Med9/Cse2 and Gal11 Modules Are Required for Transcriptional Repression of Distinct Group of Genes*

The yeast Mediator is composed of two subcomplexes, Rgr1 and Srb4, known to be required for diverse aspects of transcriptional regulation; however, their structural and functional organizations have not yet been deciphered in detail. Biochemical analyses designed to determine the subunit composition of the Rgr1 subcomplex revealed that the regulator-interacting subcomplex has a modular structure and is composed of the Gal11, Med9/Cse2, and Med10/Nut2 modules. Genome-wide gene expression and Northern analyses performed in the presence or absence of the various Mediator modules revealed a distinct requirement for the Gal11, Med9/Cse2, and Med10/Nut2 modules in transcriptional repression as well as activation. GST pull-down analysis revealed that the transcriptional repressor Tup1 binds to distinct but overlapping regions of the Gal11 module that were shown previously to be transcriptional activator binding sites. These data suggest that competition between transcriptional activators and repressors for a common binding site in the Mediator and distinct conformational changes in the Mediator induced by repressor binding may underlie the mechanism of transcriptional repression in eukaryotes. The Mediator complex was identified first as a coactivator transcriptional in the yeast Saccharo-myces cerevisiae Mediator associates tightly with to form

The yeast Mediator is composed of two subcomplexes, Rgr1 and Srb4, known to be required for diverse aspects of transcriptional regulation; however, their structural and functional organizations have not yet been deciphered in detail. Biochemical analyses designed to determine the subunit composition of the Rgr1 subcomplex revealed that the regulator-interacting subcomplex has a modular structure and is composed of the Gal11, Med9/Cse2, and Med10/Nut2 modules. Genome-wide gene expression and Northern analyses performed in the presence or absence of the various Mediator modules revealed a distinct requirement for the Gal11, Med9/Cse2, and Med10/Nut2 modules in transcriptional repression as well as activation. GST pull-down analysis revealed that the transcriptional repressor Tup1 binds to distinct but overlapping regions of the Gal11 module that were shown previously to be transcriptional activator binding sites. These data suggest that competition between transcriptional activators and repressors for a common binding site in the Mediator and distinct conformational changes in the Mediator induced by repressor binding may underlie the mechanism of transcriptional repression in eukaryotes.
The Mediator complex was identified first as a coactivator essential for transcriptional activation in the yeast Saccharomyces cerevisiae (1,2). Mediator associates tightly with RNA polymerase II (Pol II) 1 to form a Pol II holoenzyme (h-Pol II) and plays a pivotal role in diverse aspects of transcriptional regulation. The search for homologous Mediator complexes in the mammalian system identified a number of protein complexes that contain several yeast Mediator homologs that function as both positive and negative regulators of transcription in vitro (3)(4)(5)(6)(7).
The Mediator complex is required in general for the regulation of Pol II transcription. However, certain Mediator subunits function in an activator-specific manner to modulate the expression of distinct subsets of genes (8 -10), suggesting the presence of functional modules within the Mediator complex.
Differential dissociation of Mediator components by high urea treatment (11) and compositional analysis of mutant h-Pol II complexes (8,12,13) revealed that Mediator subunits with similar genetic properties form distinct modular subassemblies. Recently, the modular structure of Mediator was visualized by low resolution structural analysis (14). Electron microscopy and image processing of single particles of Mediator and h-Pol II revealed that Mediator alone assumes a compact structure; however, at high pH or in the presence of Pol II, Mediator displays an extended conformation with three domains (head, middle, and tail) forming an envelope that wraps around Pol II (14).
Because purified yeast h-Pol II in conjunction with basal transcription factors can support activated transcription in a well defined yeast transcription system (1,2), it was hypothesized that gene-specific activators communicate either directly or indirectly with Mediator to recruit Pol II to the promoter. This hypothesis was substantiated by the following observation. First, h-Pol II was shown to interact with the functional activation domains of the VP16 and Gcn4 transcriptional activator proteins (8,15,16). Subsequently, we demonstrated that a distinct Mediator module (the Gal11 module) is required for transcriptional activator binding to h-Pol II (8).
