Yin-yang 1 and glucocorticoid receptor participate in the Stat5-mediated growth hormone response of the serine protease inhibitor 2.1 gene.

A growth hormone-inducible nuclear factor complex (GHINF), affinity-purified using the growth hormone response element (GHRE) from the promoter of rat serine protease inhibitor 2.1, was found to contain Stat5a and -5b, as well as additional components. The ubiquitous transcription factor yin-yang 1 (YY1) is present in GHINF. An antibody to YY1 inhibited the formation of the GHINF.GHRE complex in an electrophoretic mobility shift assay. Furthermore, Stat5 was co-immunoprecipitated from rat hepatic nuclear extracts with antibodies to YY1. An examination of the GHRE shows that, in addition to two gamma-activated sites, it contains a putative YY1 binding site between the two gamma-activated sites, overlapping them both. Mutation of this putative YY1 site results in a decrease of GHINF.GHRE complex formation in an electrophoretic mobility shift assay and a corresponding decrease in growth hormone (GH) response in functional assays. The glucocorticoid receptor was also present in GHINF, and Stat5 co-immunoprecipitates with glucocorticoid receptor in hepatic nuclear extracts from rats treated with GH. GH activation of serine protease inhibitor 2.1 requires the unique sequence of the GHRE encompassing the recognition sites of several transcription factors, and the interaction of these factors enhances the assembly of the transcription complex.

The liver is a major target organ for growth hormone (GH) 1 action. Recent advances have led to the elucidation of much of the GH signaling pathway. GH binds to the growth hormone receptor (GHR) on the cell surface, triggering GHR dimerization, activation of Jak2 from the JAK family of tyrosine kinases, and the phosphorylation of specific tyrosine residues on the cytoplasmic domains of the GHR (1)(2)(3)(4)(5). In turn, latent cytoplasmic transcription factors, known as signal transducers and activators of transcription (STATs) are recruited to the receptor complex and become tyrosine-phosphorylated (6 -11). Subsequent phosphorylation of serine/threonine residues, dimerization, and translocation of these STATs to the nucleus is followed by binding to cognate DNA response elements of target genes, ultimately resulting in their transcriptional induction (12)(13)(14).
The serine protease inhibitor (Spi) 2.1 gene has served as a useful model for investigating the mechanism of GH action in rat liver. We delineated a GH response element (GHRE) in the promoter of Spi 2.1 that includes two ␥-activated sites (GAS), both of which are necessary for a GH response in functional assays (15). In the hypophysectomized rat, as well as in several cell lines, GH activates Stat1, -3, and/or -5 (7)(8)(9)(10)(15)(16)(17)(18)(19). Stat5, first purified as the prolactin (PRL)-responsive mammary gland factor from sheep mammary glands (20), has since been purified and cloned from mouse liver and IM-9 lymphocytes (19,21). We purified a GH-inducible nuclear factor complex, GHINF, from rat liver, in which the major DNA binding component was Stat5 (15). We also found that the GHINF complex contained additional unknown proteins. We now report the identities of the additional components and evidence for their role in the GH induction process.

EXPERIMENTAL PROCEDURES
Treatment of Rats-All animals were handled in accordance with experimental protocols approved by the University of Minnesota Institutional Committee on the Care and Use of Animals. Male rats (Harlan Sprague Dawley) were hypophysectomized at 100 -125 grams by the supplier (Harlan Sprague Dawley, Indianapolis, IN) and observed for 3 weeks to confirm growth failure. Administration of GH (Lilly) to these rats at a dosage of 150 g/100 g, body weight, led to the accumulation of activated Stat5 in liver nuclei (6). Liver nuclei from normal male rats contain variable amounts of activated Stat5, a consequence of the pulsatile nature of the GH level in the male rat (12,22). We found that treating a normal rat with GH 1 h before sacrifice consistently led to accumulation of activated Stat5 in liver nuclei. Its phosphorylation status, as determined by supershift assays with antibodies to phosphoserine or phosphotyrosine with the GHRE, is identical to that observed when a hypophysectomized rat is treated with GH. We therefore used both GH-treated normal and hypophysectomized rats for these studies.
