Negative regulatory element involved in the hormonal regulation of GlcNAc-1-P transferase gene in mouse mammary gland.

The gene encoding UDP-GlcNAc:dolichol phosphate N-acetylglucosamine-1-phosphate transferase (GPT), the enzyme that initiates the pathway for the biosynthesis of asparagine-linked glycoproteins, is ubiquitously expressed in eukaryotic cells. However, its expression in the mammary gland is developmentally and hormonally regulated; transcription of the mouse mammary GPT gene is stimulated by the lactogenic hormones, insulin, glucocorticoid, and prolactin. The involvement of cisacting elements in regulating the expression of the mouse GPT (mGPT) gene was investigated by transient transfections of various GPT promoter/luciferase (Luc) constructs into primary mouse mammary epithelial cells. A series of 5'-deletions of the GPT promoter identified a distal negative regulatory region (base pairs -1057 to -968) and deletion of this region results in enhanced hormonal induction (approximately 7-fold) with no effect on basal promoter activity. Electrophoretic mobility shift assays (EMSA) performed with nuclear extracts from different developmental stages of mouse mammary gland demonstrated that the binding activity of the nuclear proteins to the distal negative regulatory region was predominant in virgin stage as compared with pregnant and lactating stages. EMSA performed with nuclear extracts from virgin explants showed that the binding activity was markedly decreased after cultivation with the combination of the three lactogenic hormones. DNase I footprinting analysis identified two pentamer direct repeat motifs, AGGAA and GAAAC, within the negative regulatory region. EMSA competition experiments showed that mutations within the direct repeats failed to compete for binding of the nuclear proteins to labeled wild type oligonucleotide. Transcription from the promoter containing the mutated direct repeats was increased greatly, consistent with the conclusion that these motifs functions in vivo to repress GPT gene expression. These data suggest the importance of the negative regulatory region in hormonal control of mGPT gene expression in mammary gland.

The gene encoding UDP-GlcNAc:dolichol phosphate N-acetylglucosamine-1-phosphate transferase (GPT), the enzyme that initiates the pathway for the biosynthesis of asparagine-linked glycoproteins, is ubiquitously expressed in eukaryotic cells. However, its expression in the mammary gland is developmentally and hormonally regulated; transcription of the mouse mammary GPT gene is stimulated by the lactogenic hormones, insulin, glucocorticoid, and prolactin. The involvement of cisacting elements in regulating the expression of the mouse GPT (mGPT) gene was investigated by transient transfections of various GPT promoter/luciferase (Luc) constructs into primary mouse mammary epithelial cells. A series of 5-deletions of the GPT promoter identified a distal negative regulatory region (base pairs ؊1057 to ؊968) and deletion of this region results in enhanced hormonal induction (ϳ7-fold) with no effect on basal promoter activity. Electrophoretic mobility shift assays (EMSA) performed with nuclear extracts from different developmental stages of mouse mammary gland demonstrated that the binding activity of the nuclear proteins to the distal negative regulatory region was predominant in virgin stage as compared with pregnant and lactating stages. EMSA performed with nuclear extracts from virgin explants showed that the binding activity was markedly decreased after cultivation with the combination of the three lactogenic hormones. DNase I footprinting analysis identified two pentamer direct repeat motifs, AGGAA and GAAAC, within the negative regulatory region. EMSA competition experiments showed that mutations within the direct repeats failed to compete for binding of the nuclear proteins to labeled wild type oligonucleotide. Transcription from the promoter containing the mutated direct repeats was increased greatly, consistent with the conclusion that these motifs functions in vivo to repress GPT gene expression. These data suggest the importance of the negative regulatory region in hormonal control of mGPT gene expression in mammary gland.
Asparagine-linked glycosylation in eukaryotes is initiated by a stepwise, dolichol-linked assembly of Glc 3 Man 9 GlcNAc 2 -P-Pdolichol and its transfer en bloc to the nascent polypeptide in the endoplasmic reticulum, followed by co-and post-translational modifications of the tetradecasaccharide in the secretory pathway (1). The enzyme UDP-GlcNAc:Dol-P GlcNAc-1-P transferase (GPT) 1 catalyzes the first and the committed step, i.e. the transfer of GlcNAc-1-P from UDP-GlcNAc to Dol-P to form GlcNAc-P-P-Dol, in the assembly of the oligosaccharide precursor. It is, therefore, a potential target for the overall regulation of oligosaccharide assembly and protein N-glycosylation.