In addition to the essential role of Mediator complex in transcriptional activation, several lines of evidence indicate the involvement of Mediator in transcriptional repression as well. Genetic analyses revealed that several Mediator genes (such as SIN4, RGR1, and GAL11) play a role in the transcriptional repression of the HO, SUC2, and IME1 genes (17)(18)(19). Alleles of SIN4 and GAL11 were identified as suppressors of (i) an NC2 (Dr1/DRAP1) repressor mutation and (ii) defective silencing at the HMR mating-type locus in yeast (20). In addition, yeast cells lacking the MED1 gene, which encodes a subunit of the Rgr1 subcomplex, are viable but show a complex phenotype that includes partial defects in both repression and induction of GAL gene transcription (21). Finally, deletion of the HRS1 gene, which encodes a subunit of the Gal11 module, causes derepression of the GAL1, PHO5, and HSP26 genes (22), and Hrs1 protein was shown to interact with Tup1 protein in vitro (23). Taken together, these results suggest that several proteins in the Rgr1 subcomplex are required for transcriptional repression.
To decipher the modular structure of Mediator and identify the functions associated with each module, we used a ureainduced disruption assay and reconstitution experiment with recombinant proteins to analyze the protein-protein interactions among the Mediator subunits and examined the role of each subunit in transcriptional regulation with genome-wide expression analysis. CUP1 promoter region from Ϫ1 to Ϫ1000 (translation initiation site is ϩ1) was amplified using the polymerase chain reaction (PCR) (XhoI and HindIII sites are in the 5Ј-and 3Ј-end primers, respectively). A DNA fragment containing the LexA DNA binding domain, a multicloning site, and the ADH1 terminator was isolated from pEG202 (CLON-TECH, Palo Alto, CA) by using HindIII and XbaI. These fragments were introduced into the XhoI and XbaI sites of pRS313 (24) and named pCE1. The SSN6 and SRB10 coding regions were cloned by in vivo gap repair (25) and inserted in-frame into the multicloning site of pCE1 to create pCESSN6 and pCESRB10, respectively. pAJ201 (26), which had four LexA binding sites in the upstream region of the CYC1 core promoter, was used as a reporter construct. The TUP1 open reading frame was cloned by in vivo gap repair and then introduced into pGEX-4T-1 in-frame to construct the GST-Tup1 fusion construct. GST-Tup1 was expressed in bacterial strain DH5␣ and then purified. The Tup1 open reading frame was also cloned into pHAFASTBACHTb (see "Construction of Recombinant Baculoviruses") in-frame to construct an Nterminal hemagglutinin (HA) epitope-tagged Tup1 protein, and the virus expressing the HA-tagged Tup1 was screened as described in the baculovirus expression system manual (Life Technologies, Inc.). The HA-tagged Tup1 was purified with Q-Sepharose FF and HA.11 Ab matrix (BabCo, Richmond, CA).

RNA Preparation and Northern Analysis-Various
Mediator mutant cells (10,27) containing the copper-inducible repressor expression system and the reporter construct (Table I) were grown in an appropriate medium to early exponential phase at 30°C (A 600 ϭ 0.5), divided into two aliquots, and then allowed to grow for an additional 2.5 h in the presence or absence of 0.1 mM Cu 2ϩ ion. After the copper induction, the cells were shifted to 37°C and then allowed to grow for an additional 2.5 h. Total RNA was prepared as described previously (28). To prepare the probe for Northern blot analysis, the BamHI-SacI DNA fragment of LacZ and the LexA DNA binding region were isolated and 32 P-labeled by Klenow fragment in the presence of random hexanucleotide. Specifically hybridized signals were quantified using a PhosphorImager (Molecular Dynamics, Sunnyvale, CA) and the associated software.
Construction of Recombinant Baculoviruses-Recombinant baculoviruses expressing each Mediator component were created using the Bac-to-Bac baculovirus expression systems (Life Technologies, Inc.). To construct a baculovirus that expresses an N-terminal HA-tagged Mediator protein, the HA epitope sequence (YPYDVPDY, 5Ј-TACCCATAC-GACGTGCCAGACTAC-3Ј) was inserted into the NcoI and EcoRI sites in pFASTBACHTb (Life Technologies, Inc.) to construct pHAFAST-BACHTb. All the Mediator genes were cloned by in vivo gap repair, verified by nucleic acid sequencing, and introduced into pFAST-BACHTb and pHAFASTBACHTb expression vectors in-frame. Recombinant baculovirus clones for each of the Mediator proteins were isolated and further amplified until a titer of Ͼ1 ϫ 10 8 plaque-forming units/ml was achieved according to the baculovirus expression system manual (Life Technologies, Inc.).
