Interaction of Protein Inhibitor of Activated STAT (PIAS) Proteins with the TATA-binding Protein, TBP*

Transcription activators often recruit promoter-targeted assembly of a pre-initiation complex; many repressors antagonize recruitment. These activities can involve direct interactions with proteins in the pre-initiation complex. We used an optimized yeast two-hybrid system to screen mouse pregnancy-associated libraries for proteins that interact with TATA-binding protein (TBP). Screens revealed an interaction between TBP and a single member of the zinc finger family of transcription factors, ZFP523. Two members of the protein inhibitor of activated STAT (PIAS) family, PIAS1 and PIAS3, also interacted with TBP in screens. Endogenous PIAS1 and TBP co-immunoprecipitated from nuclear extracts, suggesting the interaction occurred in vivo. In vitro-translated PIAS1 and TBP coimmunopreciptated, which indicated that other nuclear proteins were not required for the interaction. Deletion analysis mapped the PIAS-interacting domain of TBP to the conserved TBPCORE and the TBP-interacting domain on PIAS1 to a 39-amino acid C-terminal region. Mammals issue seven known PIAS proteins from four pias genes, pias1, pias3, piasx, and piasy, each with different cell type-specific expression patterns; the TBP-interacting domain reported here is the only part of the PIAS C-terminal region shared by all seven PIAS proteins. Direct analyses indicated that PIASx and PIASy also interacted with TBP. Our results suggest that all PIAS proteins might mediate situation-specific regulatory signaling at the TBP interface and that previously unknown levels of complexity could exist in the gene regulatory interplay between TBP, PIAS proteins, ZFP523, and other transcription factors.

in the C termini of all PIAS proteins except PIAS␥. The function of this region is unknown (16).
Here, we show that mouse TBP interacts with ZFP523, the mouse ortholog of human ZNF76. In addition, we report the novel interaction of TBP with PIAS1, PIAS3, PIASx, and PIASy proteins. The TBP/PIAS interaction is shown to occur between in vitro translated proteins, suggesting the interaction is direct, and it is detected between endogenous proteins in nuclear extracts, suggesting it occurs in vivo. Our results suggest that PIAS proteins might modulate transcriptional signaling at the TBP interface.

EXPERIMENTAL PROCEDURES
TBP Bait Constructs and cDNA Prey Libraries-Two bait vectors were used for two-hybrid screens and confirmatory interactions: pGBKT7 (BD Bioscience) and MP34 (R. Brazas, Duke University). pGBKT7 places the Gal4 DNA binding domain upstream of the bait in the fusion protein, whereas MP34 places the DNA binding domain downstream of the bait (Fig. 1A). In both vectors, expression is from the ADH1 promoter, replication uses a 2origin, and the TRP1 marker is used for selection in yeast. The vectors also encode ampicillin resistance for selection in bacteria. cDNA fragments encoding mouse TBP fulllength (TBP-FL) and TBP N terminus (TBP-N, amino acids 1-136) were generated by PCR from a plasmid containing the predominant mouse somatic TBP cDNA (28,29) using the following primer sets (Table 1): TBP-FL, TBP-N-start primer and TBP-C-end primer; TBP-N, TBP-N-start primer and TBP-N-end primer. PCR-amplified TBP cDNA fragments were cut with SalI/NotI and ligated into SalI/ NotI-cut MP34. The TBP C terminus (TBP-C) was amplified using TBP-C-start primer and TBP-C-end primer, digested with SalI and NotI, and inserted into a pGBKT7 vector modified by digesting with BamHI, filling with Klenow, and ligating to itself to shift the reading frame one base in the ϩ1 direction, allowing for in-frame insertion of TBP-C into the modified pGBKT7 ϩ1 vector. All bait clones and vector modifications were verified by sequencing.
The pGADT7 prey plasmid (BD Bioscience) was altered to allow efficient directional cloning of oligo-(dT)-primed cDNA libraries with SalI and NotI 5Ј and 3Ј linkers, respectively, as follows. The NotI site was cut, filled with Klenow, and re-ligated to kill the site, generating pGADT7 ⌬NotI . A linker containing internal SalI and NotI sites and 5Ј BamHI and 3Ј XhoI overhangs was ligated into BamHI/XhoI-cut pGADT7 ⌬NotI . The modified prey plasmid was named pGADT7 SN .
Oligo-(dT)-primed cDNA prey libraries were constructed and inserted into the pGADT7 SN vector as follows. Total RNA was extracted and CsCl-purified (30) from embryonic day 10.5 (E10.5) wildtype C57Bl/6J whole pregnant uteri or placentas. 3 In each case, either whole pregnant uteri or placentas were obtained from four pregnant dams to generate a pool of RNA. Poly(A ϩ ) mRNA from these samples was purified using Oligo-(dT) 25 Dynabeads (Dynal Biotech ASA, Oslo, Norway) following the manufacturer's protocols. Each cDNA library was constructed using 2.5 g of poly(A ϩ ) mRNA and the Superscript plasmid system for cDNA synthesis and cloning (Invitrogen), which yields cDNAs containing 5Ј SalI and 3Ј NotI overhangs. cDNAs were ligated into SalI/NotI-digested pGADT7 SN . The whole pregnant uteri library contained ϳ3.0 ϫ 10 6 independent recombinants with 82% bearing inserts; the placental library contained ϳ2.6 ϫ 10 6 independent recombinants with 95% bearing inserts. The average insert size in both libraries exceeded 1 kb (Fig. 1B). Both libraries and all plasmids are freely available on request.
