An Upstream Regulator of the Glycoprotein Hormone α-Subunit Gene Mediates Pituitary Cell Type Activation and Repression by Different Mechanisms*

Targeting of α-subunit gene expression within the pituitary is influenced by an upstream regulatory region that directs high level expression to thyrotropes and gonadotropes of transgenic mice. The same region also enhanced the activity of the proximal promoter in transfections of pituitary-derived α-TSH and α-T3 cells. We have localized the activating sequences to a 125-bp region that contains consensus sites for factors that also play a role in proximal promoter activity. Proteins present in α-TSH and α-T3 cells as well as those from GH3 somatotrope-derived cells interact with this region. The upstream area inhibited proximal α-promoter activity by 80% when transfected into GH3 cells. Repression in GH3 cells was mediated through a different mechanism than enhancement, as supported by the following evidence. Reversing the orientation of the area resulted in a loss of proximal promoter activation in α-TSH and α-T3 cells but did not relieve repression in GH3 cells. Mutation of proximal sites shown to be important for activation had no effect on repression. Finally, bidirectional deletional analysis revealed that multiple elements are involved in activation and repression and, together with the DNA binding studies, suggests that these processes may be mediated through closely juxtaposed or even overlapping elements, thus perhaps defining a new class of bifunctional gene regulatory sequence.

Targeting of ␣-subunit gene expression within the pituitary is influenced by an upstream regulatory region that directs high level expression to thyrotropes and gonadotropes of transgenic mice. The same region also enhanced the activity of the proximal promoter in transfections of pituitary-derived ␣-TSH and ␣-T3 cells. We have localized the activating sequences to a 125-bp region that contains consensus sites for factors that also play a role in proximal promoter activity. Proteins present in ␣-TSH and ␣-T3 cells as well as those from GH3 somatotrope-derived cells interact with this region. The upstream area inhibited proximal ␣-promoter activity by 80% when transfected into GH3 cells. Repression in GH3 cells was mediated through a different mechanism than enhancement, as supported by the following evidence. Reversing the orientation of the area resulted in a loss of proximal promoter activation in ␣-TSH and ␣-T3 cells but did not relieve repression in GH3 cells. Mutation of proximal sites shown to be important for activation had no effect on repression. Finally, bidirectional deletional analysis revealed that multiple elements are involved in activation and repression and, together with the DNA binding studies, suggests that these processes may be mediated through closely juxtaposed or even overlapping elements, thus perhaps defining a new class of bifunctional gene regulatory sequence.
Expression of the gene for the common ␣-subunit of the pituitary glycoprotein hormones is restricted to two cell types within the anterior pituitary gland. In combination with the specific ␤-subunits produced by thyrotropes and gonadotropes, it forms the biologically active hormones TSH, 1 lutenizing hormone, and follicle-stimulating hormone. Many reports have described areas within the proximal promoter region of the ␣-subunit gene that are important for expression in both pituitary cell types (1)(2)(3)(4)(5)(6)(7)(8)(9) and in the placenta (3,8,10,11), where it is also a component of chorionic gonadotropin. We have recently discovered a more distal area of the 5Ј-flanking region of the mouse gene, located approximately 4 kilobase pairs upstream from the start of transcription, that directs high levels of expression of a ␤-galactosidase reporter that was restricted to the thyrotropes and gonadotropes of transgenic mice (12). Subsequent studies demonstrated that when an 859-bp KpnI-BglII fragment derived from the region between Ϫ4600 and Ϫ3700 was fused directly to a proximal ␣-subunit promoter from Ϫ341 to ϩ43, the high level of cell-specific transgene expression was maintained (13). Furthermore, we also demonstrated that the 859-bp region was also capable of stimulating proximal promoter activity in transient transfections of cells derived from thyrotropes (␣-TSH) and gonadotropes (␣-T3), both of which express the ␣-subunit gene endogenously (13). More recently, our laboratory (14) has shown that the functional interaction between the upstream enhancer and the proximal promoter in transfections of both cell lines is dependent on an intact pituitary glycoprotein hormone binding element (PGBE) located from Ϫ337 to Ϫ330, an area within the proximal area that was shown to bind a LIM homeodomain factor (15,16). In addition, interaction at the more proximal gonadotrope-specific element (GSE) from Ϫ213 to Ϫ200, which binds the steroidogenic factor SF-1 (17), is also required for the enhancer to function in gonadotrope cells but not in thyrotrope cells (14). To identify the sites of functional interaction within the upstream region, we now report that the most proximal 125 bp of the previous 859-bp enhancing area are sufficient to account for all of the proximal promoter stimulatory activity in both pituitary-derived cell types that express the ␣-subunit. We demonstrate that the native orientation of the region with respect to the proximal promoter is crucial to maintain activation. We also show that several areas within this region bind proteins present in nuclear extracts from pituitary-derived cells including somatotrope-derived GH3 cells and that the areas of interaction contain potential binding sites for known transcription factors that also bind at important functional areas within the proximal promoter (17)(18)(19). Interestingly, we further show that the same upstream region inhibits proximal ␣-promoter activity in GH3 cells. The mechanism whereby the upstream region exerts its repressive effect differs from activation in that it is not dependent on interaction at two proximal promoter binding sites and on the orientation of the upstream region with respect to the proximal region.

