Transcriptional Inhibition by Stat5

Prolactin (PRL) induces transcriptional activation of not only growth-related genes such as interferon regulatory factor-1 (IRF-1) but also differentiation-specific genes such as β-casein through a signaling cascade consisting of Janus kinases and Stat (signal transducer and activator of transcription) factors. To understand better the role of Stats in PRL signaling, we cloned rat Stat5b from a PRL-responsive T cell line Nb2. A Stat5b-specific peptide antibody was generated. In PRL receptor reconstituted COS cells cotransfected with Stat5b or Stat5a, both Stat5 proteins become tyrosine phosphorylated and bind to the IRF-1 GAS (interferon-γ activation sequence) element in a PRL-inducible manner. Unexpectedly, both Stat5b and Stat5a inhibit PRL induction of the IRF-1 promoter, but they mediate PRL stimulation of the β-casein promoter. Stat5-mediated inhibition was observed only at the native IRF-1 promoter and not at the isolated IRF-1 GAS element linked to a heterologous thymidine kinase promoter. Mutational analyses showed that the DNA binding activity of Stat5b is not required, but the carboxyl-terminal transactivation domain is essential for Stat5b to inhibit PRL induction of the IRF-1 promoter. These results suggest that Stat5b mediates inhibition via protein-protein interactions. In contrast, both DNA binding and transactivation domains of Stat5b are required to mediate PRL induction of the β-casein promoter. Furthermore, a carboxyl-terminal truncated dominant negative Stat5b can reverse Stat5b inhibition at the IRF-1 promoter. These studies suggest that Stat proteins can act as not only positive but also negative regulators of gene transcription. Further, Stat5 can modulate gene expression without binding to DNA but via protein-protein interactions.

vates JAK tyrosine kinases, which are either prebound or recruited to the receptor upon tyrosine phosphorylation. Activated JAK tyrosine kinases then phosphorylate the receptor itself and Stats that are recruited to the receptor through the interaction between the Src homology (SH2) domain on the Stats and phosphorylated tyrosine residues on the receptor. Phosphorylated Stats form homo-or heteromeric complexes, translocate into the nucleus, bind to specific DNA elements, and participate in target gene transcription. Four members of the JAK tyrosine kinase family (4), including JAK1, JAK2, JAK3, and TYK2, and seven members of the Stat family, Stat1␣/␤, Stat2, Stat3␣/␤, Stat4, Stat5a, Stat5b, and Stat6 (1-3, 5), have been described. These signaling molecules are utilized in various combinations to mediate signal transduction by the members of the cytokine receptor superfamily. Stat5a or mammary gland factor, originally cloned from the sheep mammary gland, was shown to be induced by prolactin (PRL) to activate the transcription of the milk protein ␤-casein gene, a marker for mammary cell differentiated functions (6,7).
PRL is a pituitary peptide hormone that stimulates cellular proliferation and differentiation of many different cell types (8). A role of PRL as a T cell cytokine has been elucidated (9 -11). PRL activates T cells by binding to the PRL receptor (PRL-R) which is a member of the cytokine/hematopoietin receptor superfamily (8,12). The rat Nb2 T lymphoma cell line (13) has been used extensively as a model system to study the molecular mechanisms underlying PRL signaling in T cells (14). Previous studies from our laboratory have found that PRL induces the expression of a number of early growth-related genes including interferon regulatory factor 1 (IRF-1) as part of an early activation program leading to mitogenesis (15)(16)(17). Our studies have identified the interferon-␥ activation sequence (GAS) in the IRF-1 promoter as a PRL-responsive enhancer element (18 -20). More recently, we have found that Stat1 and Stat5a become tyrosine phosphorylated and bind to the IRF-1 GAS element in a time-dependent manner in PRL-stimulated Nb2 T cells (19,21).
To understand better PRL-R-mediated mitogenic signaling in Nb2 T cells through the JAK/Stat/IRF-1 pathway, we cloned the PRL-inducible Stat5b (EMBL accession number X97541) by screening an Nb2 T cell cDNA library and generated an anti-Stat5b peptide-specific antibody (Ab). Stat5b and Stat5a are found to be equally expressed in Nb2 T cells. In PRL-R and Stat5b or Stat5a cotransfected COS cells, both Stat5b and Stat5a become tyrosine phosphorylated rapidly upon PRL stimulation and bind to the IRF-1 GAS element. Unexpectedly, this PRL-inducible interaction of Stat5b and Stat5a led to an inhibition of PRL-inducible IRF-1 promoter activity while they stimulated PRL-inducible ␤-casein promoter activity. The inhibitory effects of Stat5b on the IRF-1 promoter appear to be independent of DNA binding but involve protein-protein interactions through its COOH-terminal transactivation domain. In contrast, both the DNA binding and transactivation domains of Stat5b are required for mediating PRL induction of the milk protein ␤-casein promoter. These studies show that PRL-inducible Stat proteins can mediate stimulation or inhibition of gene transcription depending on the target promoters.