Yeast Two-hybrid System-All interactions were tested in Saccharomyces cerevisiae strain AH109 (BD Bioscience), which contains the Ade2, His3, and LacZ reporters, each under the control of a different promoter. For library transformations, a culture of AH109 containing the TBP bait construct was grown at 30°C for 48 h in liquid synthetic complete medium (SC) lacking tryptophan (SC-W) (Q-BIOgene, Irvine, CA). This culture was used to seed 300 ml of 2ϫ yeast extract/peptone/ adenine/dextrose at 5 ϫ 10 6 yeast/ml, which was determined by counting on a hemacytometer. The culture was grown at 30°C for ϳ5 h to a density of 2 ϫ 10 7 yeast/ml. Yeast were collected by centrifugation, washed once with water, and once with 100 mM LiAc and transformed with 14.4 ml of 50% PEG (average M r 3350), 2.16 ml of 1.0 M LiAc, 0.3 ml of 10 mg/ml sheared denatured salmon sperm DNA, and 120 g of the cDNA library in pGADT7 SN . Single-or two-component transformations of bait and/or prey plasmids into yeast used a standard PEG/LiAc protocol (31).  TBP-N-start  5Ј-atcgtcgactatggaccagaacaacagccttcca-3Ј  TBP-N-end  5Ј-tatgcggccgcccagagctctcagaagctggtgt-3Ј  TBP-C-start  5Ј-tatgtcgaccaccatggcgagctctggaattgtaccgcag-3Ј  TBP-C-end  5Ј-tatgcggccgcgtggtcttcctgaatccctttaa-3Ј  TBP-C160-forward primer  5Ј-atctcgagcattgcacttcgtgcaagaaatgctg-3Ј  TBP-C201-forward primer  5Ј-tatgtcgaccaccatggccaagagtgaagaacaatcc-3Ј  TBP-C210-reverse primer  5Ј-tatgcggccgcgctagtctggattgttcttcactc-3Ј  TBP-C251-forward primer  5Ј-tatgtcgaccaccatggtgctgacccaccagcagttc-3Ј  TBP-C263-reverse primer 5Ј-tatgcggccgctctggctcatagctactgaact-3Ј PIASx␤-forward primer 5Ј-tatgtcgaccatgaatatggtttctagttttagggtttc-3Ј PIASx␤-reverse primer 5Ј-tatgcggccgcctttagtccaaagagatgatgtc-3Ј PIASy-forward primer 5Ј-tatgtcgacgctagtggccaagatggc-3Ј PIASy-reverse primer Two-hybrid screens were performed on SC medium lacking leucine, tryptophan, and histidine (SC-L-W-H). Both two-hybrid screens used TBP-FL as bait to screen either the whole pregnant uteri (screen 1) or placental (screen 2) cDNA prey libraries. Primary transformants were transferred onto a new SC-L-W-H plate and clones that grew well on these plates were transferred to a SC-L-W plate, grown for 48 h at 30°C, and replica plated to higher stringency selection medium, such as SC-L-W-H ϩ 2.5 mM 3-aminotriazole, a competitive inhibitor of the His3 gene product (32) or SC-L-W-H also lacking adenine (SC-L-W-H-A). Yeast were also replica-plated to SC-L-W plates containing X-␣-gal (Glycosynth, Cheshire, United Kingdom) to identify clones that activated the LacZ reporter gene. Prey plasmids from those clones that grew under higher stringency selection and showed strong LacZ expression were isolated from yeast by glass bead lysis (33). Recovered plasmids were transformed into bacteria, clones were selected, and inserts were sequenced to determine the cDNA identity. Isolated prey plasmids were re-transformed into AH109 with the bait and grown on SC-L-W-H, SC-L-W-H ϩ 2.5 mM 3-AT, and SC-L-W-H-A plates to verify the interaction.
To identify the domain of TBP that interacted with prey proteins, TBP bait constructs in combination with the empty pGADT7 prey plasmid (autoactivation test) or with a prey plasmid containing novel or known TBP interactors were co-transformed into AH109 and plated onto SC-L-W. Individual colonies that grew on SC-L-W plates (containing both the bait and prey plasmid) were grown overnight in yeast extract/peptone/adenine/dextrose at 30°C, pelleted, and washed three times in sterile water. Yeast pellets were resuspended in 0.5 ml of water and their concentration determined by counting on a hemacytometer. Standard amounts of yeast were plated on SC-L-W-H to verify that bait plasmids were not autoactive or to test interactions of TBP subdomain baits with each prey. Prey plasmids were also tested for autoactivation by co-transforming yeast with prey plasmids and empty bait plasmid.