EXPERIMENTAL PROCEDURES
Plasmid Constructions for Enhancer-Proximal Promoter Analysis and Transient Transfection Assays-The construction of luciferase vectors containing the region from Ϫ341 to ϩ43 of the ␣-subunit promoter without and with mutated PGBE and GSE sites and the same plasmids with the 859-bp upstream enhancer fused upstream were described previously (13,14). As a first approach, the 859-bp region was subdivided by digestion at PstI restriction sites. After separation on agarose gels and purification using Qiaex II resin (Qiagen), each fragment was inserted into the PstI site of pGEM5zfϩ (Promega). Subsequent exci-sion with EcoRV and HindII generated a blunt-ended fragment that was ligated into the SmaI site of a plasmid that contained the region from Ϫ341 to ϩ43 of the ␣-subunit promoter inserted between the BamHI and HindIII sites of pSELECT (Promega). After checking by sequencing for the correct forward orientation of each PstI fragment, the fused enhancer/promoter region was excised by digestion with KpnI and HindIII and inserted between the same sites of the promoterless pA3LUC mammalian expression vector (20,21). This resulted in luciferase expression plasmids with the areas from 70 to 246, from 241 to 654, and from 694 to 833 (renumbered as 1-184 in Fig. 6, also see the area between the underlined PstI sites in Fig. 2) fused upstream of the proximal ␣-subunit promoter region from Ϫ341 to ϩ43. Progressively shorter 5Ј prime truncations of the enhancer region were generated using a PCR strategy that employed sense oligonucleotides with the following sequences: (a) 5Ј-GCCGGTACCCTGCAGGTCTGCACATA- These oligonucleotides comprise sequences originating at positions 241, 421, 552, 649, 733, 798, and 827, respectively, of the upstream 859-bp enhancer sequence, respectively (GenBank TM accession number AFF044976). The last three represent positions 85, 150, and 179 of the renumbered 210-bp region shown in Fig. 2. They were designed with a 5Ј KpnI site (italicized) to facilitate subsequent subcloning. Each oligonucleotide was used in conjunction with a common antisense 22-nucleotide amplimer that is complementary to the sequence coding for amino acids 4 -10 of luciferase and used to amplify the appropriate truncated enhancer/promoter fragment from the full-length 859-bp enhancer/promoter fusion in pA3LUC. The resulting fragments were then digested with KpnI and HindIII and reinserted between the same sites of pA3LUC. A 5Ј deletion plasmid originating at position 85 and fused to a truncated Rous sarcoma virus promoter was similarly constructed starting with the previously described 859-bp enhancer Rous sarcoma virus fusion vector (13). To generate an enhancer/promoter plasmid containing sequences that extended downstream of position 859, we first subcloned, into pGEM7zf (Promega), a fragment that extended from Ϫ4600 to Ϫ3400 that was derived by KpnI and SphI digestion from a plasmid containing ␣-subunit sequences from Ϫ5000 to ϩ43 (12). After sequencing this downstream 300-bp extension, we utilized PCR with this plasmid and a 5Ј oligonucleotide that spanned the KpnI site at Ϫ4600 along with an antisense oligonucleotide that originated 237 bp downstream of the BglII site at Ϫ3700 with 5Ј-GCGGGATCCATAAT-TCACCTTTAGGGAGG-3Ј. Incorporation of a BamHI site (italicized) allowed the amplified product to be excised with KpnI and BamHI and inserted in the forward orientation between the same sites of the pSELECT plasmid containing the Ϫ341 to ϩ43 ␣-subunit promoter sequence. The 3Ј extended enhancer/promoter fusion region was then excised with KpnI and HindIII and cloned into pA3LUC as before. An ␣-subunit proximal promoter luciferase plasmid containing the 210-bp region inserted in the opposite orientation was constructed using the following PCR strategy. An oligonucleotide was synthesized that had the same sequence as the one described above that originated at position 649 of the 859-bp enhancer sequence, except that a BamHI site was incorporated at the 5Ј end instead of a KpnI site. When this was used in PCR with an antisense oligonucleotide that spanned the downstream BglII site (at 859), it allowed the amplified product to be digested with BamHI and BglII and inserted at the BamHI site upstream of position Ϫ341 in the pSELECT plasmid containing this proximal promoter region. Because BamHI and BglII overhangs can anneal, it allowed the generation of enhancer/promoter fusions with the 210-bp region in both orientations. The reverse orientation was identified by sequencing, and the enhancer/promoter fragment was recloned into pA3LUC as described above.