GPT has been purified to apparent homogeneity from bovine mammary gland (2); the cDNAs encoding GPT enzyme have been isolated from cDNA libraries prepared from mRNA of tunicamycin-resistant Chinese hamster ovary cell lines (3,4) and mouse mammary gland (5). The GPT gene has been cloned from yeast (6), Leishmania (7), and mouse (8). These studies show that GPT is a multispan hydrophobic protein in the endoplasmic reticulum, and its catalytic site has a cytoplasmic orientation (9,10). We previously reported that the mouse GPT gene maps to the chromosome 17 (5), is encoded in nine exons spanning 7.5 kilobases (kb), and its promoter is of the housekeeping type (8). The GPT gene promoter lacks TATA box, is GC-rich, and has multiple transcription initiation sites spread over a large region. These features are consistent with the ubiquitous occurrence of N-linked glycosylation process, and the presence of GPT mRNA in many tissues examined. However, an earlier report showed that the activities of three key enzymes of the dolichol cycle, namely GPT, GDP:Man-Dol-P mannosyltransferase, and UDP-Glc:Dol-P glucosyltransferase, are developmentally regulated in mouse mammary gland (11).
The mammary gland offers a unique model for studying the regulation of protein N-glycosylation. It synthesizes and secretes only a very limited number of well characterized Nlinked glycoproteins, e.g. ␣-lactalbumin and transferrin. The gland is intensely modulated by mammogenic and lactogenic hormones as it repeatedly cycles between dormancy, growth, and differentiation upon pregnancy, lactation, and post-lactational regression throughout the reproductive life of the female (13). A recent study from our laboratory showed that GPT enzyme activity, immunoreactive GPT protein, and GPT mRNA are developmentally regulated in mouse mammary gland (12). All three parameters are low in tissues from virgin and pregnant animals, increase steadily during lactation, reaching a peak around mid to late lactation, and decline thereafter in glands of post-lactating animals. Furthermore, GPT synthesis is stimulated at the level of RNA in mouse mammary explants and primary mammary epithelial cell cultures under the synergistic influence of insulin, prolactin, and hydrocortisone (12). These studies suggest that the developmental and hormonal regulation of GPT gene expression is mediated at the transcriptional and/or post-transcriptional levels.
The aim of our studies is to identify cis-acting elements involved in the developmental and hormonal regulation of GPT gene in the mouse mammary gland. In this report we show that the pentamer direct repeat motifs, AGGAA and GAAAC, within the distal negative regulatory region of the mGPT promoter play an important role in developmental and hormonal regulation of GPT gene expression in mouse mammary gland.

EXPERIMENTAL PROCEDURES
Materials-Restriction enzymes, T 4 DNA ligase, bacterial alkaline phosphatase, and tissue culture media were from Life Technologies, Inc.; S1 nuclease and the Klenow fragment of DNA polymerase I were from Pharmacia Biotech Inc.; 32 P-labeled radioisotopes were from Du-Pont NEN; luciferase assay system, ␤-galactosidase assay system and fmol™ sequencing system were from Promega Biotech; Gene Amp DNA amplification reagent kit was from Perkin-Elmer; TA Cloning™ system was from Invitrogen Corp; oligonucleotide primers were synthesized by Oligos Etc. Inc.; female mice (C3H/HeN) were from the animal center of the National Institutes of Health.