Expression and Purification of Mediator Module Complexes from Sf9 Cells-Sf9 cells were cultured in Grace's medium containing 10% fetal calf serum (Hyclone, Logan, UT). Each recombinant virus was used for infection at a multiplicity of infection of 2. Multiple recombinant viruses were infected together at the same multiplicity of infection for coexpression experiments; Sf9 cells were infected with virus mixtures A (HA-Med1, Med4, and Med9/Cse2) and B (HA-Srb7, Med7, and Med10/Nut2) to make Med9/Cse2 and Med10/Nut2 modules, respectively. The cells were harvested by centrifugation for 5 min at 500 ϫ g after 48 h of incubation, washed once with a phosphate-buffered saline solution, resuspended in 10 ml of IP300 buffer (20 mM Tris acetate (pH 7.9), 10% glycerol, 0.1 mM EDTA, 0.2% Nonidet-400, and 300 mM potassium acetate), and incubated on ice for 30 min. Next, the lysate was disrupted by Dounce homogenization (20 strokes with a "B" pestle) and centri-fuged at 100,000 ϫ g for 1 h at 4°C. The whole-cell extract was diluted with the same volume of IP buffer (same as IP300 buffer with the exclusion of 300 mM potassium acetate) and then applied to a Q-Sepharose FF column (10 ml) equilibrated with D150 buffer (20 mM Tris acetate (pH 7.9), 0.1 mM EDTA, 0.01% Nonidet P-40, 10% Glycerol, plus 150 mM potassium acetate). The column was washed with 2 volumes of D300 buffer and then eluted with D600 buffer. Fractions containing recombinant Mediator proteins were identified by immunoblot analysis with anti-Med6 antibodies (29), dialyzed against IP150 buffer, and then incubated for 8 h at 4°C with 100 l of HA.11 Ab affinity matrix (BabCO). After binding, the resin was washed three times for 5 min at 4°C with 500 l of IP1500 buffer. The proteins bound to the resin were eluted by washing twice with IP150 buffer containing 0.5 mg/ml HA peptide (YPYDVPDYA, BabCO) at 4°C for 30 min. The purity of each eluate was analyzed by SDS-polyacrylamide gel electrophoresis followed by silver staining and immunoblotting.
Genome-wide Gene Expression High Density Oligonucleotide Array (HDA) Analysis-Wild-type and mutant yeast strains were grown to an A 600 of 0.5 in YPD at 30°C. For temperature-sensitive (ts) mutants, the cultures were diluted with an equal volume of YPD and allowed to grow for 45 min at 37°C. Total RNA was isolated, and the genomic expression profile of each strain was determined using four Affymetrix Gene-Chip arrays. Poly(A) mRNA was isolated, converted to biotin-cRNA, and hybridized to the GeneChips, which were then washed, stained, and scanned as described previously (9). Five poly(A)-tagged control RNAs were added to equal amounts of total RNA from each preparation. The levels of these controls were then used to normalize each wild-type and mutant expression profile to total RNA. Two independent experiments were performed for each wild type versus mutant comparison. Individual mRNA levels were scored if the computer algorithm analyzing the scanned results (30) returned a "Present" call in both of the two wild-type and the two mutant expression profiles for that gene or if the expression levels of the gene changed in the same direction and were greater than background levels in both wild-type and mutant comparisons. A decrease was called if an mRNA dropped more than 2-fold in both comparisons. An increase was called if an mRNA increased more than 2-fold in both comparisons.

Modular Structure of the Mediator Complex-
The observations that the Mediator has a structure that is composed of two functionally distinct subcomplexes (Srb4 and Rgr1) (11) and that the Gal11, Med2, Hrs1, and Sin4 proteins form a functional module that bind to the C-terminal region of Rgr1 (12) prompted us to examine whether Med9/Cse2 and Med10/Nut2 also form modular structures with distinct functions. To this end, we first purified h-Pol II from wild-type yeast and mutant yeast strains that carried either the med9⌬ (MED9/CSE2 deletion) or med10ts (temperature-sensitive) mutations. Subsequent immunoblot analysis of h-Pol II purified from the wildtype and mutant yeast strains revealed that the composition of h-Pol II from the med10ts strain was indistinguishable from that of wild type (data not shown). However, the h-Pol II from the med9⌬ strain was completely devoid of Med9/Cse2 and Med1, deficient in Med4, and retained all the other Mediator components but at a slightly reduced level (Fig. 1A). This result indicates that Med9/Cse2, Med1, and probably Med4 associate to form a modular structure.