Transient transfections using electroporation for ␣-TSH (3 ϫ 10 6 ), ␣-T3 (4 ϫ 10 6 ), and GH3 (4 ϫ 10 6 ) cells were performed as described previously (13,22). 20 g of ␣-subunit promoter and enhancer/promoter luciferase constructs were used for the ␣-TSH and ␣-T3 cell transfections, and 2 g of a cytomegalovirus ␤-galactosidase plasmid were included as an internal control of transfection efficiency. Transfections were carried out in duplicate with a Rous sarcoma virus promoter luciferase vector and a promoterless pA3LUC vector as positive and negative controls (20). Experiments were performed a minimum of three times with at least two preparations of each plasmid. After 16 -24 h, luciferase and ␤-galactosidase activities were measured from duplicate aliquots of freeze-thawed cytoplasmic lysates. Luciferase activities of the various enhancer/promoter constructs were normalized to the corresponding ␤-galactosidase value and expressed as the fold stimulation Ϯ S.E. of the normalized activity of the Ϫ341 to ϩ43 promoter in the various cell types.
Preparation of Nuclear Extracts and DNase I Protection Analyses-Nuclear extracts were prepared from dispersed TtT-97 thyrotropic tumors or ␣-TSH, ␣-T3, and GH3 cells as described previously (23,24). Cultured cells were placed in medium containing 10% charcoal-stripped fetal calf serum for 48 h before harvesting for extract preparation. Protein concentrations were determined by Bio-Rad DC Protein Assay (Bio-Rad) using bovine serum albumin (Roche Molecular Biochemicals) as a standard. The previously described pGEM5zf plasmid containing the 184-bp PstI fragment from position 649 to position 833 of the 859-bp enhancer region was digested with NotI and NdeI to generate a fragment for footprinting analysis that could be labeled uniquely at the upstream end using [␣-32 P]dGTP and dCTP to fill in the NotI overhang (GGCC). To generate a footprinting fragment that was labeled at the downstream position, we constructed a plasmid that contained a fragment extending from 649 to a position 100 bp 3Ј of the BglII site. This was amplified by PCR from the previously described KpnI to SphI fragment containing vector using the KpnI 5Ј tagged sense primer originating at 649 and an antisense amplimer with a 5Ј HindIII site (italicized) with the following sequence: 5Ј-GCGAAGCTTGGGG-GAAATATCACTGCATG-3Ј. As before, excision with EcoRI and MluI enabled unique filling of the MluI overhang (CGCG) with [␣-32 P]dGTP and dCTP. This 3Ј extension allowed the area immediately 5Ј of the BglII site to be displayed in an optimal area of the footprint gel. DNase I protection assays were carried out as described previously (24). Briefly, radiolabeled probes were allowed to interact with 10 g of bovine serum albumin (no extract) or 60 -70 g of nuclear extract protein derived from TtT-97, ␣-TSH, ␣-T3, or GH3 cells, subjected to DNase I digestion under defined conditions, and analyzed on a denaturing 5% polyacrylamide-8 M urea gel.

