|Home | About | Journals | Submit | Contact Us | Français|
The Mediator complex allows communication between transcription factors and RNA polymerase II (RNAPII). CDK8, the kinase found in some variants of Mediator, has been characterized mostly as a transcriptional repressor. Recently, CDK8 was demonstrated to be a potent oncoprotein. Here we show that CDK8 is a positive regulator of genes within the serum response network, including several members of the AP-1 and EGR family of oncogenic transcription factors. Mechanistic studies demonstrate that CDK8 is not required for RNAPII recruitment or promoter escape. Instead, CDK8 depletion leads to the appearance of slower elongation complexes carrying hypophosphorylated RNAPII. CDK8-Mediator regulates precise steps in the assembly of the RNAPII elongation complex, including the recruitment of P-TEFb and BRD4. Furthermore, CDK8-Mediator specifically interacts with P-TEFb. Thus, we uncovered a novel role for CDK8 in transcriptional regulation that may contribute to its oncogenic effects.
In eukaryotes, the transcription of all protein-coding genes is accomplished by a single enzyme, RNAPII. DNA-binding transcription factors and a myriad of coregulators enable activated gene transcription by RNAPII in a cell type- and signalling-specific manner. Mediator is a large protein complex found in all eukaryotes that facilitates activator-dependent transcription, yet its mechanism of action is ill defined 1,2. Mediator is considered to be an integral part of the Pre-Initiation Complex (PIC), which also includes the general transcription factors (GTFs) TFIIA, TFIIB, TFIID, TFIIE, TFIIF and TFIIH1,2. The fact that Mediator interacts directly with both transactivators and RNAPII supports the notion that it may act as a molecular bridge to facilitate activator-induced recruitment of RNAPII onto promoters3–5. However, recent genome-wide studies demonstrate that transcriptional activation at post-RNAPII recruitment steps is a widespread phenomenon in metazoans. Paused RNAPII is found at the promoters of most genes induced by stress, developmental cues and various signalling events6–9. Upon stimulus, RNAPII is released from promoters and converted into an elongation competent form by poorly understood mechanisms. P-TEFb (Positive Transcription Elongation Factor-b) is recognized as a critical elongation factor required for most RNAPII-dependent transcription10. Recruitment of P-TEFb by transactivators, the viral protein Tat or the chromatin-binding protein BRD4 stimulates elongation in a number of experimental systems11–13. CDK9, the kinase subunit of P-TEFb, phosphorylates Ser2 of the C-terminal domain (CTD) heptad repeats (YSPTSPS) of RPB1, the largest subunit of RNAPII14–16. CTD phosphorylation has been implicated in regulation of RNAPII elongation and co-transcriptional processing of nascent RNAs15,17–22. CDK7, the kinase subunit of TFIIH, can phosphorylate both Ser5 and Ser7 and mediates 5’ capping of mRNAs21–25. CDK8, a Mediator-associated kinase, can phosphorylate both Ser 2 and Ser5 in vitro, with a preference for Ser523,25, but its precise contribution to CTD phosphorylation in vivo remains to be determined. Interestingly, positive functions for Mediator at post-RNAPII recruitment steps have been documented. In mice, knock-out experiments showed that the MED23 subunit is required for expression of the EGR1 gene, yet the effects of MED23 depletion could not be merely explained by impaired RNAPII recruitment26. Within the p53 network, Mediator is strongly recruited during activation of genes carrying paused RNAPII, such as p21 and MDM2, suggesting that it acts at post-RNAPII recruitment steps at these loci27. The mechanisms by which Mediator may facilitate gene activation after transcription initiation remain to be elucidated.