Antibodies, Animals, and Nuclear Extract Preparation-Mouse antic-Myc monoclonal antibody was purchased from Amersham Biosciences; mouse anti-FLAG M2 monoclonal antibody was purchased from Sigma; goat anti-PIAS 1 (C-20) and goat anti-PIAS 1/3 (N-18) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA); horseradish peroxidase-conjugated donkey anti-rabbit antibodies were purchased from Pierce; horseradish peroxidase-conjugated sheep antimouse was purchased from Amersham Biosciences; horseradish peroxidase-conjugated rabbit anti-goat antibody was purchased from Invitrogen. Rabbit-anti-TBP N C antiserum was raised against recombinant His 6 -tagged mouse TBP amino acids 72-135. For this later antiserum, specificity was verified by Western blots using recombinant TBP, wildtype mouse nuclear extracts, and nuclear extracts from mouse cells homozygous for the tbp ⌬N mutation (34, 35) (described below, mutation eliminates antibody-reactive domain, not shown).
For nuclear extract preparation, young adult male wild-type and tbp ⌬N/ϩ mice were used. The tbp ⌬N/ϩ mice are heterozygous for a targeted mutation that exchanged the endogenous tbp gene for a version that replaced 111 amino acids within the vertebrate-specific N terminus of TBP with two copies of the FLAG epitope tag (34,35). The mutant protein contains the entire TBP CORE region of the protein and Western blots confirm that this FLAG-tagged TBP accumulates to wild-type levels in all tissues (34,35). Heterozygous animals are healthy and fertile (34). Survival of animals homozygous for this mutation to E9.5 (100%), birth (9%), and adulthood (ϳ1%) indicates that the FLAG-tagged protein is a functional replacement for most TBP activities (34).
In Vitro Translation, Transient Co-transfections, and Co-immunoprecipitations-In vitro transcription/translation reactions were performed using the "TNT" system (Promega, Madison, WI) following the manufacturer's protocols with the plasmid templates and conditions detailed in the figure legends.
For transient transfection/co-immunoprecipitation assays, TBP-FL was inserted into pCMV-HA (BD Bioscience), which fused the HA epitope tag to the N terminus of TBP. PIAS1 1-691 , PIAS1 1-452 , PIAS1 6 -691 , PIAS3 325-628 , BRF1 264 -676 , or ZFP523 202-568 were inserted into pCMV-Myc (BD Bioscience), which fused the c-Myc epitope tag to the N terminus of each protein. Human embryonic kidney 293 cells (HEK293) were plated onto 60-mm dishes at ϳ30% confluence. The next day, cells were washed once with serum-free, drug-free Dulbecco's modification of Eagle's medium (DMEM) (Mediatech, Herndon, VA). Transfection mixes contained 300 l of serum-free drug-free DMEM, 1 g of each plasmid, and 10 g of Novafector (Venn Nova, Inc., Pompano Beach, FL). Transfection mixes were added to washed cells in dishes containing 2 ml of serum-free drug-free DMEM and incubated for 5 h at 37°C, 7.5% CO 2 . After incubation, 2 ml of DMEM containing 20% newborn calf serum (Invitrogen), 4% fetal bovine serum (HyClone, Logan, UT), and 2ϫ antibiotic-antimycotic solution (Mediatech) was added to each plate. Approximately 14 h later, the medium was replaced. At 48 h post-transfection, cells were washed 2 times in icecold phosphate-buffered saline and 500 l of lysis buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1% Triton X-100, 0.1% SDS, 0.8% deoxycholic acid, 10% glycerol, 1 mM EDTA, 1 mM PMSF, 5 g/ml leupeptin-pepstatin-aprotinin, 0.15 mM NaVO 3 , and 1 mM DTT) was added to each plate. Plates were rocked for 15 min at 4°C to allow cells to detach from the plate. The contents of each plate were transferred to a 1.5-ml tube and placed on ice for 30 min with brief mixing at 5-min intervals. Lysates were clarified at 10,000 ϫ g for 20 min at 4°C. Lysates from each set of plates were pooled and a portion of each lysate (ϳ50 l) was stored at Ϫ80°C for use as transfection controls. The remaining 950 l of each lysate was pre-cleared with 20 l of protein G plus/protein A-agarose (Calbiochem, San Diego, CA) and 5 l of nonspecific antibody or antiserum for 2 h at 4°C on a rotator. Samples were centrifuged to pellet the agarose and 450 l of each supernatant was used for immunoprecipitations.