To confirm the above result, we immobilized the Mediator complex with anti-Med9/Cse2 Ab-coupled agarose beads and analyzed the polypeptides retained after a urea wash. The 1 M urea wash removed mainly the polymerase subunits associated with the Mediator complex, whereas a 2 M urea wash removed the Srb4 subcomplex from the beads without affecting the association of the Rgr1 subcomplex with the beads (Fig. 1B). When the beads were washed extensively at an even higher urea concentration (3 M), only Med1 and Med4 (in addition to Med9/Cse2) were retained on the beads.
To determine whether direct interactions exist among Med9/ Cse2, Med1, and Med4, Sf9 cells were co-infected with baculoviruses that contained Med4, Med9/Cse2, and HA-tagged Med1 expression constructs. The Sf9 lysate was then analyzed by immunoprecipitation with anti-HA monoclonal Ab-coupled agarose beads. Immunoblot and silver-staining analysis of the immunoprecipitate showed that Med4 and Med9/Cse2 proteins co-precipitated with the HA-Med1 protein via direct interaction, whereas GST protein added to the extracts remained in the supernatant ( Fig. 2A and data not shown). These observations demonstrate that Med9/Cse2, Med1, and Med4 also form a stable modular structure, as do the components of the Gal11 module.
Although the existence of the Gal11 and Med9/Cse2 modules was demonstrated with several experimental techniques, we were unable to observe a Med10/Nut2 containing module either by compositional analysis of med10ts/nut2ts h-Pol II or using the urea dissociation method (because of the relatively low binding affinity of anti-Med10/Nut2 Ab; data not shown). Therefore, we examined the physical interactions in Sf9 cells of Med10/Nut2 with the rest of the polypeptides (Med7 and Srb7) in the Rgr1 subcomplex. Immunoprecipitation of Sf9 cell extracts in which Med7, Med10/Nut2, and Srb7 were coexpressed with anti-Med10 Ab showed that Med7 and Med10/Nut2 associated directly with Srb7 in the absence of other Mediator proteins, thus demonstrating the presence of a Med10/Nut2 module (Fig. 2B). Therefore, Mediator appears to contain at least three modular structures (the Gal11, Med9/Cse2, and Med10/Nut2 modules) that are tethered independently to the Rgr1 protein in the Rgr1 subcomplex.
Differential Requirement for Each Mediator Subunit in Transcriptional Activation and Repression-To identify the distinct roles of the Med9/Cse2 and Med10/Nut2 modules in transcrip-tional regulation, we analyzed the genome-wide gene expression patterns in yeast strains that contained mutations in the MED9/CSE2 and MED10/NUT2 genes using HDAs for 6181 yeast genes. Total RNA was isolated from med9⌬/cse2⌬ and med10ts/nut2ts mutant yeast strains and their wild-type counterparts 45 min after a shift to the nonpermissive temperature, and these RNA preparations were hybridized to the HDAs. 2 Among the 5376 yeast genes, the expression levels of which were determined for the med10ts mutant strain, 2899 genes (54%) were down-regulated more than 2-fold under restrictive conditions, whereas only 56 gene were up-regulated (1%) (Table II). In contrast, the number of genes, the expression of which was affected in the med9⌬ mutant strain was quite small (384 of 4806 genes down-or up-regulated); however, a significant number of genes (209 of 384 genes affected) were upregulated (Tables II and III). Therefore, MED10/NUT2 seems to be required for the transcriptional activation of a large number of genes, whereas MED9/CSE2 is generally dispensable for transcriptional activation but required for the mediation of transcriptional repression for a subset of genes.
To examine whether the Med9/Cse2 module is involved in transcriptional repression as suggested by the HDA analysis, we used Northern analysis to examine the effect of med9⌬ mutation on the expression of genes that are repressed under a rich growth condition (FLO1, SUC2, SOL4, and HXK1) and compared their effect with that of the gal11⌬ mutation. When the cells were grown in YPD, expressions of the SOL4, SUC2, and HXK1 genes were repressed completely (SUC2) or at least several fold (SOL4 and HXK1) in wild-type cells. However, the med9⌬/cse2⌬ mutant was defective in transcriptional repression of SOL4 and HXK1 (4-and 3.5-fold increases in transcripts, respectively), whereas gal11⌬ mutants showed defects in the transcriptional repression of the SOL4 and SUC2 genes (12-and 10-fold increases, respectively; Fig. 3). In addition, Med9 is required for the transcriptional repression of FLO1, but Gal11 is dispensable for the FLO1 repression (Fig. 3). Therefore, Gal11 and Med9 proteins are required for transcriptional repression of distinct group of genes and seem to constitute repressor interaction modules of the Mediator complex.
