Med19(Rox3) Regulates Intermodule Interactions in the Saccharomyces cerevisiae Mediator Complex*

The Saccharomyces cerevisiae Mediator is a 25-subunit complex that facilitates both transcriptional activation and repression. Structural and functional studies have divided Mediator subunits into four distinct modules. The Head, Middle, and Tail modules form the core functional Mediator complex, whereas a fourth, the Cyc-C module, is variably associated with the core. By purifying Mediator from a strain lacking the Med19(Rox3) subunit, we have found that a complex missing only the Med19(Rox3) subunit can be isolated under mild conditions. Additionally, we have established that the entire Middle module is released when the Δmed19(rox3) Mediator is purified under more stringent conditions. In contrast to most models of the modular structure of Mediator, we show that release of the Middle module in the Δmed19(rox3) Mediator leaves a stable complex made up solely of Head and Tail subunits. Both the intact and Head-Tail Δmed19(rox3) Mediator complexes have defects in enhanced basal transcription, enhanced TFIIH phosphorylation of the CTD, as well as binding of RNA Pol II and the CTD. The largely intact Δmed19(rox3) complex facilitates activated transcription at levels similar to the wild type Mediator. In the absence of the Middle module, however, the Δmed19(rox3) Mediator is unable to facilitate activated transcription. Although the Middle module is unnecessary for holding the Head and Tail modules together, it is required for the complex to function as a conduit between activators and the core transcription machinery.

The Mediator complex is a conserved interface between gene-specific regulatory proteins and the general transcription apparatus of eukaryotes at transcription initiation (1). Saccharomyces cerevisiae Mediator is comprised of 25 subunits (2). Structural (3), biochemical (2), genetic (4), and genomic (5) studies have enabled the provisional assignment of yeast Mediator subunits into four distinct modules. Each of these modules is thought to have a unique role in the structure and function of the entire complex. The core functional Mediator complex was initially defined as the minimal intact complex, purified from a wild type strain, capable of facilitating activated transcription in a system reconstituted from highly purified general transcription factors (6,7). Single-particle electron microscopic analysis of core Mediator identified three areas of density that were referred to as the Tail, Middle, and Head modules of the complex (3). A fourth module, called the Cyc-C module, is found to be variably associated with the core Mediator modules and may regulate its function (8). The modular structure of the yeast complex appears to be conserved in Mediator complexes purified from metazoan cells (9,10). Biochemical and genetic studies have helped assemble a reasonable model for the composition of, and the interactions within, the structural modules of Mediator (summarized in Refs. 3 and 11; see Fig. 7A). Little, however, is known about how these modules are held together, what the functional implications of these intermodular interactions are, and how these interactions are regulated. Genetic, genomic, and functional assays in vitro have provided support for the model of the modular structure of Mediator but also raised further questions on how the modules function together.
Mediator, purified from wild type cells, has three activities in vitro: the ability to facilitate activated transcription, the ability to enhance basal transcription, and the ability to enhance phosphorylation of the C-terminal domain (CTD) 2 of RNA Pol II by TFIIH (7). Mutations in the Tail module subunits of Mediator generally lead to defects in activated transcription but have little effect on enhancement of basal transcription or CTD phosphorylation (12,13). Mutations in the Head module proteins have broad effects on all three functions in vitro (13,14). Middle module mutant Mediators, ⌬med9 and med10ts, appear to have defects in basal and activated transcription (15), although the structural composition of these complexes has not been evaluated.
A recent genomic study and a wealth of data from elegant yeast genetic screens have outlined key distinctions between the subunits in the Middle module and those in the Tail and the Head modules of Mediator. The Tail module subunits are largely encoded by nonessential genes, whereas the Middle and Head domains are encoded by roughly equal numbers of essential and nonessential genes. Genetic studies of various regulatory systems in yeast have revealed that subunits in the Tail and Head module have a positive influence on gene expression (4). In contrast, a number of genetic screens have shown that mutations in Middle, Cyc-C, and some Tail module subunits lead to derepression of heat shock factor basal transcription (16), maltose-induced genes (17), glucose-repressed genes (18), and the HO gene (19,20). Systematic gene expression microarray and clustering analyses of Mediator mutant strains found a correlation pattern among subunits of the complex (5). A strong pos-itive correlation was observed between members of Head and Tail modules, whereas a negative correlation was observed between members of these modules and the subunits of the Middle module. Given the separation of the Head and Tail  modules by the Middle module in the structural models, it is an  open question how the Head and Tail could function in concert with each other to exert a positive effect on gene expression on select sets of genes. Moreover, it is unclear how Middle module defects could lead to up-regulation of transcription, whereas these subunits are seemingly required to link two positive acting modules of the complex. A Mediator subunit that plays a key role in both activation and repression is Med19(Rox3).
