recombinase from Escherichia coli BM25.8. Plasmid DNA was transformed into E. coli DH5␣, and EcoRI inserts were subcloned into pGEM7Zfϩ (Promega Corp.) for restriction enzyme digestion and nucleotide sequencing by the chain termination method (44). DNA sequences were compared with Gen-Bank TM sequences using the BLAST protocol (45).
DNase I Footprints-Fragments containing mTSH␤ sequences from Ϫ392 to ϩ40 were single end-labeled by filling in recessed 3Ј termini as described previously (10). DNase I protection experiments were performed as described (46) Reactions containing 75 g of nuclear extract from TtT-97 tumor cells were digested with 800 ng of DNase I, while those with 25 g of bacterially derived GST-Pit-1 or 150 g of hGATA-2 transfected COS cell extracts were incubated with 30 ng of DNase I for 60 s. Control reactions contained 10 g of bovine serum albumin (Boehringer Mannheim). After heat denaturation of the deproteinized DNA at 95°C for 3 min, samples were applied to a 6% polyacrylamide sequencing gel containing 8 M urea and size-separated by electrophoresis followed by autoradiography. Film exposures were for 24 -48 h.
Gel Mobility Shift Assays-COS cells (10 6 ) were transiently transfected with the GATA-2 pMT2 vector (35) or with pSG5, an empty SV40 vector used as a mock expression control, using the calcium phosphate precipitation method (47). Extracts from the transfected cells were prepared as described (48). Isolation of TtT-97 nuclear extracts and production of mPit-1 as a glutathione S-transferase fusion in bacteria were described previously (9,41). DNA probes were radiolabeled by filling in recessed 3Ј termini with [ 32 P]dCTP using reverse transcriptase. Protein extracts were preincubated for 15 min on ice with 700 ng of salmon sperm DNA (cleaved with HincII), in a buffer containing 25 mM Tris-Cl, pH 7.9, 5 mM MgCl 2 , 1 mM EDTA, 1 mM dithiothreitol, 4 mM spermidine-HCl, and 8% Ficoll (w/v). After the addition of 25,000 cpm of the 32 P-probe, the mixture was incubated for 20 min at 15°C and electrophoresed on a 4% nondenaturing acrylamide gel. For competition experiments, 25-250 ng of double-stranded DNA was supplemented into the reaction mixture prior to probe addition. In some experiments, a rabbit antibody directed against mouse GATA-2 (gift of Dr. S. H. Orkin), rat Pit-1 (gift of Dr. R.A. Maurer), or preimmune serum was added during the last 10 min of the incubation.
Cotransfection Transactivation Assays-Transient transfections using calcium phosphate into CV-1 monkey kidney cells were performed as described previously (47). Approximately 750,000 cells were added per 100 ϫ 20-mm tissue culture dish 20 h prior to transfection. The cells were transfected with 10 g of a Ϫ392/ϩ40 mTSH␤ luciferase construct (6), 0.5 g of pCMV-Pit-1, and/or 1 g of pCMV-hGATA-2. The total amount of plasmid containing the CMV promoter was adjusted to 2 g with an empty pCMV vector. Each transfection also contained 2 g of pCMV-␤gal DNA as an internal transfection control. Each set of transfections was done in triplicate and also contained a Rous sarcoma virus promoter-luciferase plasmid transfected in parallel as a positive control and a promoterless pA 3 luc vector as a background control. After 48 h, luciferase activity was measured in a Monolight 2010 luminometer from duplicate aliquots of freeze-thaw cytoplasmic lysates (6) from the cells, while ␤-galactosidase activity was measured using a colorimetric assay (49), and these activities were compared with a standard curve of enzymatic activity. Light units were normalized to the ␤-galactosidase activity.