For co-immunoprecipitations, protein samples (either 450 l of precleared transfected cell lysate, 100 l of nuclear extract diluted to 1.0 ml with 1ϫ TBS (50 mM Tris, pH 7.5, 150 mM NaCl), or 50 l of TNT lysates from in vitro co-translation reactions diluted to 500 l with binding FIGURE 1. Yeast two-hybrid bait constructs, prey libraries, and screens. A, bait constructs. TBP-FL and TBP-N were expressed from the MP34 plasmid, which fused the Gal4 DNA binding domain downstream of TBP. TBP-C was expressed from pGBKT7 ϩ1 , which fused the Gal4 DNA binding domain upstream of TBP. N and C designate the vertebrate-specific N terminus and the paneukaryotic C terminus of TBP (TBP CORE ), respectively. B, two prey libraries were constructed and inserted into pGADT7 SN , which fused prey cDNAs downstream of the Gal4 AD. Characteristics of each library are indicated. Below is shown PCR analysis of arbitrary clones from each library using a primer pair that spans the multiple cloning site of the vector. Lane "p" contained HinfI-cut phosphate-buffered saline ϩ plasmid markers; lane "" contained HindIII/EcoRI-cut -phage DNA markers. Landmark band sizes are indicated at left of gels; asterisks denote the size of the PCR product arising from empty prey vector. C, the results of the two yeast two-hybrid screens performed are shown. In both screens, TBP-FL was used as bait to screen either the placental or pregnant uteri library for interacting proteins. TBP-interacting prey library clones were subsequently identified by sequencing.
buffer (20 mM Tris, pH 7.5, 0.1 M NaCl, 5 mM MgCl 2 , 10% glycerol, 1 mM PMSF, 5 g/ml inhibitors) were transferred to tubes containing specific or nonspecific monoclonal antibody (2 g), specific or nonspecific goat polyclonal antiserum (5 l), or anti-FLAG M2 resin (40 l, Sigma), as indicated in the figure legends. Binding reactions were incubated overnight at 4°C on a rotator and 30 -40 l of protein G plus/protein A-agarose was added to each tube (except those containing anti-FLAG m2 resin) and rotated for an additional hour at 4°C. For transfected cell lysates, the agarose was pelleted at 750 ϫ g for 1 min, washed once with ice-cold lysis buffer, four times with the first wash buffer (50 mM Tris, pH 7.5, 0.3 M NaCl, 1% Triton X-100, 10% glycerol, 1 mM EDTA, 1 mM PMSF, and 0.15 mM NaVO 3 ), and one time with a final wash buffer that contained the same components as the first wash buffer, except that the NaCl concentration was lowered to 0.1 M. For in vitro translation immunoprecipitations, washes were performed with binding buffer containing 0.1% IgePAL-CA-630 detergent (Sigma) in place of Triton X-100, and for FLAG immunoprecipitations, washes were performed with 1ϫ TBS containing 0.1% Triton X-100. Pellets were resuspended in 1ϫ loading buffer (38), boiled 5 min, and separated by electrophoresis through a 12% SDS-polyacrylamide gel. Proteins were transferred to supported nitrocellulose, probed with the indicated primary and secondary antibodies, and visualized using Supersignal-West chemiluminescence (Pierce) and x-ray film. Yeast two-hybrid screens for proteins that interact with TBP. Bait and prey plasmid combinations that were contained within each yeast clone are indicated. Sector designations correspond to those on plates. Growth of yeast on SC-L-W indicated that all clones contained both the bait and prey plasmid. Sector 1 is a negative "autoactivation" control showing that yeast bearing only the TBP bait vector and an empty prey plasmid (no insert) did not grow on SC-L-W-H, SC-L-W-H ϩ 2.5 mM 3-aminotriazole, or SC-L-W-H-A. Sector 2 was a negative interaction control that was supplied with the Matchmaker III system encoding non-interacting bait and prey proteins. Sector 3 was a positive interaction control supplied with the Matchmaker III system, and accordingly, yeast in this sector grew well under all conditions tested. In sector 4, the interaction of TFIIA and TBP is shown. Growth of this clone under all conditions tested indicated a strong interaction between TFIIA and TBP, and served as a positive control because TFIIA is known to interact with TBP (40,41). Sector 5 shows the interaction of ZFP523 with TBP. This interaction may be weaker than the TFIIA/TBP interaction because growth of yeast on the SC-L-W-H ϩ 2.5 mM 3-AT plate was reduced. However, in comparison to negative controls in sectors 1 and 2, the TBP/ZFP523 interaction was scored as strongly positive. Sectors 6 and 7 show the interaction of PIAS1 and PIAS3, respectively, with TBP. Both the PIAS clones interacted strongly with TBP as compared with control interactions in sectors 3 and 4.  1-2, PIAS1; lanes 3-4, PIAS3) or immunoprecipitated samples (lanes 5-6, PIAS1; lanes 7-8, PIAS3) were assayed using Western blots and anti-Myc antibody. PIAS1 and PIAS3 co-precipitated with TBP in the presence of the anti-TBP antibody (lanes 5 and 7, respectively), but not with the nonspecific antibody (lanes 6 and 8). Asterisks designate nonspecific cross-reactive materials. B, TBP-ZFP523 and TBP-BRF1 interactions in co-transfected cells. ZFP523 co-precipitatated with TBP using the anti-TBP antibody (lane 5), but not with the nonspecific antibody (lane 6). As a positive control, a BRF1 clone isolated from our yeast two-hybrid screens using TBP-FL as bait was included, because BRF1 is known to directly interact with TBP (42). BRF1 co-precipitated with TBP in the presence of the anti-TBP antibody (lane 7), but not with the nonspecific antibody (lane 8). C, in vitro translated TBP and PIAS1 interact in reticulocyte lysates. Transcription/translation reactions were performed in standard 50-l TNT reactions charged with 1.0 g each of NotI-linearized pGBKT7 ϩ1 -TBPFL (encodes Myc-tagged full-length mouse TBP protein) and NotI-linearized pGADT7 SN -PIAS1 6 -651 . Immunoprecipitations (IP) were performed with anti-Myc, anti-TBP-N C , anti-PIAS1/3 (N18), or a matched negative control antibody, as indicated. Western blots were loaded with immunoprecipitate or whole reticulocyte lysate corresponding to the indicated percent of the amount of extract used in the immunoprecipitation and were probed with the indicated antibody. PIAS1 co-precipitated with TBP using either the anti-Myc or anti-TBP-N C antibody (left blots); TBP co-precipitated with PIAS1 using the anti-PIAS1/3 antibody.