Localization of Proximal Promoter Activation to the Proximal 210 bp of the Upstream Region-Previous
transgenic and transient transfection data indicated that sequences required for the enhancement of the ␣-subunit proximal promoter specifically in thyrotropes and gonadotropes were present in a KpnI-BglII fragment located between Ϫ4.6 and Ϫ3.7 kilobase pairs upstream of the transcriptional start site (13). To further localize the sequences responsible for enhancement within this 859-bp region, subfragments of the KpnI-BglII fragment were fused to the Ϫ341 to ϩ43 proximal promoter region and analyzed independently for their ability to enhance in both thyrotrope-and gonadotrope-derived cells. The results of this analysis are shown in Fig. 1. A strategy that involved progressively deleting sequences from the 5Ј end demonstrated that the enhancing effect of the 859-bp area previously reported in both thyrotrope-derived ␣-TSH cells and gonadotrope-derived ␣-T3 cells (13) could be entirely accounted for by the most proximal 210 bp from 649 to 859. The lack of other areas eliciting any enhancing capability was further confirmed by showing that neither of two PstI fragments, which comprised all but the terminal 69 bp of the sequence upstream of position 654, exhibited any capacity to stimulate the proximal promoter ( Fig.  1). Because the sequences responsible for the stimulatory effect appeared to map to the region immediately 5Ј of the BglII site at Ϫ3700 (13), we thought it important to determine if sequences immediately downstream of the BglII site could further augment the previously observed enhancement. Therefore, a larger fragment from Ϫ4600 to Ϫ3400 was generated from a previously described ␣-subunit genomic clone (12), and PCR was used to amplify a fragment that extended more proximally to a position 237 bp downstream of the BglII site. This was inserted upstream of Ϫ341 in the proximal ␣-subunit promoter, and the stimulatory effect of this 3Ј augmented enhancer region was compared with the previous 859-bp area. The results of this analysis revealed no further enhancement in either cell type when the enhancer region fused to the proximal promoter-included sequences 3Ј of the BglII site (data not shown). Therefore, the sequences mediating enhancement appear to be entirely contained within the previously determined 210-bp area.
Nuclear Proteins from Pituitary Cells Bind to Areas of the 210-bp Region Corresponding to Transcription Factor Consensus Binding Sites-The nucleotide sequence of the 210-bp region that accounts for all of the enhancer activity of the upstream 5Ј region of the mouse ␣-subunit gene is shown in Fig.  2. Nucleotides 649 -859 of the previously described 859-bp enhancer region redesignated 1-210 are shown. The sequence shaded in gray denotes the location of a 60-bp sequence that is 60 -70% homologous to other sequences that occur within several reported rodent genes as revealed by a GenBank TM search. Boxed and labeled sequences correspond to consensus binding sites for several transcription factors that have been reported previously to play a role in the activity of the proximal ␣-subunit promoter (17)(18)(19) as well as in other pituitary genes (25)(26)(27)(28)(29). These include several binding sites for GATA and ETS factors. Also present are two E-box motifs (CANNTG) and single sites that could potentially interact with Pit-1, Sp1, and nuclear hormone receptor family members such as the steroidogenic factor SF-1. A proximal promoter area from Ϫ213 to Ϫ200 that binds SF-1 (the GSE) has been reported by our group (14) and by others (5,17) to be critical for both basal and enhancer-stimulated ␣-subunit promoter activity in gonadotrope cells. To begin to investigate which proteins are interacting with the 210-bp region and could perhaps be playing a role in the enhancement of the proximal promoter, we performed a comparative DNase I footprinting analysis using nuclear extracts derived from pituitary cells that express the endogenous ␣-subunit gene (TtT-97 thyrotropic tumor, ␣-TSH and ␣-T3 cells) as well as those that do not (somatotrope-derived GH3 cells). Fig. 3 shows the results of this analysis using fragments encompassing the 210-bp region that have been labeled at either the upstream (A) or downstream (B) position. In Fig. 3A, which utilized a 184-bp PstI fragment that terminates 26 bp from the proximal end of the fully enhancing region (see Fig. 2), protection from DNase I digestion can be seen in three general areas. Using DNA size standards loaded in a parallel lane (Stds) the approximate location of the footprinted areas are from 77 to 112, from 132 to 150, and from 152 to 178 on the antisense strand. No protection by any extract was evident upstream of position 77. The area from 77 to 112 appears to be similarly protected by the two