Identification, structure, and functional requirement of the Mediator submodule Med7N/31

Identification and crystallization of the Med7N/31 subcomplex

The Mediator middle module contains the subunits Med1, Med4, Med7, Med9 (Cse2), Med10 (Nut1), Med21 (Srb7), and Med31 (Soh1) (Guglielmi et al, 2004; Beve et al, 2005). To work towards the structure and function of the middle module, we purified the trimeric complex Med7/21/31 in recombinant form after bacterial co-expression of the corresponding genes and subjected it to limited proteolysis (Materials and methods). Proteolytic cleavage patterns were consistent with two flexibly linked subcomplexes, the Med7C/21 subcomplex, which we analysed earlier (Baumli et al, 2005), and a subcomplex of Med31 with the N-terminal region of Med7 (Figure 1A). We therefore co-expressed Med31 with different variants of the Med7 N-terminal region, and purified the resulting heterodimeric complexes. A complex that comprised the N-terminal region of Med7 including part of the linker (residues 1–101) was soluble, but did not crystallize, apparently due to flexibility in the proteolytically sensitive linker. A complex that lacked the entire linker and included only Med7 residues 1–83 (Med7N) crystallized. The X-ray structure of Med7N/31 was solved by selenomethionine labelling and single anomalous diffraction, and was refined to a free R-factor of 24.3% at 2.8 Å resolution (Supplementary Table I; Materials and methods).

The Med7N/31 structure reveals novel folds

The structure shows that Med31 forms a compact right-handed four-helix bundle, whereas Med7N is extended and wraps around Med31 (Figure 1B). Both folds are novel, as neither DALI nor MSD fold searches (Holm and Park, 2000; Krissinel and Henrick, 2004) produced any hits. The Med7N/31 structure is apparently conserved among eukaryotes, as hydrophobic core residues are identical or similar from yeast to human (Figure 1A), and many residues in the Med7N–Med31 interface are also conserved. The Med7N–Med31 interface is mainly hydrophobic (Figure 1), explaining why isolated recombinant Med7N or Med31 cannot be obtained in soluble form (not shown). The Med7N/31 surface reveals several conserved, non-charged regions (Figure 2A and B) that can account for interactions of Med7N/31 with other subunits of the middle module (Guglielmi et al, 2004) or other potential binding partners.

Possible CTD mimicry by Med7N

Med7N contacts Med31 mainly through two polyproline stretches (pPS1 and pPS2; Figure 1A–C), which adopt a conformation similar to a left-handed type II polyproline helix (Adzhubei and Sternberg, 1993). A deletion analysis revealed that both pPS1 and pPS2 were required for the formation of a stable complex with Med31 (Supplementary Table II; constructs 2–5). pPS2 is tightly packed onto Med31, whereas pPS1 is more accessible (Figure 1C). As the C-terminal repeat domain (CTD) of Pol II is also proline rich, we compared the Med7N stretches with known CTD peptide structures. The conformation of the CTD peptide in complex with the prolyl isomerase Pin1 (PDB 1F8A; Verdecia et al, 2000) resembles the structure of Med7N pPS1, and is also distantly related to pPS2 (Figure 2C).

The CTD interacts genetically and physically with Mediator (Thompson et al, 1993; Asturias et al, 1999; Kang et al, 2001). Although med31 deletion does not suppress the cold-sensitive phenotype caused by CTD truncations (Thompson et al, 1993), it is synthetically lethal with a truncation of the CTD to only 11 intact heptad repeats (Fan et al, 1996). We therefore speculated that the CTD may bind Med31 by replacing pPS1 and/or pPS2 of Med7N under certain conditions. We could however not investigate direct CTD binding to Med31 as isolated Med31 could not be obtained in soluble form. We therefore mutated the Med7N pPS1 and pPS2 stretches such that their sequences correspond to the CTD repeat sequence consensus (Supplementary Table II), and tested whether these Med7N mutants formed stable complexes with Med31. The pPS1 mutant, but not the pPS2 mutant, formed a stable complex with Med31 (Supplementary Table II). Crystallographic analysis of the pPS1 mutant (mutant 6; Supplementary Table II) in complex with Med31 showed that four residues of the CTD heptad repeat (15-TSPS-18) were ordered and bound to Med31, whereas the preceding three residues (12-YSP-14) were flexible (Figure 2D). However, as the Med7N/31 complex also tolerated replacement of pPS1 by the reverse CTD heptad repeat sequence (mutant 10; Supplementary Table II), and as the Med7N stretches are present in organisms lacking the canonical CTD repeats (Stiller and Hall, 2002; Bourbon, 2008), a possible CTD mimicry by Med7N and CTD binding by Med31 remains hypothetical.