Electrophoretic Mobility Shift Assays (EMSAs)-Preparation of crude rat hepatic nuclear extracts, affinity purification of GHINF, and EMSA have all been reported previously (15). Briefly, in an EMSA reaction, 5 g of hepatic nuclear extracts or ϳ2 ng of GHINF were incubated with 20 fmol of the radiolabeled probe of interest in a buffer containing 20 mM HEPES (pH 7.6), 10% glycerol, 2 mM MgCl 2 , 5 mM CaCl 2 , 0.1 mg/ml bovine serum albumin, 4% Ficoll 400, 1 mM spermidine, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, and 1 g of poly(dI⅐dC). The final KCl concentration was adjusted to 50 mM. After 30 min of incubation at 30°C, the binding mixture was loaded on to a nondenaturing, 3.2% glycerol, 5% polyacrylamide gel in a Bio-Rad Protean II system in 0.5ϫ TBE (45 mM Tris, 45 mM boric acid, and 1 mM EDTA, pH 8.3) and electrophoresed at 150 V for 2.5 h. The gel was dried and exposed to film. Quantitation of the shifted complexes on the film was determined using the Kodak Digital Science ID Image Analysis Software. In supershift assays, GHINF or hepatic nuclear extracts were incubated with the antibody under investigation for 30 min at 4°C prior to the addition of the radiolabeled probe.
Oligonucleotides-The sequences of the consensus YY1 probe, wild type GHRE, and mutant GHREs are shown in Table I. To prepare double-stranded oligonucleotides, primer P1, P2, or P3 was annealed to the appropriate oligonucleotide as stated in Table I. The annealed duplexes were then extended with the Klenow fragment of Escherichia coli DNA polymerase I and radiolabeled dCTP. The products were subsequently purified through Bio-Spin 6 columns (Bio-Rad).
Antibodies-The following antibodies were used in either "supershift" or immunoblot assays. Antibody to YY1 (C-20) (Santa Cruz Biotechnology, Inc., Santa Cruz, CA; catalogue no. sc-281 or sc-281X) is a polyclonal antibody generated against amino acids 395-414 mapping at the carboxyl terminus of human YY1. Antibody to GR (P-20) (catalogue no. sc-1002) is a polyclonal antibody generated against amino acids 750 -769 mapping at the carboxyl terminus of human GR. Antibody to Stat5a (L-20) (catalogue no. sc-1081X) is a polyclonal antibody generated against amino acids 774 -793 mapping at the carboxyl terminus of mouse Stat5a. This antibody is specific for Stat5a and is nonreactive with Stat5b. Stat5b antibody (C-17) (catalog no. sc-835 or sc-835X) is a polyclonal antibody generated against amino acids 711-727 at the carboxyl terminus of mouse Stat5b. This antibody reacts with both Stat5a and Stat5b. Antibody to phosphoserine (catalogue no. P3430) and antibody to phosphothreonine (catalogue no. P3555) were from Sigma. These two antibodies do not cross-react with each other. Antibody to serine-phosphorylated Stat5a/b (catalogue no. 06-867, Upstate Biotechnology, Inc., Lake Placid, NY) was raised against a synthetic peptide (DQAPpSPAVC, where pS represents phosphoserine) corresponding to amino acids 722-730 of human Stat5a or 727-735 of human Stat5b. This antibody cross-reacts with both mouse and rat Stat5a/b.
Immunoprecipitations and Immunoblots-Immunoprecipitations were performed with liver nuclear extracts from either hypophysectomized rats or normal rats treated with GH. Aliquots containing 100 g were incubated with 2 g of the antibody of interest in radioimmune precipitation buffer (50 mM Tris-HCl, pH 7.5, 1% Nonidet P-40, 0.1% SDS, 150 mM NaCl, 2 mM EDTA, and 0.1% sodium deoxycholate) containing, in addition, 0.2 mM phenylmethylsulfonyl fluoride, 1 g/ml aprotinin, and 10 mM NaF. The mixtures were mixed for 1 h on end over end rotation at 4°C (19). Protein A-agarose (120 g) was then added to each mixture, and end to end mixing was continued for 16 h at 4°C. The mixtures were centrifuged at 700 ϫ g for 5 min at 4°C. The resulting pellets were washed four times with radioimmune precipitation buffer, each time collecting agarose by centrifuging for 5 min at 4°C. SDS electrophoresis sample buffer (40 l) was added to each final agarose pellet, and the mixture was boiled for 90 s and centrifuged. Five l of the resultant supernatant were then loaded onto a 7.5% SDS-polyacrylamide gel. Immunoblotting was performed as reported previously (15).