Multiple variants of the human Mediator complex have been isolated that can be categorized as either ‘core’ Mediator or CDK8-Mediator based on the absence or presence of the CDK8-submodule1,28. The CDK8-submodule is composed of CDK8, cyclin C, MED12 and MED13 and has been characterized mostly as a transcriptional repressor. For example, in vitro assays show that addition of the CDK8-module shuts down transcription driven by several transactivators29. This repression is independent of CDK8 kinase activity and it is likely due to disruption of Mediator-RNAPII interactions29,30. In vivo, activation of RARα and C/EBP-β target genes correlates with a switch from CDK8-Mediator to core Mediator31,32. Additionally, the MED12 subunit mediates the recruitment of repressive histone methyl-transferases in vivo33. However, several reports have shown that CDK8 can also function as a positive regulator of gene activity. In yeast, chemical genetics experiments demonstrated that SRB10, the CDK8 ortholog, collaborates with KIN28, the CDK7 ortholog, to promote RNAPII transcription both in vitro and in vivo34. In human cells, CDK8 is recruited to specific p53 target genes exclusively during conditions of activation, and CDK8 depletion impairs induction of select genes within this network27. Recently, CDK8 was shown to positively regulate β-catenin-dependent transcription in human cells35. Furthermore, CDK8 can collaborate with GCN5L to phosphoacetylate histone H3, a mark that correlates with activation36.
A recent report provides several lines of evidence that establish CDK8 as a potent oncoprotein in colon cancer: 1) CDK8 amplification and overexpression is commonly observed in colon tumors, 2) CDK8 knock-down decreases proliferation of several colon cancer-derived cell lines, 3) CDK8 overexpression leads to focus-formation, anchorage-independent growth and tumor formation in immunodeficient animals35. These results could be explained in part by the fact that CDK8 is required for transcriptional activation by the oncogenic transcription factor β-catenin/TCF 35,37. However, only a fraction of the malignant effects of CDK8 overexpression could be blocked by a dominant negative form of TCF, suggesting that CDK8 participates in other oncogenic pathways35. Given the prominent role of growth factor signalling in cancer development, we hypothesized that CDK8 could have functions within the serum response network. Indeed, we found that human CDK8 is a potent positive regulator of immediate early genes (IEGs), which are strongly transcribed within minutes of growth factor signalling. Microarray experiments in human cells demonstrate that CDK8-dependent coactivation, rather than repression, is predominant within this gene network. These effects are direct, as CDK8 is strongly recruited to these genes during activation and CDK8 depletion leads to impaired RNAPII activity at these loci. We found that CDK8 is not required for recruitment of RNAPII to promoters and overall RNAPII promoter escape. Instead, nuclear run-on experiments demonstrate that RNAPII elongation is impaired upon CDK8 knock-down, which correlates with decreased levels of RNAPII CTD phosphorylation at Ser2 and Ser5. CDK8 regulates precise steps during the assembly of a functional elongation complex, as CDK8 depletion impairs the orchestrated recruitment of CDK7, P-TEFb and its associated factor BRD413,38. More interestingly, we show that P-TEFb interacts with both the free CDK8-submodule and CDK8-Mediator, suggesting that the CDK8-submodule may help load P-TEFb at the promoters of serum response genes. Thus, our results demonstrate a novel role for CDK8 in transcriptional regulation at post-RNAPII recruitment steps.