Expression of hGATA-2 and Pit-1 Protein Containing an Epitope Tag-For the production of hGATA-2 and mPit-1 containing an aminoterminal HA epitope, we utilized a PCR strategy and fused the coding region of each protein to a CMV promoter in the vector pCGN-2, which was a modification of the vector pCGN (50). Oct-1 sequences were removed from pCGN (kindly provided by Dr. W. Herr, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY) by digestion with XbaI and BamHI. A double-stranded linker containing a multiple cloning site was ligated between the XbaI and BamHI termini. The sense strand was 5Ј-CTAGAATTCAAGCTTGCGGCCGCTGCAGGTACCCGGG-3Ј, and the antisense strand was 5Ј-GATCCCCGGGTACCTGCAGCGGC-CGCAAGCTTGAATT-3Ј. This resulted in restriction sites for XbaI, EcoRI, HindIII, NotI, PstI, KpnI, SmaI, and BamHI. The complete coding region of human GATA-2 in a eukaryotic expression vector, pMT2, was a gift from Dr. Stuart Orkin (Harvard Medical School). The coding region was amplified by PCR with an XbaI site at its amino terminus and a NotI site at its carboxyl terminus. The sense strand primer was 5Ј-CCTTCTAGAAGTTGCCAATCTTTCACC-3Ј, and the antisense primer was 5Ј-CTTGCGGCCGCTGGTAAGGGTTTGGTC-3Ј. The PCR consisted of 25 cycles of 94.5°C denaturation for 30 s, 65°C annealing for 30 s, and 68°C elongation for 2 min using LA Taq polymerase (TaKaRa Shuzo, Kyoto, Japan) under buffer conditions recommended by the commercial supplier. The product was cleaved with XbaI and NotI and ligated to pCGN-2. This resulted in an in frame fusion of hGATA-2 with the HA epitope. Similarly, the mouse Pit-1 coding region was amplified utilizing primers containing a HindIII site at both ends and ligated into pCGN-2. Verification of the correct reading frame was done by DNA sequencing utilizing a primer in the HA coding sequence, 5Ј-ATGGCTTCTAGCTATCCTTAT-3Ј.
In Vitro Protein-Protein Binding Assays-Bacterial extracts containing the recombinant fusion GST-Pit 1 or GST alone were prepared essentially as described previously (9) with some modifications. A freshly streaked colony was resuspended in 200 l of sterile water and plated onto an LB plate containing 100 g/ml ampicillin. The lawn of bacteria was scraped from the plate and used to seed a 200-ml LBampicillin culture that was grown to an optical density (600 nm) of 0.8 and then induced with 1 mM of isopropyl-D-thiogalactopyranoside for 2 h. The bacterial cells were harvested, and the cell pellet was resuspended in 20 ml of fusion protein buffer (150 mM NaCl, 16 mM NaH 2 PO 4 , 4 mM Na 2 HPO 4 , 1% Triton X-100, 2.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, supplemented with a protease inhibitor mixture) (Complete, Boehringer Mannheim). Bacteria were lysed by sonication, and supernatant was obtained as described (51). Supernatant (30 ml) containing GST-Pit-1 or GST was allowed to bind with 0.5 ml of a 50% slurry of glutathione-Sepharose 4B (Pharmacia) for 1 h at 4°C followed by washing three times with 30 ml of fusion protein buffer. Concentration of bound protein was determined using a Bio-Rad assay and by SDS-polyacrylamide gel electrophoresis and Coomassie Blue staining.