thyrotrope cell extracts (TtT-97 and ␣-TSH), but the footprint generated by ␣-T3 nuclear extracts, although overlapping the thyrotrope footprint, appears to be less well protected for the most proximal 10 bp, suggesting that a different factor(s) present in each cell type may be interacting at this region. As shown in Fig. 2, this area contains sequences that correspond to binding sites for both E-box and Ets factors. When the sense strand labeled at the downstream position was analyzed, the same overall area was protected (Fig. 3B), but the area from 74 to 129 was subdivided into three clearly defined footprints from 74 to 86, from 92 to 105, and from 109 to 129. The first two were seen with all of the extracts, including those from GH3 cells. These encompass the upstream E-box and Ets sites. The protected area from 109 to 129, which contained a second Ets site, was observed with both thyrotropederived cell nuclear extracts (TtT-97 and ␣-TSH) but appeared not to be protected by either ␣-T3 or GH3 nuclear extracts. Two other more downstream footprints from 133 to 147 and from 152 to 178 that were also seen with the other strand were also protected by all of the extracts, including those from GH3 cells. These footprints contain a second E-box motif and potential binding sites for GATA factors, Sp1 and SF1. It is of interest that parts of the area of the enhancer that interact with the pituitary cell nuclear extracts colocalize with the 60-bp region that is repeated elsewhere in the mouse genome, shown in Fig.  2 as a gray box. A surprising observation resulting from these protein-DNA binding experiments suggests that non-␣-subunit-expressing pituitary GH3 cells may contain factors that interact with the enhancer region, and thus such factors are not confined to cell types that express the endogenous gene, i.e. thyrotropes and gonadotropes.
The 210-bp Region Represses Proximal Promoter Activity in GH3 Somatotrope-derived Cells-The finding that nuclear proteins from somatotrope-derived GH3 cells, which do not express the ␣-subunit, bind to sequences within a region that is involved in the activation of proximal ␣-subunit promoter activity in homologous ␣-TSH and ␣-T3 cells prompted us to investigate whether the upstream 210-bp area was also capable of enhancement in GH3 cells. We had previously shown that in non-pituitary CV-1 monkey kidney cells, the larger 859-bp region only modestly stimulated the proximal ␣-subunit promoter, although it was capable of enhancing the activity of a viral promoter in a cell type-independent fashion (13). Fig. 4 shows the effect of the 210-bp area on the proximal ␣-subunit promoter in GH3 cells compared with that previously seen in ␣-TSH and ␣-T3 cells. As shown before, the 210-bp area confers a 25-and 7-fold stimulation to the proximal promoter in ␣-TSH and ␣-T3 cells, respectively. However, in the non-␣-subunitexpressing pituitary-derived GH3 cells, the upstream 210-bp region dramatically inhibits the activity of the proximal pro-moter by 75% (note that the scale of the x axis in Fig. 4 is not linear). Therefore, the upstream regulatory region serves a role not only to activate ␣-subunit expression in ␣-subunit-expressing cell types but also to repress it in another pituitary cell type that does not express the endogenous gene.
Reversing the Orientation of the 210-bp Region Results in Loss of Activation but not Repression-We previously showed that the upstream region exhibited positional independence by its ability to confer enhancement when fused directly to the proximal Ϫ341 ␣-promoter in transgenic mice as well as in transiently transfected ␣-subunit-expressing cells (13). To see if the elements responsible for the stimulatory effect satisfied another criterion of a classical enhancer, that of orientation independence, the 210-bp fragment was inserted in the opposite direction and assessed for proximal promoter enhancement. Fig. 4 shows that in both ␣-TSH and ␣-T3 cells, the enhancing activity was totally abrogated by reversing the direction of the 210-bp region with respect to the proximal promoter. Thus, the elements responsible for stimulating the proximal promoter are unable to functionally interact with the proximal promoter when their orientation, with respect to those elements located proximally such as the PGBE and GSE sites, were reversed. In fact, in ␣-T3 cells, the opposite orientation of the 210-bp area consistently inhibited proximal promoter activity by 40 -50%. In contrast, reversing the orientation of the 210-bp area had no effect on its ability to inhibit the proximal ␣-subunit promoter in GH3 cells, suggesting that the mechanism of repression was orientation-independent and therefore different from that mediating the activation. (14), we demonstrated that in order for the larger 859-bp enhancer area to exert