To determine the role of CDK8 in the expression of serum responsive genes, we analyzed mRNA expression using microarray technology in HCT116 colon cancer cells where CDK8 expression was stably knocked-down. shRNA-mediated depletion of CDK8 results in an 80–90% reduction of CDK8 mRNA and total protein levels (Fig. 1a). Though CDK8 is required for proliferation of multiple colon cancer cell lines, CDK8 is not amplified in HCT116 cells and has a more modest effect on their proliferation, which enabled us to perform the experiments described in this report. Expression of CDC2L6 (CDK8-L), a CDK8 paralog also associated with Mediator28, was not affected by our shRNA (Supplementary Fig. 1a). We identified 29 transcripts that were upregulated >2-fold after 30 minutes of serum stimulation in wild type cells. Roughly 90% (26/29) of these immediate early genes (IEGs), were expressed at significantly lower levels (<2-fold) in serum-stimulated shCDK8 cells (Fig. 1b). Only 3 serum-responsive genes (THBS1, PLK2 and KLF6) were expressed at higher levels in shCDK8 cells. A number of IEGs, such as members of the AP-1 (Activating Protein 1) and EGR (Early Growth Response) families of transcription factors stood out (FOS, JUN, FOSB, JUNB and EGR1, 2, 3). Interestingly, these are potent oncogenes acting downstream of MAPK-dependent signalling that are activated by the ubiquitous serum response factor (SRF) and Ets-like transcription factors, such as ELK139. Validation of microarray results by Q-RT-PCR showed that CDK8 depletion leads to decreased expression of all genes analyzed over multiple time points (Fig. 1c and Supplementary Fig. 1b). Of note, the effect of CDK8 knock-down is more drastic during the early time points, when increased gene expression is driven by transcriptional activation and less influenced by subsequent transcriptional shut-down and mRNA degradation. The effects of CDK8 knock-down on IEG expression are not due to off-target effects of a particular shRNA sequence, because reduced expression of IEGs is also observed when targeting a different sequence within the CDK8 mRNA (Supplementary Fig. 1c).
To test whether the impact of CDK8 on IEG mRNA expression is due to a direct role of CDK8 in transcriptional activation at these loci we performed chromatin immunoprecipitation assays (ChIP) for CDK8, total RNAPII and two phosphorylated isoforms of RNAPII. ChIP analysis of CDK8 showed that prior to stimulation there is little to no CDK8 at the promoters of FOS and EGR1, 2 and 3. However, serum treatment of wild type cells produces a strong increase in CDK8 recruitment (7–14 fold) (average data for FOS and EGR1-3 is shown in Fig. 2, statistical analysis is shown in Supplementary Fig. 2). As expected, CDK8 occupancy decreases 4–9 fold in shCDK8 cells across the loci studied. ChIP analysis indicates that CDK8 is not required for the recruitment of RNAPII to all four IEGs, but does have a positive affect on CTD phosphorylation (Fig. 2). RNAPII is preloaded at the FOS proximal promoter with only a small increase in occupancy after serum treatment. However, a clear increase in RNAPII is observed at the FOS and EGR1 intragenic regions during activation, indicative of enhanced elongation. Interestingly, CDK8 knock-down produces no substantial effects on total RNAPII occupancy at any of the locations tested, before or after activation. This is more evident at the non-preloaded EGR2 and EGR3 genes, which show a sharp increase in RNAPII occupancy at their proximal promoters upon serum treatment (~4–7 fold) that is refractory to CDK8 depletion (Fig. 2). RNAPII CTD phosphorylation at Ser5 and Ser2 is linked to promoter clearance and elongation14,15. Accordingly, both phosphorylation marks increase sharply upon activation at IEGs in wild type cells. Typically, Ser5 phosphorylation peaks near the transcription start site, whereas Ser2 phosphorylation accumulates at the 3’ end of genes18. Depletion of CDK8 clearly reduces the level of Ser5 phosphorylation (S5P) at all three EGR genes, however the effects on FOS are less pronounced. More strikingly, CDK8 knock-down strongly reduces the level of phospho-Ser2 (S2P) at all four genes (Fig. 2). Importantly, CDK8 knock-down does not affect cellular levels of either mark as measured by western blot (Supplementary Fig. 3a), indicating that CDK8 effects on CTD phosphorylation are gene-specific rather than global. Of note, CDK8 knock-down does not affect mRNA expression or RNAPII activity at the housekeeping gene GAPDH (Supplementary Fig. 3b and 3c). In summary, our results indicate that CDK8 is playing a role in transcriptional activation events subsequent to recruitment of RNAPII at serum-responsive loci.