RESULTS
Murine TtT-97 Thyrotropic Tumor Cells Contain RNA Transcripts for GATA-2-Our previous studies combining scanning mutagenesis, DNase I protection, Southwestern blot studies, and functional analysis in transfected thyrotrope cells demonstrated that both Pit-1 and an unknown 50-kDa protein interacted with the proximal P1 promoter region of the mTSH␤ promoter (10). To identify this additional factor, we first examined the cis-acting sequences necessary for its binding. Inspection of the 5Ј-flanking region in this area revealed two sequences that resemble consensus sites for the GATA family of transcription factors. These sequences, AGATGC (Ϫ109 to Ϫ105) and AGATAA (Ϫ98 to Ϫ94) are boxed in Fig. 1 and are adjacent to two Pit-1 consensus sites (underlined). Mutations that disrupt either of the GATA consensus sequences, such as P1-M7 (Fig. 1) resulted in the loss of binding of the 50-kDa protein as well as a significant decrease in promoter activity in TtT-97 thyrotropes (10). Similarly, other mutations, such as P1-M3 (Fig. 1), abrogated Pit-1 binding and promoter activity due to the disruption of Pit-1 consensus sites. To determine whether thyrotrope cells contained transcripts for GATA factors, we performed Northern blot analyses under high stringency conditions with specific GATA probes. Initially we used a DBD probe derived by PCR from mouse GATA-3, which contains sequences highly conserved among all GATA family members. We detected a major transcript of 3.7 kb and a minor transcript of 3.1 kb using total RNA from two different TtT-97 thyrotropic tumors (Fig. 2A). The specificity of GATA family members was then ascertained by using distinct amino-terminal coding region probes that are unique to GATA-2, GATA-3, or GATA-4 but lack the common DBD region. A GATA-2 probe composed of coding region sequences from amino acids 20 -249, revealed the presence of abundant transcripts of 3.7 and 3.1 kb (Fig. 2B) from two TtT-97 thyrotropic tumors and in ␣TSH cells, which are thyrotrope-derived but express neither Pit-1 protein (9) nor the TSH␤ subunit (52). The pattern of these transcripts was similar to that seen with the common DBD probe ( Fig. 2A). In contrast, pituitary-derived GH 3 somatotropes, which express Pit-1, did not contain detectable GATA-2 transcripts. The blot was reprobed with ␤-actin to demonstrate that equivalent amounts of RNA were present in each lane (Fig. 2B). Next we used a probe specific for mouse GATA-3, which contained sequences for amino acids 29 -169. We detected a single abundant 2.6-kb GATA-3 transcript in ␣TSH cells and a barely detectable band of the same size in TSH␤expressing TtT-97 thyrotropes (Fig. 2C) although equivalent amounts of RNA were present as accessed by reprobing the filter with ␤-actin (Fig. 2C, bottom). The filter from Fig. 2B was then reprobed with a mouse GATA-4-specific probe that contained 116 bp of 5Ј-noncoding sequence and coding sequences for amino acids 1-117. A faint band of about 4 kb was found in ␣TSH RNA (Fig. 2D), while, in contrast, no GATA-4 transcripts were detected in poly(A) ϩ RNA from either TtT-97 thyrotropes or GH 3 somatotropes.
Additionally, we used the common DNA-binding domain probe to isolate recombinant GATA cDNAs from a TtT-97 library constructed in EX-LOX under high stringency washing conditions. From approximately 750,000 primary phage, we isolated six GATA recombinant clones; all contained mouse sequences highly homologous to human GATA-2 and not to other GATA family members. These data suggest that GATA-2 is the major GATA transcript present in TSH␤-expressing thyrotropes, although a faint signal for GATA-3 was detectable in these cells.
TtT-97 Thyrotrope Cells Contain GATA-2 Protein-Next we determined whether GATA-2 protein was present in the TtT-97 thyrotropes. Since expression of full-length GATA-2 is toxic to bacteria (21), we overexpressed human GATA-2 in COS cells using the vector pMT2 (35) to serve as a positive control. As a negative control, we transfected COS cells with pSG5, an empty SV40 promoter construct that was lacking a cDNA se-quence. In Western blots using a rabbit polyclonal antibody directed specifically against GATA-2, we detected a single GATA-2 band of 50 kDa from an extract of the hGATA-2 transfected COS cells (Fig. 3, lane 2), which migrated to a similar position as a protein in nuclear extracts from TtT-97 thyrotropic tumors (Fig. 3, lane 3). In contrast, no band was detected in the mock transfected COS cells (Fig. 3, lane 1). This demonstrates that protein in addition to mRNA for GATA-2 are present in the TSH␤-expressing thyrotropic tumor cells.