In addition to the MED9/CSE2 and GAL11 results shown here, certain alleles of other Mediator components (SIN4 and RGR1) have been implicated in transcriptional repression (18,19,22,31,32) 4A). In the absence of copper, LexA-Srb10 and LexA-Ssn6 were expressed at a barely detectable level from the CUP1 promoter; therefore a moderate level of basal expression of the reporter gene (LacZ) was detected in all of the strains tested (Fig. 4B). Upon induction with copper, the highly expressed repressor fusion proteins were recruited to the reporter gene promoter (CYC1 promoter fused to LacZ), which resulted in a 9-fold repression of LacZ transcription in wild-type cells. The inactivation of MED10/NUT2 activity did not affect the transcriptional repression of LacZ by either of the repressor fusion proteins. However, when the transcriptional repressors were overexpressed in the absence of the Med9/Cse2 or Gal11 proteins, neither LexA-Srb10 nor LexA-Ssn6 was able to repress LacZ expression. These results suggest that Med9/Cse2 and Gal11, but not Med10/Nut2, are required directly for transfer of the repression signal from gene-specific transcriptional repressor proteins to the basal transcription machinery. Therefore, the Med9/Cse2 and Med10/Nut2 modules seem to play major roles in the transcriptional repression and activation processes, respectively, whereas the Gal11 module is required for both types of gene regulation.
Transcriptional Repressor Binding Module of Mediator-Our previous studies (8,10) and the genome-wide gene expression analysis presented here show that two Mediator modules (the Gal11 and Med10/Nut2) are required for distinct aspects of transcriptional activation. The Gal11 module functions as binding sites for transcriptional activator proteins, whereas the Med10/Nut2 module functions at the post-activator binding stages to modify and relay the activation signal to the basal transcription machinery (8). The observation that transcriptional repression also requires two different Mediator modules (the Gal11 and Med9/Cse2 modules) suggests that a similar mechanism may work to mediate the signal between repressors and the basal transcription machinery. To test this idea, we analyzed the physical interactions between transcriptional repressor Tup1 and Mediator modules using a GST pull-down assay. The Med9/Cse2 and Med10/Nut2 modules and the Srb4 subcomplex were purified with anti-HA Ab-coupled beads, and the HA-agarose eluates were applied to GST-or GST-Tup1agarose beads. The protein bound to the GST and GST-Tup1 beads were subjected to immunoblot analysis with anti-HA antibodies that recognize the HA-tagged versions of Med1, Srb7, and Srb4 in the Med9/Cse2 and Med10/Nut2 modules and the Srb4 subcomplex, respectively. This analysis revealed that none of the Mediator components tested bound to the Tup1 protein (Fig. 5A). However, when we tested Rgr1 and each of the Gal11 module proteins for Tup1 interaction using the GST pull-down assay, we detected strong Tup1 binding to the Rgr1 and Med2 in addition to the previously reported Hrs1 (23) but not to the Sin4 and Gal11 proteins (Fig. 5B). Therefore, Tup1 interacts with the Mediator mainly through the Rgr1 and Gal11 module proteins, as do certain transcriptional activator proteins (VP16, Gal4, and Gcn4).
Although some of the individual Mediator subunits required for Tup1 interaction (Rgr1 and Med2) are distinct from those for the transcriptional activator proteins we tested (VP16, Gal4, and Gcn4), the fact that both types of transcriptional regulators interact with Mediator through the Gal11 module is intriguing. In particular, the interaction of Hrs1 both with Gcn4 and Tup1 suggests that certain transcriptional activator and repressor proteins may share Mediator binding sites. To investigate this possibility, we deciphered the transcriptional regulator binding regions of Hrs1 using a series of GST-fusion proteins that contained various Hrs1 fragments in GST pulldown experiments. The GST pull-down assays revealed amino acids 180 -343 of Hrs1 interact specifically with Gcn4, and amino acids 83-179 interact specifically with Tup1. The Nterminal region of Hrs1 that includes amino acids 1-82 interacted with both Gcn4 and Tup1 proteins (Fig. 5C). These results demonstrate that Gcn4 and Tup1 each utilize distinct but overlapping regions of Hrs1 as the Mediator binding targets.