Med19(Rox3) was originally identified in a search for mutants increasing aerobic expression of the CYC7 gene (21). Analysis of purified Mediator by mass spectrometry conclusively demonstrated that Med19(Rox3) was a subunit of Mediator (22). Med19(Rox3) was initially found to be encoded by an essential gene product; however, a strain made as part of the genome wide deletion library lacks MED19(ROX3) and is viable (EUROSCARF). med19(rox3) mutants have been shown to be defective for activation of Gal4 (23) and Gcn4 (24)-induced genes. On the other hand, med19(rox3) mutants also lead to derepression of heat shock (16), glucose-repressed (18), and HO genes (19). Gene expression microarray studies of a med19(rox3) deletion strain (25) and a med19(rox3) truncation strain (5) have shown that the expression of broad sets of genes are both up-and down-regulated. To gain further insight into the unusual functional properties of Med19(Rox3), we purified Mediator complexes from the ⌬med19(rox3) strain. Consistent with the Middle module phenotypes caused by med19(rox3) mutants, we found that Med19(Rox3) was critical for the stable association of the Middle module with the complex. In contrast to current models of Mediator structure, we found that release of the Middle module of Mediator left an intact and stable complex between Tail and Head module subunits. Characterization of this complex using binding and transcription assays in vitro has given us further insight into the coordination of Mediator structure and function.

MATERIALS AND METHODS
Yeast Strains-The ⌬med19(rox3) strain was obtained from EUROSCARF (accession number Y03119). To construct the MED18(SRB5)-FLAG-tagged ⌬med19(rox3) strain, the NatR marker was swapped for the KanMX4 marker in an otherwise wild type MED18(SRB5)-FLAG strain (SHY349). The strain yLM40 was made by transforming SHY349 (26) with an EcoRI fragment of the plasmid p4339 (27) containing the NatR maker and selecting for nourseothricin resistance. A triple FLAG epitope tag was placed at the C terminus of MED18(SRB5) in the ⌬med19(rox3) strain, by amplifying the 3Ј end of a previously FLAG-tagged copy of MED18(SRB5) from the genomic DNA of the yLM40 strain using the following primers: 5Ј-GGA-GGGTTCCTTTTAAAAGCA-3Ј and 5Ј-GAAGCAAATTGC-CAAACA-3Ј. The PCR product was then used to transform the ⌬med19(rox3) strain, and transformants were selected for nourseothricin resistance. The correct integration of the FLAG tag was confirmed by PCR and immunoblotting for FLAGtagged Med18(Srb5) in the strain yLM45.
Protein Purification-The untagged ⌬med19(rox3) complexes were prepared by modification of the methods used for conventional purification of the wild type Mediator complex (28). Specifically, the ⌬med19(rox3) strain was grown at 26°C to an A 600 of 2 in YPD (4% glucose). The lysis, as well as the Bio-Rex 70 and DEAE-Sepharose chromatography were carried out as previously described for the wild type complex (28). The hydroxyapatite column was run as previously described; however, there were two distinct peaks containing Mediator proteins instead of one. The first peak, containing only Middle module subunits, co-eluted at ϳ75 mM potassium phosphate, whereas a second peak containing subunits from all three modules co-eluted at ϳ125 mM potassium phosphate. The peak containing only Middle module subunits was pooled and applied to a Mono-Q 5/5 column (GE Biosciences) as previously described, except that a 26 column volume linear gradient was run from 200 to 1500 mM potassium acetate. The peak of Middle module subunits co-eluted at ϳ925 mM potassium acetate. This peak of Middle module subunits was pooled and applied to a Heparin 5PW column (TosoHass) at 200 mM potassium acetate. The Middle module subunits all flowed through the Heparin column. The flow-through fractions were pooled, concentrated, and used directly for immunoblots.
