Preparation and topology of the Mediator middle module

Purification of endogenous middle module

The med19Δ yeast strain (BY4741; MATa, his3Δ1, leu2Δ0, met15Δ0, ura3Δ0, YBL093C::clonNAT) was obtained from Stephan Jellbauer (Gene Center Munich) and C-terminal TAP-tags were introduced on Med7, Med15 or Med18, respectively, using a kanMX4-marker by means of vector pYM13 (30). Yeast cultures were cultivated and the protein complexes purified using tandem affinity purification as described (15).

Preparation of recombinant middle module

Monocistronic vectors were cloned by standard procedures. Bi- or tricistronic vectors were generally created by introducing additional ribosomal binding sites by means of PCR into pET- and pCDF-vectors (Novagen) as described (13,16) and illustrated in Figure 2. A detailed listing of all vectors used and created for this study can be found in Supplementary Table S1. For expression, E. coli BL21 (DE3) RIL cells (Stratagene) were transformed with one to three plasmids and grown in LB medium at 37°C to an optical density at 600 nm (OD₆₀₀) of 0.6. Expression was induced with 0.5 mM IPTG for 16 h at 18°C. Cells were lysed using a high-pressure homogenizer (Avestin) (large-scale protein purification) or by sonication (Branson) (small-scale purification).

The quaternary complex Med7/10/21/31-His₆ was expressed from two plasmids encoding Med10, Med7 and Med21 in a tricistronic pET21b vector (pSB104) and Med31-His₆ in a pET24d vector (pSB102). Four liters of E. coli culture were harvested and lysed in buffer A (50 mM Tris pH 8.0, 150 mM NaCl, 10 mM β-mercaptoethanol) containing protease inhibitors. After centrifugation, imidazole was added to a final concentration of 20 mM to the supernatant and loaded onto a 3 ml Ni-NTA gravity flow 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. The 4-subunit middle module was further purified by anion exchange and gel filtration chromatography.

The Med4/9 complex was coexpressed from a bicistronic pET21 vector (pSB118) using 2 l of E. coli culture. Cells were harvested and lysed in buffer B (20 mM Tris pH 8.0, 150 mM NaCl, 10 mM β-mercaptoethanol) containing protease inhibitors. Insoluble cell debris were removed by centrifugation and the proteins purified by ammonium sulfate precipitation and anion exchange chromatography. Ammonium sulfate precipitation was achieved by the gradual addition of saturated (20°C) ammonium sulfate solution up to 30% (v/v). After centrifugation, the pellet was resuspended and purified on a MonoQ 10/100 column (GE Healthcare) in buffer C (50 mM Tris pH 8.0, 100 mM NaCl, 10 mM β-mercaptoethanol) using a linear gradient of 20 column volumes (CVs) from 100 mM to 1 M NaCl. A six-subunit middle module comprising Med4/7/9/10/21/31 was obtained by assembling the two purified subcomplexes. Assembly was performed at 20°C on a rotating wheel with a 1.3 molar excess of Med4/9. Contaminants, excess Med4/9 and unassembled 4-subunit middle module were separated using anion exchange chromatography (buffer C, 20 CV from 100 mM to 1 M NaCl). Following concentration in a 100 kDa MWCO spin concentrator (Amicon Ultra, Millipore), the 6-subunit middle module was purified to homogeneity by gel filtration chromatography using a Superose 6 size exclusion column (GE Healthcare) with buffer A.

The complete middle module comprising Med1/4/7/9/10/21/31 was expressed from the cotransformed vectors pSB104 (encoding Med10, Med7 and Med21), pTK26 (encoding His₆-Med31) and pTK114 (encoding StrepII-tagged Med1, Med9 and Med4). After cotransformation of two vectors, cells were made competent again in order to transform the third vector. Six liters of E. coli culture were harvested and lysed in buffer D [50 mM Tris pH 8.0, 150 mM NaCl, 10% (v/v) glycerol, 10 mM β-mercaptoethanol] containing protease inhibitors. After centrifugation, the complex was purified on 3 × 1 ml Strep‐Tactin MacroPrep (IBA) gravity flow columns according to the manufacturer’s instructions. Following elution by d‐Desthiobiotin (IBA) addition, the sample was concentrated in a 100-kDa MWCO spin concentrator and subjected to gel filtration using a Superose 6 size-exclusion column (GE Healthcare) with buffer A.