Med7N/31 is required for normal yeast growth in vivo

Med7 is essential for yeast viability in rich medium (Myers et al, 1998). To investigate which Med7 region is required for viability, we carried out complementation studies in yeast. We generated plasmids expressing full-length Med7, Med7N, and Med7N plus the linker part, which was not present in the Med7C/21 structure (Baumli et al, 2005) or a Med7C variant. Plasmids were introduced into a med7Δ strain rescued by a MED7-encoding URA3 plasmid. Complementation was observed with full-length MED7 or MED7C, which allows for loss of the URA3 plasmid and growth on media containing 5-fluorotic acid (5-FOA; Figure 3A). As the remaining linker residues (102–113) present in the Med7C/21 structure are not conserved, we infer that only Med7C is essential for viability, whereas Med7N and the linker are not. A strain that lacked the region encoding Med7N (med7NΔ) exhibited a slow-growth phenotype, similar to a med31Δ strain and a double deletion strain med7NΔ/med31Δ (med7N/31Δ) (Figure 3B; Supplementary Table III).

Med7N/31 is essential for transcription in vitro

To investigate whether the Med7N/31 subcomplex is required for activated transcription in vitro, we used a plasmid-based in vitro transcription assay with a HIS4 core promoter and a single upstream Gal4-binding site (Ranish et al, 1999). A nuclear extract from strain med7N/31Δ was defective in basal transcription (not shown) and in transcription activated with the Gal4–VP16 fusion protein or with a fusion protein of the Gal4 DNA-binding domain with the Gal4 activation helix (Gal4-Gal4AH) (Figure 3C, lane 1; Supplementary Figure 1, lane 1). Addition of tandem affinity purification (TAP)-purified Mediator restored transcription (Figure 3C, lane 2), providing a positive control. Transcription could also be restored by the addition of recombinant Med7N/31, demonstrating that the crystallized complex is functionally active and can act in trans without being covalently tethered to the Mediator (Figure 3C, lanes 3–5; Supplementary Figure 1, lane 2). To investigate whether loss of Med7N/31 leads to the dissociation of other subunits from Mediator in vivo, we purified Mediator from deletion strains med7NΔ, med31Δ, and med7N/31Δ with the use of a TAP tag fused to Med18 (Materials and methods). After purification from these deletion strains, the remainder of Mediator stayed intact (Figure 3F). These results show that Med7N/31 forms a peripheral subcomplex on Mediator that is required for activated transcription in our in vitro assays.

Med7N/31 is a functional Mediator submodule in vivo

To investigate the function of the Med7N/31 subcomplex in vivo, we carried out comparative microarray-based gene expression profiling with the yeast deletion strains med31Δ, med7NΔ, and med7N/31Δ. Compared with wild-type cells, the expression profiles of the three deletion strains showed similar overall changes illustrated by high Pearson's correlations (Figure 4A–D; Table I; Supplementary Table IV). A majority of changes in mRNA levels that are induced by deletion of both Med7N and Med31 are already observed when either Med7N or Med31 is lacking from cells. This shows that Med7N/31 serves as a functional submodule of the Mediator in vivo at a majority of affected genes. Some differences between profiles for the med7NΔ, med31Δ, and med7N/31Δ strains at selected genes (Figure 4C) indicate subunit-specific functions that may result from additional protein interactions (unpublished data). As the majority of affected genes is downregulated (Figure 4D), the Med7N/31 submodule has a predominantly positive function, consistent with previous expression data on a med31Δ strain (van de Peppel et al, 2005).