The hydroxyapatite peak containing subunits from all three modules was pooled and applied to a Mono-Q 5/5 column as previously described, except that a 26-column volume linear gradient was run from 200 to 1500 mM potassium acetate. Instead of a single peak of Mediator proteins, a peak of Head and Tail module subunits co-eluted at ϳ740 mM potassium acetate, and a separate peak containing Head, Middle, and Tail modules co-eluted at ϳ880 mM potassium acetate. The peak of Head and Tail module subunits was pooled and applied to a Heparin 5PW column as described (28). The Head and Tail module subunits co-eluted from the Heparin column at ϳ410 mM potassium acetate. The peak fractions were pooled and used directly for immunoblots. The peak from the Mono-Q column containing subunits from all three modules was pooled and applied to a Heparin 5PW column as described (28). The Head, Middle, and Tail module subunits co-eluted from the Heparin column at ϳ370 mM potassium acetate. The peak fractions were pooled, concentrated, and used directly for immunoblots. Purification of untagged wild type Mediator, as well as the general transcription factors, was carried out as previously described (28).
Affinity Purification of ⌬med19(rox3)-The MED18(SRB5)-FLAG-tagged wild type and ⌬med19(rox3) strains were grown at 26°C to an A 600 of 2 in YPD (4% glucose) and an extract prepared by the a blender/liquid N 2 lysis method (29). The extract was dialyzed against FLAG buffer (20 mM Hepes-KOH, pH 7.6, 0.01% Nonidet P-40, 10% glycerol, 300 mM KOAc, 1 mM dithiothreitol, and protease inhibitors) and adjusted to a final KOAc concentration of 300 mM by the addition of FLAG buffer with no KOAc. Approximately 20 mg of dialyzed extract was added to 200 l Anti-FLAG M2 agarose beads (Sigma), incubated for 2h at 4°C with rotation, and washed four times with 10 ml of FLAG buffer. Mediator was eluted with FLAG buffer containing 100 g/ml 3ϫ FLAG peptide (Sigma). To evaluate the stability of the wild type and ⌬med19(rox3) Mediators, the Med19(Rox3) Regulates Mediator Structure above protocol was modified to add a wash with 10 ml of FLAG buffer containing 1 M urea subsequent to the 4 ϫ 10-ml washes with FLAG buffer. Before the FLAG peptide elution, a 10-ml wash with FLAG buffer followed the wash with 1 M urea.
To obtain a highly purified preparation of the ⌬med19(rox3) complex containing only Head and Tail module subunits, a lysate was prepared from the MED18(SRB5)-FLAG-tagged ⌬med19(rox3) strain. Starting from this lysate, a complex containing only Head and Tail module proteins was purified as described above for the untagged complex. After the Mono-Q step, the peak fractions containing the Head-Tail (HT) complex were pooled, dialyzed against FLAG-buffer, and adjusted to a final salt concentration of 300 mM KOAc. This sample was than applied to the FLAG-agarose beads, which were washed and eluted as described above.
Transcription and Kinase Assays-Basal and activated transcription reconstituted from purified yeast general transcription factors was measured using the G-less cassette assay as previously described (7) with the following modifications. The final salt concentration in the reaction buffer was 165 mM KOAc, and the reactions were preincubated for 10 min in the absence of nucleotides followed by a 60-min reaction time after the addition of the nucleotides. The general transcription factors used for all of the reactions in this study came from identical aliquots of a single preparation. The kinase assays with purified Mediator complexes, TFIIK, and GST-CTD were performed as previously described (30).
Activator and RNA Pol II Binding Assays-Binding of wild type and ⌬med19(rox3)-HT purified Mediator complexes to immobilized GST activator fusion proteins was carried out as previously described (31). To assay the binding of core RNA Pol II to wild type and ⌬med19(rox3)-Head-Middle-Tail (HMT) Mediator, a Med18(Srb5)-FLAG-tagged version of each of these complexes were immobilized on FLAG-agarose as described above. The immobilized Mediator beads were equilibrated in FLAG buffer with 200 mM KOAc. Equal amounts of purified core RNA Pol II was added to the wild type and ⌬med19(rox3)-HMT Mediator beads. After incubation for 2 h at 4°C with rotation, the supernatant was collected, and the beads were washed four times with FLAG buffer with 200 mM KOAc. Wild type or ⌬med19(rox3)-HMT Mediator and any associated RNA Pol II was eluted with FLAG buffer containing 100 g/ml 3ϫ FLAG peptide (Sigma).