Cloning of Rat Stat5b cDNA-A 237-bp reverse transcription-polymerase chain reaction DNA corresponding to the sheep Stat5a SH2 domain (6) was used to screen a cDNA library prepared from the PRL-dependent rat Nb2-11C cell line (16). Three Stat5a and seven overlapping Stat5b cDNA clones were obtained. After rescreening the Nb2 cDNA library using a 518-bp polymerase chain reaction DNA corresponding to the 5Ј most available Stat5b sequence (clone 3), a full-length Stat5b cDNA clone (clone 4) was obtained and sequenced using the Sequenase version 2.0 DNA sequencing kit (U. S. Biochemical Corp.). The 3.6-kb Stat5b cDNA was released with EcoRI and subcloned into the pRc/CMV and pcDNA3.1(Ϫ) mammalian expression vectors (Invitrogen, San Diego). The 3.7-kb rat Stat5a cDNA was released with EcoRI from the pBK/CMV vector (23) and recloned into the pRc/CMV vector to have the same vector background.
Generation of Anti-Stat5b Peptide Antibodies-A 10-amino acid peptide (DSQWIPHAQS, amino acids 777-786) corresponding to the unique COOH-terminal sequence of Stat5b (containing an added Lys at the COOH terminus) was synthesized and coupled to a multiple antigenic peptide resin (Peptide Synthesis Core Laboratory, Baylor College of Medicine). Two rabbits were immunized with 1 mg of this peptide plus complete adjuvant and boosted with the same dose plus incomplete adjuvant six times. Antisera were collected and stored at Ϫ20°C.
In Vitro Transcription and Translation (TNT) of Stat5b-Stat5b and Stat5a were transcribed and translated in vitro using the T3 TNTcoupled reticulocyte lysate systems (Promega, Madison, WI) with 40 Ci of [ 35 S]methionine as described previously (25). 5 l of each reaction mixture was resolved by 7.5% SDS-polyacrylamide gel electrophoresis and analyzed by autoradiography to examine the size of Stat5b and Stat5a or used for immunoprecipitation with anti-Stat5b and anti-Stat5a Abs.
Transient Transfection and CAT Assays-2 ϫ 10 5 /well of COS or 293 fibroblast cells was seeded in six-well tissue culture plates overnight in DMEM containing 10% fetal bovine serum. Transient transfections were performed using LipofectAMINE (Life Technologies, Inc.) as described by the manufacturer. Briefly, cells were rinsed twice with serum-free DMEM and incubated with a LipofectAMINE-DNA mixture in 1 ml of serum-free DMEM. After 5 h, 2 ml of DMEM with 1% horse serum (ICN-Flow Laboratories, Mclean, VA) was then added. For Western blot assays and EMSA, each well contained 3 g of Nb2 PRL-R (26), 3 g of either Stat5b, Stat5a, or vector control, and 15 l of Lipo-fectAMINE. At 24 h after transfection, cells were stimulated with or without 100 ng/ml ovine PRL (NIDDK-oPRL-20) for 30 min, harvested by scraping, washed once with ice-cold phosphate-buffered saline containing 1 mM sodium orthovanadate, and either lysed for Western blot analysis or used for whole cell extraction for EMSA. For CAT assays, 2 g of Nb2 PRL-R, between 0.05 and 2 g of Stat (Stat5b, Stat5a, Stat5bEE, Stat5bVVVI, Stat5b⌬40C, or vector control), 0.2 g of IRF-1-CAT (1.7-kb, 205-bp, 136-bp, and 112-bp IRF-1-CAT 19), 0.2 g of three copies of the IRF-1 GAS in thymidine kinase (TK)-CAT (3C GAS-TK-CAT), or 1 g of 2.3-kb ␤-casein-CAT (28) were used. At 24 h after transfection, cells were stimulated with or without 100 ng/ml PRL for another 24 h. Cell extracts were prepared by four cycles of freezingthawing in 0.25 M Tris, pH 7.4, normalized by protein concentration and assayed for CAT enzyme activity as described previously (18), or prepared by using a Reporter lysis buffer (Promega) and assayed as described by the manufacturer. In our experiments, extracts normalized to equal amounts of protein or equal expression of CMV-␤-galactosidase gave similar results (18). Briefly, between 2 and 5 g of cell extracts, 50 l of 3 mg/ml acetylcoenzyme A (Sigma) or 5 l of 5 mg/ml n-butyrylcoenzyme A (Promega), and 3 l of [ 14 C]chloramphenicol (50 mCi/mmol, NEN Life Science Products) were used and assayed for 4 h at 37°C. CAT activity was analyzed by thin layer chromatography or liquid scintillation counting. CAT conversion measured by thin layer chromatography was quantitated further by using the Betascope 603 blot analyzer (Betagen, Mountain View, CA). Fold induction by PRL was calculated as described (18) and plotted by using the CA-Cricket Graph III (Computer Associates International, Inc., Islandia, NY).