Med7N/31 regulates a subset of genes

Gene ontology (GO) analysis of genes with significantly changed mRNA levels revealed an over-representation of genes involved in sulphate/methionine metabolism, iron homoeostasis, and transport functions (Table I; Supplementary Table IVB). A group of genes involved in telomere maintenance was also over-represented, consistent with the observation that a med31Δ strain exhibits shortened telomeres (Askree et al, 2004). The differential expression profiles for Med7N/31 were largely distinct from the previously published profiles (Larivière et al, 2008) of the tail subunit deletion strains med2Δ and med3Δ, and correlated only moderately with those of strains med8CΔ, med18Δ, and med20Δ, which lack parts of the head submodule Med8C/18/20 (Figure 4A and B). However, both the Med7N/31 and the Med8C/18/20 submodules regulate genes responsible for siderophore transport and conjugation. Thus, the Med7N/31 submodule is required not only for the regulation of a specific subset of genes but also those genes that depend on other parts of the Mediator.

To identify transcription factors that regulate the differentially expressed genes, we used the transcription factor to gene promoter mapping matrix from the YEASTRACT website (Teixeira et al, 2006), and performed Fisher's exact test with the R/Bioconductor software (Gentleman et al, 2004) (Supplementary Table IVF). This analysis revealed that the differentially expressed genes contained over-represented target genes for certain transcription factors, including Cst6, Aft1, Msn2, Msn4, Yap family factors, Met28, Met31, Met32, Hsf1, and Pdr1. Consistent with the GO analysis, Msn2, Msn4, and Yap family members are involved in stress response, Cst6 regulates genes involved in telomere maintenance, and Aft1 is required for activating the low iron response regulon through Mediator (van de Peppel et al, 2005). The mRNA levels for these factors were unaltered in the investigated strains, except for Met28 and Msn4. Met28 was downregulated for Med7N/31 deletions and upregulated for Med2/Med3 deletions (Supplementary Table IVG), explaining corresponding negative and positive responses, respectively, of target genes involved in methionine and cysteine biosynthesis (Supplementary Table IVD).

Overlapping and specific functions of Mediator submodules

To investigate the deregulation of genes induced by disruption of Med7N/31, we analysed the growth of several yeast strains with deletions in Mediator genes (Supplementary Table III) on various selective agar plates (Table II). The phenotypes of Med7N/31 deletions correlated well with results from transcription profiling. In the absence of methionine, these deletion strains grew poorly (Figure 4E), consistent with the observed downregulation of genes involved in methionine/sulphate metabolism. In the presence of siderophores, several deletion strains took up more complexed iron, as indicated by a red colour (Figure 4E), consistent with higher expression of siderophore transport genes (some likely regulated by Aft1). Sensitivities of the deletion strains towards osmotic and cell wall stress were consistent with reported impaired cell wall metabolism of Mediator subunit deletion strains of Schizosaccharomyces pombe (Linder et al, 2008). A decreased sensitivity of the med31Δ and med7N/31Δ strains for the translation inhibitor paramomycin can also be explained by impaired cellular uptake or transport (Figure 4E). Most deletion strains were further impaired in their response to temperature, oxidative, and DNA repair stresses (Table II). Also observed were inositol^– phenotypes that are often indicative of defects in the general transcriptional apparatus (Hampsey, 1997).

This comparative phenotyping of Mediator deletion strains showed that the med8CΔ and med18Δ strains exhibited the most severe growth defects (Table II), possibly reflecting an important function of these subunits in the initiation complex assembly (Ranish et al, 1999). Deletion of subunits Med1, Med5, or Med9 was less detrimental. In agreement with earlier studies (van de Peppel et al, 2005), Cdk8 behaves partially contrasting with the phenotyping assays. On the transcriptome level, analyses of Med7N/31 and Med8C/18/20 submodule deletion profiles with med2Δ and med3Δ profiles (Supplementary Table IVC and D) reveal different overlaps in enriched GO terms. Some GO terms are exclusively enriched for specific Mediator submodules (e.g. telomere capping for Med7N/31), whereas other GO terms are affected for two different submodules (e.g. siderophore transport for Med7/31 and Med8C/18/20 or serine family amino-acid catabolism with Med8C/18/20 and Med2/3), and other GO term groups are affected by all investigated mutations (e.g. response to pheromone). Such overlap in the phenotyping assays (Table II) and the GO analysis (Table I; Supplementary Table IVC and D) shows that the characterized non-essential parts of Mediator are generally required for regulating metabolic sensing, stress response, and amino-acid biosynthesis pathways.

