The Pit-1β Domain Dictates Active Repression and Alteration of Histone Acetylation of the Proximal Prolactin Promoter*

A critical problem in current molecular biology is to gain a detailed understanding of the molecular mechanisms by which related transcription factor isoforms with identical DNA sequence specificity mediate distinct transcription responses. Pit-1 and Pit-1β constitute such a pair of transcription factor isoforms. Pit-1 enhances the Ras signaling pathway to the prolactin promoter, and Pit-1β represses basal prolactin promoter activity as well as Ras signaling to the prolactin promoter in pituitary cells. We have previously demonstrated that the β-domain amino acid sequence dictates the transcriptional properties of Pit-1β. Here, we show that five hydrophobic β-domain residues are required for Pit-1 isoform-specific repression of Ras signaling, and we demonstrate that sodium butyrate and trichostatin A, pharmacological inhibitors of histone deacetylation, as well as viral Ski protein, a dominant-negative inhibitor of recruitment of N-CoR/mSin3 histone deacetylase complexes, specifically reverse β isoform-specific repression of Ras signaling. Moreover, we directly demonstrate, with a chromatin immunoprecipitation assay, that the Pit-1β isoform alters the histone acetylation state of the proximal prolactin promoter. This differential analysis of Pit-1/Pit-1β isoform function provides significant insights into the structural determinants that govern how different transcription factors with identical DNA sequence specificity can display opposite effects on target gene activity.

Most transcription factors are members of extended families defined by conserved structural motifs, typically in the DNAbinding domain, yet differing in other domains, especially the transactivation domain (TAD). 1 A number of transcription factors are expressed as a set of proteins derived from a single gene via alternative promoter usage or splicing events that result in virtually identical transcription factor isoforms that, nonetheless, can mediate distinct responses (e.g. PR-A versus PR-B; TR␤1 versus TR␤2; Oct 2.1 versus Oct 2.5; Ets-1 versus Ets-1⌬VII; and Pit-1 versus Pit-1␤) (1)(2)(3)(4)(5)(6). The molecular mechanisms, by which related transcription factor isoforms with identical DNA sequence specificity mediate distinct transcription responses, remain an area of active investigation.
Pit-1 is a pituitary-specific POU homeodomain transcription factor that governs both anterior pituitary cell identity and hormone gene expression (reviewed in Ref. 7). Pit-1 occurs in vertebrates, including humans, as two principal splice isoforms (Fig. 1). The ␤ isoform arises from an alternative splice-acceptor sequence at the end of the first intron resulting in a 26amino acid insertion, the ␤-domain, at position 48 in the TAD, between the first and second exons, TAD1 and TAD2 ( Fig. 1) (5,6). The ␤ isoform differs from Pit-1 only in the TAD and displays identical DNA sequence specificity with respect to the prolactin (PRL) promoter (5) but has dramatically different transcriptional properties than Pit-1, and these differences are dictated by the unique amino acid sequence of the ␤-domain TAD insertion (8). The ␤-domain insertion causes Pit-1␤ to act as a pituitary-specific repressor of both of basal transcription of the rat (r) PRL and of Ras signaling to the rPRL promoter gene (reviewed in Ref. 8). Additionally, the ␤-domain blocks functional interaction with Ets-1 (9) and functional interaction with the thyroid hormone and retinoic acid receptors (10) in nonpituitary cells. Nevertheless, this same 26-amino acid insertion endows Pit-1␤ with even greater potency with regard to mediation of protein kinase A signaling to the rPRL promoter (8,11).
In this article, we focus on identifying key ␤-domain residues that are responsible for the ␤ isoform-specific repression of Ras signaling to the rPRL promoter, as well as identifying a mechanism for this repression. We have utilized an epitope-scanning approach to replace sequentially 6 amino acid blocks of the ␤-domain in order to identify a limited subset of functionally important residues. Replacement of each of these residues with alanine identified five hydrophobic residues that are required for the ␤-domain to act as a transcriptional repressor of Ras signaling to the rPRL promoter. Moreover, we demonstrate that the ␤-domain does not simply disrupt TAD structure but functions as an active repression domain, which modifies the acetylation state of the proximal PRL promoter in a manner dependent upon an N-CoR/mSin3-containing histone deacetylase complex (HDAC). Thus, analysis of the Pit-1/Pit-1␤ isoform pair provides significant insight into the structural determinants of transcription activation versus repression mediated by two nearly identical transcription factor isoforms.

EXPERIMENTAL PROCEDURES
Cell Culture-Monolayer cultures of HeLa human cervical carcinoma cells and GH 4 T2 rat pituitary tumor cells (12) were maintained in Dulbecco's modified Eagle's medium, 15% horse serum, 2.5% fetal bovine serum, and 50 g/ml penicillin and streptomycin at 37°C in 5% CO 2 . The medium was changed 16 -18 h before each transfection. Cells used for transfections were harvested at approximately 60 -80% confluence using 0.05% trypsin and 0.5 mM EDTA.