Intriguingly, our ChIP analysis did not detect significant differences in RNAPII occupancy within IEGs intragenic regions, yet clear effects on CTD phosphorylation and mRNA steady-state levels were observed. CTD phosphorylation is known to affect both RNAPII elongation and cotranscriptional RNA processing. To investigate the impact of CDK8 depletion on these events, we first measured the levels of unprocessed primary transcripts using PCR primers amplifying intronic regions. Clearly, CDK8 depletion causes a sharp decrease in the levels of FOS and the EGR pre-mRNAs produced upon serum stimulation (Fig. 3a), suggesting that the CDK8-dependent effects on CTD phosphorylation result in a defect in the rate of transcription rather than co-transcriptional RNA processing. To test this, we performed nuclear run-on assays in conditions that prevent re-initiation. Serum starved cells were treated with serum for 10 minutes before active elongation complexes were stalled and nuclei isolated. Transcription was then allowed to progress for either 5 or 60 minutes before nuclei were lysed. Analysis of primary transcripts generated during the run-on shows that at the 5 minute time point RNA levels of FOS, EGR1-3 are 2–20 fold higher in control cells as compared to shCDK8 cells; however, by the 60 minute time point the RNA levels are either equivalent or much more similar (Fig. 3b). This provides strong evidence that CDK8 is positively affecting the actual rate of transcription. Although RNAPII still elongates in CDK8-depleted cells, it does so at a slower pace.
In agreement with the fact that CDK8 plays no role in RNAPII recruitment, we found that CDK8 depletion did not affect recruitment or activation of the key transactivators acting at the FOS promoter. The SRF-ELK1 complex is constitutively bound to the chromatin of its target genes and is activated by MAPK-dependent phosphorylation39. Phosphorylation of the C-terminal domain of ELK1 has been shown to be important for the allosteric stimulation of histone acetyl-transferase (HAT) activities40. CDK8 depletion did not affect the levels of chromatin bound SRF or ELK1 (condensed data is shown Fig. 4, statistical analysis is shown in Supplementary Fig. 4). Likewise, we did not observe differences in ELK1 phosphorylation or overall histone acetylation (AcH4) (Fig. 4a). Expectedly, CDK8 did not affect the chromatin-bound levels of Pre-Initiation Complex (PIC) components involved in RNAPII recruitment, such as TBP or TFIIB (Fig. 4a).
It has been demonstrated that phosphorylation of ELK1 triggers association with the MED23 subunit of Mediator and that MED23 is required for activation of certain ELK1-target genes41. Our ChIP analysis confirms that MED23 is recruited to FOS and EGR1 upon serum stimulation (Fig. 4b). Interestingly, the levels of chromatin-bound MED23 are decreased by roughly 50% in CDK8 knock-down cells, suggesting that CDK8 helps stabilize Mediator association at these loci. This notion is supported by the fact that two other Mediator subunits, MED1 and MED12, also show decreased association to the FOS and EGR1 promoters upon CDK8 depletion (Fig. 4b). Whereas CDK8 knock-down may have a modest effect on MED12 stability, it does not decrease the cellular pools of MED23 and MED1 (Supplementary Fig. 3a).
Given the clear effects of CDK8 knock-down on RNAPII CTD phosphorylation and elongation rates at IEGs, we next focused on complexes containing CTD-kinases. We found that recruitment of both CDK7 and CDK9 to the FOS and EGR1 loci is impaired upon CDK8 knock-down (average data for FOS and EGR1 is shown in Fig. 5, statistical analysis is shown in Supplementary Fig. 5). In contrast to CDK7 and CDK9, the recruitment of other factors regulating RNAPII pausing and elongation, such as NELF, DSIF and FACT, is not affected by CDK8 knock-down. NELF and DSIF (a dimer of SPT4 and SPT5) are key regulators of RNAPII pausing at promoters42,43. Upon promoter escape, SPT5 associates with RNAPII and functions as a positive elongation factor43,44. FACT (a dimer of SSRP1 and SPT16) is another positive regulator of elongation required for transcription-coupled nucleosome remodelling45. Our ChIP analysis demonstrates that NELF-A and SPT5 occupy the FOS promoter carrying preloaded RNAPII. Upon activation, NELF-A remains at the transcription start site whereas SPT5 associates with the intragenic region. In contrast, SPT16 is recruited strongly to the gene body upon serum treatment (Fig. 5). Occupancy profiles for all three factors are unaffected by CDK8 depletion during both basal and stimulated conditions. These results indicate that CDK8 orchestrates key events in the formation of a functional elongation complex, namely, recruitment of CDK7 and CDK9 and subsequent RNAPII CTD phosphorylation.