its maximum stimulatory effect on the proximal ␣-subunit promoter, interaction of proteins present in ␣-subunitexpressing cells with proximal promoter elements must be preserved. Mutation of the proximal PGBE site, which leads to the disruption of binding most likely of a LIM homeodomain protein (15,16), resulted in a dramatic reduction of enhancing capacity in ␣-TSH cells and complete abrogation of proximal promoter stimulation in ␣-T3 cells. However, mutation of the more proximal GSE site, which was shown to affect the binding of a gonadotrope cell protein as well as the steroidogenic factor SF1, reduced the effect of the upstream enhancer only in ␣-T3 cells (14). We now wanted to see if repression by the upstream region was also dependent on interaction at these proximal elements. The same constructs with the 859-bp region fused to the wild type or mutated Ϫ341 to ϩ43 proximal promoter were transfected into GH3 cells, and the results of these studies are shown in Fig. 5. The larger 859-bp area also inhibits the proximal promoter to a similar degree as the extreme 3Ј 210-bp region. The data in Fig. 5 also further demonstrate that mutation of either the PGBE or GSE site still resulted in the ability of the upstream region to inhibit proximal promoter activity in GH3 cells. This further supports the notion that to mediate repression of ␣-subunit promoter activity in GH3 cells, the upstream region functionally interacts with the proximal promoter through a different mechanism than does activation in thyrotrope-and gonadotrope-derived cells. Also presented in Fig. 5 are data that show that although the upstream regulatory region inhibits the ␣-subunit promoter in GH3 cells, a viral promoter was stimulated 2.4-fold, which was similar to that seen previously in ␣-TSH, ␣-T3, and CV-1 cells (13).

Repression in Somatotrope Cells Is Not Dependent on Intact Proximal PGBE or GSE Sites-In a previous report
Both Activation and Repression Are Dependent on Multiple Sites-To further dissect the specific elements within the 210-bp area that might be contributing to both activation and repression, we devised 5Ј and 3Ј deletion strategies that pro- FIG. 4. The 210-bp region represses proximal promoter activity in GH3 cells. 20 g of a construct containing the 210-bp region fused in either the forward or reverse orientation (denoted by the 3 or 4, respectively) to the ␣-subunit proximal promoter region from Ϫ341 to ϩ43 or one with the proximal region alone was transfected as before into ␣-TSH or ␣-T3 cells or into 4 ϫ 10 6 GH3 cells along with 2 g of a ␤-galactosidase transfection control. The figure shows the fold stimulation in ␣-TSH and ␣-T3 cells and the inhibition of the proximal promoter activity (set to 1) in GH3 cells derived from the normalized enhancer/promoter and promoter alone luciferase activities Ϯ S.E. for n determinations. Note that the x axis below 1 has been expanded to display both stimulation and inhibition on the same graph.
gressively eliminated the footprinted areas which contained the putative transcription factor binding sites shown in Fig. 2 and also shown schematically in Fig. 6A. Fig. 6B shows that the most 5Ј 85 bp, which contain two GATA sites and a Pit-1 motif, contribute to neither activation in ␣-TSH or ␣-T3 cells or repression in GH3 cells of the proximal ␣-subunit promoter. However, the truncated region still continues to stimulate a viral promoter 2-4 fold in all cell types. This area was also not footprinted by the pituitary cell nuclear extracts (Fig. 3), except at the most 3Ј 10 bp. However, removal of an additional 65 bp (to 150) which harbors two E-boxes, two Ets sites, and a GATA site dramatically decreased enhancement to only 2-fold in ␣-TSH cells and actually inhibited the proximal promoter by 50% in ␣-T3 cells. In GH3 cells, an inhibition of proximal ␣-subunit promoter activity of only 25% was seen as a result of removing the 5Ј 150 bp. Deletion of an additional 28 bp, leaving only the terminal 32 bp which contains two GATA motifs, resulted in a fragment that had no activating or repressing effect on the proximal promoter in any cell type. Although the terminal 32 bp had no activity alone, when they were deleted in the context of the 210-bp region (1-184), reduced stimulatory capability in both ␣-TSH and ␣-T3 cells (from 20-to 12-fold and from 6-to 3-fold, respectively) was observed. Inhibition in GH3 cells was maintained, but only to 50%. These data suggest that both activation and repression of the proximal ␣-subunit promoter can be mediated by only the terminal 125 bp of the previously described upstream region from Ϫ4600 to Ϫ3700 and that within that area, the elements that mediate activation in ␣-TSH cells may differ from those in ␣-T3 cells, and, in turn, these may also differ from those that confer GH3 cell-dependent repression.