Plasmids-The rat PRL promoter luciferase expression vector, pA 3 PRL luc, contains the firefly luciferase coding region under the control of a 498-bp fragment (Ϫ425 to ϩ73) of the rPRL promoter downstream of three polyadenylation termination sites in pA 3 luc (12). Plasmid pSV Ras contains the T24 bladder carcinoma Harvey Ras valine 12 mutant oncogene (Ras  ) under control of the SV40 early promoter. Plasmids pRSV HA Pit-1 and pRSV HA Pit-1␤ express N-terminally hemagglutinin (HA)-tagged Pit-1 and Pit-1␤ under the control of the RSV promoter (8), and plasmids pCGN2-Pit-1 and pCGN2 Pit-1⌬TAD express N-terminally HA-tagged Pit-1 and Pit-1 deleted for its TAD (amino acids 1-80) under the control of the CMV promoter, and they were the generous gift of Dr. David F. Gordon (University of Colorado Health Sciences Center, Denver, CO). Plasmid pAPR EtsZ encodes the DNAbinding domain (amino acids 334 -466) of human c-ETS-2 fused to LacZ under the control of the actin promoter (13). Plasmid pRSVt3 v-Ski contains the avian Sloan-Kettering virus ski oncogene under the control of the RSV promoter (14) and was the generous gift of Dr. Edward Stavnezer (Case Western Reserve University, Cleveland, OH). Plasmid DNAs were prepared by Qiagen (Qiagen Inc., Chatsworth, CA) columns and quantified by fluorimetry.
Luciferase Assays-Transient transfections were performed in triplicate, in at least three separate experiments. After incubation for 24 h, cells were harvested with phosphate-buffered saline containing 3 mM EDTA, pelleted, and resuspended in 100 mM potassium phosphate buffer, pH 7.8, 1 mM dithiothreitol. Cells were lysed by three cycles of freeze-thawing and by vortexing between thaws. Cell debris was pelleted by centrifugation for 10 min at 10,000 ϫ g at 4°C, and the supernatant was used for subsequent assays. Luciferase activity in the supernatant was assayed as described previously (12). Samples were measured in duplicate using a Monolight 2010 Luminometer (Analytical Luminescence Laboratories, San Diego, CA). Relative light units for each transfection were calculated by normalizing for total protein. Protein assays were performed according to the method of Bradford (16) using commercially available reagents (Bio-Rad). Results are expressed as the fold activation of the rPRL promoter Ϯ S.E. for at least three experiments, each in triplicate.
Western Blot Analysis of HA-tagged Pit-1 Proteins-Transient transfections were performed as above. Cells were harvested with phosphatebuffered saline containing 3 mM EDTA, pelleted, and resuspended in a triethanolamine/SDS solubilization buffer (55 mM triethanolamine, 111 mM NaCl, 2.2 mM EDTA, and 0.44% SDS) (17). Lysed extracts were passed through a 25-gauge needle seven times. The protein content of each extract was assayed according to the method of Lowry et al. (18), using commercially available reagents (Bio-Rad).
Equal amounts (100 g) of protein from each extract were separated on 15% SDS-polyacrylamide gels and transferred to Immobilon-P (polyvinylidene difluoride) membrane (Millipore, Bedford, MA). The HAtagged Pit-1 proteins were detected with a mouse monoclonal anti-HA primary antibody (BAbCO, Richmond, CA), secondary sheep antimouse HRP-conjugated antibodies (Santa Cruz Biotechnology Inc., Santa Cruz, CA), and ECL media (Amersham Pharmacia Biotech). Dilutions of 1:1,000 of the primary anti-HA monoclonal antibody and of 1:10,000 of the secondary sheep anti-mouse antibody preparation were used.
Chromatin Immunoprecipitation Studies-Chromatin immunoprecipitation (ChIP) assays were performed according to the protocol for the acetyl-histone H4 ChIP Assay Kit (Upstate Biotechnology, Lake Placid, NY), as modified by Lambert and Nordeen (19). Transient transfections were performed as above. Twenty four hours after transfection, 2 ϫ 10 7 GH 4 cells were cross-linked by addition of formaldehyde into the medium at a final concentration of 1% and incubated for 15 min at room temperature. Cells were washed with ice-cold phosphate-buffered saline and resuspended in 500 l of ChIP Lysis buffer (1% SDS, 10 mM EDTA, 50 mM Tris-HCl, pH 8.0, with protease inhibitors). The lysates were sonicated utilizing a Branson Sonifier 450 at power setting 2 with three 10-s pulses at duty cycle 90 and diluted to 3 ml with ChIP dilution buffer (0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris-HCl, pH 8.0, 167 mM NaCl). 1 ml of each sample was precleared by incubating with 80 l protein A-agarose beads for 30 min at 4°C with rotation. 5 l of anti-acetyl histone H4 antibody (Upstate Biotechnology, Lake Placid, NY) was added, and immunoprecipitation was done overnight at 4°C with rotation. Immune complexes were collected with 60 l of protein A-agarose and washed once with 1 ml each of the following buffers in sequence: Low Salt Wash Buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl, pH 8.0, 150 mM NaCl), High Salt Wash Buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl, pH 8.0, 1500 mM NaCl), LiCl Wash Buffer (250 mM LiCl, 1% Nonidet P-40, 1% sodium deoxycholate, 1 mM EDTA, 10 mM Tris-HCl, pH 8.0); and twice with TE (10 mM Tris-HCl, pH 8.0, 1 mM EDTA). Immune complexes were eluted, and cross-links were reversed by heating at 65°C and subjected to proteinase K treatment. DNA was recovered by phenol/ chloroform extraction followed by ethanol precipitation and was used as a template for PCR (25 cycles) using pA3 Ϫ425 PRL Luc promoter-specific commercially synthesized deoxyoligonucleotides (Life Technologies, Inc.) that contain a PRL promoter-specific sequence, GCCTTTCTTTATGTTTT-TGGC, and a luciferase-specific sequence, GACTCAAGATGTCAGTCAGC. In addition, internal control PCRs were performed with pSV Ras plasmidspecific commercially synthesized deoxyoligonucleotides (Life Technologies, Inc.) that contain an SV40 promoter-specific sequence, GCATCTCAATT-AGTCAGC, and a Ha-Ras-exon-1-specific sequence, ACCAGCTTATATTC-CGTC. Control reactions were performed to ensure that all PCR assays took place in the linear range of response to input DNA. PCR products were separated by agarose gel electrophoresis, and bands were imaged and quantified on an Alpha Imager 2000 Gel Documentation System (Alpha Innotech, San Leandro, CA).