Disruption of Mediator Complex by a Transcriptional
Repressor Protein-The observation that the Mediator binding sites for a transcriptional activator (Gcn4) and a transcriptional repressor (Tup1) partially overlap suggests that competition between the two types of transcriptional regulators for a common binding site in the Mediator complex may play a significant role in mechanisms of transcriptional regulation. To test this hypothesis, we challenged the Gcn4-Mediator interaction with increasing amounts of Tup1 protein or bovine serum albumin as follows. Purified Mediator complex was first bound to GST-Gcn4 fusion protein beads. After extensive washing, increasing amounts of Tup1 or bovine serum albumin protein were added to the beads, and the amount of Mediator complex retained in the beads or released to the supernatant was monitored by immunoblot assay with anti-Srb4 Ab. No detectable change in the amount of Mediator bound to GST-Gcn4 was observed by the addition of bovine serum albumin protein (Fig.  6). In contrast, the addition of a 10-fold excess of Tup1 protein caused a significant reduction in the amount of Mediator bound to the GST-Gcn4 beads along with a significant increase in the amount of Mediator proteins released to the supernatant. This result suggests a mechanism in which a transcriptional repressor blocks the access of a transcriptional activator protein to Mediator complex to repress transcription.

DISCUSSION
Modular Structure of Mediator Complex-Mediator was identified first as a coactivator that enables the basal transcription machinery to respond to traditional gene-specific transcriptional activator proteins (33,34). Subsequent biochemical purification of the Mediator complex revealed its role as an intermediary molecule between transcriptional regulators and Pol II (1). The intermediary role of Mediator was also shown by genetic studies that revealed the existence of two classes of Mediator genes: one that encodes Pol II-interacting proteins and another that interacts with gene-specific transcriptional factors (11,18,19,31,35).
The structural organization of the Mediator complex also reflects the two aspects of Mediator function, as suggested by several experiments, some of which are in this study. First, the co-disappearance of several polypeptides from the Mediator complex in med9⌬/cse2⌬, and gal11⌬ strains indicates that a Number of genes that gave significant hybridization signals above background.
b The number of genes that are up-regulated or down-regulated in each Mediator mutant strain was scored. these subunits are connected to the rest of the complex via the Med9/Cse2 or Gal11 proteins. Second, high concentration urea treatment of Mediator complex immobilized on beads revealed not only the Srb4 and Rgr1 subcomplexes but also the Med9/ Cse2 module. Third, coimmunoprecipitation from Sf9 cells of Mediator subunits that belong to the same module demonstrates that the Mediator modules or subcomplexes can be assembled with recombinant proteins. Lastly, electron microscopy and image processing of Mediator revealed the modular structure of the complex, which is composed of head, middle, and tail domains (14). In particular, the absence of the tail domain in the sin4⌬ Mediator indicates that the tail domain corresponds to the Gal11 module. Because the Gal11 module is connected to the C-terminal region of Rgr1, the tail and middle domains must constitute the Rgr1 subcomplex. Therefore, the Srb4 subcomplex appears to form the head structure, which is connected to Pol II. An extensive genome-wide two-hybrid screen in yeast (36) also identified physical interactions among the subunits of each Mediator subcomplex (Srb5-Med8 and Srb6-Med11 interactions for the Srb4 subcomplex; Med7-Med4, Med7-Srb7, and Med9/Cse2-Med4 interactions for the Rgr1 subcomplex), which are in good agreement with the physical interactions identified in this study.
Among the modular structures of the Mediator complex, only two components of the middle domain (the Rgr1 and Med10/ Nut2 module proteins) and Med6 of the head domain are conserved evolutionarily (37). Even in the Rgr1 protein, only the N-terminal region of the protein is conserved. The C-terminal region, where the species-specific tail domain is connected, shows a high sequence divergence (5). Therefore, the head and tail domains of the Mediator complex seems to have evolved to accommodate the diversity and specificity of higher eukaryotic transcriptional regulators and Pol II, whereas the conserved middle domain contains the basic architecture of the Mediator complex.