Molecular Cloning of Rat Stat5b-Our previous studies have
shown that PRL stimulation of Nb2 T cells leads to a biphasic transcriptional induction of the growth-related IRF-1 gene (15,16,18,19). Studies from our laboratory (19,21) and others (29 -32) have indicated that Stat1 and Stat5 are involved in PRL signaling to the IRF-1 gene and other target genes in different tissues. The initial goal of this study was to determine how Stat5 may be involved in the PRL signaling pathway to the IRF-1 gene in T cells. However, at that time Stat5 had not been cloned from rodents. Accordingly, we screened a rat Nb2 T cell cDNA library using a 237-bp reverse transcription-polymerase chain reaction DNA obtained from the rat mammary gland, which corresponds to the SH2 domain of the sheep Stat5a, as a probe (23). Several rat Stat5a and Stat5b cDNA clones were obtained. One of these Stat5b clones contains 110-bp 5Ј-noncoding, 2,358-bp coding, and about 1,200-bp 3Ј-noncoding sequences (EMBL accession number X97541; data not shown). Sequence analysis of this full-length Stat5b cDNA shows 90% homology at the nucleotide level and 96% homology at the amino acid level between rat Stat5b and rat Stat5a (23) and 99.2% homology at the amino acid level between rat and mouse Stat5b (33). However, four amino acid (Ala 296 , Val 328 , Ala 489 , and Pro 690 ) differences are observed between our Stat5b sequence and that cloned recently from the rat liver (Leu 296 , Ser 328 , Arg 489 , Arg 690 ) (24). These differences are most likely due to sequencing error (Ala 296 (GCT) to Leu (CTG); Val 328 (GTC) to Ser (TCG); Ala 489 (GCA) to Arg (CGA); Pro 690 (CCC) to Arg (CGC)), as our rat sequences at these positions are identical to those of mouse Stat5b (33). As also found in the mouse and human, rat Stat5b 1) has a PCEPA five-amino acid insertion (amino acids 687-691) at the end of the highly conserved SH2 domain and immediately upstream of the critical tyrosine residue (Y699); 2) has eight distinct amino acids at the carboxyl terminus; and 3) is seven amino acids shorter than rat Stat5a.
Analysis of Stat5b Protein-To assess the protein size of rat Stat5b, Stat5b was transcribed and translated using T3 RNA polymerase in the TNT-coupled rabbit reticulocyte lysate system as described previously (25). A single predominant [ 35 S]methionine-labeled protein of approximately 95 kDa was observed in the Stat5b TNT lysate (Fig. 1A, lane 1) but not in the pBluescript SKϪ vector control lysate (Fig. 1A, lane 3). Although Stat5b is only seven amino acids shorter than Stat5a, it resolves as a slightly smaller band than the in vitro translated Stat5a protein (ϳ96 kDa) (Fig. 1A, lane 2). To analyze Stat5b protein expression further, a polyclonal rabbit anti-Stat5b peptide Ab was raised against a unique 10-amino acid peptide corresponding to the carboxyl terminus of Stat5b. The specificity of anti-Stat5b Ab was determined by its ability to immunoprecipitate TNT-Stat5b protein specifically but not in vitro translated Stat5a protein (data not shown). This anti-Stat5b Ab specifically immunoblotted recombinant Stat5b protein produced in control and PRL-stimulated Stat5b-transfected COS cells (Fig. 1B, lanes 1 and 2) but not Stat5atransfected (Fig. 1B, lanes 3 and 4) or control vectortransfected (Fig. 1B, lanes 5 and 6) COS cells. The anti-Stat5b Ab also immunoprecipitated native Stat5b protein from Nb2 T cells (Fig. 1C, lane 1). The previously reported affinity-purified anti-Stat5a1 peptide Ab did not recognize TNT-Stat5b protein (data not shown) but did immunoprecipitate Stat5a protein from Nb2 T cells (Fig. 1C, lane 2). Both Stat5b and Stat5a proteins are expressed equally in Nb2 T cells, as shown by immunoprecipitation with a commercially available anti-Stat5 mAb directly against the commonly shared SH3 and SH2 domains of the Stat5 proteins (Fig. 1C, lane 3).