Plasmid Construction-All promoter deletion plasmids were obtained by subcloning SmaI-and XhoI-cut products of the polymerase chain reaction (PCR) in the corresponding sites of pGL2-Basic Luciferase vector (Promega). In all PCR amplification reactions, the template was the genomic subclone Eco.4 DNA (8). The upstream primers were complementary to the 15 base pairs (bp) of the 5Ј-end of the desired deletion flanked by a SmaI site and six additional nucleotides in the 5Ј-end. The downstream primer was complementary to the 5Ј noncoding sequence (bp Ϫ15 to Ϫ5) of the mouse GPT gene flanked by a XhoI site and six additional nucleotides in the 5Ј-end. Mutated constructs were obtained by PCR-based site-directed mutagenesis method (14). Briefly, two primary PCR reactions were performed to produce two overlapping DNA fragments bearing the same mutation using four primers; the outermost two primers were wild type and the inner two primers contained indicated mutations (Fig. 5B). The upstream and downstream outermost primers were complementary to the 5Ј noncoding sequences bp Ϫ1161 to Ϫ1147 and bp Ϫ15 to Ϫ5, respectively, of the mouse GPT gene flanked by either a SmaI site (upstream) or a XhoI site (downstream) and six additional nucleotides in the 5Ј-end. The upstream outermost primer and the downstream inner primer were used as one pair and the upstream inner primer and downstream outermost primer as another pair in the primary PCR amplifications. Excess primers were removed by gel electrophoresis. Small aliquots of purified, primary PCR products were then mixed, denatured, reannealed, and subjected to five cycles of PCR amplification without primers to allow the two fragment to recombine and extend. The resulting products then served as templates in the secondary PCR amplifications, using only the outermost two primers. The sequences of all constructs were verified by DNA sequencing using the fmol™ sequencing system.
DNase I Footprinting-The probe used was double-stranded DNA fragment from bp Ϫ1161 to Ϫ 984 (FP1). The FP1 was end-labeled in coding strand with [␣-32 P]dATP using the Klenow fragment of DNA polymerase I. Different amounts of nuclear extracts were incubated in a 20-l reaction mixture containing 10 mM Tris (pH 7.5), 100 mM NaCl, 2.5 mM MgCl 2 , 0.5 mM EDTA, 1 g of poly(dG-dC)⅐poly(dG-dC), 6.5% glycerol, and 1 mM DTT and 50,000 cpm end-labeled probe for 20 min at room temperature. After the incubation, 0.99 units of DNase I (Pharmacia) were added, and digestion was carried out at room temperature for 60 s. The reaction was stopped by the addition of 60 l of 4 ϫ DNase stop solution containing 768 mM NaOAc, 128 mM EDTA, and 0.65% SDS. The products were extracted by phenol/chloroform (1:1) and then separated by electrophoresis on 8% polyacrylamide, 8 M urea gels. The gel was then dried and subjected to autoradiography. G ϩ A reaction was performed by the Maxam and Gilbert method (16).
Electrophoretic Mobility Shift Assays-The double-stranded FP1 fragment and WT oligonucleotide (Ϫ1050 to Ϫ1004 bp) were used as probes. WT was end-labeled with [␥-32 P]ATP using T 4 polynucleotide kinase. The binding conditions for EMSA were same as described for the footprinting experiments except that 2 g of nuclear extract was added in each reaction. After 20 min of incubation at room temperature, the resulting complexes were subjected to electrophoresis in 5% polyacrylamide gels in 0.5 ϫ TBE buffer. Competition experiments were performed by adding the unlabeled oligonucleotides before the addition of the labeled probe. The DNA-protein complexes were visualized by autoradiography.
Cell Cultures and Transfection-For primary cell culture, mammary epithelial cells were isolated from mammary glands of pregnant mice by enzymic dissociation followed by differential centrifugation as described previously (17). 1.5 ϫ 10 6 cells were plated on a 35-mm plastic culture dish coated with Matrigel (0.3 ml/dish, Collaborative Research, Inc.). Cells were cultured at 37°C in 95% air, 5% CO 2 in 3 ml of Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, antibiotics, and the indicated hormones. Cells were transfected with 20 g of an appropriate plasmid DNA and 5 g of pSV-␤-galactosidase plasmid/ dish, using the calcium phosphate-precipitation method (18). Unless stated otherwise, the final concentration of added hormones was as follows: porcine insulin (I), 5 g/ml; hydrocortisone (H), 1 g/ml; and bovine prolactin (P), 5 g/ml. Medium was changed every other day. Cells were cultured for 5 days and harvested by treatment with 0.3% collagenase, 1% dispase. Cell extract was prepared by three cycles of freeze-thawing in 0.25 M Tris/HCl (pH 7.8) followed by centrifugation at 16,000 ϫ g for 20 min. For explant culture, explants were prepared from mammary glands of virgin mice and cultured as described (19).