Deletion of Med19(Rox3) Leads to Dissociation of the Middle Module of Mediator, Leaving an Intact Mediator Subcomplex
Consisting of Head and Tail Modules-We undertook a conventional purification of complexes containing Mediator subunits from a ⌬med19(rox3) strain to determine the structural impact of deleting the Med19(Rox3) subunit from Mediator. We followed the purification by immunoblotting with antibodies against subunits in all three modules of core Mediator. From the ⌬med19(rox3) lysate through the first two ion exchange columns (Bio-Rex70 and DEAE-Sepharose), the strain showed an identical fractionation pattern to wild type (Fig. 1A). Fractionation of the DEAE pool containing ⌬med19(rox3) Mediator on hydroxyapatite, however, showed a pattern distinct from that of wild type. Instead of a single peak, we observed an initial peak of Middle module subunits, followed by a second peak containing subunits from all three modules (Fig. 1B). To further characterize the peak containing only Middle module subunits, we pooled these fractions and observed that the Middle module subunits co-elute over a Mono-Q column and all flow through a Heparin column (Fig. 1A). The flow through fractions were pooled, concentrated, and compared with wild type Mediator by immunoblotting. This Middle module complex (⌬med19 (rox3)-M) contained stoichiometric amounts of Middle module subunits (compared with wild type) and levels of Head and Tail module subunits that were below detection (Fig. 1D).
The second peak on the hydroxyapatite, containing subunits from all three modules, was pooled and further fractionated on a Mono-Q column (Fig. 1A). In contrast to wild type Mediator, two distinct peaks of Mediator eluted from the Mono-Q column. The first peak contained subunits from only the Head and Tail modules of the complex, whereas the second peak contained subunits from all three modules (Fig. 1C). Both of these peaks were individually pooled and applied to a Heparin column. Fractionation of the first pool (⌬med19(rox3)-HT) on Heparin resulted in a peak of Head and Tail module subunits co-eluting from the column (Fig. 1A), whereas fractionation of the second pool (⌬med19(rox3)-HMT) on Heparin resulted in a peak of subunits from all three modules co-eluting from the column (Fig. 1A). These two peaks were individually pooled, concentrated, and compared with wild type Mediator by immunoblotting. The analysis of the (⌬med19(rox3)-HT) complex showed stoichiometric amounts of Head and Tail module subunits (compared with wild type) and levels of Middle module subunits that were below detection (Fig. 1D). The analysis of the (⌬med19(rox3)-HMT) complex showed stoichiometric amounts of subunits from all three modules (compared with wild type) (Fig. 1D). Immunoblotting with an antibody against Med19(Rox3) confirmed that this subunit was absent from all Mediator preparations from the mutant strain (data not shown). Three distinct ⌬med19(rox3) Mediator complexes exist after purification. The first is a largely intact complex (⌬med19(rox3)-HMT), missing only the deleted subunit. The second and third (⌬med19(rox3)-M and ⌬med19(rox3)-HT) appear to result from the dissociation of the Middle module complex from the ⌬med19(rox3)-HMT Mediator. The existence of a stable ⌬med19(rox3)-HT complex was surprising; to further characterize the composition of the complex, we took an affinity purification approach.
To purify the ⌬med19(rox3)-HT complex to homogeneity, we grew cells from a Med18(Srb5)-FLAG-tagged wild type and ⌬med19(rox3) strain and followed the conventional purification protocol, as described above, through the Mono-Q stage. After separation on Mono-Q, the fractions containing the ⌬med19(rox3)-HT complex were pooled and subjected to a Med19(Rox3) Regulates Mediator Structure FEBRUARY 23, 2007 • VOLUME 282 • NUMBER 8 final purification step on FLAG-agarose. Immunoblotting of the pure ⌬med19(rox3)-HT complex shows that an affinity tag on the Head module could pull down stoichiometric amounts of Tail module proteins as compared with the wild type complex ( Fig. 2A). Immunoprecipitation of Head subunits by an antibody against a Tail subunit (␣-Med2) confirmed this result (data not shown). A silver stain analysis of purified ⌬med19(rox3)-HT complex shows that the complex appears to contain only Head and Tail module proteins in stoichiometric quantities (Fig. 2C). This result suggests that it is unlikely that additional "non-Mediator" proteins had associated with the complex and "compensated" for the absence of the Middle module subunits. We cannot, however, rule this possibility out given the differential staining of proteins by silver. Contrary to current models, the Head and Tail modules of Mediator can directly form a stable complex in the absence of a Middle module.