GATA-2 and Pit-1 Can Bind Independently to the Proximal mTSH␤ Promoter-To determine whether GATA-2 is capable of binding to the proximal P1 region, we used a probe containing sequences from Ϫ117 to Ϫ88 and full-length GATA-2 expressed in COS cells in electrophoretic mobility shift analysis. In a previous Southwestern blot, this probe detected a 50-kDa protein but did not detect Pit-1 (33 kDa) in thyrotrope cells (10) although it contained a consensus Pit-1 site at Ϫ107 to Ϫ101 (Fig. 1). A single major complex formed when we incubated the probe with extracts from hGATA-2 overexpressed in COS cells (Fig. 4A, lane 3) but not with mock transfected cells (lane 2). A similar comigrating complex was formed with TtT-97 thyrotropic tumor nuclear extract (lane 4). Specificity of GATA-2 binding to the proximal P1 site was demonstrated by showing that the GATA-2 complex was effectively competed with increasing amounts (25-250 ng) of homologous duplex oligonucleotide (Fig. 4B, lanes 3-5). However, up to 250 ng (2500-fold molar excess) of the same fragment containing the P1-M7 mutation, which disrupts the GATA consensus sequence, failed to compete for GATA-2 binding to the wild type probe (Fig. 4B,  lanes 6 -8). Specificity of the hGATA-2 and TtT-97 complex was demonstrated by its ability to be disrupted with a GATA-2specific polyclonal antibody (Fig. 4, C and D, lane 3) but not with preimmune rabbit serum (lane 4), indicating that GATA-2 FIG. 1. Schematic of the wild type and mutated P1 region of the mouse TSH␤ promoter. The nucleotide sequence of the sense strand is shown from Ϫ140 to Ϫ80 with the P1 footprinted region depicted at the top by a bracket. Two consensus Pit-1 binding sites are underlined, while two sequences with homology to GATA consensus elements are boxed. Two site-specific mutations, P1-M3 and P1-M7, are shown below the wild type sequence.

FIG. 2. Detection of GATA transcripts by Northern blot analysis in murine thyrotropes.
Twenty g of total RNA (A) or five g of poly(A) ϩ RNA (B, C, and D) from TtT-97, ␣TSH, or GH 3 cells were size-fractionated on denaturing agarose gels, transferred to nitrocellulose, and hybridized with a 32 P-labeled mGATA common DNA-binding domain probe (A) and with specific cDNA probes for mGATA-2 (B), mGATA-3 (C), or mGATA-4 (D). Blots in B and C were reprobed with a 250-bp mouse ␤-actin cDNA, which is shown below. The same filter as in B was reprobed in D. The positions of 32 P-labeled -HindIII fragments used as size standards are shown.
in the TtT-97 extract can bind to the proximal P1 site.
We next performed additional gel mobility shift assays using the entire P1 region probe, from Ϫ144 to Ϫ74 that contains both the GATA-2 and Pit-1 sites. Each factor formed a distinct complex (a and b, respectively, Fig. 5A, lanes 2 and 3). When both proteins were combined in vitro, we detected both single complexes, but in addition, a slower migrating complex (c, Fig.  5A, lane 4), consistent with both factors binding to the same DNA fragment, was evident. Since GATA-2 and Pit-1 are present in TtT-97 thyrotrope cells, the c complex most likely represents a ternary complex of both proteins interacting with the DNA probe. That the additional complex contains both proteins was confirmed by using 32 P-labeled probes containing either the P1-M7 or P1-M3 mutation in EMSA studies. With the P1-M7 probe, we show that GATA-2 fails to interact, whereas Pit-1 can form the same b complex seen with the wild type probe (Fig. 5B, lanes 2 and 3). However, when both proteins were present in the binding reaction, only the Pit-1-containing b complex was detected. As expected, the additional c complex seen with the wild type probe failed to form (Fig. 5B, lane 4). Conversely, using a probe containing the P1-M3 mutation, we found that the GATA-2 a complex formed, but the Pit-1 b complex was not detected (Fig. 5C, lanes 2 and 3). Again as would be predicted, when both proteins were combined, the slower migrating c complex also was not detected (Fig. 5C, lane  4). Thus, the slower migrating complex requires an intact GATA-2 site as well as a Pit-1 site, which suggests that it is probably due to both proteins simultaneously binding to the mTSH␤ P1 region to form a ternary complex.