Cooperation of Med7N/31 and TFIIS

As Med31 interacts genetically with the elongation factor TFIIS (Krogan et al, 2003; Malagon et al, 2004; Collins et al, 2007; Guglielmi et al, 2007), we tested whether Med7N/31 cooperates with TFIIS during activated transcription in vitro. Indeed, transcription in a med7N/31Δ nuclear extract that was rescued by recombinant Med7N/31 could be enhanced four-fold by the addition of recombinant TFIIS using Gal4–VP16 as an activator (Figure 3D). A stimulatory effect of TFIIS, although only about two-fold, was also seen for basal and activated transcription in wild-type (WT) extracts (Figure 3E). Addition of TFIIS to a med7N/31Δ extract resulted in a signal even in the absence of recombinant Med7N/31, although TFIIS was not limiting in the extracts, as a three-fold higher concentration did not further increase transcription (Figure 3D, lanes 3 and 4). TFIIS also stimulated transcription activated with the alternative activator Gal4–Gal4AH, although only about two-fold (Supplementary Figure 1). In contrast, transcription in a med18Δ extract was not enhanced at all by TFIIS addition (Figure 3D, lanes 8 and 9) and a med18Δ extract that was rescued by the addition of recombinant Med18/20 was not further stimulated by the addition of TFIIS (Figure 3D, lane 10). These results show that TFIIS can partially compensate for the loss of Med7N/31, and that highest transcript levels required the addition of both TFIIS and Med7N/31. These observations were specific for the med7N/31Δ extract, as this effect was not obtained in corresponding experiments with a med18Δ extract. The results are further consistent with an important role for TFIIS during initiation (Guglielmi et al, 2007; Kim et al, 2007).

To investigate whether TFIIS and Med7N/31 cooperate at specific genes in vivo, we determined the gene expression profile for a yeast strain that lacks the gene for TFIIS, dst1 (Supplementary Table IVA, B, E and F). Only 162 mRNA levels were significantly altered (at a two-fold cutoff and a false discovery rate of P<0.05; Supplementary Table IVA), consistent with a lack of a phenotype of the dst1Δ strain in rich medium (Clark et al, 1991). Pearson's correlation and GO analysis did not reveal any significant overlap between dst1Δ and med7N/31Δ profiles (Figure 4B; Supplementary Table IV). Consistently, we could not detect a direct physical interaction between TFIIS and recombinant Med7N/31 (not shown). Instead, analysis of synthetic lethality partners suggests that Med31 and TFIIS are functionally connected through the SWR1 complex that is involved in exchange of histone H2A by the variant H2A.Z (Supplementary Figure 2A). Similar correlations were also found for the Mediator subunits Med1, Med9, and Med20 (not shown). Additionally, the Mediator kinase module is a highly significant phenotypic suppressor of core Mediator subunits and TFIIS (Supplementary Figure 2B). Taken together, the functional cooperation of Med7N/31 and TFIIS is apparently not gene specific in vivo, but may generally occur during initiation, and could further be linked to Swr1-dependent histone variant exchange near transcription start sites (Santisteban et al, 2000; Zanton and Pugh, 2006; Zlatanova and Thakar, 2008).