Native mass spectrometry

The buffer exchange for the native MS analysis was performed using 10 kDa cut-off membrane spin columns (Millipore, England) by six sequential concentration/dilution cycles against 200 mM ammonium acetate pH 6.8. The concentration for the analysis was 5 µM (assuming intact protein complexes). Analysis of the intact protein complex, as well as the n-propanol and DMSO measurements were carried out on a LCT mass spectrometer (Waters, UK). Needle voltage was set to 1250 V, cone voltage varied between 100 and 150 V. Tandem mass spectrometry was performed on a modified Q-ToF; here, needle voltage was set to 1300 V and cone to 150 V. For further details see ref. 31 and references within. Ion mobility mass spectrometry (IM-MS) was carried out on a Synapt HDMS (Waters, UK) (32). The source pressure was set to 6.9 mbar, the pressure in the trap was 3.5 × 10⁻² mbar, 0.7 mbar in the ion mobility separation (IMS) cell and 2 × 10⁻⁶ mbar in the ToF. The wave height in the IMS cell was fixed on 11.3 V and the wave velocity set to 250 m/s. The gas used in the trap was xenon and nitrogen in the IMS cell. Needle voltage was 1200 V and cone voltage 150 V. The bias value was set to 20 V, trap and transfer collision energy to 12 V. Cross-section calculations were done as described by ref. 33. The average volume of global proteins and protein complexes was calculated according to ref. 34.

Limited proteolysis analyses

Limited proteolysis time courses were performed to identify durations suitable for obtaining medium-size fragments. Digests were performed using 20–50 µg protein complex with 0.2 µg of chymotrypsin (Sigma C3142) in buffer A supplemented with 4 µM CaCl₂ by incubation at 37°C for 1–60 min. The reactions were stopped by the addition of SDS sample buffer and were heated immediately to 95°C for 10 min. For experiments depicted in Figure 4, purified 3-, 4- and 6-subunit middle module complexes were subjected to limited proteolysis for 10 min using chymotrypsin and after stopping the reaction by addition of a protease inhibitor mixture, loaded onto a Superose 6 gel filtration column (GE Healthcare). For proteolysis, 1 µg of sample was used for 3- and 4-subunit, and 2 µg for the 6-subunit middle module. Bands of interest in individual peaks were analyzed on SDS-PAGE after TCA precipitation by MS and Edman-sequencing with a Procise 491 sequencer (Applied Biosystems) following transfer to PVDF membranes.

Coexpression and copurification pull-down assays

Coexpression was performed as described above using 50 ml cultures for StrepII-affinity purifications or 7 ml cultures for Ni-NTA purifications. Cell lysates were clarified by centrifugation and copurification pull-down assays performed in batch. StrepII-affinity purifications were performed using 25 µl of Strep‐Tactin MacroPrep (IBA). The clarified lysates were incubated with the beads at 4°C on a rotating wheel for 1 h. Beads were washed with 3 × 1 ml of buffer A and the samples eluted by d-Desthiobiotin. Ni-NTA purifications were performed using 40 µl of MagneHis beads (Promega). The clarified lysates were incubated with the beads at 4°C on a rotating wheel for 15 min. Beads were washed with 2 × 1 ml of buffer A and the samples eluted by 400 mM Imidazole in buffer A. Samples were analyzed on SDS-PAGE.

Small-angle X-ray scattering

Small-angle X-ray scattering (SAXS) data were collected at the X33-Beamline (EMBL/DESY, Hamburg, Germany). Scattering patterns from 6-subunit Mediator middle module solutions comprising Med4/7/9/10/21/31 were measured in buffer A at 5 mg/ml two times in order to exclude potential radiation damage errors. The ATSAS software package (35) was used for data processing.

Endogenous Mediator middle module

It was previously reported that the middle module is lost from Mediator that is purified from a med19Δ strain under stringent conditions (36). To purify middle module from a med19Δ strain, we introduced a C-terminal tandem affinity purification (TAP) tag on the middle module subunit Med7. TAP purification resulted in pure middle module without the need for the previously used urea dissociation (36), indicating that the tag further destabilizes the Mediator. Fusing a TAP-tag with Med15 or Med18 did not lead to purification of individual parts of Mediator (Supplementary Figure S1). The high yield and purity of the preparation from the med19Δ/MED7-TAP strain (Figure 1C) enabled us to determine unambiguously by MS (see ‘Materials and Methods’ section) that the endogenous yeast Mediator middle module comprises the subunits Med1, Med4, Med7, Med9 (Cse2), Med10 (Nut2), Med21 (Srb7) and Med31 (Soh1), consistent with previous description (19,21,27,37).