Plasmid Constructions and Functional Assays-The construction of Spi 2.1 (Ϫ275/ϩ85) into the HindIII/PstI sites of the parent CAT reporter plasmid pCAT(An), designated here as Spi-A-CAT, and that of mutations in place of the wild type GHRE sequence in Spi-A-CAT were described previously (15). Based on our observation that the insertion of the Xba site at Ϫ116/Ϫ111 does not disrupt the GH response, we constructed long PCR primers extending from this site to 15 base pairs (bp) 5Ј of the putative YY1 recognition site. Within these primers, we inserted short mutated sequences into or upstream of the putative YY1 recognition site. Primers incorporating mutations R, P, K, L, and Q, listed in Table I, were used to generate PCR products extending from Ϫ116 to Ϫ275. These were subsequently ligated, together with PCR products extending from Ϫ116 to ϩ85, into pCAT(An). All mutations were confirmed by TaqFs dye terminator cycle sequencing (Applied Biosystems, Foster City, CA) performed by the Microchemical Facilities at the University of Minnesota.
Functional assays in primary rat hepatocytes were carried out as described previously (15). Briefly, primary hepatocytes were isolated from male Harlan Sprague Dawley rats (180 -240 g) using the collagenase perfusion method and plated on 35-mm culture dishes at a concentration of 1.2 ϫ 10 6 cells/plate. After a 4-h attachment period, transfection was performed using Lipofectin reagent (Life Technologies, Inc.) in modified Williams E medium with 27.5 mM glucose for 12-14 h. 500 g/ml of Matrigel (Life Technologies) was then added to the medium, and the hepatocytes were cultured for 48 h in the presence or absence of 0.5 g/ml GH. At the end of 48 h, the cells were harvested and lysed in 1ϫ Reporter lysis buffer (Promega Corp., Madison, WI) for chloram- YY1 and GR Participate in Stat5-mediated GH Action phenicol acetyltransferase (CAT) assays. Results of the assays were expressed as percentage conversion of chloramphenicol to its acetylated forms as determined by phosphor screen autoradiography (Molecular Dynamics Corp., Sunnyvale, CA). Each experiment was repeated three times with freshly isolated hepatocytes. At least two different plasmid preparations of each construct were tested in these experiments.

GHINF Contains Serine Phosphorylated Stat5a and
Stat5b-The GHINF complex was purified from liver nuclear extracts of GH-treated rats using affinity chromatography on a GHRE sequence (Ϫ147/Ϫ103) from the Spi 2.1 promoter. The GHINF complex contained two bands with apparent molecular masses of ϳ93 and ϳ70 kDa. The ϳ93-kDa band was previously shown to contain Stat5 that was tyrosine-phosphorylated in response to GH treatment. To determine what forms of Stat5 were present in this band and to further examine their phosphorylation states, we used antibodies to Stat5a, Stat5b, phosphoserine, and phosphothreonine in GHRE binding assays. Fig. 1a demonstrates that the GHINF⅐GHRE complex can be partially supershifted with an antibody specific to Stat5a. An antibody recognizing both Stat5a and Stat5b (Stat5b antibody) supershifted the GHINF⅐GHRE complex. In addition to containing phosphorylated tyrosine residues, both Stat5a and Stat5b isoforms are serine-phosphorylated, since a phosphoserine antibody supershifted the GHINF⅐GHRE complex (Fig.  1b). Similarly, an antibody specific for serine-phosphorylated Stat5a/b supershifted the entire GHINF⅐GHRE complex (data not shown). In contrast, these isoforms of Stat5 are not threonine-phosphorylated, since a phosphothreonine antibody did not alter the GHINF⅐GHRE complex (Fig. 1c).
