Cooperative activity of cdk8 and GCN5L within Mediator directs tandem phosphoacetylation of histone H3

Abstract

The human Mediator complex is generally required for expression of protein-coding genes. Here, we show that the GCN5L acetyltransferase stably associates with Mediator together with the TRRAP polypeptide. Yet, contrary to expectations, TRRAP/GCN5L does not associate with the transcriptionally active core Mediator but rather with Mediator that contains the cdk8 subcomplex. Consequently, this derivative 'T/G-Mediator' complex does not directly activate transcription in a reconstituted human transcription system. However, within T/G-Mediator, cdk8 phosphorylates serine-10 on histone H3, which in turn stimulates H3K14 acetylation by GCN5L within the complex. Tandem phosphoacetylation of H3 correlates with transcriptional activation, and ChIP assays demonstrate co-occupancy of T/G-Mediator components at several activated genes in vivo. Moreover, cdk8 knockdown causes substantial reduction of global H3 phosphoacetylation, suggesting that T/G-Mediator is a major regulator of this H3 mark. Cooperative H3 modification provides a mechanistic basis for GCN5L association with cdk8-Mediator and also identifies a biochemical means by which cdk8 can indirectly activate gene expression. Indeed our results suggest that T/G-Mediator directs early events-such as modification of chromatin templates-in transcriptional activation.

Figure 1

TRRAP/GCN5L associates with human Mediator. (A) TRRAP/GCN5L and cdk8-Mediator are enriched in the P0.5 M fraction; core Mediator is enriched in the P1 M fraction. Each lane (PFT, P0.3, P0.5, P1 M) contained 11 μg total protein. (B) Silver-stained gel (7% acrylamide) of glycerol-gradient fractions from VP16 affinity purification. The identities of the subunits are indicated at the right; arrow indicates the obvious ‘unknown' co-migrating polypeptide. (C) Mediator purified from the P0.5 M fraction (Input, lane 1) was incubated with an anti-TRRAP or anti-Med26 antibody resin. Following extensive washing with 0.5 M KCl HEGN, bound proteins were eluted with TRRAP peptide antigen. Mediator remained bound and was eluted from the anti-TRRAP antibody resin only (lanes 2 and 3). A full-colour version of this figure is available at The EMBO Journal Online.

Figure 2

Purification of T/G-Mediator. (A) Western blots against partially purified column fractions. Each lane contained 24 μg total protein. (B) Purification protocol used for isolation of T/G-Mediator. (C) Silver-stained gel (7% acrylamide) showing glycerol gradient fractions from the purification outlined in (B). Subunit identities are shown at the right. (D) Western blots against various subunits from the T/G-Mediator sample. Numbers in parenthesis represent the approximate molecular weight of each protein. Subunits in red are specific to STAGA, whereas Med26 is specific to core Mediator. A full-colour version of this figure is available at The EMBO Journal Online

Figure 3

T/G-Mediator contains the cdk8 submodule. (A) Representative 2D classes of cdk8-Mediator and core Mediator. Scale bar: 150 Å. (B) Silver-stained gradient (5–15% acrylamide) gel showing the T/G-Mediator sample used for EM analysis. Note that only prominently staining bands are identified and smaller, poorly staining polypeptides representing other consensus Mediator subunits are not labelled (Sato et al, 2004). Asterisk: insulin (used for TCA precipitation). (C) EM analysis indicates that T/G-Mediator sample has a size and shape consistent with cdk8-Mediator. Representative 2D classes are shown; each class (1–4) was obtained by averaging ∼150 aligned single-particle images. Scale bar: 150 Å. A full-colour version of this figure is available at The EMBO Journal Online

Figure 4

T/G-Mediator cooperatively phosphoacetylates histone H3. (A) Recombinant core histones alone (lane 1), T/G-Mediator alone (lane 2) or T/G-Mediator together with recombinant core histones (lane 3) were incubated with ATP and acetyl CoA. Reactions were then probed with an antibody specific to the doubly modified S10/K14 H3. (B) ATP enhances T/G-Mediator acetylation specificity for H3. Acetylation assay with T/G-Mediator and core histones in the absence (lane 2) and presence (lane 4) of ATP. Lane 1 and 3: loading reference for lanes 2 and 4. Quantitation of the [¹⁴C]-acetyl label is shown to the right of lanes 2 and 4. (C) T/G-Mediator-catalysed acetylation of H3K14 is reduced in H3S10A mutant. Equivalent amounts of T/G-Mediator were incubated with wild-type (WT) or mutant (S10A) H3 tails in the presence of ATP and [¹⁴C]-acetyl CoA. Representative data are shown; quantitation of the [¹⁴C]-acetyl label is shown beside each radiolabelled band, with WT normalized to 100. The value shown for S10A is the average of four independent experiments (σ=8.9). Representative loading controls (silver stain) are shown for each lane. A full-colour version of this figure is available at The EMBO Journal Online

Figure 5

TRRAP and cdk8 are co-recruited to the promoter regions of active human genes; cdk8 knockdown depletes global levels of phosphoacetylated H3. (A) p53-dependent activation of CDKN1A in HCT116 cells. cDNA samples were prepared after addition of 10 μM Nutlin-3 to the culture media. Values are expressed as fold induction over untreated cells after normalization to 18S rRNA. (B) Serum-induced activation of egr1 in HCT116 cells. cDNA samples were prepared after serum replenishment, following 24 h of serum withdrawal. (C) ChIP analysis of the CDKN1A locus. HCT116 cells were treated with vehicle or 10 μM Nutlin-3 for 8 h before harvest and extract preparation. IPs were performed with antibodies against pol II, TRRAP and cdk8. ChIP-enriched DNA was analysed by Q–PCR with four different amplicons (A–D) covering different regions of the CDKN1A locus. (D) ChIP analysis of the egr1 locus. Serum-starved HCT116 cells were stimulated with serum for 30 min before harvest. ChIPs were performed as in (C). Error bars represent standard deviations on triplicate experiments. (E) HCT116 cells were transduced with control or anti-cdk8 shRNAs and analysed by Western blot. Quantitation of knockdown is reported as % reduction from control cells (^*average of two experiments; % reduction measured was 92/92 for cdk8, 86/75 for H3S10P/K14Ac, (−10)/16 for nucleolin (loading control)).

Figure 6

In vitro transcription on chromatin templates. (A) Unlike core Mediator, T/G-Mediator does not directly activate transcription. Core Mediator is required to reconstitute activator-dependent transcription in vitro on chromatin templates (compare lanes 1–4); T/G-Mediator blocks core Mediator-dependent activation in a dose-dependent manner (compare lanes 4–7); T/G-Mediator is unable to activate transcription even in the presence of activator (lanes 8–11). (B) PARP-1 does not activate T/G-Mediator. Titration of PARP-1 (20, 50 or 120 ng) in the presence of T/G-Mediator and Sp1 (lanes 6–8). Quantitation of RNA transcripts is shown in the bar graph below each lane.