RESULTS
The ␤-Domain Does Not Simply Disrupt TAD Function to Repress Ras Signaling-We have previously demonstrated that the ␤-domain insertion converts Pit-1 from a co-activator to a repressor of Ras signaling to the rPRL promoter (9,20). Moreover, we have shown that the amino acid sequence of the ␤-domain dictates this repression (8,11). One possible mechanism for ␤-domain-specific repression of Ras signaling would be that ␤-domain residues disrupt secondary or tertiary structures within the Pit-1 TAD that are important for Ras signaling. A prediction of this hypothesis would be that a particular Pit-1 TAD structure should be necessary for Ras signaling.
In order to test this prediction, HA-tagged wild-type Pit-1, Pit-1␤, and Pit-1 ⌬TAD, which is deleted for amino acids 1-80 ( Fig. 1), were introduced into GH 4 pituitary cells by electroporation in the presence of a rPRL promoter-driven luciferase reporter and pSV Ras (Fig. 2). As documented previously, cotransfection with wild-type Pit-1 constructs enhanced the Ras response from 3-fold in its absence to 10 -13-fold in its presence, and co-transfection of the Pit-1␤ isoform not only failed to enhance the Ras response but actually reduced it by more than one-third. The deletion of the Pit-1 TAD, on the other hand, did not significantly interfere with enhancement of Ras signaling by Pit-1 but, in fact, enhanced the response to 9-fold. These data thus demonstrate that the TAD itself is not required for Pit-1 mediation of Ras signaling to the rPRL promoter.
Epitope-scanning Mutagenesis of the Pit-1 ␤-Domain-In order to identify ␤-domain residues that are functionally important for repression of Ras signaling, we took a two-step approach. In order to identify small regions required for repression, we sequentially altered overlapping 6 amino acid blocks of the ␤-domain (Table I). Individual residues in functionally important regions identified were then subjected to alanine-scanning mutagenesis (see below).
We specifically chose the AU1 epitope (21) to replace 6 amino acid stretches of the ␤-domain because it did not affect the ability of mutant Pit-1␤ to function as a transcription factor when part of a ␤-domain substitution mutant (8). Each mutant ␤-domain is of the same size and in the same position as the wild-type ␤-domain and differs from the wild type by at most 6 amino acids. The mutant and wild-type constructs were each N-terminally tagged with an HA epitope (8). Thus, all constructs expressed proteins that contained the same epitope in the same relative position to allow for their detection by Western blot analysis regardless of possible alterations of protein structure by the ␤-domain substitutions.
Expression of Epitope-scanned Pit-1␤ Proteins-It has been previously shown that wild-type Pit-1 and Pit-1␤ constructs express protein at different levels in transient transfection experiments (6,8). In order to exclude the effect of differences in protein expression level on transcription potency, we carried out a series of transfection experiments to find levels of input DNA that would yield similar levels of protein expression from the wild-type and mutant Pit-1 vectors. In a preliminary experiment, 10 g of each of the pRSV HA Pit-1 constructs were introduced into HeLa nonpituitary cells and GH 4 pituitary cells by electroporation. Extracts from transfected cells were separated by SDS-PAGE, and Western blot analysis was used to determine the level of Pit-1␤ protein expression (data not shown). Pit-1␤ was expressed at lower levels than Pit-1 as observed previously (8); ES1, ES2, and ES5 were expressed at levels similar to but slightly higher than Pit-1␤; ES6 was FIG. 1. Structural organization of Pit-1 isoforms and constructs. Top, Pit-1 with its TAD, POU-specific, and POU homeodomains and their amino acid end points. P B and HD B represent POUspecific and POU-homeodomain basic domains; ␣1-4 and ␣1-3 represent their ␣-helices; Hinge represents the region between the TAD and the bipartite DNAbinding domain; FL represents the 15amino acid flexible linker between the POU-specific and POU-homeodomains. Middle, Pit-1␤ with the 26-amino acid ␤-domain insertion in its TAD. Bottom, the location of the ⌬TAD mutation is shown, a deletion of amino acids 2-80. expressed at levels similar to but slightly lower than Pit-1␤; whereas ES3 and ES4 were expressed at higher levels than Pit-1␤. Relative expression levels of Pit-1␤ proteins did not differ between cell lines (data not shown). Guided by these results, varying amounts of each of the mutant constructs were then introduced into GH 4 pituitary cells by electroporation, and DNA doses that result in similar levels of protein expression were identified as follows: 10 g of Pit-1, 25 g of Pit-1␤, 20 g of ES1, 20 g of ES2, 10 g of ES3 and ES4, 20 g of ES5, and 30 g of ES6 (Fig. 3A). In the vector-only lane, protein migrating in the Pit-1 range of 30 -33 kDa was not detected. Examination of the relative amounts of wild-type versus mutant Pit-1␤ reveals that the levels of all mutant constructs (lanes 4 -9) were roughly equivalent to that of wild-type Pit-1␤ (lane 3). ES6 (lane 9), however, was expressed at a somewhat lower level than the other constructs. The DNA doses noted above were used for subsequent experiments.