The bromodomain protein BRD4 has been recently implicated in recruitment of P-TEFb to several human genes38,46. BRD4 interacts with both acetylated histones and P-TEFb, thus linking chromatin modifications to elongation control. Importantly, several reports described an interaction between BRD4 and Mediator, yet the relevance of this interaction has not been defined38,47–49. Our ChIP analyses demonstrate that BRD4 is strongly recruited to FOS and EGR1 upon serum stimulation with a profile similar to that of CDK9 and Mediator. Interestingly, BRD4 recruitment is significantly decreased upon CDK8 knock-down. Strikingly, our detailed analysis of histone acetylation at FOS and EGR1 indicates that BRD4 recruitment is most likely due to its interaction with Mediator rather than its association with the acetyl-lysines previously described to be recognized by its double bromodomain. Hargreaves et al concluded that BRD4 detects histone H4 acetylated at lysines 5, 8 and 1246. In contrast, Zippo et al concluded that BRD4 binds instead histone H4 acetylated at lysine 16 and histone H3 acetylated in lysine 938. Our ChIP analyses using antibodies against tetra-acetyl H4 (lysines 5, 8, 12, 16), acetyl-H4-K8 (H4 K8Ac), acetyl-H4-K12 (H4 K12Ac), acetyl-H4-K16 (H4 K16Ac) and acetyl-H3-K9 (H3 K9Ac) demonstrate that none of these marks is affected by CDK8 knock-down and that the occupancy profiles of these marks do not resemble that of BRD4 (Figs. 4a, ,5a5a and Supplementary Figs. 5 and 6).
Our results indicate that many of the effects of CDK8 knock-down on RNAPII activity at post-recruitment steps could be explained by decreased P-TEFb recruitment. To test for the impact of CDK9 activity on IEG activation, we employed the pharmacological inhibitor flavopiridol10. Pre-treatment of cells with flavopiridol (150 nM) prevents serum-induced mRNA accumulation of FOS, EGR1-3 (Fig. 6a). ChIP analysis demonstrates that CDK9 inhibition does not decrease the levels of promoter-bound RNAPII at IEGs carrying pre-loaded RNAPII (e.g. EGR1) or those displaying induced RNAPII recruitment (e.g. EGR2) (Fig. 6b). Clearly, inhibition of CDK9 by flavopiridol has a more pronounced effect on RNAPII occupancy at intragenic regions than CDK8 depletion, which could be explained by the fact that some amount of P-TEFb is still recruited to these loci in CDK8-knock-down cells. However, a small but significant amount of intragenic RNAPII is still detected in flavopiridol-treated cells. These RNAPII complexes presumably elongate at such low rates that prevent significant mRNA accumulation. Expectedly, flavopiridol treatment completely blocks RNAPII CTD phosphorylation in Ser2. Flavopiridol also blocks Ser5 phosphorylation in intragenic regions, but has no effect on promoter-proximal Ser5 phosphorylation. Overall, these results suggest that decreased CDK9 activity at IEGs upon CDK8 depletion could explain the post-recruitment defects observed in CDK8 knock-down cells.