To more precisely define the binding site for GATA-2 within the P1 region of the mTSH␤ promoter, we performed DNase I protection studies. Using a single end-labeled promoter fragment, we detected a protected area from Ϫ133 to Ϫ100 when the fragment was incubated with a bacterially produced GST-Pit-1 fusion protein (Fig. 6, lane 1) when compared with the control reaction incubated with BSA (lane 5). COS cells expressing human GATA-2 produced an overlapping footprint (Ϫ118 to Ϫ88) that extended several bp closer to the transcriptional start site (lane 2). This extended footprint pattern is coincident with the extension produced by TtT-97 thyrotropic tumor extracts from Ϫ133 to Ϫ88 (Fig. 6, lanes 3 and 4) and is identical to the previous TtT-97 footprint found using the P1-M3 probe (10). Thus, GATA-2 and Pit-1 can account for the same pattern of protection exhibited by TSH␤-expressing nuclear extracts. These data strongly suggest that GATA-2 is the 50-kDa protein that binds to the promoter at the proximal P1 site.   Pit-1 and GATA-2 Functionally Interact to Stimulate mTSH␤ Promoter Activity-We next determined the functional consequences of co-transfecting GATA-2 and Pit-1 on a mTSH␤ promoter construct. Using the region of the mTSH␤ promoter from Ϫ392 to ϩ40 fused to a luciferase reporter, we cotransfected CMV-directed expression vectors for hGATA-2 and mPit-1 into heterologous monkey kidney CV1 cells. The functional consequences of cotransfecting GATA-2 and Pit-1 are shown in Fig. 7A. Neither GATA-2 nor Pit-1 alone appreciably stimulated promoter activity (1.2-and 1.8-fold, respectively) from the wild type promoter when compared with the empty plasmid control. However, the combination of GATA-2 with Pit-1 showed an 8.5-fold stimulation of promoter activity. The stimulation did not occur when the TSH␤ luciferase construct contained mutations that abrogated binding to GATA-2 (P1-M7) or Pit-1 (P1-M3) (Fig. 7A). Thus, we have mapped the functional cooperativity of GATA-2 and Pit-1 to specific sequences within the 5Ј-flanking P1 region of the mTSH␤ promoter.
To ensure that each protein was being expressed at similar levels in these cotransfection studies, we tagged them with an HA epitope at their amino terminus (68). Using an antibody specific for the HA epitope on Western blots, we detected equivalent amounts of the appropriately sized HA-tagged proteins in extracts from cells that were transiently transfected with either GATA-2 or Pit-1 alone (Fig. 7B, lanes 2 and 3) or in combination (lane 4). Control cell extracts transfected with a CMV vector lacking cDNA sequences (Fig. 7B, lane 1) demonstrated only a nonspecific (NS) band that reacted with the HA antibody and was also present in the other transfected cells. Functional synergism of a similar magnitude was also seen in the presence of both HA-tagged proteins (data not shown).
Physical Interaction of Pit-1 with GATA-2-Our transfection studies indicated that Pit-1 can functionally synergize with GATA-2 to activate mTSH␤ promoter activity. This synergistic effect, which required the participation of both factors, suggested a possible direct protein-protein interaction between the two factors as the mechanism responsible for the functional cooperativity. To address this possibility, bacterial GST fusion proteins of rPit-1 or GST alone were immobilized on glutathione-Sepharose beads and used in binding assays with in vitro transcribed and translated GATA-2 that had been radiolabeled with [ 35 S]methionine. As a positive control, we used 35 S-labeled cEts-1, which has been shown to physically interact and functionally synergize with Pit-1 on the rat prolactin promoter (51). Additionally, we used a radiolabeled 252-amino acid truncation  were transiently transfected in duplicate by the calcium phosphate precipitation method. Transfections contained 10 g of mTSH␤ promoter-luciferase reporter vector, 1 g of pCMV-hGATA-2 and/or 0.5 g of pCMV-mPit-1 and were adjusted to a total of 2 g with a pCMV empty vector lacking cDNA sequences. In addition, each transfection included 2 g of pCMV-␤gal as an internal transfection control. Parallel reactions contained a promoterless pA3luc control as well as a RSV-luciferase positive control vector. Transfections were performed with the wild type promoter luciferase sequence (top, n ϭ 12), or contained the P1-M7 (middle, n ϭ 4) or P1-M3 (bottom, n ϭ 3) mutation (Fig. 1). Activity is shown as -fold activation of the value obtained using the mTSH␤ promoter-luciferase reporter (with S.E.) cotransfected with 2 g of the CMV empty vector. B, detection of