To determine the composition of the ⌬med19(rox3) Mediator under mild conditions, we purified the complex directly from lysate of the MED18(SRB5)-FLAG-tagged ⌬med19(rox3) strain on FLAG-agarose. During the conventional purification of the ⌬med19(rox3) Mediator, the complex is subjected to several highly stringent conditions that could lead to dissociation of weakly interacting subunits or modules. Immunoblotting of Mediator subunits from affinity-purified Med18(Srb5)-FLAGtagged wild type and ⌬med19(rox3) complexes show the presence of stoichiometric amount of subunits from all three mod-ules, as compared with conventionally purified wild type Mediator (Fig. 2B). Hence, it appears as if the largely intact ⌬med19(rox3)-HMT Mediator is the predominant form of the complex under mild conditions. This procedure was an effective means to purify the ⌬med19(rox3)-HMT complex to homogeneity. Silver stain analysis confirmed the purity and stoichiometry of the ⌬med19(rox3)-HMT Mediator complex purified under mild conditions (Fig. 2C).
Absence of Med19(Rox3) Destabilizes Attachment of Middle Module in Intact Mediator Complex-Because a majority of the ⌬med19(rox3) Mediator under mild conditions appears to be present in a largely intact form, we sought to determine the origin of the ⌬med19(rox3)-HT complex. We hypothesized that the absence Med19(Rox3) leads to the destabilization of the intact complex and that a stringent step during the conventional purification resulted in the partial displacement of Middle module from the complex. To test this idea we immobilized both Med18(Srb5)-FLAG-tagged wild type and ⌬med19(rox3)-HMT Mediator on FLAG-agarose and washed the complex with a buffer containing 1 M urea. Earlier studies have shown that the an immobilized wild type Mediator is resistant to 1 M urea but starts to dissociate at concentrations greater than 2 M (32). After treatment with 1 M urea, the Med18(Srb5)-FLAG-tagged wild type and ⌬med19(rox3)-HMT Mediator were eluted with FLAG peptide and subjected to immunoblotting. As shown previously, the wild type Mediator remained intact after this treatment (Fig. 3). However, treatment of the ⌬med19(rox3)-HMT complex with 1 M urea displaced the Middle module, leaving an intact complex of Head and Tail module proteins (Fig. 3). This result not only shows that the absence of Med19(Rox3) destabilizes the association of the Middle module with Mediator but also shows that the remaining ⌬med19(rox3)-HT complex is stable under conditions as stringent as 1 M urea.
The Middle Module of Mediator Is Required for Activated and Basal Transcription in Vitro-Because it appears that the Middle module is not required for holding the Tail and Head module together, it was of interest to determine whether the absence of the module impaired the function of Mediator, using in vitro transcription assays. Using equimolar amounts of purified wild type, ⌬med19(rox3)-HMT, and ⌬med19(rox3)-HT Mediator complexes (Fig. 2C), we monitored enhanced basal transcription, as well as Gcn4p and Gal4-VP16 activated transcription, using a system reconstituted with purified general transcription factors. Although the ⌬med19(rox3)-HMT was able to enhance basal transcription (Fig. 4, lane 7 versus lane 1), the level of enhancement was ϳ3-fold lower than wild type (Fig.  4, lane 4 versus lane 1). Higher amounts of ⌬med19(rox3)-HMT did not compensate for this defect (data not shown). Both Gcn4p and Gal4-VP16 were able to utilize the ⌬med19(rox3)-HMT complex as a co-activator. The fold activation by Gcn4 using the ⌬med19(rox3)-HMT Mediator (Fig. 4, lanes 7 and 9) was comparable with the wild type complex (Fig. 4, lanes 4 -6), whereas activation by Gal4-VP16 was reduced 2-fold (Fig. 4,  lanes 7 and 8). In both cases the total amount of transcript is lower because of the defect in basal transcription.