Activation of Stat5b by PRL in COS Cells-To study the transactivation potential of Stat5b, we first examined its activation by PRL after transient transfection into COS cells. COS cells were chosen because these cells have been used successfully for analysis of PRL-inducible ␤-casein and ␤-lactoglobulin promoter activation (7,(33)(34)(35)(36). In this and subsequent transfections, PRL-R reconstituted COS cells were generated by transient transfection with the Nb2 PRL-R along with Stat5b, Stat5a, or empty vector control. To determine whether Stat5b can be phosphorylated by PRL stimulation, Stat5b phosphorylation was examined by anti-phosphotyrosine mAb immunoblotting after stimulation with PRL for 30 min. Stat5b as well as Stat5a became highly phosphorylated only after PRL stimulation ( Fig. 2A, lanes 4 and 6). Both Stat5b (Fig. 2B, lanes 3 and  4) and Stat5a (Fig. 2B, lanes 5 and 6) proteins were expressed abundantly after transfection into COS cells. In control cells transfected with empty vector alone, no Stat5b or Stat5a phosphorylation ( Fig. 2A, lanes 1 and 2) or steady-state proteins (Fig. 2B, lanes 1 and 2) were detected, suggesting that COS cells express very low, if any, Stat5b or Stat5a proteins.
Stat5b Binds to the IRF-1 GAS Element-The IRF-1 gene is one of the immediate-early genes induced by PRL stimulation in T cells (16). Studies from our laboratory have shown that PRL stimulates both Stat1␣ and Stat5a to bind to the IRF-1 GAS element in Nb2 T cells (19,21) and Stat1 to bind to the IRF-1 GAS in PRL-R reconstituted COS cells (20). We next examined whether Stat5b or Stat5a can bind to the IRF-1 GAS element in a PRL-dependent manner. Whole cell extracts prepared from transiently transfected COS cells were employed in EMSA, using a 27-bp oligonucleotide containing one copy of the IRF-1 GAS element as a probe (19,21). In vector-transfected control cells, PRL stimulated the rapid formation of an inducible complex (Fig. 3, A and B, lanes 1 and 2), which was more evident in extracts prepared from a parallel experiment (Fig. 3,  A and B, lanes 14). This PRL-inducible complex contains an endogenous Stat1-like factor as it can be supershifted by an anti-Stat1 mAb (Fig. 3, A and B, lanes 13) but not by preimmune sera (Fig. 3, A and B, lanes 14). The band at the top of the gel is nonspecific as it was observed with preimmune sera (Fig.  3A, lanes 13 and 14). Signals from the endogenous Stat1 complex are faint in this and subsequent EMSAs (Fig. 3, A and B, ϩPRL lanes) because of the short exposure times of the autoradiograms. In both Stat5b (Fig. 3A, lanes 3 and 4) and Stat5a (Fig. 3B, lanes 3 and 4) transfected COS cells, PRL treatment induced the formation of a major complex that can be competed away by increasing concentrations of cold IRF-1 GAS DNA (Fig. 3, A and B, lanes 5 and 6) but not by mutant IRF-1 GAS (Fig. 3, A and B, lanes 7) or an unrelated DR1 DNA (Fig. 3, A  and B, lanes 8). This complex contains Stat5b or Stat5a as it can be supershifted in a dose-dependent manner by anti-Stat5b Abs (Fig. 3A, lanes 9 -11) or anti-Stat5a1 Abs (Fig. 3B, lanes  9 -11) but not by preimmune sera (Fig. 3, A and B, lanes 12). Although some faster migrating bands were also competed by excess cold IRF-1 or DR1 DNA, they are not PRL-inducible and are not further studied. These studies show that recombinant Stat5b and Stat5a can be induced by PRL stimulation to bind to the IRF-1 GAS element.
Stat5b and Stat5a Inhibit PRL-inducible IRF-1 Promoter in COS Cells-Having established that transfected Stat5b or Stat5a is expressed abundantly and can be tyrosine phosphorylated (Fig. 2) and bind to the IRF-1 GAS (Fig. 3) in a PRLinducible manner, we next determined whether Stat5b and Stat5a can functionally transactivate the IRF-1 promoter. The 1.7-kb IRF-1-CAT, containing the maximally responsive IRF-1 promoter and flanking DNA (18,19), was cotransfected into the reconstituted COS cell system. First, a dose response employing empty vector control or no exogenous vector (data not shown) showed that PRL stimulated a 3-fold induction of the 1.7-kb IRF-1 promoter in COS cells (Fig. 4B). Lower concentrations of Stat5b and Stat5a DNA did not alter PRL stimulation of the IRF-1 promoter (Fig. 4A). Unexpectedly, with increasing concentrations of Stat5b and Stat5a DNA, Stat5b and Stat5a not only did not augment PRL-inducible IRF-1 promoter activity, but instead they both inhibited PRL induction of IRF-1 promoter activity to the level observed in uninduced cells (Fig.  4A). The inhibitory effect (2-3-fold reduction) of Stat5b and Stat5a on the 1.7-kb IRF-1 promoter was reproducibly observed in a number of independent experiments (Fig. 4B).