Luciferase and ␤-Galactosidase Assays-Luciferase and ␤-galactosidase enzyme assays were performed according to the manufacturer's instructions (Promega). Luciferase activity was normalized for transfection efficiency relative to that obtained with the reference ␤-galactosidase plasmid pSV-␤-galactosidase control vector.

RESULTS
The sequence of the mGPT promoter revealed the presence of several potential binding sites for transcription factor such as Sp1 and AP2, sequences similar to the half-palindromic CTF/ NF1 element, and a 13-bp direct repeat motif (8). Using a promoterless Luc reporter vector, deletion mutations were targeted systematically to eliminate these potential binding sites to determine whether the sites and/or any other cis-acting elements were involved in developmental and hormonal regulation of GPT gene expression. The Luc constructs containing the deleted promoter fragments were transiently transfected in the PMME cells and the basal and induced activities were determined (Fig. 1). Deletion from bp Ϫ1489 to Ϫ264 had little effect on the basal level of Luc activity. When the 5Ј deletion reached to bp Ϫ175, the basal level of Luc activity dropped to half, indicating that the sequence between bp Ϫ264 and Ϫ175 is important for the basal promoter activity. Since the major transcription initiation sites were located in this region (8), the possibility that removal of these sites caused the reduction of promoter activity could not be ruled out.
Hormonally induced Luc activity was increased dramatically by the removal of the sequence between bp Ϫ1057 and Ϫ968 and construct B7 (ϳ7-fold) was one of the most active constructs analyzed. The hormone inducibility was greatly reduced when the sequence between bp Ϫ264 and Ϫ92 was deleted ( Fig. 1). Again, the decrease in induction activity could result from the removal of either positive regulatory element(s) or the major transcription initiation sites in this region. These results demonstrate that a distal fragment from Ϫ1057 to Ϫ968 behaved as a negative regulatory region and is crucial for induction of promoter activity by the lactogenic hormones.
The region involved in the negative regulation contained a 13-bp direct repeat sequence, TGACTGGGAGGAA, at bp Ϫ1052 to Ϫ1039 and bp Ϫ1033 to Ϫ1020. Since some transcription factors act through binding to direct repeat sequences (such as vitamin D, thyroid hormone, and retinoic acid response elements, AGGTCA) (20), it was speculated that this direct repeat sequence was involved in binding a trans-acting factor that down-regulates GPT expression. To examine the possible presence of trans-acting factors that interact with the negative regulatory region, EMSA were performed using a 178-bp DNA fragment FP1 that encompasses the negative regulatory region and nuclear extracts prepared from mouse mammary glands of different developmental stages. As shown in Fig. 2A, FP1 formed a major shifted band with the nuclear proteins from virgin mouse mammary gland. The complex was abolished by addition of 50-fold molar excess of either unlabeled FP1 (data not shown) or a much shorter oligonucleotide WT (lane 3). In contrast, when a unrelated oligonucleotide FP2 was used as a nonspecific competitor, it failed to abolish the binding (lane 2). In addition, the amount of the complex increased with increasing protein concentration (lanes 4 -6), sug-gesting that it was a result of specific interactions. The binding activity to the FP1 was highest in virgin stage (Fig. 2B, lane 1) and was lowest in lactating stage (lane 3). The binding activities were intermediate at pregnant stage (lane 2) and returned to a higher level in post-lactating stage (lane 4). Results from EMSA experiment involving mixing of nuclear extracts from the virgin and lactating tissues showed the same binding pattern