Function of Repressor-specific Modules of Mediator-The modular structure of the Mediator complex suggests that each module is responsible for distinct functions. Transcriptional activation requires the Gal11 and Med10/Nut2 modules for the activator binding and post-activator binding processes, respectively. Similarly, the Tup1 repressor also requires two separate modules, one for binding to Mediator and the other for postrepressor binding processes (transfer of the repressive signal to inhibit Pol II transcription of target genes). Intriguingly, the Gal11 module seems to be a common binding site for transcriptional activator and repressor proteins. Therefore, a simple competition for Mediator-binding sites may occur between transcriptional activator and repressor proteins and may ultimately determine the mode of transcriptional regulation. In addition to the Tup1-Mediator interaction shown here, biochemical analysis showed that Sfl1 coimmunoprecipitates from yeast cell extracts with Srb9, Srb11, Sin4, and Rox3 (38). Sfl1 is a known transcriptional repressor and regulates SUC2 transcription through a repressor-binding site located immediately 5Ј to the TATA box. Therefore, the direct interaction between Required for pre-mRNA splicing 3 STB2 Sin3p-binding protein 3 YDR111C Similar to alanine aminotransferase 3 a The fold change in gene expression in the mutant as compared to its isogenic wild-type strain.

FIG. 3. Transcriptional repression defects in Mediator mutant yeast strains. Northern blot analysis of Mediator mutants is shown.
To determine the effect of Mediator mutations on the transcriptional repression, wild-type (wt), med9/cse2 null, and gal11 null strains were grown in YPD at 30°C to the early exponential phase and then shifted to 37°C for another 2.5 h. Nucleic acid probes for Northern blotting are shown at the right.
Sfl1 and Mediator may be the main mechanism of transcriptional repression for SUC2.
If both transcriptional activator and repressor proteins interact with Mediator through the Gal11 module, how can activators and repressors regulate transcription in opposite directions? One of the mechanisms by which regulatory proteins activate or inhibit transcription is through their interaction with additional coactivator or corepressor complexes. In addition to its interaction with Mediator, the Ssn6-Tup1 repressor complex interacts with histones H3 and H4 in chromatin to form a repressive chromatin structure over the promoter region, and the interaction of Ssn6-Tup1 with Mediator may block the access of transcriptional activator proteins to the promoter (39). Therefore, it would be interesting to determine whether this repressor-Mediator complex interacts with additional repressor complexes such as the histone deacetylation complex to repress target gene expression.
Our results suggest another mechanism for transcriptional repression. Although transcriptional activator and repressor proteins may share some binding sites in the Mediator complex, the precise collection of Mediator binding surfaces required for each type of regulatory protein are distinct. Therefore, in addition to the competition between activator and repressor proteins for binding to Mediator, the binding of a transcriptional repressor to a distinct surface of the Gal11 module may cause a specific conformational change in the Mediator complex that results in transcriptional repression. This may result in the dissociation of the Mediator complex from other transcriptional factors, which in turn destabilizes the preinitiation complex formation at the promoter region as shown by the decreased occupancy of TATA-binding protein and h-Pol II at the Tup1-bound promoters (40).
Contrary to the most common function of the Gal11 module in the binding of both types of transcriptional regulators, the Med9/Cse2 module seems to be involved mainly in transcriptional repression. Because transcriptional repressor proteins we tested do not interact directly with the Med9/Cse2 module, we hypothesized that it functions at the post-repressor binding Expression of the LexA-repressor fusion proteins was controlled by a copperinducible promoter (CUP1). The effect of repressor binding on CYC1 promoter activity was monitored. B, Mediator mutant strains containing the copper-inducible LexA-Srb10 (YSJW10, YSJ910, YSJG1110, and YSJ1010) or LexA-Tup1 (YSJWS6, YSJ9S6, YSJG11S6, and YSJ10S6) expression systems along with a LacZ reporter gene construct driven by the CYC1 core promoter and a LexA binding site were cultured in the presence (ϩ) or absence (Ϫ) of 0.1 mM copper ion and then shifted to the nonpermissive temperature. mRNA was prepared from each strain and analyzed by Northern blot with the specific probe indicated at the right. As a control, the ACT1 transcript (AIP1) level is shown. wt, wild type. stages, as does the Med10/Nut2 module during transcriptional activation, or Med9/Cse2 may serve as a binding site for a different type of transcriptional repressor protein. Identification of the precise function of the Med9/Cse2 module will be necessary to determine the mechanism by which Mediator complex participates in the transcriptional repression process.