PART II: CHARACTERIZATION OF CKM-CMED BINDING IN THE

5.2.1 CKM is not a part of the transcription pre-initiation complex

Having isolated CKM, and CKM-cMed, we asked whether these complexes participate in the transcription pre-initiation complex (PIC). To approach this question, we set up an immobilized template assay (ITA) using a DNA scaffold containing a promoter sequence modeled on the yeast HIS4 promoter, with a TATA-box and a transcription start site (TSS), as a platform for PIC assembly. Biotinylation on the 3’ end allowed immobilization of the template on streptavidin beads, and an EcoRV restriction site downstream of the promoter sequence allowed elution by endonuclease cleavage. CKM(A or KD) in the presence or absence of ATP were incubated on the immobilized template together with purified cMed and purified PIC components; RNAPII and the general transcription factors TFIIA, TFIIB, TFIIE, TFIIF and the TFIID subunit TBP. Excess factors were washed away, and the DNA template was eluted by restriction digestion from the beads. The elutions were then analyzed by SDS-PAGE to investigate what remained bound to promoter DNA (figure 5.8).

CKM alone showed no background binding to the DNA template, but some unspecific binding was seen in the presence of cMed. In the presence of the PIC components, cMed bound stoichiometrically to the PIC, but CKM binding did not exceed background levels, indicating that CKM is excluded from the PIC, whereas cMed binds to the PIC. Moreover, this was so, irrespective of whether active kinase-containing complex CKM(A) or the dead mutant CKM(KD) were used, demonstrating that this effect is predominantly a steric, structural effect, rather than an effect mediated by Cdk8-kinase activity.

5.2.2 CKM directly competes with RNAPII for Mediator binding

To investigate whether the CKM-cMed and the cMed-RNAPII complexes are mutually exclusive, we performed an in vitro competition assay with purified CKM, cMed and RNAPII. In this assay, we took advantage of the MBP-tag on the CKM complex to immobilize it on amylose beads. Having determined that the exclusion of the CKM from the PIC is a structural effect, we used CKM(KD). After binding of CKM to beads, overstoichiometric amounts of cMed were added to ensure binding saturation. The beads were then copiously washed to remove any unbound cMed.

Finally, wash buffer with increasing concentrations of RNAPII was added, and the washes were analyzed by western blot analysis using an antibody against Med17 (alternatively called Srb4 in yeast), one of the subunits of cMed. Presence in the washes of increasing amounts of Med17, hence cMed, correlating to increasing added RNAPII is an indication that RNAPII competes with CKM for cMed binding (figure 5.9). The first lane is a buffer control, where no RNAPII was added. As an additional control, the same experiment was carried out, without addition of cMed.

The rationale behind this control was to ensure that the observed effect is not simply due to trace amounts of co-purifying cMed subunits from the RNAPII endogenous purification, resulting in an increasing signal as RNAPII is added in increasing concentrations, or from a cross-reactivity of the antibody to any of the other experimental components. This experiment provides corroborating evidence in

support of the notion that cMed can interact with the CKM or with RNAPII, but not with both simultaneously, and allowed us to identify RNAPII as the PIC component responsible for preventing the incorporation of the CKM into the PIC.

5.2.3 A shared binding interface of CKM and RNAPII on Mediator accounts for their mutually exclusive binding behavior

To investigate the mode of binding of CKM to cMed, we performed crosslinking-mass spectrometric analysis (XL-MS) of the CKM-cMed complex. To do that, CKM and cMed were mixed together in a 1:1 ratio, dialyzed into lower salt to allow the complex to form, and then crosslinked with 1 mM bis(sulfo)succinimidyl suberate (BS3), after having titrated the crosslinker concentration to a concentration just below complete crosslinking. The crosslinked complex was then run on a native gel to separate the CKM-cMed complex from the two component sub-complexes. CKM-cMed was extracted from the gel, trypsin digested, enriched for crosslinked peptides and subjected to mass spectrometric analysis. The analysis was carried out with a false discovery rate (FDR) cutoff of 1%. The generated crosslinking network (figure 5.10 A), showed extensive crosslinking within the CKM and cMed subcomplexes, but also several high confidence crosslinks between CKM and cMed.

To validate the interaction map, crosslinks between cMed subunits, for which the structure is already known, were mapped onto the structure. Crosslinks used for validation are shown on the interaction map in black, together with the measured atomic distances (figure 5.10 A; 5.11 A). The appearance of many crosslinks within the expected distance constraints confirmed the validity of the crosslinking network (figure 5.11 A).

In addition to crosslinks within cMed, several high confidence crosslinks between CKM and cMed were found. These crosslinks were mapped onto the cMed structure

(figure 5.10 B), and are indicated by red spheres, and corresponding red lines on the interaction network. Crosslinks that could not be mapped exactly, due to their presence in flexible regions that were not visible in the structure, were mapped to the closest structured residue and are indicated by pink spheres, and corresponding pink lines on the interaction network. This means that pink spheres indicate approximate crosslinking locations, and red spheres indicate exact locations, whereas both pink and red spheres refer to high confidence crosslinks. Dashed lines connected to the spheres indicate the complementary residues on CKM. Crosslinks for which more than one unique match was found are emphasized in bold face.

An overwhelming majority of crosslinks between CKM and cMed fall on the RNAPII-binding face of cMed, with a larger number of crosslinks to the cMed middle module, than to the head. In particular, a clustering of crosslinking sites to the CKM subunits Med12 and Med13 is seen on the knob and hook domains of cMed. On the other hand, Cdk8 and CycC contact the head module spine and moveable jaw, respectively.

In figure 5.11 B, the RNAPII-cMed interaction within the PIC is shown. RNAPII is rendered with low opacity to allow seeing through the polymerase into the polymerase-Mediator interfaces. These interaction interfaces are between the Rpb4/7 stalk of the polymerase and arm/spine domain of the cMed head module, between the polymerase dock and the moveable jaw of the cMed head module (made up of the Med18/Med20 heterodimer), and an additional transient interaction between the polymerase foot and the mobile plank domain (Med4/Med9) of the cMed middle module (not shown). These interfaces contain conserved residues, and suggest a conserved nature of the RNAPII-mediator interaction across species. The presence of crosslinks directly on, and in close spatial proximity to, these interfaces, as shown in figure 5.11 B, is unequivocal proof of mutually exclusive binding.

Moreover, it provides a structural framework to explain the observed biochemical competition behavior, by demonstrating that RNAPII and CKM share an overlapping binding surface on cMed.