as virgin nuclear extract alone (data not shown), suggesting that lactating extracts did not contain any factor(s) that inhibit the DNA-protein complex formation. Since the adipocytes comprise the most mass in the virgin tissue, whereas the epithelial cells predominate in the lactating tissue (13), we examined the possibility that alterations in cell population were responsible for the results relating to the developmental changes. Fat cells were removed from mammary tissues of different developmental stages, and the resulting epithelial enriched cell fractions were used to prepare nuclear extracts for EMSA experiments. The results were similar to those obtained from the intact glands (Fig. 2C). Possible variations in the efficiency of nuclear isolation and extraction were examined by EMSA using consensus oligonucleotide sequence for known transcription factor, mammary gland factor (MGF). MGF activity in mice is strongly increased at the end of pregnancy and is maintained at high levels during lactation (38). As shown in FIG. 1. Functional analysis of the mouse GPT promoter in PMME cells. A, a schematic representation of the GPT promoter with the position of the putative regulatory elements: É, CAAT box; f, AP-2; E, Sp1; å, CTF/NF1; 3, 13-bp direct repeat; q, imperfect GC box; ࡗ, core sequence for CAAT box. The multiple transcription start sites are shown by the bent arrows, with the major site (nucleotide Ϫ202) being indicated by the dark arrow. B, structure and Luc activity of the GPT/Luc deletion constructs. Names of the various GPT promoter/Luc constructs used and the deletion start and end points relative to the adenosine (ϩ1) of the translation initiation codon ATG are listed to the left. BV, pGL2-Basic luciferase vector. Twenty g of the indicated GPT/Luc construct were co-transfected with 5 g of pSV-␤-galactosidase plasmid into PMME cells. Twenty hours after transfection, cells were cultured for 5 days in the presence of either I (control) or IHP. The Luc activity was determined and normalized to ␤-galactosidase activity measured in the same sample. The value show the percent Luc activity relative to IHP-treated B7 construct (100%). Thin lines represent the S.E.; n ϭ 3-6. Fig. 2D, MGF activity as determined by binding to DNA was very low in the virgin and post-lactating tissues and reached its highest level in the lactating tissue. These results suggest that developmental changes in the level of DNA binding activity are not due to the difference in the efficiency of nuclear extraction in our experiments.
The effect of lactogenic hormones on the FP1 binding activity in mammary epithelial cells was examined in mouse mammary explant culture system. Earlier studies have demonstrated that the combination of insulin (I), hydrocortisone (H), and prolactin (P) stimulate GPT activity (11) and GPT gene expression in cultured mammary gland (12). The results in Fig. 3 show that the hormone treatments, IH, IE (E, epidermal growth factor), and IP did not cause significant changes in the binding activity over control (T 0 , lane 2). On the other hand, the combination of all three lactogenic hormones, IHP (lane 6), greatly reduced the binding activity.
To localize the protein binding sites, DNase I footprinting analysis was performed using nuclear extracts from the virgin tissue and FP1 probe. The nuclear extracts protected two regions (Fig. 4). The sequence of region I is 5Ј-AGGAAGAAACT-TGAC-3Ј, between bp Ϫ1029 and Ϫ1044, and the sequence of region II is 5Ј-AGGAAACCCACAAGC-3Ј, between bp Ϫ1010 and Ϫ1025 (Fig. 5A). Their positions overlap with the 13-bp direct repeat sequence. Comparison of these two sequences revealed the presence of two direct repeats AGGAA and GAAAC. The AGGAA direct repeat is separated by 14 bp, and the GAAAC direct repeat is separated by 11 bp.