DISCUSSION
In this report, we have further characterized the sequences present in a 859-bp upstream region of the mouse ␣-subunit gene that are capable of directing specific pituitary cell-dependent activation of the proximal promoter. We have localized the enhancer function to the most proximal portion of the 859-bp region and showed that sequences immediately downstream did not contribute any further stimulatory effect. Utilizing a DNase footprinting approach, we showed that this area interacts with proteins present in thyrotrope-and gonadotropederived cells. Some of these protected areas contain consensus recognition sites for transcription factors that play a role in the activity of the proximal ␣-subunit promoter or other pituitary promoters in transiently transfected pituitary-derived cell lines. These include binding sites for GATA factors that contribute to proximal ␣-subunit promoter activity in gonadotropes (18) and also synergize with Pit-1 to activate the TSH␤ promoter in thyrotrope cells (30). A perfect consensus nuclear hormone response element, which could potentially bind SF-1, and two E-box sequences are located within the footprinted areas. Similar motifs in the proximal promoter have been shown to be important for ␣-subunit proximal promoter activity in ␣-T3 cells (5,19). Finally, sequences potentially capable of binding Pit-1, Sp1 and Ets factors, which contribute to growth hormone and prolactin promoter activity (25)(26)(27)(28)(29), are also located in the functional upstream area. Pit-1 protein is not present in ␣-TSH cells (31) and therefore does not appear to be required for ␣-subunit expression.
The mechanism whereby an enhancer located Ͼ3000 bp distant can act on proximal elements to influence transcriptional efficiency represents an intriguing problem in understanding gene expression. We previously showed that the ability of the enhancer to functionally interact with the proximal promoter was dependent on a proximal site (PGBE) that interacted with a LIM homeodomain protein (15,16). Such LIM domain-containing proteins have been shown to participate in proteinprotein interactions through adaptor molecules referred to as LIM domain-binding (Ldb) proteins (32, 33). Bach et al. (34) cloned a similar cofactor, termed clim, from mouse pituitary FIG. 5. Mutation of either the proximal PGBE or GSE site has no effect on repression by the upstream region in GH3 cells. 20 g of a luciferase plasmid containing the ␣-subunit proximal region from Ϫ341 to ϩ43 or one with the same region bearing a mutated (X), PGBE (P), or GSE (G) proximal site (all with or without the 859-bp enhancer region fused upstream) was electroporated into GH3 cells as described in the legend to Fig. 4. A similar vector containing a truncated Rous sarcoma virus viral promoter with or without the 859-bp region was also included. The figure shows the fold stimulation or inhibition of the viral promoter or proximal promoter activity, respectively, Ϯ S.E. for n determinations, calculated from the normalized luciferase activities as described in Fig. 4. gland using a two-hybrid screen with the LIM domain of the pituitary restricted P-Lim/Lhx3 homeodomain factor and showed that it conferred transcriptional synergy between P-Lim and P-OTX, another homeodomain protein on the proximal ␣-subunit promoter. Interestingly, a Drosophila homolog of these Ldb cofactors termed Chip was identified in a genetic screen designed to identify factors that facilitate communication between remote enhancers and promoters (35). In related studies, Wadman et al. (36) demonstrated that the nucleotide sequence preferred by erythroleukemia cell complexes, which contained Ldb, possessed an arrangement of E-box and GATA motifs similar to that present between positions 132 and 148 of the upstream enhancer (Fig. 2). It is intriguing to speculate that a Ldb factor present in thyrotropes and gonadotropes, but not in somatotropes, mediates the enhancer/promoter functional interaction that stabilizes the complex of a helix-loophelix protein and a GATA factor. GATA-2 has been reported to be expressed early in pituitary gland development (37) and to cooperate with Pit-1 in the cell-specific expression of the TSH␤ subunit in thyrotropes (30). It also contributes to proximal ␣-subunit promoter activity in of ␣-T3 gonadotropes (18).