Preparation of recombinant Med7N/31

The trimeric complex Med7/21/31His was expressed from two plasmids encoding Med7/21 and Med31His that were co-transformed into Escherichia coli BL21 (DE3) RIL cells (Stratagene). The complex was purified by Ni-NTA and anion exchange chromatography and subjected to limited proteolysis using either chymotrypsin or trypsin. The reaction was stopped by the addition of a protease inhibitor mixture, loaded onto a Superose 6 gel filtration column (GE Healthcare) and individual peaks were analysed on SDS–PAGE after TCA precipitation. The gene encoding S. cerevisiae Med31 was cloned into a pET-28b vector (Novagen) using the NdeI/NotI restriction sites, resulting in a thrombin-cleavable N-terminal hexahistidine tag. DNA encoding Med7 residues 1–83 was cloned into a pET21b vector through the EcoRI and SalI restriction sites. E. coli BL21 (DE3) RIL cells (Stratagene) were co-transformed with the two plasmids and grown in LB medium at 37°C to an optical density at 600 nm (OD_{600 nm}) of 0.6. Expression was induced with 0.5 mM IPTG for 16 h at 18°C. Selenomethionine labelling was carried out as described (Budisa et al, 1995; Meinhart et al, 2003). For protein purification, cells were lysed using a high-pressure homogenizer (Avestin) in buffer A (50 mM Tris pH 8.0, 150 mM NaCl, and 10 mM β-mercaptoethanol). After centrifugation, imidazole was added to a final concentration of 20 mM to the supernatant and loaded onto a 3 ml Ni-NTA column (Qiagen) equilibrated with buffer A containing 20 mM imidazole. The column was washed with 20 column volumes (CVs) of buffer A containing 20 mM imidazole and eluted with buffer A containing 200 mM imidazole. Following overnight cleavage with thrombin, proteins were purified by anion exchange chromatography (Mono Q). The column was equilibrated with buffer B (50 mM Tris pH 8.0, 100 mM NaCl, and 10 mM β-mercaptoethanol), and proteins were eluted with a linear gradient of 20 CVs from 100 mM to 1 M NaCl in buffer B. After concentration, the sample was applied to a Superdex-200 size exclusion column (GE Healthcare) equilibrated with buffer C (50 mM MES pH 6.5, 150 mM NaCl, and 10 mM β-mercaptoethanol).

X-ray structure determination

For crystallization, pure Med7N/31 was concentrated to 50 mg/ml, and 5 mM DTT was added. Crystals were grown at 20°C in hanging drops over reservoirs containing 50 mM MES pH 5.6, 1.8 M Li₂SO₄, and 10 mM MgCl₂. Crystals were harvested by gradually adding ethylene glycol to a final concentration of 10% (v/v) and were flash-cooled in liquid nitrogen. Diffraction data were collected at 100 K on a MARCCD 225 and on a PILATUS 6M detector at the Swiss Light Source (SLS), Villigen, Switzerland (native data set and mutant data set, respectively) and on a MARCCD 165 detector at the Berliner Elektronenspeicherring-Gesellschaft für Synchrotronstrahlung m.b.H. BESSY (SAD data set). Diffraction data for selenomethionine-labelled Med7N/31 was processed using XDS and XSCALE (Kabsch, 1993). Program SOLVE (Terwilliger and Berendzen, 1999) identified six selenium sites that were used for phasing. Solvent flattening, two-fold non-crystallographic symmetry averaging, and initial model building was done with RESOLVE (Terwilliger, 2003). The resulting electron density map allowed for manual building of most Med7N and Med31 using COOT (Emsley and Cowtan, 2004). The model was refined using conjugate gradient minimization in the programs CNS (Brunger et al, 1998), REFMAC (McCoy et al, 2007), and Phenix (McCoy et al, 2007). The asymmetric unit contained two Med7N/31 heterodimers that deviate only at the C termini of the proteins. The mutant structure was solved by molecular replacement using Phaser (McCoy et al, 2007) and processed accordingly. The structures and diffraction data of wild-type Med7N/31 and the mutant complex Med7 (1–83) (12.18 YSPTSPS)/Med31 (mutant 6; Supplementary Table II) have been deposited in the Protein Data Bank under accession codes 3FBI and 3FBN, respectively.

Yeast strains and growth assays

The heterozygous MED7/med7Δ strain was obtained from Euroscarf (Frankfurt, Germany) and transformed with the cognate gene clone plasmid pYCG_YOL135C (containing a URA3 marker). Diploids were sporulated and tetrads were dissected on YPD plates. To assess functionality, pAL-med7_1–83, pAL-med7_1–101, pAL-med7_101–222, pAL-med7_1–222, or pRS315-med7_101–222 were transformed into the MED7 shuffle strain. Transformants were re-streaked onto 5-FOA and YPD plates. Yeast knockout strains were obtained from Euroscarf. All strains exhibited a BY background and were, with the exception of med2Δ and med18Δ, MATa. Deletion of the N-terminal region of Med7 (residues 1–83) was obtained by homologous recombination after transforming only the C-terminal part (residues 84–222) fused to a ClonNAT marker into either WT or med31Δ strains. Additionally, a C-terminal TAP tag on Med18 was introduced using an URA3 marker into wild-type, med31Δ, med7NΔ, and med7N/31Δ strains. Phenotype analyses of Mediator subunit deletion mutants were performed from cultures grown to stationary phase. Cells were diluted to an OD_{600 nm} of 1.0, washed, and spotted in serial dilution onto plates. Assays were mostly performed as described (Hampsey, 1997). For synthetic defined (SD) (-met) and SD (-SO₄²⁻) and for 6-azauracil plate assays, strains with methionine or uracil auxotrophy were transformed with pRS411 or pRS316 plasmids, respectively. Siderophore uptake assays were performed with SD plates containing 500 μM bathophenanthroline disulphonic acid and 10 μM of siderophores.