Identification of TBP-interacting Proteins by Yeast Two-hybrid-To
identify mouse proteins that interacted with TBP, we screened both a placental cDNA library (screen #1) and a whole pregnant uteri cDNA library (screen #2) with mTBP-FL in the MP34 bait plasmid (Fig. 1). MP34-mTBP-FL did not grow on SC-L-W-H medium in the presence of an empty prey plasmid (Fig. 2, sector 1), confirming that the bait was not autoactive.
In screen 1 (placenta library), ϳ2.5 ϫ 10 6 primary transformants (yeast containing both bait and prey plasmids) (Fig. 1C) were plated onto SC-L-W-H medium. Fifty-eight colonies grew and all were transferred to a SC-L-W-H plate containing X-␣-gal. Thirty-seven clones grew well and expressed LacZ. These were transferred to a SC-L-W plate, grown for 48 h, and replica plated to SC-L-W-H-A medium to test for combined expression of both the HIS3 and ADE2 reporter genes. In this round of selection, nine clones grew well. After isolating the prey plasmid from each, we re-tested the interactions by co-transforming new yeast with the cDNA-containing prey plasmid and TBP-FL bait plasmid. All nine clones re-grew on SC-L-W-H ϩ X-␣-gal medium and all expressed LacZ (not shown).
In screen 2 (whole pregnant uteri library; Fig. 1C), of ϳ0.6 ϫ 10 6 primary transformants, ϳ100 colonies grew on SC-L-W-H plates, of which 41 were selected for further analysis. These were transferred to SC-L-W-H-A medium containing X-␣-gal. Thirteen clones grew well and expressed LacZ following retransformation.
Sequence analysis of the prey cDNAs from screen 1 revealed that seven of the nine clones encoded PIAS1, one clone encoded PIAS3, and one encoded the ␣-subunit of transcription factor IIA (TFIIA), a known TBP-interacting protein (39 -41). Analysis of TBP-interacting prey cDNAs from screen 2 identified three more PIAS1 clones. We also obtained three clones encoding BЈ-related factor 1 (BRF1), another known TBP interactor (39 -42). Additionally, we identified ZFP523, the mouse homologue of a recently identified novel TBP interacting protein, hZNF76 (11). The remaining six clones were all unique, although no obvious physiological connection to TBP was evident. The interac-  bottom panels), which selects for the interaction between bait and prey. In sectors 1-10 and 13-16, all baits were expressed from the MP34 plasmid and all truncated PIAS1 preys expressing the indicated amino acids were expressed from the pGADT7 SN plasmid. Clones in sectors 11 and 12 were positive and negative controls, respectively, supplied with the Matchmaker III system. All 4 isolates of PIAS1 6 -491 (clone 1) grew at a slower rate and lower density compared with the more robustly growing clones in sectors 2 and 3. Therefore, the bottom panel shows the same plate photographed above that had been allowed to grow for an additional 3 days to reveal growth of the more slowly growing clone in sector 1. Sectors 7 and 9 were the longest and shortest PIAS1 clones, respectively, that were isolated directly from yeast two-hybrid screens using TBP-FL as bait. Sectors 8 and 10 were controls confirming that PIAS1 453-651 and PIAS1 6 -651 , respectively, were not autoactive. The absence of growth of clones 4 -6 and 13 on SC-L-W-H suggested that these clones lacked the TBP interaction domain. Four independent clones of all PIAS1 truncations were tested and each gave the same result; one representative set is shown. None of the truncated PIAS1 fragments tested were able to grow when supplied with the empty bait plasmid (not shown), confirming that the prey plasmids were not autoactive. B, deletion analysis of the TBP interaction domain of PIAS1. The full-length PIAS1 protein is represented at the top, including the location of known domains. PIAS1 amino acids expressed by each cDNA clone in pGADT7 SN are shown at the left. The six different PIAS1 clones isolated from two-hybrid screens as TBP-interacting proteins are indicated. Approximate relative two-hybrid growth rates with the TBP-FL bait vector are represented by ϩϩ, ϩ, and Ϫ, at the right. Results suggested that the TBP-interacting region of PIAS1 was between amino acids 452 and 492, which included the Ser/Ac domain.
tions of TBP with TFIIA, ZFP523, PIAS1, and PIAS3 in the yeast twohybrid system is shown in Fig. 2.