Although the pattern of protection of the 210 bp fragment by GH3 cell nuclear extracts qualitatively resembled that generated by extracts from expressing ␣-TSH and ␣-T3 cells, the 210-bp region dramatically repressed the proximal ␣-subunit promoter in GH3 cells. Inhibition of ␣-subunit gene expression by the upstream region in GH3 cells was perhaps not unexpected, as the flanking region between Ϫ4600 and Ϫ3700 was able to restrict the function of the Ϫ341 ␣-promoter, which was active in all pituitary cells, to only thyrotropes and gonadotropes of transgenic mice (13). The current transfection data support our previous notion that both the upstream region and the proximal area between Ϫ381 and Ϫ341 (13) can exert pituitary cell restrictive properties, and this is accomplished in somatotropes through a functional interaction of the upstream region with the proximal region to suppress ectopic activity. Because P-Lim, which interacts at the PGBE site (15,16) that is required for maximal enhancer function (14), is present in GH3 cells (38), it is highly likely that the inhibitory effect is also being mediated through interference with binding at the PGBE site. However, our data are not consistent with such a mechanism as repression in GH3 cells by the upstream regulator is still seen when the PGBE site is mutated, suggesting that functional interaction at a different proximal site or sites via a different mechanism is involved in repression.
Another piece of evidence that supports a different mechanism of action for the repressive effect is the demonstration that fusing the 210-bp area to the Ϫ341 ␣-promoter in the reverse orientation completely abolishes its ability to behave as an enhancer in ␣-TSH and ␣-T3 cells, whereas it was equally effective as a repressor in GH3 cells in either orientation (Fig.  4). This suggests that the factor(s) interacting with upstream elements to mediate repression, unlike those that confer activation, can interact functionally with proximal elements, irrespective of their physical or spatial geometry. Interestingly, in ␣-TSH cells, the reversed enhancer merely abolished activation, whereas in ␣-T3 cells, it resulted in the acquisition of a 50% inhibition of proximal promoter activity. Because the level of activation in ␣-T3 cells was consistently less than that in ␣-TSH cells (7-fold versus 20-fold), it is tempting to speculate that ␣-TSH cells and ␣-T3 cells both contain an enhancer binding activator(s), but that ␣-T3 cells also contain an inhibitory factor(s) that competes with the activator(s) to lower the overall level of enhancement. When the influence of the activator(s) is nullified by reversing the orientation, the inhibitor(s) now becomes dominant, resulting in the repressive effect seen in ␣-T3 cells.
The results of these studies may have important implications regarding the emergence of differentiated pituitary cell phenotypes from a common precursor in Rathke's pouch precursor cell that expresses the ␣-subunit (37). Our previous transgenic data (12) showed that the onset of ␣-enhancer-driven transgene expression occurred by embryonic day 9.5 in Rathke's pouch, which parallels the temporal and spatial expression of the endogenous ␣-subunit gene (39,40). This suggested that the factor(s) required for activation is already present and initiates ␣-subunit gene expression at this early stage of development, before any hormone-producing cells have emerged. As the developmental program progresses, an inhibitory factor(s) is expressed in cells destined to develop into the somatotrope lineage. However, the presomatotropes also cease expression of the activator(s), and the inhibitor(s) is thus able to effectively silence the endogenous ␣-subunit gene. This idea is supported by the finding that certain naturally occurring pituitary tumors that probably arise from premature undifferentiated cells continue to express both the ␣-subunit and growth hormone (41,42).
The deletion studies to further define the elements responsible for activation and repression within the 210-bp region show that the 5Ј most 85 bp are not involved in either effect, but the remaining 125 bp harbor multiple cell type-specific areas involved in both activation and repression. In ␣-T3 cells, activation was mapped to the 65 bp between 85 and 150, and, similar to what was seen with the reverse orientation when enhancement was nullified by the removal of this region, the remaining 60 bp now developed inhibitory properties. In contrast, the 3Ј terminal 60-bp region, which lacks the E-box/ GATA composite site described earlier, was still activated by about 2-fold in ␣-TSH cells but was only modestly inhibited in GH3 cells. These data underscore the complexity of this novel upstream regulator of the ␣-subunit gene which mediates different influences on the cognate proximal promoter region depending on the nuclear factor content of the pituitary cell type in which it is being assayed. To our knowledge, no other gene regulatory region exhibits these opposing cell-specific influences through functional elements in such close proximity to one another. There are many reported examples of classical enhancer regions that are both position-and orientation-independent and interact to stimulate expression from proximal promoter regions of genes such as Pit-1 (43) and ␣-globin (44). Other elements such as the neuron restrictive silencer element have been reported to silence expression from brain-specific genes in non-neural tissue (45). Since the ␣-subunit regulatory region appears to carry out both activation and repression through the same or closely associated elements by different mechanisms, we believe that this unique regulatory region will represent a new class of control region important for governing complex temporal and spatial patterns of gene activation and repression.