As mentioned above, direct repeats of the hexamer AGGTCA can serve as response elements for vitamin D, thyroid hormone, or retinoic acid (20 -22). The specificity of the response appears to reside in the spacing between the hexamers. To test whether the direct repeats found in mouse GPT gene are responsible for the observed DNA-protein interaction, a 46-bp double-stranded oligonucleotide probe WT (bp Ϫ1050 to Ϫ1004) containing these sites along with oligonucleotides containing the mutated sites in the direct repeats were used in EMSA competition experiments (Fig. 5B). As shown in Fig. 5C, the WT probe formed a complex with the nuclear extracts from the virgin tissue and the formation of this complex was efficiently competed by excess unlabeled WT oligonucleotide (lanes 2 and 3). Oligonucleotides MM1 and MM2, which had the mutations at the both sites of the direct repeats AGGAA and GAAAC, respectively, failed to compete, suggesting that these sites were critical to binding (Fig. 5C, lanes 4 and 5). Since the right half-sites of these two pentamers were overlapped, it was difficult to differentiate their effect on the binding. Therefore, mutations were introduced only at the half-sites. Both oligonucleotides M1 and M1-2, which were mutated at the left and right half-site of the direct repeat AGGAA, respectively, showed some competition (Fig. 5C, lanes 6 and 7), whereas MML mutated at both left half-sites of AGGAA and GAAAC had weaker effect on the binding (lane 8). In contrast, oligonucleotide M2 with mutations at the left half-site of the direct repeat GAAAC competed the binding substantially (lane 9). These results suggested that both direct repeats are involved in DNA-protein binding and AGGAA plays a more important role. To evaluate the effect of spacing between the direct repeats on the binding, oligonucleotides with deletions and insertions between the direct repeats were used in EMSA experiment (Fig.  5B). SM2 and SM4 with two and four base deletions, and SMI3 with three base insertions between the two footprints, com-FIG. 2. Developmental profiles of nuclear factors that specifically bind to sequence ؊1161 to ؊983 of mGPT promoter. Doublestranded FP1 fragment was end-labeled by the Klenow reaction. A, increasing amounts (2, 4, and 6 g) of protein from virgin nuclear extracts were used (lanes 4 -6); lane 1, no extract (NE); lanes 2 and 3, 50-fold excess of unlabeled oligonucleotide FP2 (nonspecific, bp Ϫ776 to Ϫ623) and WT (specific, detailed in the legend to Fig. 5), respectively, were used as competitors. B, 2 g of protein from nuclear extracts of virgin (V), pregnant (P), lactating (L), and post-lactating (PL) mouse mammary glands were used (lanes 1-4). C, 2 g of protein from nuclear extracts of adipocyte-depleted fraction isolated from virgin (V), pregnant (P), and lactating (L) mammary glands were used (lanes 1-3) in electrophoretic mobility shift assays. D, end-labeled oligonucleotide for MGF (nucleotides Ϫ104 to Ϫ84, 5Ј CGTAGACTTCTTGGAATTGAA3Ј) and 2 g of protein from nuclear extracts of virgin (V), pregnant (P), lactating (L), and post-lactating (PL) mouse mammary glands were used (lanes 2-4). Lane 1, no extract (NE). peted the binding completely (Lanes [11][12][13], indicating that spacing is not important. However, MS4, deleted four bases within the footprint I, failed to compete efficiently (lane 10), whereas MS2-5 deletion with two bases within the footprint I and three bases between the two footprints competed the binding completely (lane 14). These data suggest that sequences within the footprints are important for DNA-protein contact.
The results from the EMSA experiments indicated the ability of the GPT promoter sequences between bp Ϫ1050 and Ϫ1004 to form a specific complex with binding proteins in the nuclear extracts of virgin tissue. In order to assess the functional importance of these DNA-protein interaction, 12 GPT/Luc expression constructs, containing the mutations used in the competition EMSA experiments, were prepared. These constructs were transfected into PMME cells in parallel with the unmutated GPT/Luc construct WT (Fig. 6). The WT construct produced the same pattern of transcriptional activity as observed in previous experiments (compare Fig. 6 to Fig. 1, B4). The MM1 and MM2 constructs had dramatically increased induction activity. M1, M2, and MMl constructs also showed much higher induction activity than the WT, indicating that mutations at half-site of AGGAA pentamer alone could partially relieve the repressive effect of the negative regulatory region. M2 construct showed only half of the induction activity compared to the M1, consistent with the competition EMSA results that AGGAA is more important than GAAAC. All spacing mutant constructs had no effect except for MS4, and MS4 caused an increase in induction activity. These results suggest that the DNA-protein complex between the factor in the virgin extracts and the sequence containing the direct repeats suppressed the expression of the GPT gene.