A role for Mediator in recruitment of TFIIH has already been described in other systems50, which could explain the reduced association of CDK7 to IEG promoters upon CDK8 depletion. However, it is unclear how CDK8-Mediator may impact P-TEFb recruitment. At least two variants of the human Mediator complex exist within the cell nucleus1. CDK8-Mediator complexes contain the CDK8-submodule and are devoid of the MED26 subunit, which is unique to ‘core’ Mediator, whereas the MED1 subunit is found in both. We have previously reported that a large fraction of the CDK8-submodule exists separate from core Mediator51. Serendipitously, research efforts to identify factors interacting with Mediator showed that P-TEFb associates specifically with CDK8-Mediator. In these experiments, antibodies against CDK8 or MED1 were used to immunoprecipitate Mediator from nuclear extracts. After a series of high-salt washes, bound proteins were eluted and identified using MudPIT methodology. Interestingly, multiple peptides for CDK9 and cyclin T were detected in the CDK8 IP, but not in the MED1 IP (Ebmeier and Taatjes, unpublished results). These results were confirmed by quantitative western blot experiments (Fig. 7b). Expectedly, subunits of the CDK-submodule (MED12, MED13 and Cyclin C) are enriched in the CDK8 IP, whereas MED26 was not detected in these CDK8 IP samples. In agreement with several independent reports38,47–49, we also observed an association of BRD4 with Mediator. Interestingly, and in agreement with Wu and Chiang, we find that BRD4 associates with Mediator regardless of the presence of the CDK8-submodule48. Importantly, quantitative western blotting clearly shows enrichment of the P-TEFb subunits CDK9 and Cyclin T1 in the CDK8-Mediator sample (Fig. 7b).
In order to investigate if P-TEFb interacts with CDK8-Mediator and/or the free CDK8-submodule, we tested for the presence of CDK9 in fractions enriched for these complexes via biochemical fractionation. Using purification methods previously described51, we found that CDK9 copurifies with both CDK8-Mediator and the free CDK8-module (Fig. 7c). These results suggest that the CDK8-module may facilitate association of P-TEFb with the rest of Mediator to regulate CTD phosphorylation and transcriptional elongation at specific gene loci such as the IEGs studied here.
The CDK8-submodule of Mediator is a pleiotropic regulator of gene expression. On one hand, this protein complex has been characterized as a transcriptional repressor that inhibits RNAPII by multiple mechanisms, including: 1) kinase-independent allosteric regulation of Mediator-RNAPII interactions29; 2) kinase-dependent inactivation of TFIIH via phosphorylation of its cyclin H subunit52 ; 3) gene silencing via recruitment of histone methyl-transferases33. In contrast, several reports indicate that CDK8 is a positive regulator of gene activity in other scenarios. In yeast, SRB10 promotes RNAPII activity in vitro and in vivo34. In human cells, CDK8 depletion impairs activation of genes responsive to p53 and β-catenin27,35. Additionally, CDK8-Mediator complexes can perform histone modifications associated with gene activation36. In this report, we provide evidence that CDK8 is a direct positive regulator of transcription within the serum response network. We found that CDK8 is dispensable for RNAPII recruitment to promoters carrying pre-loaded RNAPII (FOS, EGR1) as well as promoters displaying inducible RNAPII recruitment (EGR2, EGR3). Furthermore, CDK8 depletion did not produce a drastic impairment in RNAPII escape from promoter regions, as significant amounts of RNAPII were detected throughout the intragenic regions of these genes upon activation. Instead, decreased IEG expression in CDK8 knock-down cells is explained by the appearance of slower elongation complexes containing hypophosphorylated RNAPII. The findings in this report represent a marked advance in our understanding of how CDK8 promotes gene activity, but also prompt more general questions about the role of CTD phosphorylation, CTD kinases and Mediator in mammalian gene expression control.
Here, we establish a strong correlation between CDK8 status, CTD phosphorylation and RNAPII elongation rates. Although some effect of CDK8 depletion on IEG RNA processing can not be fully discarded, our results indicate that most of the negative effects of CDK8 knock-down on IEG expression are explained by decreased elongation rates. In nuclear run-on experiments, CDK8 depletion decreases elongation rates by ~2-3-fold at the FOS locus and >10-fold at the EGR loci. Importantly, these differences are attenuated when additional time is allowed for elongation complexes derived from CDK8-depleted cells to proceed, with the FOS locus showing equivalent RNA synthesis and the EGR loci reaching up to 30–60% of wild type levels by 60 minutes. Interestingly, the impact of CDK8 on CTD phosphorylation is more pronounced at the EGR loci, further supporting a link between CTD phosphorylation and elongation rates. The notion that decreased transcription, rather than impaired processing, causes lower IEG mRNA expression in CDK8-depleted cells is reinforced by the fact that these cells accumulate lower levels of pre-mRNAs.