Recombinant middle module

We previously used heterologous coexpression of physically associated Mediator subunits in E. coli to obtain milligram quantities of Mediator subcomplexes of up to three subunits (13,15,16). Here, based on commercial pET- and Duet vectors (Novagen), we created poly-cistronic expression constructs with additional ribosomal binding sites between middle module subunit open reading frames (Figure 2). For expression of a 4-subunit complex comprising Med7, 10, 21 and 31, we constructed a tricistronic pET21b vector encoding Med10, 7 and 21 (Figure 2B) and a pET24d vector encoding C-terminally hexahistidine-tagged Med31. These vectors were cotransformed into E. coli BL21(DE3) RIL cells (Stratagene) and after coexpression, the 4-subunit complex could be purified (Figure 2B, ‘Materials and Methods’ section). To obtain a 6-subunit complex that comprised also subunits Med4 and Med9, these two subunits were coexpressed separately using a bicistronic pET21b vector. A partially purified Med4/9 heterodimer was assembled with the pure 4-subunit complex and the resulting 6-subunit complex was purified to homogeneity. To obtain complete 7-subunit middle module, three vectors were cotransformed into E. coli BL21(DE3) RIL cells, the tricistronic vector encoding Med10, 7 and 21, a vector encoding N-terminally hexahistidine-tagged Med31, and a vector encoding Med4, Med9 and N-terminally StrepII-tagged Med1 (Figure 2C, ‘Materials and Methods’ section). After coexpression of the seven proteins from these three vectors, the 7-subunit module was purified in two steps, including a StrepII-affinity step (Figure 2 C, ‘Materials and Methods’ section).

Subunit stoichiometry

To investigate the subunit stoichiometry in the recombinant middle module and its subcomplexes, we used native MS that allows determination of molecular weights of entire protein complexes (38,39). The detected masses were 191 650 ± 60 Da for the complete 7-subunit middle module (191 335 Da expected), 124 760 ± 50 Da for the 6-subunit complex (124 471 Da expected) and 75 840 ± 30 Da for the 4-subunit complex (75 620 Da expected) (Figure 3 and Supplementary Figure S2A). We calculated expected masses taking into account that the N-terminal methionines of Med4, 7, 10, 21 and 31 were lacking (Edman sequencing data, not shown). The slightly higher experimental masses can be explained by incomplete desolvation of the complexes (40). The MS analysis also revealed that the 4- and 6-subunit complexes tend to dimerize. This is likely due to exposed hydrophobic surfaces, since addition of DMSO eliminated the dimers from the recorded MS spectra (Supplementary Figure S2B). These results reveal that only a single copy of each subunit is present in the complexes and establish the equimolar subunit stoichiometry of the middle module.

Module topology

We next investigated the topology of the middle module by MS analysis. We observed spontaneous dissociation of subunits and subcomplexes in the buffer used for the native MS experiments. This effect could be enhanced by addition of DMSO or n-propanol (Supplementary Figure S2). Dissociation of the 7-subunit module revealed that Med1 could detach from the module separately, to result in the 6-subunit complex. Dissociation of the 6-subunit complex revealed a dimer of Med4/9, confirming the interaction between those two subunits. Dissociation of the 4-subunit complex resulted in the trimers Med7/10/31 and Med7/21/31, and in the dimers Med7/31 and Med7/21.Overall, Med7 never dissociated from the middle module, showing it is an architectural subunit, whereas Med9 and Med10 were least stably attached. To further analyze the module topology, we investigated fragmentation of the 4-, 6- and 7-subunit complexes by tandem MS. Generally consistent with solution dissociation experiments, Med9 and Med10 readily dissociated from the complexes, whereas Med7 never did, except to a low extent when the complete middle module was used. Taken together, the dissociation and fragmentation experiments are consistent with earlier structural and interaction analysis of the middle module and the existence of heterodimers Med7N/31, Med7C/21 and Med4/9 within intact middle module (13,15,19).

Exposed regions

To detect exposed and flexible protein regions, middle module complexes were subjected to limited proteolysis using chymotrypsin (see ‘Materials and Methods’ section). In an attempt to separate stable fragments, we stopped proteolysis by addition of protease inhibitor at certain time points, and subjected the samples to gel filtration. Peak fractions were TCA-precipitated and analyzed by SDS-PAGE and Edman sequencing (Figure 4). This approach could identify stable subcomplexes suitable for crystallization, such as Med7N/31 and Med7C/21, which result from proteolysis of a Med7/21/31 trimer (Figure 4A) (13,15).