Akin to the ⌬med19(rox3)-HMT complex, the ⌬med19 (rox3)-HT complex was able to enhance basal transcription  Fig. 2B. To obtain better resolution of its components, the Mediator complexes were separated on an 8 and 12.5% SDS-polyacrylamide gels, and the boundary of the two protein gels is marked with asterisk. The molecular mass markers are shown on the left. FEBRUARY 23, 2007 • VOLUME 282 • NUMBER 8 (Fig. 4, lane 10 versus lane 1), albeit again at ϳ3-fold lower than wild type (Fig. 4, lane 4 versus lane 1). In contrast to the ⌬med19(rox3)-HMT complex, however, the ⌬med19(rox3)-HT was unable to support activated transcription by either Gal4-VP16 or Gcn4 (Fig. 4, lanes 10 -12). Individually, the Med19(Rox3) subunit of Mediator contributes to the ability of Mediator to enhance basal transcription, although the ⌬med19(rox3) complexes still retain some of this activity. The absence of the Middle module subunits has no further impact upon this impaired enhancement of basal transcription observed with the ⌬med19(rox3)-HMT complex. However, displacement of the Middle module, beyond the absence of Med19(Rox3) alone, has a major impact of the ability of the complex to serve as a co-activator.

Med19(Rox3) Regulates Mediator Structure
⌬med19(rox3) Mediators Retain Activator Binding but Have a Decreased Affinity for RNA Pol II-Mediator associates with at least two components of the transcriptional apparatus: activators and core RNA Pol II. Previous studies have demonstrated the binding of Mediator to the VP16 and Gcn4 activa-tion domains (13,31). A more recent study demonstrated that a Gcn4 activation domain could pull down a largely intact Mediator complex from a ⌬med19(rox3) extract (33). This finding was consistent with our observation that the ⌬med19(rox3)-HMT could support activated transcription. We performed binding assays to determine whether the purified ⌬med19(rox3)-HT complex, which was defective for activated transcription, could still bind the activator VP16 with the purified complex. As previously shown (31), wild type Mediator binds to a GST-VP16 but not an inactive mutant activation domain (GST-VP16(⌬456, F442P)) fusion protein immobilized on glutathione-Sepharose (Fig. 5A). Similarly the purified ⌬med19(rox3)-HT complex was found to bind to immobilized GST-VP16 but not GST-VP16(⌬456, F442P) (Fig. 5A). Consistent with previous work identifying the Tail module as the primary point for activator contact (13,(33)(34)(35), the ⌬med19(rox3)-HT complex is still capable of activator binding. The activated transcription defect of the ⌬med19(rox3)-HT complex does not originate from an inability to bind activators.
A second physical interaction that may be critical to Mediator function is its association with core RNA Pol II through the CTD of the large subunit. To test whether the absence of Med19(Rox3) had an impact on the ability of Mediator to bind RNA Pol II, we immobilized Med18(Srb5)-FLAG-tagged wild type and ⌬med19(rox3)-HMT Mediator on FLAG-agarose and assayed for RNA Pol II binding. Immunoblots show that purified RNA Pol II is retained by the immobilized wild type Mediator but that under identical conditions the ⌬med19(rox3)-HMT Mediator does not retain a detectable amount of RNA Pol II (Fig. 5B). Absence of the Med19(Rox3) subunit from Mediator impairs binding of the complex to RNA Pol II. Similar assays show, not surprisingly, that the ⌬med19(rox3)-HMT and ⌬med19(rox3)-HT Mediator also have a decreased affinity for the CTD of RNA Pol II (data not shown).
⌬med19(rox3) Mediator Is Unable to Stimulate Phosphorylation of the CTD of RNA Pol II by TFIIH-Wild type Mediator enhances phosphorylation of the CTD by the kinase domain of TFIIH (6,7) and is also a target of this kinase activity on at least two subunits (30,36). To determine whether the decreased affinity of the ⌬med19(rox3) Mediator for RNA Pol II and the  . Basal and activated transcription in a purified system with wild type, ⌬med19(rox3)-HMT, and ⌬med19(rox3)-HT Mediator complexes. Equimolar amounts of purified wild type, ⌬med19(rox3)-HMT, and ⌬med19(rox3)-HT Mediator complexes (as shown in Fig. 3) were added to transcription reactions reconstituted with purified TBP, TFIIB, RNA Pol II, TFIIF, TFIIE, and TFIIH. The purified activator proteins Gcn4 and Gal4-VP16 were added to the reconstituted reactions to assess activated transcription from the pGCN4:CG-and pJJ470:CG-DNA templates respectively as previously described (7).