Epitope-scanning Pit-1␤ Proteins Retain Variable Transcription Function-Epitope-scanning mutagenesis could have induced alterations in the three-dimensional structure of the mutant Pit-1␤s such that they could no longer activate transcription under any circumstances, and such a result would preclude the identification of functionally important ␤-domain residues. To address this problem, we utilized an isoforminsensitive HeLa transcription reconstitution system, in which the Pit-1␤ isoform retains basal transcription potency on the rPRL promoter (8).
The wild-type and mutant Pit-1/Pit-1␤ constructs were introduced into HeLa nonpituitary cells with an rPRL promoter-driven luciferase reporter, and their ability to transactivate target promoter activity was measured. Pit-1␤ displayed a stronger effect on transcription of the target promoter compared with Pit-1 (210-versus 102-fold, respectively) (Fig. 3B). All but one (ES2) of the epitope-scanning mutants resulted in diminished basal activity, compared with that of wild-type Pit-1␤ (Fig. 3B), yet were expressed at levels comparable to the wild-type Pit-1␤ (Fig. 3A). In particular, the almost 90-fold difference in transcription activation potency between ES2 (347-fold) and ES3 (5-fold) cannot be explained by any difference in protein expression and is likely due to a difference in intrinsic activity of the proteins. These data identify ␤-domain amino acids altered by the ES3 mutation but not by the ES2 mutation, PKCL, as contributing to the basal transcription potency of Pit-1␤. Nevertheless, the key point is that the epitope-scanning ␤-domain mutants are all capable of transactivating the rPRL promoter, albeit to a varying degree.
Two Regions of the ␤-Domain Are Required for Repression of Ras-stimulated rPRL Promoter Activity in Pituitary Cells-In HeLa nonpituitary cells, both Pit-1␤ and Pit-1 activate transcription. However, in GH 4 pituitary cells, the two isoforms have opposite transcriptional effects with regard to Ras signaling to the rPRL promoter; Pit-1 enhances and Pit-1␤ represses the oncogenic Ras Val-12 response (9,22). Moreover, we have previously demonstrated that the amino acid sequence of the ␤-domain dictates this repression (8) (Fig. 2). In order to identify small regions of the ␤-domain that are functionally important in repressing the Ras response, the Pit-1␤ mutant epitopescanning constructs were introduced into GH 4 pituitary cells by electroporation in the presence of a rPRL-driven luciferase reporter with and without pSV Ras (Fig. 4).
As documented previously, co-transfection of a Pit-1 construct enhanced the Ras response from 3-fold in its absence to 9-fold in its presence, whereas co-transfection of the Pit-1␤ isoform not only failed to enhance the Ras response but reduced it to 50% below that achieved by Ras alone. This inhibitory effect of Pit-1␤ on Ras signaling to the rPRL promoter in pituitary cells defines the stereotypical repressor phenotype of Pit-1␤ in this assay (8). Therefore, we tested the Pit-1␤ epitopescanning mutants for their ability to reproduce the repressor   (Fig. 4), despite displaying variable basal transcriptional potencies (Fig. 3B) and equivalent protein expression levels (Fig. 3A). The other three mutations, ES2, ES4, and ES5, switched the Pit-1␤ repressor phenotype such that each no longer repressed but instead enhanced the Ras response, just as did Pit-1. These data identify two separate regions of the ␤-domain required for repression of the Ras signaling (Table II).
Alanine-scanning Mutagenesis of the Pit-1 ␤-Domain-We next sequentially substituted alanines for nine implicated residues that reversed the Pit-1␤ phenotype whenever changed (Table II). Alanine-scanning mutagenesis (23) was employed because alanine is an uncharged amino acid whose small CH 3 side chain should have minimal effects on protein secondary structure. Moreover, since there are no alanines in the ␤-domain, each alanine substitution is distinct from the endogenous ␤-domain sequence (Table II). Alanine-scanning mutagenesis has been used to probe structure-function correlates in several systems (reviewed in Ref. 24). All constructs were N-terminally HA-tagged to allow identification of DNA doses that result in similar levels of protein expression.