Limited proteolysis of the 4-subunit complex revealed not only protease cleavage within regions predicted to form flexible loops, but also cleavage in some helical regions of Med21 (known structure) and Med10 (predicted) (Figure 4B, and C). Two subcomplexes were detected based on Med7N/31 and Med7C/21 that both contain N-terminally truncated Med10 (Figure 4C). Proteolysis of the 6-subunit complex also revealed Med7C/21 and Med7N/31 in different gel filtration fractions, but now associated with different truncated variants of Med10 (Figure 4D). Consistently, native MS analysis of an early 7-subunit complex preparation revealed an intact middle module with a truncated form of Med10 that lacked 66 N-terminal residues (data not shown). Med9 showed only one chymotrypsin cleavage site and was detected either as full-length subunit (star in Figure 4D) or as an N-terminally shortened variant (star and triangle peak in Figure 4D). Med4 was cleaved at multiple sites, but preferentially from the C-terminus.

This analysis supports a submodular architecture of the middle module based on the Med7N/31 and Med7C/21 core subcomplexes connected via a flexible Med7 linker. Med4, 9 and 10 are associated with both core subcomplexes, but their proteolytic fragments were not sufficient to hold these subcomplexes together. The C-terminal region of Med4, the N-terminal region of Med9, and the N-terminal region of Med10 are not stably bound to the core subcomplexes and may be involved in contacts with other Mediator modules or external factors. Consistently, yeast complementation assays revealed that the flexible Med4 C-terminal residues 194–250 are required for viability, whereas the N-terminal residues 1–66 are not (Supplementary Figure S3).

Intra-module subunit interactions

To further elucidate the subunit interactions within the middle module, we used coexpression pull-down experiments (Figure 5A and Supplementary Table S2). Our biochemical analysis generally confirmed previous subunit interactions based on yeast two hybrid (Y2H) assays and immunoprecipitation (19), but additionally allowed us to distinguish weak and strong subunit interactions, and to detect interaction domain boundaries (Figure 5B and C). Important findings from this analysis were the following. First, Med7N/31 neither bound Med7C/21 nor Med4/9. For interaction of Med4/9 to any Med7-containing complex, the Med7 linker between Med7N and Med7C was required. Second, the Med4/9 heterodimer binds to Med1. Med4/9 heterodimer formation neither required the 18 N-terminal residues of Med4 nor the predicted loop in Med9 comprising residues 19–63, consistent with the prior observation that Med9 contains two domains (compare 41). Third, Med10 and Med4 interacted weakly and Med21 stabilized this interaction. The C-terminal part of Med10 was sufficient for binding the Med7C/21 subcomplex. Med10 was also found to stably bind to Med14, confirming previous data (19) and showing that Med14 (1-259) is sufficient to establish a connection of middle module towards Mediator tail module. Finally, reported Y2H interactions between Med21 and Med31 were not confirmed. Stable expression of Med9 always required coexpression with Med4, but was not possible with Med1 or Med7, suggesting that Med9 has no direct interaction with these. Together with published data, this analysis establishes the middle module subunit interaction network (Figure 5C).

Module shape

To investigate the overall shape of the middle module in solution we used light scattering and SAXS. Static light scattering analysis of the 6-subunit complex revealed a hydrodynamic radius of 66 Å, 34% larger than that of Med7C/21, which exists as a heterotetramer in solution (compare Supplementary Table S3). SAXS analysis showed that the 6-subunit middle module was partially aggregated at the high concentrations required for the analysis (Figure 6A and Supplementary Table S3). Thus, we could not reliably determine the radius of gyration and could not calculate low-resolution structural models. Nevertheless, the scattering curve shows clearly that the middle module exhibits an elongated shape (compare 42, Figure 1, e.g. model body 10). The Kratky-plot further indicates that the protein complex is folded and consists of a few globular areas (Figure 6B). Unfortunately, analysis of the 7-subunit middle module by SAXS was hampered by susceptibility of the Med1 subunit to degradation and difficulties in complex concentration.

To add support to the results obtained in solution, we used ion mobility (IM) MS analysis. IM-MS can provide the collisional cross section (CCS), which corresponds to an averaged lateral cut through the protein (43). We applied IM-MS to the three recombinant Mediator middle module complexes (Figure 6C and Supplementary Table S4). Calculation of calibrated CCS revealed conformational flexibility in all complexes, consistent with our inability to crystallize any of these complexes and published in silico analysis data (44). This becomes more clear even when looking at an averaged drift time analysis (Figure 6D and Supplementary Figure S2C). Compared to standard proteins of similar mass (34), the most abundant conformations for the 4- and 6-subunit complexes represented extended structures. The conformation of the complete 7-subunit middle module was, however, more compact (Figure 6C). The average calculated densities for the 4-, 6- and 7-subunit complexes were 0.48, 0.50 and 0.56 g/cm³ respectively. Thus, three different experimental methods are consistent with an elongated shape of the middle module, which is further consistent with published cryo-electron microscopic data (7,9,10).