hemagglutinin-tagged GATA-2 and Pit-1 proteins expressed in transiently transfected CV-1 cells. Whole cell extracts (100 g) from CV-1 cells transfected with 2 g of pCMV control vector lacking cDNA sequences (lane 1, CMV empty), 1 g of pCMV-HA-tagged hGATA-2 (lane 2), 0.5 g of pCMV-HA-tagged mPit-1, or 1 g of hGATA2 plus 0.5 g of mPit-1 HA-tagged expression vector were resolved on a 10% acrylamide-SDS gel and transferred to a polyvinylidene difluoride membrane. The blot was probed with a 1:1000 dilution of a mouse monoclonal anti-HA-peroxidase antibody followed by chemiluminescent detection. The positions of prestained rainbow protein standards are shown on the right. of cEts-1, termed cEts-1⌬5-6, in which the Pit-1-interacting domain has been deleted (51) as a negative control. Equal amounts of 35 S-labeled cEts-1, hGATA-2, or cEts-1⌬5-6 were incubated with immobilized GST-Pit-1 or GST alone. As shown in Fig. 8A, cEts-1 was able to interact with GST Pit-1 (lane 3) but showed no detectable binding to GST alone (lane 2). GATA-2 interacted with GST Pit-1 (Fig. 8B, lane 3) and again showed no detectable binding to GST alone (lane 2). In contrast, truncation of cEts-1⌬5-6, which removes its interacting domain with Pit-1 did not interact with either GST alone or with GST Pit-1 (Fig. 8C, lanes 2 and 3). Thus, these studies demonstrate that Pit-1 can physically interact with GATA-2 and may provide a molecular mechanism for their functional cooperativity on the mTSH␤ promoter.

DISCUSSION
The absolute restriction of thyrotropin ␤-subunit expression to the pituitary thyrotrope suggests that its transcription is governed by an exclusive interaction between the TSH␤ gene promoter and cell-specific transcription factors. To date, little information is available about which factors are necessary and sufficient to allow high levels of basal transcription in these cells. Our observations define the participation of two different classes of transcription factors to synergize on a closely spaced response element within the proximal P1 promoter region and demonstrate a functional ternary protein-DNA complex involving the pituitary-specific factor, Pit-1. Such composite elements have been described for a number of interacting transcription factors in other systems (53)(54)(55) and provide a mechanism for their combinatorial action within a regulatory network. The analysis of the mTSH␤ gene described in this report provides evidence for the participation of a zinc finger factor, GATA-2, with a POU homeodomain partner, Pit-1, on a such a composite element and may be a general mechanism in which other GATA family members can functionally and physically interact with other homeodomain partners.
A number of lines of evidence point to the fact that the pituitary-restricted factor, Pit-1, is necessary but not sufficient to allow basal transcription of the mTSH␤ gene. Pit-1 is expressed in thyrotropes, lactotropes, and somatotropes of the anterior pituitary (13,14), while TSH expression is restricted solely to the thyrotrope cell population. While it is clear that Pit-1 is critical for thyrotrope ontogeny, the role of Pit-1 in governing TSH␤ gene expression in thyrotropes is less well understood. Several reports have shown that when using a transfection reconstitution system, Pit-1 alone fails to transactivate TSH␤ promoter activity using 1.2 kb of 5Ј-flanking DNA in either pituitary or nonpituitary cells (9,56). In contrast, a marked enhancement of TSH␤ promoter activity resulted when other pituitary factors were cotransfected with Pit-1 such as P-lim (57) or a unique thyrotrope-specific Pit-1 isoform, termed Pit-1T (58,59). However, the role of P-lim in TSH␤ activation is not clear, since it is also present in GH 3 somatomammotrope cells, which do not express endogenous TSH␤. It is unlikely to interact with the P1 region, which we demonstrated to be of importance in the present study, since P-lim clearly does not bind to the Ϫ120 to Ϫ60 region, which contains the P1 region, unless the LIM domain was deleted (57). Moreover, such a truncated form of this protein lacking the LIM domain has not been demonstrated in thyrotropes. Thus, it is unlikely that P-lim accounts for the extended footprint found with thyrotropic nuclear extracts. In this report, we provide evidence that GATA-2 is the predominant GATA factor in thyrotrope cells, can functionally synergize with Pit-1 on juxtaposed sites in the P1 region, and in vitro can form a protein-protein interaction. This combinatorial interaction likely participates in TSH␤ gene expression in thyrotropes.