DISCUSSION
The present work is part of a study focused on the identification of cis-acting elements involved in the developmental and hormonal regulation of GPT gene in the mouse mammary gland. GPT has a ubiquitous function in eukaryotic cells, and GPT gene has several characteristics of a housekeeping promoter. Such promoters often lack both TATA and CCAAT boxes, are GC-rich, and contain one or several Sp1 binding sites. The housekeeping genes are expressed in all cells, and their expression is not subject to control by environmental factors such as hormones (23). The mGPT gene is unusual in that its promoter contains characteristics of housekeeping genes but its activity is modulated during the developmental state of the mammary gland and by hormones.
Removal of consensus sequences for several potential transcription factors within the promoter region of mGPT gene indicated that these sequences, e.g. a CCAAT box (Ϫ1487), two Sp1 binding sites (Ϫ754 and Ϫ25), and an AP-2 binding site (Ϫ577), were not important for expression of the GPT gene in transient transfections of PMME cell assays. On the other hand, a region from bp Ϫ1057 to Ϫ968 was identified to be critical for developmental and hormonal regulation of GPT FIG. 4. DNase I footprinting analysis of sequence ؊1161 to ؊983 of mGPT promoter. Double-stranded FP1 was labeled on the coding strand and incubated either with increasing amounts (5, 10, and 20 g) of protein from virgin nuclear extracts or without the extracts. After the binding reaction, the mixture was subjected to DNase I digestion and electrophoresed on 8% denaturing polyacrylamide gel. Letters indicate introduced mutations, * indicates deleted nucleotides, and ࡗ indicates inserted nucleotides. C, 30-fold excess competitor oligonucleotides were incubated with 2 g of protein from nuclear extracts of virgin mouse mammary gland prior to addition of labeled WT oligonucleotide. Competitors used are indicated above appropriate lanes. gene in mouse mammary gland. This region contained a 13-bp direct repeat sequence, which was a potential site for binding a trans-acting factor. Deletion of this region resulted in the enhancement of hormonally inducible promoter activity. EMSA experiments using a DNA fragment targeted to this region revealed that a nuclear protein(s), a trans-acting repressor, could specifically bind to this region. The binding activity changed during the development of the mammary gland such that it was high in virgin and post-lactating glands, intermediate in pregnant gland, and low in lactating gland. The binding activity was also hormonally regulated since mammary explants cultured with lactogenic hormones, IHP, showed greatly reduced DNA-protein complex formation. Our data are consistent with the results from previous studies that GPT enzyme activity is developmentally and hormonally regulated in the mammary gland. It is high in the lactating gland (11,12), as well as in the mammary explants cultured in the presence of IHP.
DNase I footprinting analysis identified that the presence of two pentamer direct repeats, AGGAA and GAAAC, within the protected sequences. EMSA competition experiments, using a 46-bp oligonucleotide containing this direct repeat motifs and oligonucleotides with mutations at the direct repeats, confirmed that the direct repeat AGGAA is largely responsible for the observed DNA-protein interaction and the direct repeat GAAAC is also involved. A dramatic increase in hormonally inducible transcription was observed after introducing the direct repeat mutations into Luc reporter constructs. This is consistent with the view that failure of the repressor, identified by in vitro EMSA, to bind in vivo resulted in the enhancement of transcription activity. The finding that two protected regions separated by 4-bp were identified by footprint analysis and that only one major band was detected by EMSA suggested that the repressor probably binds to this negative regulatory element as a dimer. The nature of this DNA repressor remains to be studied.