GHINF Contains YY1-Examination of the GHRE sequence reveals a potential YY1 binding site, ATCCATGTT, located between the two GAS and overlapping them both (Table I). A comparison of this sequence with the consensus sequence for YY1 binding, (C/g/a)(G/t)(C/t/a)CATN(T/a)(T/g/c), with the uppercase letters denoting preferred bases (23), indicates that this is a lower affinity YY1 binding site. Previously, we showed that UV-cross-linking of GHINF led to the appearance of two bands, migrating to ϳ93 and ϳ70 kDa respectively, on SDSpolyacrylamide gel electrophoresis (15). We had shown that the 93-kDa band contained Stat5. We now show that the ϳ70-kDa species is YY1. YY1, with a molecular mass of 48 kDa and a highly charged N terminus, migrates in an anomalous fashion to ϳ70-kDa on SDS-polyacrylamide gel electrophoresis (24). Utilizing an antibody to YY1, we tested for the presence of YY1 in GHINF by means of EMSA and immunoblotting. Fig. 2a shows that YY1 antibody diminishes the formation of the GHINF⅐GHRE complex in EMSA. The extent of this inhibition is similar to that observed with a consensus YY1 binding site, indicating that YY1 is present in GHINF. This is further confirmed on an immunoblot of purified GHINF (Fig. 2b), which contains a band migrating at ϳ70 kDa that is immunoreactive with YY1 antibody.
To examine whether the presence of YY1 in GHINF was merely a contaminant of the purification process and to assess whether YY1 might associate with Stat5 in a state-specific manner, antibodies to YY1 were used for immunoprecipitation of hepatic nuclear extracts from hypophysectomized rats and normal rats treated with GH. Fig. 3a shows an immunoblot of these immunoprecipitates probed with Stat5b antibody. The YY1 immunoprecipitate from a normal rat treated with GH contains Stat5 (lane 2), since this band co-migrates with Stat5 in hepatic extracts and GHINF (lanes 3 and 4). Hepatic nuclear extracts from hypophysectomized rats contain little Stat5, and YY1 antibody immunoprecipitates from these extracts contained little Stat5 (lane 1). A faster moving and rather diffuse band in lanes 1 and 2 is an artifact of YY1 immunoprecipita-tions, since the same band was also seen on an identical blot probed with YY1 antibody (Fig. 3b, lanes 1 and 2). This blot also showed that YY1 immunoprecipitates from both hypophysectomized and normal rats treated with GH contain YY1 (Fig. 3b,  lanes 1 and 2). These results indicate that YY1 is associated with Stat5 during the response to GH.
Effects of Mutations of GHRE on GHINF Binding-GHINF binds synergistically to the two GAS in GHRE (15). One hypothesis, given the above data, is that YY1 may play a role in promoting this synergistic binding. To test this hypothesis, we studied the effects of altering the sequence or spacing between these two GAS on GHINF binding. EMSA was carried out with hepatic nuclear extracts from a hypophysectomized rat treated with GH (Fig. 4). There was no formation of the GHINF⅐GHRE complex when the spacing between the two GAS was changed FIG. 2. GHINF contains YY1. a, effect of the addition of YY1 antibody (␣YY1) on GHINF binding. Supershift assays were carried out with ϳ2 ng of affinity-purified GHINF and radiolabled GHRE (lanes 1 and 2). Control supershift assays were performed with normal rat hepatic nuclear extracts and a radiolabeled YY1 consensus sequence probe (lanes 4 and 5). Lane 3 contained free YY1 probe. b, an immunoblot of normal rat liver crude extract (Cr, 25 g) and affinity-purified GHINF (ϳ12 ng), probed with an antibody to YY1 (␣YY1). Molecular mass size markers are indicated on the right.