To localize further a minimal region through which Stat5b and Stat5a may be exerting their inhibitory effects, a series of IRF-1 promoter deletion constructs (19) was used (Fig. 4B). The 205-bp IRF-1 promoter construct contains one copy of the GAS element which has been shown also to act as a PRL-responsive enhancer element (19). Both Stat5b and Stat5a inhibited the PRL-inducible expression of the 205-bp IRF-1 promoter (2-3fold reduction) in a manner similar to that observed for the 1.7-kb IRF-1 promoter. Further deletion of the IRF-1 promoter to 136 bp still produced a reduced but consistent PRL induction  (ϳ2-fold induction), which was also inhibited by both Stat5b and Stat5a. However, IRF-1 promoter deletion to 113 bp, which eliminated the GAS element, abolished PRL stimulation of the IRF-1 promoter. Without a PRL-inducible response, no inhibition of the IRF-1 promoter could be observed when cotransfected with either Stat5b or Stat5a. These studies confirm that the IRF-1 GAS element is required for PRL induction of the IRF-1 promoter and suggest that Stat5b and Stat5a inhibit PRL-inducible IRF-1 promoter activity possibly through the IRF-1 GAS element.
To examine whether the IRF-1 GAS element alone is sufficient in mediating Stat5 inhibition, three copies of the IRF-1 GAS element were linked to a minimal heterologous TK promoter (20) and tested for a response to Stat5 inhibition. A reproducible 10-fold PRL induction of the 3C GAS-TK-CAT was observed (Fig. 4C), presumably mediated by activated endogenous Stat1-like factors (20). The TK-CAT parent vector was not responsive to PRL stimulation (data not shown). However, addition of either Stat5b or Stat5a did not inhibit the PRL induction of this synthetic GAS-TK promoter reporter. These studies suggest that the IRF-1 GAS element, when isolated from the native IRF-1 promoter, can still mediate a PRL response, but it alone is not sufficient to mediate Stat5 inhibition of this PRL response. In other words, Stat5-mediated inhibition of the IRF-1 promoter is only observed when the GAS element is present in the context of the native IRF-1 promoter.
In contrast, under the same reconstituted COS cell system, both Stat5b and Stat5a reproducibly mediated PRL activation of the 2.3-kb ␤-casein promoter (Fig. 4D). The ␤-casein promoter was unresponsive to PRL stimulation in empty vectortransfected control cells but was stimulated 5-6-fold by PRL in the presence of transfected Stat5b or Stat5a, as also observed by others (7,33,37). These studies show clearly that Stat5b and Stat5a can inhibit PRL induction of the IRF-1 promoter while activating PRL induction of the ␤-casein promoter in transfected COS cells. The inhibitory effects of Stat5b and Stat5a on the IRF-1 promoter were also reproducible in a different 293T fibroblast cell system (Fig. 4E). These results show that Stat5 inhibition of PRL induction of the IRF-1 promoter is not a COS cell-specific event.
Stat5b-mediated Inhibition at the IRF-1 Promoter Does Not Require DNA Binding-Next, we asked whether Stat5-mediated inhibition of the IRF-1 promoter is caused by competition with endogenous Stat factors for binding to the IRF-1 promoter or by competition with another factor(s) that is/are necessary for PRL induction of the IRF-1 promoter. To distinguish between these two possibilities, specific residues within the DNA binding domain of Stat5b were mutated (Fig. 5A). Within the DNA binding domain of Stat factors, the VTEE and SLPVVVI residues are highly conserved and are critical for Stat-DNA interactions (38,39). The EE (amino acids 437 and 438) or the VVVI (amino acids 466 -469) of Stat5b were mutated to ala- nine (EE to AA, or VVVI to AAAA), and the Stat5bEE and Stat5bVVVI mutants were tested for their ability to bind DNA and to regulate promoter activity. Whole cell extracts were prepared following treatment with PRL and used in EMSA with the IRF-1 GAS DNA as a probe. PRL stimulated an inducible Stat5b complex (arrow) in wild-type Stat5b-transfected cells (Fig. 5B, lanes 3 and 4) but not in empty vectortransfected control cells (Fig. 5B, lanes 1 and 2). The Stat5bEE mutant showed increased basal binding to the IRF-1 GAS which was induced further by PRL treatment (Fig. 5B, lanes 5  and 6). However, the Stat5bVVVI mutant failed to bind to the IRF-1 GAS (Fig. 5B, lanes 7 and 8), although equal levels of wild-type and mutant Stat5b were expressed and tyrosine phosphorylated in response to PRL stimulation in the transfected cells (data not shown). These studies show that the VVVI but not EE residues in the DNA binding domain of Stat5b are critical for binding to the IRF-1 GAS.