Expression of Alanine-scanning Pit-1␤ Proteins-Ten g of RSV HA Pit-1, 25 g of RSV HA Pit-1␤, and 25 g each of the mutant constructs were introduced by electroporation into HeLa nonpituitary cells (data not shown) and GH 4 (Fig. 5A) pituitary cells. Extracts from transfected cells were separated by SDS-PAGE, and Western blot analysis was used to determine the level of Pit-1␤ protein expression; the levels of all mutant constructs were roughly equivalent to that of wild-type Pit-1␤. AS3, AS5, and AS6 were expressed at a somewhat higher level than the other constructs, with AS3 displaying a second band of slightly slower mobility. Relative expression levels of Pit-1␤ proteins did not differ between cell lines (data not shown). These DNA doses were used for all further experiments.
Alanine-scanning Pit-1␤ Proteins Retain Transcription Function-All of the alanine-scanned Pit-1␤ proteins retained basal transcriptional potency in the isoform-insensitive HeLa

FIG. 4. Two small regions of the ␤-domain mediate repression of Ras-stimulated rPRL promoter activity in pituitary cells.
Mutant and wild-type pRSV HA Pit-1 constructs were introduced into GH 4 pituitary cells by electroporation with 3 g of pA3 PRL luc-425 and 5 g of pSV Ras. Plasmid pRSV HA Pit-1 plasmid DNA amounts were adjusted for equal protein expression (see Fig. 3). Total pRSV plasmid amount was maintained constant with pRSV ␤-globin DNA. After 24 h, cells were harvested and total light units measured (see "Experimental Procedures").

␤-Domain Modulates Interaction with Histone Deacetylase
reconstitution assay (Fig. 5B). In contrast to the epitope-scanning (ES) proteins, for which transcriptional potency was variable (Fig. 3B), the alanine-scanning (AS) proteins displayed basal activity comparable to wild-type Pit-1␤ (ϳ140-fold) (Fig.  5B). The higher transcription potency of AS6 (478-fold) or the lower transcription activity of AS4 (ϳ67-fold) cannot be explained by extreme levels of protein expression, since AS6 and AS4 are expressed at levels equivalent to the other AS mutants and wild-type Pit-1␤ (Fig. 5A). In addition, AS3, despite the novel slower-migrating band, activated the rPRL promoter to the same extent as did wild-type Pit-1␤. Again, the key point is that the alanine-scanning ␤-domain mutants are all capable of transactivating the rPRL promoter in an isoform-insensitive assay.
Five Hydrophobic ␤-Domain Residues Mediate Repression of the rPRL Promoter Ras Response-In order to identify specific ␤-domain residues that are required for repression of Ras signaling, the alanine-scanning constructs were assessed for retention of the Pit-1␤ repressor phenotype in GH 4 pituitary cells. Again, Pit-1 enhanced the Ras response from 6-fold without Pit-1 to 13-fold with Pit-1, whereas the Pit-1␤ isoform inhibited the Ras response by 50% (Fig. 6A). Four of the alanine-scanning proteins, AS3, AS4, AS7, and AS9, acted like wild-type Pit-1␤ and repressed the Ras response (Fig. 6A). Despite the aberrantly sized band seen with AS3 (Fig. 5A), it had no effect on repression, and AS3 repressed the Ras response at least as well as Pit-1␤ (Fig. 6A). Five mutations, AS1, AS2, AS5, AS6, and AS8, abrogated ␤-domain mediated repression of the Ras response. Two of these mutations, AS5 and AS8, eliminated the ␤-domain-mediated repression and restored the 6-fold Ras response but did not allow further enhancement of the Ras effect (Fig. 6A). In contrast, three mutations, AS1, AS2, and AS6, switched Pit-1␤ into an enhancer of Ras signaling, such that they functioned essentially the same as Pit-1 (ϳ13fold) (Fig. 6A). These data identify five ␤-domain amino acids (leucine 7, isoleucine 8, tyrosine 17, phenylalanine 18, and methionine 20) (Fig. 6B) that are required for repression of the Ras response of the rPRL promoter.
Pharmacological Inhibitors of Histone Deacetylation Reverse the ␤-Domain-dependent Repressor Phenotype-Two distinct models could explain ␤-dependent repression of Ras signaling to the rPRL promoter. In one model, the Pit-1␤ isoform acts as a classic dominant-negative inhibitor of Ras signaling, by binding to the Pit-1 site (FP IV) of the rPRL promoter composite Pit-1/Ets-1 Ras response element required for the Ras effect (9,22). In the second model, the ␤-domain imparts the repressor phenotype by modulating functional interaction with, or the function of, an HDAC, and thus altering the acetylation state of the PRL promoter.