Synergistic interaction involving Pit-1 with other transcription factors on a variety of pituitary hormone genes has also been reported. On the growth hormone gene promoter, Pit-1 has been shown to synergize with Zn-15 (60), the thyroid hormone receptor (61), C/EBP (62), and P-OTX (63). Additionally, Pit-1 can form functional partners with Ets-1 (64 -66), the estrogen receptor (67, 68), P-lim (57), Oct-1 (69), and P-OTX (63) on the prolactin gene promoter and with NZF-1 on the Pit-1 enhancer/promoter (70). The sites of synergy with Pit-1 have been examined for several of these interactions. Synergy with Zn-15 grossly mapped to the amino-terminal transactivation domain of Pit-1 (60), while a smaller Pit-1 deletion, lacking amino acids 72-100, eliminated synergism with thyroid hormone receptor without affecting independent transcriptional activation on the rat growth hormone gene (61). In contrast, the site of interaction of Pit-1 with the estrogen receptor mapped to amino acids 45-72 (71), while the interaction with Oct-1 mapped to the POU homeodomain, consisting of amino acids 210 -273 (69). However, levels of expressed protein from these Pit-1 mutant constructs compared with wild type Pit-1 were not reported, and loss of function could be due to inefficient expression. Thus, Pit-1 may contain or provide a number of distinct domains, each of whose function or accessibility is dependent on the specific cis-acting promoter sequence and its synergistic transcription partner. Future studies will define the synergistic domains of Pit-1 with GATA-2 on the TSH␤ promoter, as well as the specific sites of protein-protein interaction. Such factor interactions with Pit-1 may represent a more general mechanism whereby a tissue-restricted homeodomain factor can recruit widely expressed factors to a particular composite element and integrate high levels of tissue-specific gene expression, as has been proposed by Gutierrez-Hartmann's group (64,65).
The GATA family of transcription factors are also somewhat tissue-restricted in their expression. GATA-1 was the first member of the family to be discovered, and its expression is generally restricted to erythroid cells, mast cells, and megakaryocytes, where it binds to a consensus sequence (A/ T)GATA(A/G) by virtue of its two zinc finger domains and plays a pivotal role in erythropoiesis (72). GATA-2 is expressed in a wide variety of tissues; GATA-3 is most abundantly expressed in T lymphocytes, endothelial cells, and in the developing nervous system; GATA-4 is generally restricted to the heart and gonads (32). Recently, Steger et al. (30) have demonstrated that GATA transcription factors play a role in pituitary ␣-subunit gene expression. They detected GATA-2 and a mouse GATA-4related protein in gonadotrope-derived ␣T3 cells and showed that both GATA-2 and -3 can activate ␣-subunit gene promoter FIG. 8. In vitro binding of GATA-2 to Pit-1. Binding assays were performed as described under "Experimental Procedures." Aliquots (20-l packed volume) of glutathione-Sepharose beads bound to 1 g of GST or GST- Pit-1 (A, B, and C, lanes 2 and 3) were incubated with equal amounts of in vitro translated 35 S-labeled cEts-1 (68 kDa, panel A), hGATA-2 (50 kDa, panel B), or amino-truncated cEts-1⌬5-6 (28 kDa, panel C). Lane 1 in each panel shows 10% of the amount of each radiolabeled protein (Input) added to the binding reactions. activity 3-fold. The GATA site responsible for this activity mapped just upstream of the cyclic AMP-responsive site within the proximal 180 bp of the 5Ј-flanking region. In TSH␤-expressing thyrotropic tumor cells, a majority of the GATA transcripts were GATA-2, since the pattern of hybridization was identical when using a GATA-2-specific probe or with a probe only containing the conserved DNA-binding domain. Additionally, using a common DNA-binding domain probe from mGATA-3 in a cDNA library screening, all of the clones we isolated contained GATA-2 sequences. There was, however, a low but detectable level of GATA-3 expression in these cells, and we cannot rule out its contribution in activating the TSH␤ promoter. However, we failed to detect any GATA-4 transcripts in TtT-97 thyrotropes, although a nonabundant GATA-4 transcript was detected in non-TSH␤-expressing ␣TSH cells. Finally, it is interesting to speculate that GATA-2 may also transactivate the ␣-subunit gene promoter in thyrotrope cells where it may coordinately regulate both subunits of thyrotropin.