CTD impacted its ability to enhance CTD phosphorylation, we performed kinase assays with purified components. We have previously shown that the enhancement of CTD phosphorylation by TFIIH can be completely recapitulated in a reaction containing purified Mediator, TFIIK (the kinase module of TFIIH) and a GST-CTD fusion protein (30). We repeated these assays for the purified wild type, ⌬med19(rox3)-HT, and ⌬med19(rox3)-HMT Mediators. Analysis of the products of the kinase reactions on SDS-PAGE showed that neither the ⌬med19(rox3)-HT nor the ⌬med19(rox3)-HMT Mediator could enhance the CTD kinase activity of TFIIH (Fig. 6, lanes 2,  7, and 8). These findings appear to correlate with the reduced binding affinity of the ⌬med19(rox3) Mediators for RNA Pol II.

DISCUSSION
This work advances the understanding of the modular structure of Mediator, and provides new insight on the essential function of the Middle module subunits as well as the paradox-ical role of Med19(Rox3). Based on a set of largely self-consistent biochemical, genetic, and structural information, a working model for the modular assembly of Mediator subunits has been proposed (Refs. 3 and 11 and Fig. 7A). A reasonable inference from this model is that the subunits in the Middle module serve to hold the Head and Tail module together to form an intact Mediator. It might be further inferred that the requirement for the Middle module subunits encoded by essential genes originated from a need to hold the Head and Tail together. The ability to purify an intact complex of Head and Tail module proteins from a ⌬med19(rox3) strain, however, indicates that the Middle module subunits are dispensable for the interaction of Head and Tail, at least in the absence of Med19(Rox3). Several modifications to the current model can be proposed to integrate our new finding; among the most straightforward are the two following alternatives. The first alternative expands the role of the Med14(Rgr1) subunit. Although the C terminus of Med14(Rgr1) gives the protein the properties of a Tail subunit (13), the genetic interactions of the subunit (16 -18, 20) and the fact that it is encoded by an essential gene could lead one to consider it a Middle module subunit. Because Med14(Rgr1) is retained in our ⌬med19(rox3)-HT complex, it is possible that Med14(Rgr1) serves, in part, as a Middle subunit and spans the Head and Tail modules and holds them together (Fig. 7B). To explain how the absence of Med19(Rox3) may destabilize the presence of the Middle module in Mediator, we have placed it on the interface of the Head and Middle modules. A previous study (32) demonstrated an interaction between recombinant Med19(Rox3) and Med17(Srb4), but a similar recent study was unable to detect this interaction (37). These findings are suggestive of a weak interaction between Med19(Rox3) and the Head module, which would be consistent with Med19(Rox3) being stabilized by both the Head and Middle modules. Whether separate domains of Med19(Rox3) make unique contacts with other Mediator subunits or serve separate functions is unclear. Genetic studies, however, have shown that virtually all phenotypes of Med19(Rox3) in both the positive and negative regulation of transcription appear to be contained in the final ϳ100   FEBRUARY 23, 2007 • VOLUME 282 • NUMBER 8 amino acids of the C terminus of the 220-amino acid protein (5,21,23). The placement of Med19(Rox3) in this model and its genetic interactions (16,18,19) are more consistent with it being characterized as a Middle module subunit (Fig. 7, B and C) rather than a Head subunit (32).

Med19(Rox3) Regulates Mediator Structure
A second alternative to the previous model (Fig. 7A) that could explain the stability of the ⌬med19(rox3)-HT complex would be the existence of direct interactions between Head and Tail subunits, in addition to those interactions facilitated by the Middle module (Fig. 7C). This model is not necessarily mutually exclusive from the model presented in Fig. 7B. Single-particle electron microscopic structure analysis of the core Mediator complex shows a small spatial separation between the Head and Tail modules (38). At the level of resolution of these structures, it would be difficult to rule out an interaction between the Head and Tail module. In fact, an intriguingly high positive correlation between gene expression profiles for deletion mutants of the Tail (⌬med2, ⌬med3, ⌬med14(gal11)) and Head (⌬med18(srb5), ⌬med20(srb2)) subunits (5) suggests a closer physical association between these subunits than inferred from the previous model (Fig. 7A). An additional possibility is that the absence of Med19(Rox3) might also lead to a stabilization of the proposed Head-Tail interaction, even while it destabilizes the interaction of the Middle module with the Head-Tail complex. Med19(Rox3) exerting both stabilizing and destabilizing effects on the complex would be consistent with its dual role in both activation and repression observed in the cell (16,18,19,23,24).