FIG. 3. YY1 associates with Stat5. a, an immunoblot of YY1 antibody immunoprecipitates (Ipptn YY1) of hepatic nuclear extracts probed with antibody to Stat5b (␣Stat5b). Lanes 1 and 2 contain immunoprecipitates from extracts from a hypophysectomized rat (H) and a normal rat treated with GH (NϩGH), respectively. Lanes 3 and 4 contain extracts from a normal rat treated with GH that have not been immunoprecipitated with YY1 antibody and GHINF, respectively. Molecular mass markers are indicated on the left. The arrow on the right denotes the location of Stat5 with an apparent molecular mass of ϳ93 kDa. b, a similar immunoblot probed with YY1 antibody (␣YY1). Molecular mass markers are indicated on the left. The arrow on the right denotes the migration of YY1 on SDS-polyacrylamide gel electrophoresis. On both blots, there was a nonspecific band that migrated at ϳ88 kDa.

YY1 and GR Participate in Stat5-mediated GH Action
from the native spacing of 6 bp to 2 bp (M2) or 4 bp (M4) (lanes  1 and 2). Within the wild type spacing of 6 bp, mutating only 2 bp critical to YY1 binding (CC 3 AA) led to a significant decrease in GHINF binding (R, lane 4), when compared with the wild type GHRE (lane 3). Placement of the putative YY1 binding site on the opposite strand (P) led to little change in GHINF binding (lane 5) when the differences in specific activities of the probes were taken into consideration (Table II). However, increasing the spacing between the two GAS to 11 or 13 bp, thus changing the phase of the GAS (S, K, and L), led to noticeable decreases in the intensities of GHINF binding. Mutation of the TAA of the 5Ј GAS to TCA (Q) led to a further decrease in GHINF binding (lane 9). The decreases in GHINF binding to these probes were particularly noteworthy, since the specific activities of these probes were at least 2-fold higher than that of GHRE (Table II). Therefore, either mutating the putative YY1 binding site or changing the phase of the GAS led to significant decreases in GHINF binding.
Effects of Mutations of GHRE on GH Induction of Spi 2.1 Promoter Activity-Spi 2.1 promoter fragments (Ϫ275/ϩ85), incorporating the sequences of mutations K, P, R, L, and Q in place of wild type sequences, were cloned into a CAT reporter plasmid and tested for their responses to GH over a range from 5 to 500 ng/ml in primary hepatocytes (Fig. 5). The activity of Spi-P-CAT, with the YY1 recognition site on the opposite strand compared with wild type, was somewhat lower than that observed for wild type Spi-A-CAT in the entire range of GH concentrations tested. Their response curves, however, were very similar. Mutation of only 2 bp within the putative YY1 site, Spi-R-CAT, led to a dramatic decrease of the GH response. For the entire range of GH concentrations in which Spi-A-CAT had a positive response (12.5-500 ng/ml), the response of Spi-R-CAT was only 10 -17% of that of Spi-A-CAT. Increasing the spacing between the GAS but retaining the YY1 binding site, Spi-K-CAT, led to a diminished GH response and a shift of the response curve considerably to the right, since there was no response at GH concentrations of 12.5 or 25 ng/ml. For GH concentrations of 50 -500 ng/ml, the response of Spi-K-CAT was only 15-25% of that of Spi-A-CAT. A combination of these two alterations, mutation of 2 bp within the putative YY1 binding site and an increase of the spacing between the two GAS, Spi-L-CAT, led to an almost total abrogation of the GH response over the entire range of concentrations tested. Spi-Q-CAT, with an intact YY1 site but only a 4 out of 6 match in its 5Ј GAS sequence to that of the 3Ј GAS, was likewise unresponsive to GH. Thus, the synergistic binding of two dimers of Stat5, the length of spacing between these two Stat5 binding sites, and the sequence of that spacing are all critical factors in the ability of GHRE to respond to GH.