For verification, the interactions between TBP and proteins identified by two-hybrid screens were tested by co-immunoprecipitation in cell culture. Both PIAS1 and PIAS3 co-precipitated with TBP when coexpressed in HEK293 cells (Fig. 3A). Additionally, ZFP523 and BRF1, both known TBP interactors, co-precipitated with TBP (Fig. 3B).
In Vitro Co-translated PIAS1 and TBP Co-immunoprecipitate-The PIAS/TBP interaction was identified using yeast, which contain homologues of most mammalian transcription machinery components. Verification used transfected human cells, which contain the mammalian machinery. It was possible that, in both systems, PIAS and TBP did not interact directly, but rather, PIAS proteins assembled into a stable TBPcontaining complex via interactions with other endogenous TBP-interacting proteins. To distinguish these possibilities, we co-translated the two proteins in vitro using a reticulocyte-based system and tested whether PIAS1 and TBP co-immunopreciptated. Immunoprecipitation of either protein pulled-down the other; however, nonspecific antibody did not pull-down either (Fig. 3C). Thus, the PIAS/TBP interaction does not require other nuclear proteins, but rather, is very likely the result of direct PIAS/TBP contacts.
Identification of the TBP-interacting Domain on PIAS1-Sequence analysis of the 10 TBP-interacting PIAS1 clones isolated in the twohybrid screens revealed that six were of differing lengths. The longest clone encoded amino acids 6 to 651 (PIAS1 6 -651 ); the shortest encoded amino acids 453 to 651 (PIAS1 453-651 ) (Figs. 2 and 4). Other clones encoded amino acids 135-651, 359 -651, 430 -651, and 439 -651 of PIAS1 (Fig. 4B). The single PIAS3 clone encoded PIAS3 amino acids FIGURE 5. PIAS1 and PIAS3 interact with the TBP CORE . A, TBP-FL, -N, and -C baits were tested for autoactivation and for two-hybrid interactions with short (amino acids 453-651) and long (amino acids 6 -651) clones of PIAS1 and with a clone of PIAS3 (amino acids 325-628). As controls, interactions of the TBP-FL and the N-and C-terminal TBP subdomains with TFIIA, a known TBP CORE interactor (40,41), are shown. Serial dilutions of yeast containing the plasmid combinations indicated were spotted onto SC-L-W (left column) and SC-L-W-H (right column) plates. Growth of clones on the SC-L-W plate verified that yeast contained both the bait and prey plasmid. In the top right panel, the absence of growth indicated that none of the TBP bait clones were autoactive when supplied with the empty prey plasmid. The bottom right panel is a control showing that TFIIA interacted with TBP-FL and TBP-C, but not with TBP-N. Right panels [2][3][4] show that all PIAS clones interact with both TBP-FL and TBP-C, but not with TBP-N. The bottom rows of right panels 2, 3, 4, and 5 are autoactivation controls for each prey plasmid, none of which grew on SC-L-W-H, and thus were not autoactive when supplied with the empty bait plasmid. B, truncated versions of the TBP CORE were constructed to more accurately identify the subdomain that is required for interaction with PIAS1 6 -651 . Only the TBP-C 135-263 protein (sector 1), encoding approximately the first two-thirds of TBP-C, was able to maintain the interaction with PIAS1. This suggested that an extended conformation of TBP-C is required for the interaction with PIAS1. FIGURE 6. Endogenous PIAS/TBP interaction. Co-immunoprecipitation (IP) of endogenous PIAS1 and TBP from mouse nuclear extracts were performed with the indicated antibodies on nuclear extracts prepared from livers and spleens harvested from wildtype or tbp ⌬N/ϩ mice. A, ⌬N-TBP co-immunoprecipitates with PIAS1. Nuclear extracts from wild-type (ϩ/ϩ) or heterozygous (⌬N/ϩ) mice were analyzed by Western blotting with anti-FLAG antibody before or after immunoprecipitation with anti-PIAS1 antibody. Results showed that the ⌬N/ϩ nuclear extracts contained a single major anti-FLAGreactive protein (⌬N-TBP) and that this protein co-immunoprecipitated with PIAS1. B, PIAS1 co-immunoprecipitates with ⌬N-TBP. Nuclear extracts were immunoprecipitated with anti-FLAG or an irrelevant negative control mouse monoclonal antibody and probed Western blots with goat anti-PIAS1 (C20) antibody. Results showed that a portion of the endogenous PIAS1 co-precipitated with endogenous FLAG-tagged TBP CORE from heterozygous but not wild-type nuclear extracts; the nonspecific antibody did not bring down PIAS1. 325-628. Because libraries were oligo-(dT)-primed, all clones used their natural stop codon.