Despite the fact that CDK8 is a CTD-kinase and that its depletion leads to hypophosphorylated CTD and slower elongation complexes, a simple interpretation of these results is complicated by the fact that recruitment of two other CTD-kinases, CDK7 and CDK9, is impaired by CDK8 depletion. Until recently, the prevalent view on CTD phosphorylation maintained that CDK7 was the major Ser5-kinase and CDK9 the predominant Ser2-kinase. This view originated with the early characterization of CDK7 and the CDK9-like kinases in yeast16,53. However, recent chemical genetic experiments in yeast and human cells showed that inhibition of CDK7 kinase activity does not result in global defects in Ser5 phosphorylation24,54,55; but CDK7 may instead be the major kinase for Ser7, an alternate phospho-acceptor site within the CTD24,56. Interestingly, CDK9 may be a relevant Ser5-kinase in certain contexts57,58. CDK8 is known to phosphorylate both Ser2 and Ser5 in vitro23, but its in vivo contributions remain to be elucidated. Interplay between the CTD kinases is supported by the fact that Ser5 phosphorylation by CDK7 ‘primes’ the CTD for subsequent Ser2 phosphorylation by CDK959. Yeast Mediator was shown to be required for TFIIH recruitment to specific promoters and our data suggest that a similar interaction between CDK8-Mediator and TFIIH may also take place at IEGs50. Additionally, our data indicate that P-TEFb association with the RNAPII machinery is also facilitated by CDK8-Mediator and we provide evidence of a physical interaction between the two complexes.
Though much is known regarding the critical role of P-TEFb in elongation control, little is known about the mechanisms by which this complex is recruited to allow for signalling- and gene-specific regulation. Several modes of P-TEFb recruitment have been described, including direct interactions with DNA-binding proteins or with the bromodomain protein BRD413,38,48,49,60. Interestingly, BRD4 has been shown repeatedly to interact with Mediator and our results confirm this interaction. Furthermore, our results agree with those of Wu and Chiang who found that BRD4 associates with both CDK8-Mediator and core Mediator48. In contrast, we find that P-TEFb associates preferentially with CDK8-Mediator and the free CDK8-submodule. These observations suggest that BRD4 and P-TEFb interact with Mediator by different means.
Early evidence that Mediator regulates RNAPII activity at post-recruitment steps was demonstrated by Wang et al26. Using MED23 null mice, they found that abolishing Mediator recruitment to the EGR1 locus drastically impairs transcriptional activation without affecting binding of TFIIA, TFIID, histone acetylation, methylation or association of the chromatin remodelling factor BRG1. Importantly, the defects in EGR1 transcription observed in MED23−/− cells could not be merely explained by reduced RNAPII association, as a significant fraction of RNAPII remained associated with the promoter in mutant cells. Thus, they concluded that Mediator affected both recruitment and post-recruitment steps. Given that RNAPII occupancy remains unaffected in CDK8 knock-down cells, we have been able to more clearly define the contribution of Mediator to post-recruitment steps. A significant fraction of MED23 remains associated with the promoter upon CDK8 depletion, which may explain why we do not observe any recruitment defects. Instead, most effects of CDK8 knock-down on IEG expression can be explained by impaired P-TEFb recruitment. This notion is supported by the fact that CDK9 inhibition by flavopiridol resembles (albeit with more penetrance) the effects of CDK8 knock-down. Given past results that demonstrate mutually exclusive association of CDK8-submodule and RNAPII with Mediator28,29, our data further suggest dynamic association of the CDK8-submodule at the promoter of active serum response genes. Taken together, these results invoke a mechanism involving CDK8-submodule exchange at the promoter to regulate RNAPII and P-TEFb association within the transcriptional apparatus at post-recruitment steps.