To determine how these mutations affect Stat5b function, the Stat5bEE and Stat5bVVVI mutants were tested for their ability to regulate the 1.7 IRF-1 (Fig. 5C) or 2.3-kb ␤-casein (Fig. 5D) promoters. Both Stat5b mutants were as effective as wild-type Stat5b in inhibiting PRL induction of the IRF-1 promoter (Fig. 5C), regardless of whether they can (Stat5bEE) or cannot (Stat5bVVVI) bind DNA. These studies clearly show that the DNA binding domain of Stat5b is not required for Stat5b-mediated inhibition at the IRF-1 promoter. In contrast, the Stat5bVVVI mutant lost its ability to activate the ␤-casein promoter, whereas the Stat5bEE mutant, which still binds DNA, retained its ability to stimulate the ␤-casein promoter (Fig. 5D). These studies demonstrate that the DNA binding activity of Stat5b is required for mediating PRL induction of the ␤-casein promoter but is dispensable for Stat5b inhibition of the IRF-1 promoter.
The COOH-terminal Transactivation Domain of Stat5b Is Required for Inhibition of the IRF-1 Promoter-It has been shown that the COOH-terminal domain is critical for Stat factor functions (37,39,40). We next asked whether the COOH-terminal transactivation domain of Stat5b is involved in mediating its inhibitory effects at the IRF-1 promoter. Stat5b⌬40C, which lacks the last 40 amino acids, is a naturally occurring splice variant isolated from rat liver (24) (Fig. 6A). Stat5b⌬40C protein was expressed to levels comparable to those of wild-type Stat5b, but unlike the wild-type Stat5b, Stat5b⌬40C was tyrosine phosphorylated without PRL stimulation after transfection into COS cells (data not shown). Consistent with this observation, Stat5b⌬40C bound to the IRF-1 GAS element strongly even in the absence of PRL stimulation (Fig. 6B, lane 5), and this binding was enhanced further by PRL stimulation (Fig. 6B, lane 6) (more evident with a shorter exposure, data not shown). No PRL-inducible endogenous Stat1 binding was observed in this experiment (Fig. 6B) since a very short exposure time was used. These experiments show that COOH-terminal Stat5b deletion did not abolish but instead showed enhanced binding to the IRF-1 GAS element.
To examine whether Stat5b⌬40C can still inhibit the IRF-1 promoter, Stat5b⌬40C was cotransfected with the PRL-R into COS cells along with either 1.7-kb IRF-1-CAT (Fig. 6C) or 2.3-kb ␤-casein-CAT (Fig. 6D). Interestingly, Stat5b⌬40C lost its ability to inhibit the IRF-1 promoter compared with wildtype Stat5b (Fig. 6C). These results show that Stat5-mediated inhibition at the IRF-1 promoter requires its COOH-terminal transactivation domain and is consistent with the interpretation that the inhibitory effects of Stat5 involve protein-protein interactions via its COOH-terminal domain. On the other hand, Stat5b⌬40C was unable to stimulate the ␤-casein promoter compared with wild-type Stat5b (Fig. 6D) as also shown by others (37). These studies confirm that the COOH-terminal transactivation domain of Stat5b is critical for function, that is, either transcriptional repression of the growth-related IRF-1 promoter or transcriptional activation of the differentiationspecific ␤-casein promoter.
Stat5b⌬40C Suppresses Wild-type Stat5b Function-It is  1 and 2), wild-type Stat5b (lanes 3 and 4), Stat5bEE mutant (lanes 5  and 6), or Stat5bVVVI mutant (lanes 7  and 8). Whole cell extracts with or without PRL stimulation were used for EMSA as described in Fig. 3. Asterisk (*), weak signal for Stat1-like complex at the IRF-1 GAS. Panel C, 1.7-kb IRF-1-CAT. COS cells were cotransfected as described in Fig. 4B, except that 1 g of Stat5bEE or Stat5bVVVI mutant was used (n ϭ 3). known that Stat factors lacking the COOH-terminal transactivation domain can act as dominant negative mutants, presumably by forming inactive heterodimers with wild-type Stats (37,41,42). We next asked whether Stat5b⌬40C can block wild-type Stat5b function at the IRF-1 (Fig. 7A) or ␤-casein (Fig. 7B) promoter. Again, wild-type Stat5b inhibited PRL induction of the IRF-1 promoter as seen previously (Figs. 4 -6). Interestingly, increasing concentrations of Stat5b⌬40C reversed wild-type Stat5b-mediated inhibitory effects at the IRF-1 promoter in a dose-dependent manner. On the other hand, increasing concentrations of Stat5b⌬40C blocked wildtype Stat5b-mediated PRL induction of the ␤-casein promoter as also shown previously by others (37). These results demonstrate that the inhibition of PRL induction of the IRF-1 promoter is Stat5-dependent and that the COOH-terminal truncated dominant negative mutant Stat5b⌬40C can reverse Stat5b function at either the IRF-1 or the ␤-casein promoter. DISCUSSION We cloned the rat Stat5b cDNA from Nb2 T cells, generated anti-Stat5b-specific peptide Ab, and constructed various Stat5b expression vectors. Using these reagents, we report the following findings. 