GHINF Contains GR-Stat5 can associate with GR when both are overexpressed in COS7 cells (25). In addition, a role for GR in the Stat5-mediated PRL induction of the ␤-casein gene has been reported (26). To determine if GHINF purified from livers of rats after GH treatment also contains GR, we used an antibody to GR to probe an immunoblot of purified GHINF (Fig. 6a). A ϳ93-kDa band that is immunoreactive with GR antibody was detected in purified GHINF. Of note, GR and Stat5 have similar molecular masses and migrate to similar positions on SDS-polyacrylamide gel electrophoresis (26). To assess whether GR associates with Stat5 during a GH response, we carried out GR antibody co-immunoprecipitations of hepatic nuclear extracts from hypophysectomized rats and normal rats treated with GH. Fig. 6b shows an immunoblot of these immunoprecipitates probed with an antibody to Stat5b. In the immunoprecipitate from the hepatic nuclear extract from a normal rat treated with GH, there is a band migrating to ϳ93 kDa that is immunoreactive with Stat5b antibody (lane 2). This band is absent in liver nuclear extracts from hypophysectomized rats (lane 1). Similar results were obtained when Stat5 antibody immunoprecipitates were probed with GR antibody (not shown). Thus, GR is associated with Stat5 during a GH response. DISCUSSION Previously, we have shown that the affinity-purified GHINF complex contains Stat5 and that it is tyrosine-phosphorylated (15). We now show that both isoforms of Stat5, Stat5a and Stat5b, are present in GHINF and that they are, in addition to being tyrosine-phosphorylated, serine-phosphorylated. Tyrosine and serine residues of Stat5 are sequentially phosphorylated during a GH response (14). While tyrosine phosphorylation is necessary for Stat5 binding to GHRE, the role of serine phosphorylation is unknown. Stat5a mRNA is found predominantly in mammary tissues during lactation, whereas Stat5b message is found at similar levels in both liver and mammary tissues (21).
Stat5 binds synergistically to two GAS in GHRE from the Spi 2.1 promoter, and the migration of the Stat5⅐GHRE complex is more retarded than that of a Stat5 dimer bound to a single GAS (15), consistent with a higher order complex composed of two interacting STAT dimers (27). Similar cooperative binding to neighboring sites has been reported for Stat1, -4, and -6 (28). Many STAT proteins bind weakly, or not at all, to individual native sites. Interactions between dimeric molecules bound on The ratio for mutation P complex intensity to that of wild type GHRE was calculated from a lighter exposure. EMSAs carried out with these probes (between three and ten times for each probe) gave very similar results.  Table I. adjacent sites serve to stabilize the overall binding (29) and are dependent on the distance between them. An introduction of 5 or 7 additional bp between them in GHRE, causing them to be out of phase with each other (S, K, or L), led to a significant reduction in the formation of the Stat5 tetramer complex, when compared with that of the wild type GHRE. This suggests that a different conformation of Stat5 factors is bound in these cases. The higher threshold and 75% diminution of the GH response of Spi-K-CAT in functional assays indicate that this conformation is not as effective as that of the wild type.
In addition to spacing, Stat5 binding to GHRE is dependent on the sequence of that spacing. This 6-bp sequence, TCCATG, constitutes the core of the potential YY1 9-bp recognition sequence of ATCCATGTT. This sequence in either the top or the bottom strand (GHRE and P) led to strong Stat5 binding in EMSA and similar GH responses in functional assays. Mutating 2 bp of this sequence, critical for YY1 binding (CC 3 AA in mutation R), led to a significant decrease in Stat5 binding in EMSA and a large decrease in GH response in functional assays. The response of Spi-R-CAT to GH was only 10 -17% of that of Spi-A-CAT over the entire range of GH concentrations tested. An identical mutation in a lengthened spacing between the two GAS (Spi-L-CAT) led to an almost total abrogation of the GH response. These results indicate that YY1 or another factor that recognizes this sequence promotes the synergistic binding of Stat5 to the two GAS in GHRE, an event that is necessary for Spi 2.1 to respond to GH. The observation that both Spi-R-CAT and Spi-L-CAT, containing only a mutation of 2 bp critical to YY1 binding (CC 3 AA), responded weakly to GH strongly suggests that YY1 is the factor involved. Taken together, these results suggest that YY1 enhances the GH response of Spi 2.1 by promoting the assembly of an active conformation of the Stat5 tetramer⅐GHRE complex.