Stabilization of a Head-Tail interaction might also explain our observation that the ⌬med19(rox3)-HMT complex had a decreased affinity for core RNA Pol II. This finding was surprising in light of studies showing that both the Head module and Middle-Tail module can independently bind core RNA Pol II (32,39). If removal of Med19(Rox3) stabilized Head-Tail interactions, it is possible that the mutant Mediator would be locked in a "closed" conformation and unable to form the "open" structure that exposes the RNA Pol II binding surfaces and envelops the polymerase (38). It is unlikely that the Med19(Rox3) subunit itself is directly required for interaction with RNA Pol II, because individual modules lacking Med19(Rox3) have been shown to bind the polymerase (32,39). The decreased binding of ⌬med19(rox3) Mediator to RNA Pol II may also impact the function of the mutant complexes in transcription.
Our studies showed that the removal of the individual Med19(Rox3) subunit and the removal of the entire Middle module of Mediator had differential effects on basal and activated transcription. The absence of either Med19(Rox3) alone or the entire Middle module both led to a comparable decrease in Mediator enhanced basal transcription. Because both complexes retain some ability to enhance basal transcription, it is interesting to note that the complete absence of the Middle module did not lead to a defect beyond that caused by the absence of Med19(Rox3) alone. It is possible that the decreased ability of ⌬med19(rox3) complexes to bind RNA Pol II is correlated with their decreased ability to enhance basal transcription, as well as their inability to enhance phosphorylation of the CTD by TFIIH. Given the decreased affinity for RNA Pol II, it was surprising that the fold activation supported by the ⌬med19(rox3)-HMT complex was comparable with wild type Mediator. The above results suggest that the effect of activators on the core transcription machinery through Mediator may not The assignment of subunits is as previously described (3,11), except for the assignment of Med5(Nut1) as a Tail subunit (41). B, an alternative model in which Med14(Rgr1) spans the Head and Tail domain and Med19(Rox3) stabilizes the interaction of Middle module subunits with the rest of the complex. Med14(Rgr1) is colored both yellow and green to represent that it has properties of both a Tail and Middle module subunit. Med19(Rox3) is also reassigned as a Middle module subunit. C, a second alternative model in which direct interaction between Head and Tail domain subunits could hold together the complex in the absence of the Middle module. Again, Med19(Rox3) could stabilize the interaction of Middle module subunits with the rest of the complex. Med19(Rox3) is also reassigned as a Middle module subunit. be dependent on strong Mediator-RNA Pol II interactions. The second notable result of the in vitro transcription assays was that the removal of the Middle module from ⌬med19(rox3) Mediator led to the abrogation of activated transcription. This defect was not a result of lack of activator-Mediator binding, as the Gcn4 (33) and VP16 (this work) activation domains retain their affinity for the ⌬med19(rox3) complexes. Our analyses of both the composition and the function of the ⌬med19 (rox3)-HT complex suggests that the Middle module is more than a structural bridge between Tail and Head but plays an important role in the transmission of information from activators to the core transcription machinery.
Studies of the ⌬med19(rox3) mutant in vivo demonstrate that it has severe defects in the induction of high levels of transcription of several genes controlled by the Gal4 (23) and Gcn4 (24) activator. Because our work and others (33) suggests that the ⌬med19(rox3)-HMT is the form of the complex in vivo, it is not entirely clear why this mutation gives more severe defect in activated transcription in vivo than it does in our minimal in vitro system. Our observation that the overall amount of transcription is lower, because of defects in basal transcription, may explain this phenomenon in part. A second important consideration is that these same studies in vivo also showed that other co-activators, such as SAGA, are critical for the induction of these genes and that their recruitment was often interdependent with Mediator (40). Because our current in vitro system does not reflect this interdependence, there may be further activation defects in the ⌬med19(rox3) Mediator that we have not yet observed. A second question that remains to be resolved is why mutations in MED19(ROX3) and other Middle module subunits lead to depression of specific genes in vivo. It might have been expected that a Mediator without a Middle module would be constitutively active. Our results show that this is not the case in vitro. This finding leaves open the possibility that the derepression of genes in Middle module mutants may arise from components not currently in our purified system. Relief of inhibition by the Cyc-C module (8) and/or chromatin (20) are among the mechanisms that could account for the derepression observed in med19(rox3) mutants in vivo.