The shortest PIAS1 clone isolated from the two-hybrid screens (PIAS1 453-651 ) contained the last 199 amino acids of PIAS1. To more precisely identify the domain that interacted with TBP, we constructed more 5Ј-truncated versions, which encoded amino acids 484 -651, 555-651, and 598 -651, and 3Ј-truncated versions of PIAS1, which encoded amino acids 6 -452, 6 -491, 6 -562, and 6 -605 (Fig. 4B). Each PIAS1 mutant was tested for interaction with TBP-FL in the two-hybrid system (Fig. 4A). Only clones containing the region from 453 to 491 interacted with TBP, as indicated by growth on SC-L-W-H selective medium (Fig. 4A). Yeast that contained the PIAS1 6 -491 prey grew less well (Fig. 4A, bottom panel, sector 1), suggesting that this truncation might have weakened but not ablated the TBP interaction domain. Sec-tors 7-10 are TBP interaction and autoactivation controls for PIAS1 6 -651 and PIAS1 453-651 , respectively. Our data suggest that the TBP interaction domain of PIAS1 requires the 39 amino acids from positions 453 to 491, which includes the Ser/Ac domain (Fig. 4B).
PIAS1 and PIAS3 Interact with the TBP CORE -To determine whether the PIAS clones that were isolated in two-hybrid screens interacted with the conserved C-terminal TBP CORE or the vertebrate-specific N terminus (43), we co-transformed yeast with the following baits and preys: TBP-N/PIAS1 453-651 , TBP-C/PIAS1 453-651, TBP-N/PIAS1 6 -651 , TBP-C/PIAS1 6 -651 , TBP-N/PIAS3 325-628 , and TBP-C/PIAS3 325-628 . Growth of all clones on SC-L-W verified the presence of both the bait and the prey plasmid. As controls, we tested the interactions of TBP-N or TBP-C with TFIIA␣ (Fig. 5A) (40,41). Neither the TBP bait clones nor the PIAS and TFIIA prey For other family members, dots represent amino acids that are identical to PIAS1. Comparison of the PIAS protein sequences revealed that the deduced TBP-interacting domain of PIAS1 (denoted in bold) is highly conserved in PIAS3 and PIASx␤, with the greatest amino acid identity occurring within the Ser/Ac domain. PIASy contains a similar, albeit more extended Ser/Ac domain (27). Shaded amino acids represent the first amino acid encoded by the shortest 5Ј-truncated PIAS1 clone (PIAS1 453-651 ) that interacted with TBP and the single PIAS3 clone (PIAS3 325-628 ) that was isolated in yeast two-hybrid screens described above. The boxed amino acid in PIAS1 represents the last amino acid encoded by the most 3Ј-truncated mutant (PIAS1 6 -491 ) that interacted with the TBP and defined the C-terminal boundary of the deduced TBP interaction domain of PIAS1.
clones were autoactive (Fig. 5A), confirming the validity of the assay. We found that all PIAS clones and the TFIIA clone interacted with the TBP CORE (indicated by growth on SC-L-W-H in Fig. 5A, right panels), but not with the TBP N terminus (Fig. 5A).
We also constructed truncated TBP CORE mutants to further define the domain of TBP that interacted with PIAS1. Only one of our mutants, TBP-C 135-263 , encoding approximately the first two-thirds of the TBP-CORE , allowed growth with PIAS1 6 -651 (Fig. 5B, sector 1).
Association of Endogenous PIAS1 and TBP in Mouse Nuclear Extracts-Our results indicated that PIAS1 could interact with the TBP CORE both in vivo under conditions of overexpression (two-hybrid and transient transfection) and in reticulocyte extracts. To gain insights into whether endogenous PIAS1 and TBP proteins in normal cells might also interact, we tested whether we could detect physical association of the endogenous proteins in mouse nuclear extracts. Mice bearing a targeted mutation at the tbp locus that replaces most of the vertebrate-specific N-terminal domain with two copies of the FLAG tag (tbp ⌬N mice) provided a unique resource in which a functional TBP CORE expressed at wild-type levels is epitopetagged (34,35).
Nuclear extracts were prepared from livers and spleens of adult wildtype and heterozygous (tbp ⌬N/ϩ ) mice. Immunoprecipitation using anti-PIAS1 antibody brought down TBP⌬N protein from heterozygous nuclear extracts; absence of an anti-FLAG antibody-reactive band in wild-type extracts confirmed the specificity of the FLAG antibody/tag combination (Fig. 6A). Nuclear extracts were then immunoprecipitated with the anti-FLAG antibody or a negative control antibody and Western blots were probed with anti-PIAS1 antibody (Fig. 6B). The anti-FLAG antibody, but not the control antibody, brought down PIAS1 protein from heterozygous but not wild-type extracts. These results indicated that a portion of the endogenous PIAS1 in mouse liver and spleen nuclear extracts was complexed with TBP⌬N.