YY1 is a ubiquitously expressed, zinc finger-containing factor with homology to the GLI-Krü ppel family of proteins (30). It participates in different processes associated with gene transcription, including both activation and repression. It binds to, among others, TAFII55-TATA-binding protein and cAMP-response element-binding protein-binding protein (31). The sequences flanking either side of the YY1 core "CAT" binding site can be quite variable, presenting opportunities for its site to overlap with other DNA binding sites. For example, YY1 and the serum response factor binding sites overlap in the c-fos promoter (32). Competition between these factors has been proposed as one of the mechanisms by which YY1 represses transcription (30,33). In the case of ␤-casein, where the YY1 site overlaps a GAS in the prolactin response element, PRL acts by overcoming the repression of YY1 (34,35). In other cases, YY1 mediates transcription by bending the DNA to promote contacts between other proteins that interact within the promoter and enhancer domains (33,36,37). We now show that YY1, in sharing overlapping sites with Stat5 at each end of its binding site in GHRE, acts by enhancing transcription of Spi 2.1 during a GH response. We suggest that it does so by facilitating or stabilizing the formation of an active complex of two Stat5 dimers bound to adjacent GAS.
Given the significance of the YY1 recognition site in the formation of the Stat5 tetramer complex in EMSA and in Spi 2.1 functional assays, it is puzzling that YY1 binding to GHRE is not demonstrable in EMSA. We have previously performed cold competitions with unlabeled YY1 on GHINF binding. The addition of a 25-fold molar excess of unlabeled YY1 did not affect GHINF binding; nor did it affect the ability of the antibody of YY1 to ablate GHINF binding. In hepatic nuclear extracts, the addition of a 25-fold molar excess of unlabeled FIG. 5. Comparison of responses of GHRE mutation CAT constructs to GH in primary hepatocytes. Spi-A-CAT contains the wild type GHRE. Spi-K/ L/P/Q/R-CATs contain GHRE mutations K/L/P/Q/R, respectively, listed in Table I. The GH concentrations tested ranged from 5 to 500 ng/ml. CAT activities were calculated as percentage conversion of chloramphenicol to its acetylated forms. The values shown are representative of four separate experiments. In each experiment, the ratio of each mutation construct's activity to that of Spi-A-CAT was within Ϯ10% of that shown in this figure. The relative relationships of activities shown in this figure were seen for all experiments.
FIG. 6. GR associates with Stat5. a, an immunoblot of affinitypurified GHINF probed with GR antibody (␣GR). The arrow on the left denotes migration of GR to ϳ93 kDa. Molecular mass markers are indicated on the right. b, an immunoblot of GR antibody immunoprecipitates of liver extracts from a hypox rat (H) or a normal rat treated with GH (NϩGH) (lanes 1 and 2, respectively), probed with antibody to Stat5b (␣Stat5). Lanes 3 and 4 contain affinity-purified GHINF and extracts from a normal rat treated with GH (NϩGH), respectively. Molecular mass markers are indicated on the right. GHRE did not affect YY1 binding. However, not all binding of transcription factors to their cognate sequences on target genes is demonstrable in EMSA. For example, in acute-phase reactions, the presence of Stat3 is usually confirmed by its binding to a high affinity, mutated sis-inducible element but not as easily to the wild type sequences of lower affinity that occur in the native promoters of acute-phase reactants (18). Similarly, we show that it is possible to enhance YY1 binding to GHRE in EMSA by placing an additional G or C in the immediate 5Јflanking region of the 9-bp YY1 recognition site (Q and K).
The possibility exists that the involvement of YY1 with Stat5 occurs primarily via a protein/protein interaction and that its actual binding to DNA within the GHRE plays only a secondary role. Immunoprecipitation studies of rat hepatic nuclear extracts with YY1 antibody show that YY1 is associated with Stat5 during GH stimulation. However, functional studies of mutation of the putative YY1 binding site (Spi-R-CAT) indicate that the YY1 recognition sequence is necessary to achieve a level of GH response equivalent to that of the wild type. Therefore, YY1 binding to DNA most likely occurs during the formation of the GHINF⅐GHRE complex. YY1 exhibits very rapid on/off rates, less than 10 s in vitro (23), and several examples of it serving as an initiator binding protein during the formation of the basal transcription complex have been reported (38 -40). Hence, YY1 binding to the GHRE may occur as a transient interaction during the process of Stat5 binding.