In vitro transcription assays

Nuclear extracts were prepared from 3 l of culture as described by the Hahn laboratory (www.fhcrc.org/labs/hahn). Plasmid transcription was performed essentially as described (Ranish and Hahn, 1991). Transcription reactions were carried out in a 25 μl volume. The reaction mixture contained 100 μg nuclear extract, 150 ng of pSH515 plasmid, 1 × transcription buffer (10 mM HEPES pH 7.6, 50 mM potassium acetate, 0.5 mM EDTA, and 2.5 mM magnesium acetate), 2.5 mM DTT, 192 μg of phosphocreatine, 0.2 μg of creatine phosphokinase, 10 U of RNase inhibitor (GE Healthcare), and 100 μM nucleoside triphosphates. For activated transcription, 150 ng of Gal4–VP16 or 200 ng of Gal4–Gal4AH was added. The reaction was incubated at room temperature for 40 min and then stopped with 180 μl of 100 mM sodium acetate, 10 mM EDTA, 0.5% sodium dodecyl sulphate, and 17 μg of tRNA/ml. Samples were extracted with phenol–chloroform and precipitated with ethanol. Transcripts were analysed by primer extension essentially as described (Ranish and Hahn, 1991). Instead of the ³²P-labelled lacI oligo, 0.125 pmol of a fluorescently labelled 5′-FAM-oligo was used. Quantification was performed with a Typhoon 9400 and the ImageQuant Software (GE Healthcare).

Gene expression profiling analysis

All experiments were performed in synthetic complete medium with 2% glucose. For microarray analysis, three independent colonies were used for inoculation, and overnight cultures were diluted in fresh medium to OD_{600 nm}=0.1 (25 ml cultures, 160 r.p.m. shaking incubator 30°C). Cells were harvested by centrifugation (4000 r.p.m., 3 min, 20°C) at OD_{600 nm}=0.8. Total RNA was prepared after cell lysis using a mixer mill (Retsch) and subsequent purification using the RNAeasy kit (Qiagen). The total RNA preparation was treated on-column with DNase (Qiagen). All following steps were conducted according to the Affymetrix GeneChip Expression Analysis Technical Manual (P/N 702232 rev. 2). Briefly, one-cycle cDNA synthesis was performed with 1 μg of total RNA. In vitro transcription labelling was carried out for 16 h. The fragmented samples were hybridized for 16 h on Yeast Genome 2.0 expression arrays (Affymetrix), washed and stained using a Fluidics 450 station, and scanned on an Affymetrix GeneArray scanner 3000 7G. Data analysis was performed using R/Bioconductor (Gentleman et al, 2004). S. pombe probes were filtered out prior to normalization with the GCRMA algorithm (Wu et al, 2004). Linear model fitting and multiple testing correction using an empirical Bayes approach was performed using the LIMMA package (Smyth, 2004). Differentially expressed genes were defined as having an adjusted P-value smaller than 0.05 and an estimated fold change of at least 2.0 (calculated as the fold change of the average expression in the triplicate measurements). Previously published transcriptome profiles (Larivière et al, 2008) were included in further analyses by matching genes and comparing fold changes. Over-represented biological processes for genes with significant expression changes were determined using the topGO package (Alexa et al, 2006). Clustering was calculated using TIGR MeV application (Saeed et al, 2003). Microarray data were submitted to the ArrayExpress database (http://www.ebi.ac.uk/microarray) under accession number E-MEXP-1916.