Two transcripts for GATA-2 have also been reported for the human protein. Structural analysis of human genomic DNA clones encoding GATA-2 coupled with RNA blot analysis demonstrated that the observed molecular heterogeneity is due to alternative use of two polyadenylation consensus sequences, which are 612 bp apart within exon 6 (73). The sizes of the mouse GATA-2 transcripts that we detected (3.7 and 3.1 kb, Fig. 2) in thyrotrope RNA also differ by 600 nucleotides and are also most likely the result of alternative polyadenylation sites. It is of interest that ␣TSH cells, a thyrotrope-derived cell line that expresses the ␣-subunit gene but not TSH␤, contains transcripts for GATA-2 and GATA-3. However, although this cell line contains Pit-1 transcripts, Pit-1 protein was not detected (9). On the other hand, GH 3 cells, which express growth hormone and prolactin but not TSH␤, contain Pit-1 transcripts and protein but lack transcripts for GATA-2. Moreover, the 50-kDa protein, identified as GATA-2 in the present study, was not detected in GH 3 extracts by Southwestern blot analysis with the proximal P1 probe (10). These data suggest that both factors must be present in the same cell for TSH␤ gene expression. Studies are ongoing in our laboratory to stably express modulated levels of Pit-1 protein in ␣TSH cells or GATA-2 in GH 3 cells and determine whether endogenous TSH␤ expression can then be detected. Of course, additional factors that interact with GATA-2 and/or Pit-1 may also be required for full TSH␤ reconstitution.
The role of GATA-2 in pituitary thyrotrope development has yet to be determined. Recently, GATA-2 has been shown to be essential for normal mouse development. The targeted disruption of the GATA-2 gene resulted in the death of the embryos by embryonic day 10 or 11. This was due to a severe anemia that developed because of the critical role that GATA-2 plays in the regulation of genes controlling growth factor responsiveness or the proliferative capacity of early hematopoietic cells (74). Unfortunately, thyrotrope function could not be assessed in this model, since TSH␤ markers are first detected in the mouse pituitary gland during embryonic days 12.5-14.5 (13,75). During pituitary development, a transient population of TSH␤expressing thyrotropes appears in the rostral tip and is present before Pit-1 is expressed (75). These cells do express thyrotrope embryonic factor (56), but it is yet to be determined whether GATA-2 is present and plays a role in TSH␤ expression within this initial thyrotrope population as we have shown in mature thyrotropes. One way to address the importance of GATA-2 in thyrotrope development would be to disrupt the gene only in thyrotrope cells using the Cre recombinase system (76) fused to a thyrotrope-specific promoter. We have previously demonstrated that a 4.6-kb promoter fragment of the mouse ␣-sub-unit promoter will direct high levels of gene expression solely to thyrotropes and gonadotropes and at an appropriate time during pituitary development (77). Future studies will use these molecular tools to ascertain the role of GATA-2 within thyrotropes in vivo.
In summary, we have shown the requirement for at least two different classes of transcription factors to regulate mTSH␤ gene expression. Both GATA-2 and Pit-1 can bind independently to the P1 region of the promoter, form a heteromeric complex with DNA, and functionally synergize to activate TSH␤ promoter activity. In addition, these two factors can directly interact with each other in vitro. These factor interactions may be critical in defining cell-specific expression for this unique pituitary gene.