The role of YY1 in the GH activation of Spi 2.1 in the liver thus appears to be different from its role in the PRL activation of ␤-casein in the mammary gland. In the case of ␤-casein, YY1 represses the expression of ␤-casein in the absence of PRL, and mutation of its recognition site led to hormone-independent induction. Thus, PRL acts by relieving the repression of YY1 (34,35,41). For Spi 2.1 expression, however, mutation of the YY1 site (Spi-R-CAT) did not result in hormone-independent induction. Instead, it led to a significant attenuation of the GH response in the entire range of GH concentration tested. The differences in the placement of the YY1 recognition sites in the prolactin response element and GHRE may explain the variations in their functional significance. In the prolactin response element, the YY1 recognition site is 5Ј of both GAS sequences, and the 3Ј GAS alone can support the PRL response of ␤-casein promoter in functional assays (25). In the GHRE, however, the YY1 site occurs between the two GAS and overlaps them both. Moreover, both GAS are necessary to support the Spi 2.1 GH response in functional assays. Thus, the unique placement of the YY1 site in GHRE probably leads to YY1-induced cooperation that facilitates interactions among Stat5 dimers and other components of the transcription machinery, thereby enhancing transcriptional activation.
GR has been shown to associate with Stat3 during interleukin-6-mediated cellular responses (42) and with Stat5 during PRL activation of ␤-casein expression (25). In our previous studies in the intact rat, we found that Spi 2.1 gene expression is greatly reduced by hypophysectomy and can be restored to only 40% of its normal level by the administration of GH alone. Full restoration requires the synergistic action of GH, thyroxine, corticosterone, and dihydrotestosterone (43). Furthermore, in primary hepatocyte cultures, the GH response of Spi 2.1 is highly dependent on the presence of dexamethasone in the culture medium (44). Therefore, it is not surprising that we now find that GR is associated with Stat5 in the liver, both by its presence in purified GHINF and by its co-immunoprecipitation with Stat5 from hepatic nuclear extracts of normal rats treated with GH. There are several half-glucocorticoid response elements in the Spi 2.1 promoter (45). One of them occurs within GHRE and is situated between the two GAS, overlap-ping the 3Ј GAS (Table I). It is not known whether GR actually binds to this site, since the addition of GR antibody in EMSA studies of GHINF binding did not lead to the formation of a supershifted complex, nor did it prevent the formation of the GHINF⅐GHRE complex (data not shown). Recently, it has been shown that during PRL stimulation, GR acts as a ligand-dependent coactivator that is independent of the specific DNAbinding domain of GR (26). Together with Stat5, GR forms a molecular complex that cooperates in the induction of transcription of the ␤-casein gene. Specific DNA binding function of the GR is apparently not required for this cooperation, since GR variants that lack active ligand binding domains are able to cooperate with Stat5 in the induction of a ␤-casein-luciferase construct in COS cells (25). Moreover, GR antibody was able to immunoprecipitate a Stat5⅐GR complex from nuclear extracts that had been incubated with a radiolabeled wild type Stat5-DNA binding site. No complex was precipitated when a radiolabeled mutated Stat5 binding site was used (46). Our results therefore are consistent with the concept that the involvement of GR with Stat5 is probably one of protein/protein interaction that does not depend upon its binding to a glucocorticoid response element.
In addition to Stat5a/b, we have now demonstrated that the GHINF complex also contains YY1 and GR. There is increasing evidence that several activators may act together to facilitate transactivation mediated by a single DNA-binding transcription factor (47). The evidence presented here suggests that, during a GH response, Stat5 becomes both tyrosine-and serine-phosphorylated and probably associates with GR. Following translocation to the nucleus, it also associates with YY1. By virtue of the unique placement of its binding site, we suggest that YY1 promotes the assembly of an active conformation of the Stat5 tetramer⅐GHRE complex. This conformation, interacting with other components of the transcriptional machinery, leads to efficient transcription of Spi 2.1.