To distinguish between these two models for ␤-domain-dependent repression, we first tested the role of HDACs using sodium butyrate and trichostatin A. Both sodium butyrate and trichostatin A have been widely used as pharmacological inhibitors of histone deacetylase activity (25,26) and are being considered for clinical use in managing certain cancers (reviewed in Ref. 27). We used EtsZ, a recombinant dominantnegative inhibitor, as a control. EtsZ consists of an Ets DNAbinding domain fused to LacZ and prevents transactivation of target promoters by competitively binding to Ets-binding sites (EBS) on DNA (13). Specifically, EtsZ blocks the Ras response of the rPRL promoter by interfering with endogenous Ets-1 binding to the rPRL promoter composite Pit-1/Ets-1 Ras response element (20,28). EtsZ was therefore expected to be insensitive to treatment with histone deacetylase inhibitors.
In the absence of histone deacetylase inhibitors, Ras activated rPRL promoter activity by 4-fold, whereas Pit-1␤ and EtsZ repressed the Ras response by 50 and 66%, respectively (Fig. 7). However, in the presence of 5 mM sodium butyrate, Pit-1␤ enhanced the 5-fold activation by Ras to 16-fold, whereas EtsZ-mediated repression was unchanged. Addition of 100 ng/ml trichostatin A also switched Pit-1␤ into an enhancer of Ras signaling. Trichostatin A and Pit-1␤ increased the 7-fold activation by Ras alone to 28-fold but had no effect on EtsZ, which continued to strongly repress the Ras response. Thus, both sodium butyrate and trichostatin A reversed ␤-domainmediated repression of the Ras response yet had no effect on repression by EtsZ, the competitive inhibitor of EBS binding. Moreover, similar effects were seen with 50 and 200 ng/ml trichostatin A (data not shown). These results implicate histone deacetylation in the mechanism of ␤-domain-dependent repression of Ras signaling and demonstrate that the mechanisms by which the ␤-domain and the simple dominant-negative inhibitor, EtsZ, repress Ras signaling are distinct.
v-Ski Protein Reverses ␤-Domain-dependent Repression-Pit-1 has recently been shown to interact with both a co-repressor complex that contains an N-CoR/mSin3 HDAC and a signaldependent co-activator complex (29). The Sloan-Kettering virus Ski oncogene, v-ski (30), is a dominant-negative viral form of mammalian c-Ski protein and a specific inhibitor of transcriptional repression mediated by N-CoR/mSin3-containing HDACs. We examined the effect of v-Ski on Pit-1␤-mediated repression of the Ras response, in order to determine whether the Pit-1␤ repressor phenotype is mediated by an N-CoR/ mSin3-containing HDAC.
In the absence of v-Ski, Ras activated rPRL promoter activity by 3-fold, whereas Pit-1␤ and EtsZ reduced it (Fig. 8). However, co-transfection of pRSV t3 v-Ski (5 g) switched Pit-1␤ from a repressor to an enhancer of the Ras response, increasing the 4-fold activation by Ras to 9-fold. EtsZ-mediated repression was unchanged. In addition, 2-and 10-g doses of pRSV t3 v-Ski reversed the Pit-1␤ repressor phenotype (data not shown) without affecting the EtsZ-mediated inhibition. Thus, v-Ski, like the pharmacological histone deacetylation inhibitors, sodium butyrate and trichostatin A, blocks the repressive effect of the ␤-domain yet has no effect on repression by EtsZ.
Pit-1␤ Alters the Acetylation State of the Proximal PRL Promoter-In order to test directly the hypothesis that the Pit-1 ␤-domain represses PRL promoter expression by altering the level of histone deacetylation of the proximal PRL promoter, we performed a ChIP assay. The ChIP assay has been used by a number of laboratories to probe the effects of hormones, DNA-binding factors, and transcription cofactors on the acetylation state of chromatin (reviewed in Ref. 19). In the assay, total cellular extracts are reversibly cross-linked, and DNA bound to acetylated histone proteins is immunoprecipitated with antibodies directed against acetylated histones. The cross-links are reversed and DNA-precipitated, and DNA sequences of interest are amplified by PCR. Changes in the amounts of specific PCR products reflect changes in the amount of acetylated histone bound to the DNA sequence.
We examined the effect of Pit-1 and Pit-1␤ on the acetylation state of the proximal PRL promoter in two independent ChIP assays with internal control experiments to verify that the biological effect in question, repression of Ras activation of PRL promoter activity, had occurred (see "Experimental Procedures"). In addition we utilized the SV40 promoter of the pSV Ras plasmid as a control promoter to demonstrate that Pit-1␤dependent changes in histone deacetylation were specific for the proximal PRL promoter and not global changes in histone deacetylation.
In two independent ChIP assays, Pit-1 increased the amount of proximal PRL promoter associated with acetylated histone H4 by 2-3-fold, whereas Pit-1␤ decreased the amount of proximal PRL promoter associated with acetylated histone H4 by 60% (Fig. 9). Neither Pit-1 nor Pit-1␤ had an appreciable effect on the amount of SV40 internal control promoter associated with acetylated histone H4. In parallel control experiments to verify Pit-1␤ repression of Ras signaling, Pit-1 enhanced the Ras response from 2-to 4-fold, whereas the Pit-1␤ isoform inhibited the Ras response by 80% (data not shown). Thus, the presence of the ␤-domain endows Pit-1␤ with the ability to increase histone deacetylation in a target promoter-dependent manner.