TBP Interacts with PIASx and PIASy-Amino acid alignment of all known isoforms of PIAS proteins from all four pias genes (16) revealed that all shared the TBP interaction domain ( Fig. 7 and data not shown) identified here (Fig. 4, A and B). The remainder of the C-terminal region shows little identity between family members. In fact, PIASy does not contain ϳ100 amino acids that are found in the other PIAS family members. Based on the amino acid conservation in the TBP interaction domain between all PIAS proteins, we wished to determine whether proteins from the piasx and piasy genes also functionally interacted with TBP. Full-length cDNAs for PIASx␤ and PIASy were isolated by reverse transcriptase-PCR and were inserted into the two-hybrid prey vector (pGADT7 SN ) to test for interaction with either MP34-TBP-FL or the empty bait plasmid in two-component two-hybrid assays. Both proteins interacted strongly with TBP-FL; neither prey was autoactive when tested with empty bait (MP34) vector (Fig. 8).
In conclusion, our data suggest that TBP interacts with ZFP523 and with all members of the PIAS family of proteins through the conserved TBP CORE . Using PIAS1 as a family representative, we show that that PIAS/TBP interaction involves a conserved 39-amino acid region within the PIAS C-terminal region and very likely occurs through direct contacts between PIAS and TBP proteins. Moreover, using nuclear extracts from mice bearing an epitope-tagged TBP CORE , we present evidence suggesting that the endogenous proteins interact as well. These results lend insights into the gene-regulatory interplay between TBP, ZNF523, PIAS proteins, and other transcriptional regulators, as discussed below.

DISCUSSION
TBP is a central player in transcription initiation whose activity is controlled by interactions with activators, repressors, and other transcriptional regulatory proteins. In agreement with the recently reported interaction between TBP and human ZNF76 (11), we show that mouse ZFP523, which shares 92% amino acid identity with ZNF76, interacts with TBP (Figs. 2 and 3B). The clone of ZFP523 we isolated in our screen was a partial cDNA that encoded the last 366 amino acids of the 568amino acid protein. This clone contained the entire GARD domain (amino acids 362-444), which is the region of the human protein shown to interact with TBP (11).
Most of the TBP-interacting clones isolated in our two-hybrid screens encoded PIAS1 or PIAS3 proteins (Fig. 2). Previously, regulation of gene transcription by the PIAS family proteins has been shown to occur via interactions with factors that are upstream of TBP in the transcriptional initiation process, some of which have also been shown to directly interact with TBP. These include the general co-activator CBP/p300 (44), the transcriptional factors MSX2 (15), p53 (25,45), p73 (46), and others. To our knowledge, this is the first report of a direct interaction between PIAS proteins and a component of the basal transcription machinery.
The interaction between TBP and PIAS proteins involves the conserved C-terminal core of TBP and a C-terminal 39-amino acid region found in common between PIAS1, PIAS3, PIASx, and PIASy proteins. Although the functions of the more highly conserved N-terminal region of PIAS proteins, including its SUMO E3-ligase activity, are well characterized, the C-terminal region is less well understood (21). The C-terminal regions of individual PIAS proteins have been shown to mediate protein-protein interactions with some regulators (21). However, alignment shows that only a region within the TBP-interacting domain described here (Fig. 4B) is shared among all seven PIAS proteins (Fig. 7). Our interpretation is that all PIAS proteins share the property of interacting directly with TBP; the remainder of the PIAS C-terminal domains determine the interactions that distinguish the activities of each family member from the others.
Another activity shared by all PIAS proteins is that they function as SUMO E3-ligases through amino acids in the conserved N terminus (12,FIGURE 8. Interaction of PIASx and PIASy with TBP. Two-hybrid assays were performed as in previous figures. PIASx␤  and PIASy 1-507 interacted with TBP-FL by yeast two-hybrid analysis (sectors 1 and 3, respectively) and neither was autoactive when co-expressed with the empty bait plasmid (sectors 2 and 4, respectively). 18,24,27). PIAS1-dependent sumoylation of the ZNF76 prevents ZNF76 from interacting with TBP (11). Our data suggest that the interaction of the PIAS1 C terminus with TBP could also influence the interaction of ZNF76 with TBP. The localization of SUMO E3-ligase activity and TBP-binding activity to opposite ends of PIAS proteins suggests that PIAS proteins might "dock" at TBP and sumoylate transcription factors at the promoter. Further investigation of the interactions between TBP, PIAS proteins, and transcription factors like ZNF76 will be required to determine which interactions are cooperative, which are antagonistic, and how these proteins interact at the TBP interface to regulate the expression of specific target genes.