1) Both Stat5b and Stat5a are equally expressed in Nb2 T cells. 2) Both Stat5b and Stat5a are highly expressed and become tyrosine-phosphorylated in response to PRL stimulation in reconstituted COS cells. 3) Both Stat5b and Stat5a bind to the IRF-1 GAS in a PRL-inducible manner. 4) Unexpectedly, both Stat5b and Stat5a inhibit PRL-inducible growthrelated IRF-1 promoter activity, but they stimulate differentiation-specific ␤-casein promoter activity in COS cells. 5) Stat5 inhibits only the native IRF-1 promoter but not the isolated IRF-1 GAS element linked to the heterologous TK promoter. 6) The inhibitory effects of Stat5b on the IRF-1 promoter are independent of its DNA binding activity and are mediated through its COOH terminus. 7) Both DNA binding and carboxy transactivation domains of Stat5b are required for PRL induction of the ␤-casein promoter. 8) The COOH-terminal truncated Stat5b can reverse Stat5b-mediated transcriptional inhibition at the IRF-1 promoter.
Stat5a was originally cloned from sheep mammary gland as a positive mediator for PRL induction of ␤-casein gene transcription (6,7). Murine, rat, and human Stat5a and Stat5b (43) have been cloned and are found to be conserved evolutionarily. Both Stat5b and Stat5a are activated by tyrosine phosphorylation in response to many different cytokines (24, 44 -50), including PRL (Fig. 2) (30 -33, 51). Stat5 is also stimulated by other signals such as epidermal growth factor (52), plateletderived growth factor (53), angiotensin (54), and even extracellular matrix (55), and by oncoproteins such as v-abl (56), v-mpl (50), and Bcr-abl (57). Recently, a Stat-like factor was identified in Drosophila (58) and in Dictyostelium (59), both of which share the most homology with Stat5 and are involved in regulating cell fate in early development as well as cell proliferation. Thus, Stat5 appears to mediate diverse biological functions, ranging from cellular differentiation and proliferation to transformation.
Our studies show unexpectedly that Stat5b and Stat5a inhibit the PRL induction of the IRF-1 promoter, but they stimulate the ␤-casein promoter (Fig. 4). These results suggest that Stat proteins can act as not only positive regulators but also negative regulators of gene transcription depending on the target promoter. Stat5a has also been shown to mediate PRL induction of the ␣ 2 -macroglobulin promoter in differentiated rat granulosa cells (51). Although all three genes are PRLinducible, IRF-1 is stimulated by PRL during a mitogenic response in T cells, whereas ␤-casein and ␣ 2 -macroglobulin genes are stimulated by PRL as end points of differentiation in epithelial cells. It is noted that all of these promoters contain multiple responsive elements that are regulated by other tran-  1 and 2), Stat5b (lanes 3 and 4), or Stat5b⌬40C (lanes 5 and 6) as described in Fig. 3. Whole cell extracts with or without PRL stimulation were used for EMSA as described in Fig. 3. Panel C, 1.7-kb IRF-1-CAT. COS cells were cotransfected as described in Fig. 4B scription factors (18, 19, 51, 60 -62), which together determine the overall transcriptional activity of the promoters.
In PRL-R reconstituted COS cells, PRL stimulates IRF-1 promoter activity presumably through the recruitment of endogenous JAK and Stat factors that primarily signal to the critical GAS element (Fig. 4B) (20). Our studies show that Stat1 acts as a positive mediator of PRL signaling to the IRF-1 promoter in reconstituted COS cells (20). However, Stat5 and Stat1 are found in different PRL-inducible complexes, suggesting that they do not form heterodimers as also observed previously (19, 51, 60 -62). How then do Stat1 and Stat5 regulate the IRF-1 promoter in an opposite manner? One simple explanation is that Stat5b or Stat5a interferes with the action of endogenous Stat factors such as Stat1 by competing for binding to the GAS element. Alternatively, overexpressed Stat5 factors compete for a putative factor(s), such as a nearby DNA-binding protein or a non-DNA-binding coactivator protein, which is essential for activation of the IRF-1 promoter. Our data clearly favor the second mechanism because the Stat5bVVVI mutant, which cannot bind DNA (Fig. 5B), still inhibits PRL induction of the IRF-1 promoter (Fig. 5C). Thus, although Stat5 can bind to the IRF-1 GAS in vitro (Figs. 2, 5, and 6), Stat5-mediated inhibition of the IRF-1 promoter is independent of DNA binding in vivo. This observation suggests that the inhibitory effect of Stat5 is mediated via protein-protein interactions at the IRF-1 promoter.