DISCUSSION
Although a number of functionally distinct transcription factor isoforms with identical DNA specificities have been identified (1)(2)(3)(4)(5)(6)31), the molecular mechanisms by which structural alterations outside of DNA-or ligand-binding domains specify altered transcriptional function remains a relatively unstudied subject. The work presented here shows that the naturally occurring ␤-domain insertion converts the Pit-1␤ isoform into a transcriptional repressor by modulating functional interaction with histone deacetylases.
Only two other transcription factor splice isoforms, both nuclear receptors, are known to alter distinct interactions with FIG. 6. Five hydrophobic ␤-domain residues mediate repression of Rasstimulated rPRL promoter activity in pituitary cells. A, mutant and wild-type pRSV Pit-1 constructs were introduced into GH 4 pituitary cells by electroporation with 3 g of pA3 PRL luc-425 and 5 g of pSV Ras. Plasmid pRSV HA Pit-1 plasmid DNA amounts were adjusted for equal protein expression (see Fig. 5). Total pRSV plasmid amount was maintained constant with pRSV ␤-globin DNA. After 24 h, cells were harvested and total light units measured (see "Experimental Procedures"). B, the five hydrophobic residues leucine 7, isoleucine 8, tyrosine 17, phenylalanine 18, and methionine 20, which are required for ␤-domain-dependent repression of Ras signaling. co-activators or co-repressors. Alternate exon usage in the TR␤2 isoform of TR␤ introduces a novel N terminus that reverses a functional interaction with N-CoR, thus altering the mechanism of ligand-independent repression (32). The HNF4␣2 splice isoform of nuclear receptor hepatocyte nuclear factor 4␣ (HNF4␣) contains a 10-amino acid insertion in the inhibitory F-domain, which abrogates an F-domain block of interaction with the GRIP1 co-activator (33). Thus, the Pit-1/ Pit-1␤ isoform pair represents the first example of differential interaction with either co-repressors or co-activators by transcription factor splice isoforms outside of the nuclear receptor family.
Transcriptional repressors can act through either passive or active mechanisms (reviewed in Refs. 34 and 35). Passive transcriptional repressors act by blocking such functions as nuclear localization or DNA binding that are necessary for transcriptional activation but that are not directly involved in nuclear signal transduction. Such repressors as Drosophila knirps protein, (36), mammalian ATF-2 (37), and Drosophila I-POU protein (38) can function as passive repressors. EtsZ, the recombinant dominant-negative construct used in this article, also acts as a passive binding site competitor in the context of the rPRL FPIV composite Ras response element (28). Active repressors interfere with the molecular mechanism of signal transduction in an activator-specific manner by blocking activation functions of specific transcription factors, recruiting corepressors, or by sequestering co-activators. Examples of active FIG. 9. Pit-1␤ alters the acetylation state of the proximal PRL promoter. Plasmid pA3 PRL luc-425 (3 g), pSV Ras (2 g), and either pRSV-Pit-1 (10 g) or pRSV-Pit-1␤ (25 g) were introduced into 10 100-mm plates of GH 4 (2 ϫ 10 7 ) pituitary cells by electroporation. After 24 h, eight plates were harvested for chromatin immunoprecipitation assays, and two plates were harvested for luciferase assays (see "Experimental Procedures"). A, a representative ChIP assay is shown; Blank lanes, negative control that had no template for the PCR; B, two independent ChIP assays are shown, expressed as relative amounts of target and control promoter DNA associated with acetylated histone H4 (in arbitrary densitometric units). repressors include YY1 repression of CREB-mediated transcription (39), KRAB-KAP-1 repression (40), the t(8;21) AML-1/ETO fusion protein (41), Ume 6 (42), TR (43), PLZF protein (44), and Laz3/BCL6 oncoprotein (45).
The data presented here (Figs. 7-9) clearly demonstrate that the ␤-domain endows the Pit-1␤ isoform with the properties of an active repressor. Pit-1␤ blocks Ras signaling through a mechanism quite distinct from that of EtsZ, the artificially constructed passive binding site competitor. Moreover, by four separate methodological approaches (site-specific mutagenesis, pharmacological inhibitors, co-expression of a dominant-negative effector, and chromatin immunoprecipitation assay), we show that five hydrophobic residues present as a bipartite element within the ␤-domain and that the ␤-domain mediates repression of the Ras response by modulating the acetylation state of the proximal PRL promoter in a manner dependent upon an N-CoR/mSin3-HDAC with Pit-1. The ability of two pharmacological inhibitors of histone deacetylase activity, trichostatin A and sodium butyrate, to reverse ␤-domain repression of Ras signaling to the rPRL promoter is consistent with such a mechanism for ␤-domain repression. In fact, sodium butyrate and trichostatin A appear to phenocopy some of the epitope-and alanine-scanning mutations of the ␤-domain and convert Pit-1␤ from a repressor to an enhancer of Ras signaling. By contrast, neither of these inhibitors affected the dominant-negative EtsZ inhibitory response (Fig. 7). This specificity of the pharmacological inhibitors for the ␤-domain-dependent repressor phenotype supports the validity of this approach and our interpretation of the data.
