The human CDK8 subcomplex is a molecular switch that controls Mediator coactivator function

Abstract

The human CDK8 subcomplex (CDK8, cyclin C, Med12, and Med13) negatively regulates transcription in ways not completely defined; past studies suggested CDK8 kinase activity was required for its repressive function. Using a reconstituted transcription system together with recombinant or endogenous CDK8 subcomplexes, we demonstrate that, in fact, Med12 and Med13 are critical for subcomplex-dependent repression, whereas CDK8 kinase activity is not. A hallmark of activated transcription is efficient reinitiation from promoter-bound scaffold complexes that recruit a series of pol II enzymes to the gene. Notably, the CDK8 submodule strongly represses even reinitiation events, suggesting a means to fine tune transcript levels. Structural and biochemical studies confirm the CDK8 submodule binds the Mediator leg/tail domain via the Med13 subunit, and this submodule-Mediator association precludes pol II recruitment. Collectively, these results reveal the CDK8 subcomplex functions as a simple switch that controls the Mediator-pol II interaction to help regulate transcription initiation and reinitiation events. As Mediator is generally required for expression of protein-coding genes, this may reflect a common mechanism by which activated transcription is shut down in human cells.

Figure 1.

Purified factors used for reconstituted transcription on chromatin templates.

Figure 2.

The CDK8 submodule represses transcription independently of its kinase activity. (A) Silver-stained gel showing the purified CDK8 subcomplex. (B) Kinase assays with wild-type (4wt) or kinase-dead (4kd) CDK8 subcomplexes with a known CDK8 substrate, the Rpb1 CTD. (C) Reconstituted transcription on chromatin templates. Each reaction followed the same time line (see Supplemental Fig. 3) and contained identical amounts of template, NTPs, TFIIA, IIB, IID, IIE, IIF, IIH, and pol II; Sp1, Mediator, and submodule were added as shown. (Lane 9) Peptide elution buffer used for CDK8 subcomplex purification had no impact on transcription. Representative data are shown; the graph summarizes data (mean and standard error) from multiple experiments for each experimental condition (n = 7, 7, 5, 12, 4, 4, 4, and 2 for lanes 2–9, respectively).

Figure 3.

Additional Mediator overcomes repression by the CDK8 submodule; Med12/13 are key to submodule-dependent repression. (A) Reactions were completed as described for Figure 2C, except TFIIH or Mediator was doubled in concentration for lanes 4 or 5 and pol II was titrated to 1.5-, two- and threefold higher concentration above standard in lanes 7–9. Representative data are shown; the graph summarizes data (mean and standard error) from multiple experiments for each experimental condition (n = 5, 5, 2, 2, and 5 for lanes 2–6, respectively). (B) Experiments were completed as described for Figure 2C. CDK8 submodules added as shown. (2WT) CDK8/cyclin C binary complex; (4WT) CDK8 subcomplex; (K8/CC/12 or K8/CC/13) three-subunit complexes with Med12 or Med13, respectively; (12/13) Med12/Med13 binary complex (n = 6, 4, 5, 6, 2, 1, 2, 1, 6, 5, and 4 for lanes 3–13, respectively).

Figure 4.

The CDK8 submodule binds directly to core Mediator via Med13. (A) Binding assays with anti-Med26 immobilized core Mediator. Western blots against indicated subunits were performed on anti-Med26 peptide eluates, except for positive control (shown in lane 3). (Lane 1) Anti-Med26 resin only. (Lane 2) Immobilized core Mediator only. (Lane 3) Positive control (CDK8 subcomplex). (Lane 4) Anti-Med26 resin + CDK8 subcomplex. (Lane 5) Immobilized core Mediator + CDK8 subcomplex. (B) 2D classes generated from EM analysis of core Mediator–CDK8 submodule assembly experiment (see the text). Panels 1 and 2 reflect the presence of free CDK8 submodules and core Mediator, respectively. Panel 3 shows a 2D class resembling CDK8-Mediator, suggesting stable interaction between the recombinant CDK8 submodule and core Mediator. Bar, 100 Å. (C) Mediator-binding assays. The complete four-subunit CDK8 subcomplex (4WT) or the partial subcomplexes (as shown) were immobilized and incubated with a partially purified fraction containing Mediator. Peptide-eluted material was analyzed by Western blot against the subunits indicated at the right.

Figure 5.

The human CDK8 submodule interacts with the Mediator leg/tail domain. (A) 3D structure of the recombinant CDK8 submodule, rendered to 600 kDa. (B) Modeled interaction of CDK8 subcomplex with core Mediator. Orientations of core Mediator (semitransparent blue) and the CDK8 subcomplex (red) are based on docking each structure within the entire CDK8–Mediator EM map. The overlaid structures reveal how the hook domain might dovetail with the Mediator leg domain upon interaction. (C) Overlay of docked CDK8 submodule (semitransparent red) onto the CDK8–Mediator structure (white). The submodule overlays are oriented as in A and B. The core Mediator structure is shown for reference and specific domains are indicated (* leg is called the tail domain in the yeast Mediator). Note the CDK8–Mediator and core structures have been published and are bound to VP16 (Taatjes et al. 2002). (D) Core Mediator (blue) docked within the CDK8–Mediator (white mesh) EM map. (E) Independent 3D reconstruction of the three-subunit Med12/CDK8/cyclin C complex (green), rendered to 350 kDa. Docked structures within the entire CDK8 subcomplex (semitransparent red) are also shown. Docking for B–E was completed with Chimera (Pettersen et al. 2004).

Figure 6.

The CDK8 submodule can repress transcription reinitiation. (A) Time line for transcription. The different points (t1, t2, t3, t4) represent times at which the CDK8 subcomplex was added to the reaction. (B) Reconstituted transcription reactions were completed as described for Figure 2C. CDK8 subcomplexes (wild-type or kinase-dead) were added at various time points as shown. Representative data are shown; the graph summarizes data (mean and standard error) from multiple experiments for each experimental condition (n = 5, 2, 2, 2, 2, 5, 2, 2, 2, 2, and 5 for lanes 3–13). Note these data, like those in Figures 2 and 3, represent multiround transcription. (C) Single-round transcription. Experiments were completed as in B except for lanes 3–5, where Sarkosyl was added 90 sec after NTP addition to block transcription reinitiation. The CDK8 submodule was added 3 min prior to NTPs in lane 4, and together with GTFs/Mediator in lane 5 (n = 2, 2, 2, and 3 for lanes 2–5).

Figure 7.

A model for transcriptional regulation by the CDK8 subcomplex. See main text for details. Scaffold PIC components (e.g., TFIID and IIH) are shown transparent with the CDK8–Mediator because their association may be destabilized in this context. The precise architecture of the PIC is not completely known, and the organization shown here is for illustrative purposes only. Individual PIC components are shown at the same relative scale.