Recent studies have shown that the COOH-terminal regions of Stat proteins are uniformly involved in transactivation of target promoters (37,39,63). Studies with naturally occurring COOH-terminal truncated splice variants as well as deletion constructs have demonstrated that COOH-terminal deletions generally act as dominant negative mutants to block wild-type Stat functions, presumably by engaging Stats in nonfunctional heterodimers or by forming nonfunctional homodimers that bind more persistently and strongly to DNA (37). Our data demonstrate that the COOH-terminal deleted dominant negative mutant Stat5b⌬40C can no longer inhibit the IRF-1 promoter (Fig. 6C). Instead, dominant negative Stat5b efficiently antagonizes wild-type Stat5b inhibition of the IRF-1 promoter (Fig. 7A), presumably by forming inactive heterocomplexes that then release inhibition of the IRF-1 promoter. These studies suggest that the COOH terminus is required for Stat5b to function as a transcriptional repressor at the IRF-1 promoter, perhaps through an interaction with nearby DNA-binding and/or non-DNA-binding coactivator proteins.
Recent studies have shown that Stat factors can interact with inducible as well as basal transcription factors (40,64,65) to regulate promoter activity. Stat5 can interact with the glucocorticoid receptor, and this complex promotes PRL activation of the ␤-casein promoter but inhibits the glucocorticoid re-sponse at the mouse mammary tumor virus long terminal repeat promoter (66). Stat1 can interact with the coactivators CBP and p300 and thereby prevent AP-1/ets factors from activating the macrophage scavenger receptor promoter (67). These combined studies illustrate a novel way in which Stat factors can modulate promoter activity, that is, by forming complexes with other DNA-binding factors thereby preventing their interactions with target elements or by competing for non-DNA-binding coactivators. These studies further strengthen our observation that Stat factors can function not only as transcriptional activators but also as transcriptional inhibitors depending on the target promoter.
In addition to the GAS element, other DNA elements also contribute to the full transcriptional response of the IRF-1 promoter to PRL stimulation. An Sp1 and a Yi site (68) are found upstream of the GAS element in the native IRF-1 promoter. Deletion analyses (Fig. 4B) as well as mutations 2 of these sites reduced the overall PRL-inducible IRF-1 promoter activity. Interestingly, the inhibition by Stat5 was only observed at the native IRF-1 promoter but not at the isolated IRF-1 GAS element when it is linked to a heterologous TK promoter (Fig. 4C). Other studies have shown that multimerization of a regulatory element can alter its specificity, inducibility, and even generate an opposite transcriptional response (69). In this regard, it is possible that the multimerized GAS elements cannot recruit IRF-1 promoter-specific factors or coactivators and thereby cannot mimic the Stat5 inhibitory response observed at the native IRF-1 promoter. Additionally, overexpressing Stat3 did not inhibit but further enhanced PRL stimulation of the IRF-1 promoter in transfected COS cells (data not shown). These observations support the notion that Stat5, but not other overexpressed Stat factors, exerts a specific inhibitory effect at the IRF-1 promoter.
Our previous studies showed that PRL induces biphasic transcription of the IRF-1 gene first during G 1 and again during the G 1 /S transition in Nb2 T cells (15,18,19). Our recent studies identified Stat1 as the major component and Stat5a as a minor component of the G 1 IRF-1 GAS complex in Nb2 T cells (19 -21). In light of the current studies, it is possible that Nb2   7. Dominant negative Stat5b-⌬40C antagonizes the function of wild-type Stat5b. Panel A, 1.7-kb IRF-1-CAT. COS cells were transiently cotransfected with the Nb2 PRL-R, 1.7-kb IRF-1-CAT, 1 g of wild-type Stat5b, and 0, 0.5, 1, 2, or 4 g of Stat5b⌬40C as described in Fig. 4B (n ϭ 3). Panel B, 2.3-kb ␤-casein-CAT. COS cells were cotransfected as described in panel A, except that 2.3-kb ␤-casein-CAT was used (n ϭ 3). The fold induction Ϯ S.E. by PRL stimulation for each CAT construct in the presence of the various Stat vectors is shown. n, number of independent experiments. tein tyrosine kinases (70 -73) and protein tyrosine phosphatases (74 -76), participate in the biphasic signaling to the IRF-1 promoter in Nb2 T cells also remain to be elucidated. Our studies show that the initial signaling components employed by the PRL-R to mediate a mitogenic versus differentiative response may be similar, but at least some of the specificity of the final biological response will be determined by the presence of other DNA-binding factors, coactivators, and promoter-specific DNA elements at the different target promoters.