The use of v-Ski as a molecular tool allowed us to address the following two concerns raised by the pharmacological inhibitor experiments used here: 1) the possibility that these inhibitors had nonspecific effects, and 2) that the pharmacological approach cannot distinguish between different HDAC subtypes. The c-Ski protein is a normal component of an HDAC subtype, and it serves to recruit N-CoR and mSin3 (14,46). The oncogenic form, v-ski, lacks the mSin3-binding domain and inhibits transcriptional repression by Rb, Mad, and TR␤ (14,46). Thus, v-Ski is a specific inhibitor of HDAC-mediated transcriptional repression and provides independent evidence for the specificity of the ␤-domain repression mechanism.
The ChIP assays allowed us to demonstrate directly that the Pit-1␤ isoform causes a decrease in acetylation of histones associated with the proximal PRL promoter in the presence of activated Ras. The lack of change in the acetylation state of the control promoter shows that the transcriptional effects of the Pit-1␤ isoform are not global in nature but specific to Pit-1 target genes.
Recently it had been shown that an N-CoR/mSin3-containing-HDAC is known to associate with the POU domain of Pit-1 and that the balance of interaction between this HDAC and signal-specific co-activators with Pit-1 determines the transcriptional activity of Pit-1 (29). Taken together with these data, our findings that the Pit-1 ␤ isoform modifies the acetylation state of the proximal PRL promoter and that dominantnegative v-Ski blocks the ability of the ␤-domain to repress Ras activation of the proximal PRL promoter suggest a model in which the ␤-domain modulates an already existing functional interaction between Pit-1/Pit-1␤ and N-CoR/mSin3 HDAC. The exact mechanism by which the ␤-domain may modulate this functional interaction remains to be determined.
The fact that the Pit-1␤ isoform only acts as a repressor in pituitary cells, and actually is a superior mediator of protein kinase A signaling to the rPRL promoter in nonpituitary cells (8,11), suggests that the ␤-domain functions through a signalspecific (and possibly even cell-type-specific) rather than a glo-bal mechanism. It is important, then, to note that Pit-1 functions in Ras signaling as part of a multicomponent complex consisting of signal-specific CBP/p300 co-activators and an N-CoR/mSin3 co-repressor (29), as well as Ets-1, a prototypical member of the ETS transcription factor family. Ets-1 and Pit-1 functionally interact via the (Pit-1)/EBS (Ets-1) composite DNA element in the rPRL promoter (9,22,28), and Ets-1 is a direct nuclear target of the Ras/mitogen-activated protein kinase signaling pathway through phosphorylation of a conserved threonine 82 residue (9,47,48). Moreover, Ets-1 interacts with the same region of CBP as Pit-1 (29,49). Therefore, the specific mechanism of ␤-domain-dependent repression may take place in the context of a Pit-1⅐Ets-1 complex in balanced interaction with co-activator and co-repressor complexes.
It is of interest, then, that the ␤-domain blocks synergy between Pit-1 and Ets-1 in HeLa nonpituitary cells (reviewed in Ref. 8). Our recent finding that the ␤-domain can interact physically with Ets-1 in vitro (50) raises the possibility that the ␤-domain might function by blocking the ability of Ets-1 to respond to Ras signaling, perhaps by blocking access to the crucial threonine 82 mitogen-activated protein kinase phosphorylation site. Thus, the ␤-domain may function by modulating the ability of either a single component of the Pit-1⅐Ets-1 complex or a novel interaction surface generated by the complex to interact with co-repressor(s).
The role of the ␤-domain in basal transcriptional activity of Pit-1␤ is unclear at this time. For example, Pit-1 lacks the ␤-domain yet retains basal transcription function in the HeLa reconstitution assay (Figs. 3A and 5A), and complete substitution mutations of the ␤-domain also retain basal transcription activity (11). The ES and AS mutations have differential effects on the basal transcriptional potency of Pit-1␤ (Figs. 3A and 5A), despite the similarity of their overall levels of protein expression (Figs. 3A and 5A). It is possible that the ES mutations that resulted in decreased basal transcriptional activity of Pit-1␤ identify critical residues required for Pit-1␤ to act as a transcriptional activator. However, the fact that these same mutations resulted in unambiguous repression or enhancement of Ras signaling in the GH 4 pituitary cell assay of the repressor phenotype suggests that these residues are not necessary for repression of Ras signaling. However, one must be cautious interpreting the HeLa cell reconstitution data because the basal transcriptional effects seen in this assay could be due to subtle structural differences caused by the larger size of the epitope-scanning mutations versus the alanine-scanning mutations.
The data presented in this report advance our understanding of the repressor function of Pit-1␤, showing that the ␤-domain represses Ras signaling by modulating the acetylation state of the proximal prolactin promoter in a manner dependent upon the function of an N-CoR/mSin3-containing HDAC. Taken together, our results suggest that the ␤-domain might induce changes in either or both of the Pit-1 or Ets-1 three-dimensional structures that alter the balance of POU-domain interactions with co-activators (CPB) and co-repressors (N-CoR/ mSin3-containing HDACs), in a manner reminiscent of the HNF4␣2 isoform (33). However, an interaction with the N-CoR⅐mSin3 complex containing HDAC, through a signalspecific cofactor, cannot be ruled out.