Negative regulatory mechanisms appear to be quite common in eukaryotic cells and negative elements are often associated with cell-specific gene expression (32). These promoters contain basal positive elements that bind ubiquitous transcription factors, which function in different cell types. Therefore, negative control mechanisms are required for their repression in these different cell types (33). On the other hand, ubiquitous expression of a promoter can be achieved by activation of different elements by nuclear proteins in different cell types (31). Al-though GPT gene is expressed in all cells, its expression in the mammary gland is regulated in a tissue-specific manner by lactogenic hormones during the development. Our data suggest that high levels of the repressor in the virgin tissue is responsible for the repression of promoter activity. Low levels of the repressor and the presence of possible mammary gland-specific positive factors in the lactating tissue confer the promoter full responsiveness to the hormonal induction. Therefore, developmental and hormonal regulation of GPT gene in mammary gland most likely involves a combination of both negative and positive elements that are modulated by changing in hormone milieu. Such promoter structure is correlated with the regulation of the GPT gene and the functions of the GPT enzyme, which participates not only in the constitutive pathway of N-linked glycosylation but also in the tissue-specific, hormonally regulated pathway to synthesize large amounts of N-linked glycoproteins such as ␣-lactalbumin in the lactating mammary gland. The ␤1,4-galactosyltransferase (␤4-GT) gene is another example of a widely expressed gene whose expression involves negative regulatory element (29,30). Both GPT and ␤4-GT have housekeeping function to participate in biosynthesis of N-linked glycoproteins. It was proposed that differential initiation from the two promoters of the ␤4-GT gene, a distal housekeeping promoter for the 4.1-kb transcript and a proximal mammary cell-specific promoter for the 3.9-kb transcript, is responsible for regulation of ␤4-GT enzyme levels. Binding of repressor(s) to the negative regulatory element could account for the repression of transcription from the 3.9-kb start site in all somatic tissues except the lactating mammary gland. Tissue-specific transcriptional regulation of genes that encode other enzymes involved in N-linked glycoprotein biosynthesis, e.g. ␣1,3-galactosyltransferase (34), ␣2,6-sialyltransferase (35,36), and N-acetylglucosaminyltransferase I (37), also have been reported.
In GC-rich, Sp1 binding site-containing promoters, the ubiquitous promoter activity is due to the ubiquitous expression of the Sp1 protein (24). Binding of Sp1 has been shown to be critical for transcription initiation (25,26). In the case of mGPT promoter, removal of the two perfect GC boxes did not affect the basal and inducible promoter activities, suggesting that these elements are inactive. However, the involvement of Sp1 could not be ruled out, because there are three imperfect copies of GC boxes with a 1-bp mismatch between bp Ϫ264 and Ϫ92. In addition, the transfection experiments indicated that this region may contain positive regulatory elements for both basal FIG. 6. Transient expression of GPT/Luc mutated constructs in PMME cells. Mutated oligonucleotides used in competition EMSA were used as primers to generate mutated GPT/Luc constructs, which contain sequence Ϫ1161 to Ϫ5 of the GPT gene. The transfection experiments were preformed as described in Fig. 1. The Luc activity was determined and normalized to ␤-galactosidase activity measured in the same sample. The values are presented as the percent Luc activity relative to IHPtreated MM1 construct (100%). Thin lines represent S.E.; n ϭ 3. and inducible promoter activity. Furthermore, two copies of CCAAT box core sequence, CAAT, and two sequences similar to the half-palindromic CTF/NF1 element are also present in this region. A unique half-site CTF/NF1-binding protein has been identified in lactating mammary glands (27,28) and implicated in the regulation of murine ␤4-GT gene (29,30) and several milk protein genes (27,28). CCAAT box-binding proteins have been shown to be the major factors involved in basal promoter activity of the cytosolic aspartate aminotransferase gene. This gene, like mGPT gene, is also ubiquitously expressed but regulated by hormones in a tissue-specific manner (31). The roles of these elements in regulation of mGPT gene are currently under investigation.
The nuclear protein(s) that binds to the negative regulatory element is present in all tissues examined, although the position of the shifted band differed a little among different tissues (data not shown). Relative higher levels were found in thymus, spleen, ovary, and virgin mammary gland. Among these tissues, thymus, ovary, and virgin mammary gland are targets of prolactin action. Although the mechanism of prolactin signal transduction is not completely understood, MGF has been shown to be a Stat (signal transducer and activator of transcription) protein and confer the prolactin response in ␤-casein gene (38 -40). MGF binding site was found in the promoters of the genes of the most abundantly expressed milk proteins, namely caseins (41). However, MGF binding site is not present in the promoter of mGPT gene. It is possible that lactogenic hormones exert their influence partially through reduction of the binding of the repressor to the negative regulatory element in the mGPT promoter. Further purification and characterization of the repressor will be necessary to understand its mode of action and to determine whether it plays a more general role in transcriptional regulation of TATA-less housekeeping genes.