CDK8 is a potent oncogene in colon cancer. CDK8 overexpression promotes cell proliferation, anchorage-independent growth, and tumor growth in xenografts35. Part of these effects can be attributed to the fact that CDK8 promotes transcription mediated by the oncogenic transcription factor β-catenin/TCF35. Here, we report that CDK8 is a potent positive regulator of the serum response network, which has been implicated in multiple tumorigenic phenotypes. The precise contributions of the β-catenin and serum response networks to the overall oncogenic effects of CDK8 overexpression await further experimentation. Whereas the transformation-inducing effects of β-catenin can be effectively blocked by overexpression of a dominant negative version of its binding partner TCF35, the equivalent task for the serum response network is not straightforward. Upon serum stimulation, ERKs translocate to the nucleus and activate a myriad of transcription factors, including multiple members of the Ets family, MYC and C/EBP-β39. Possible functional redundancy within the Ets family will impose a combinatorial genetic approach to determine their precise interaction with CDK8 during tumor development. Nonetheless, the fact that many MAPK-regulated transcription factors and their target genes have well demonstrated roles in cancer development will make this a worthy task that would lead to a detailed understanding of the oncogenic effects of CDK8 overexpression.
HCT116 cells were maintained in McCoys 5A media supplemented with 10% (v/v) fetal bovine serum (Hyclone) and antibiotic/antimycotic mix (Gibco-Invitrogen). HEK293FT cells were grown in DMEM media supplemented with 10% (v/v) fetal bovine serum and antibiotic/antimicotic mix (Gibco-Invitrogen). For western blot, all Q-RT-PCR assays and ChIP, cell cultures at 50–70% confluency were washed twice with PBS, starved in serum free media for greater than 40 hours and then treated with serum-containing media for the indicated times. For flavopiridol treatments (150 nM), the drug was added 1 h prior to serum stimulation.
Stable clones constitutively expressing a short-hairpin RNA targeting CDK8 (shCDK8) were generated using the lentiviral-based shRNA delivery vector pLL3.7. The CDK8 mRNA sequence targeted by the shCDK8 is 5’-GGATGATAAAGACTATGCT-3’. Viral particles were produced in HEK293FT cells and used to transduce HCT116 cells, which were subsequently cloned in the presence of 1 μg ml−1 G418 (Sigma). For transient CDK8 knock-down, cells were starved two hours prior to transfection, with control siGlo siRNA (Dharmacon) or CDK8 specific siRNA (Invitrogen), using Lipofectamine (Invitrogen) per the manufacturers protocols. 40 hours post-transfection, cells were treated with complete media for the indicated times before harvesting for Q-RT-PCR analysis. The sequence targeted by the CDK8 siRNA is 5’-CCCAATAAAGCGAATTACCTCAGAA-3’.
These assays were performed as described in58. For western blots, 10μg of total protein extract were loaded onto SDS-PAGE gels, transferred to PVDF membranes and blotted with the appropriate antibodies. Antibody information can be found in Supplementary Table 1.
Total RNA was isolated using the RNeasy Extraction Kit (Qiagen) and cDNA was generated using the iScript cDNA Synthesis Kit (BioRad). Relative mRNA levels, normalized to 18S rRNA, were determined using SYBR Green I chemistry in an ABI Prism 7900 HD RT-PCR machine (Applied Biosystems). See Supplementary Table 2 for primer sequences.
Serum starved HCT116 cells were left untreated or treated with 10% (v/v) FBS McCoy’s 5A for 30 minutes. RNA was isolated as described above for Q-RT-PCR. Duplicate biological samples were treated according to manufacturers protocols and hybridized to the Human Gene 1.0 ST Arrays (Affymetrix). The raw data from the *.CEL files was imported into SpotFire (TIBCO), normalized using Robust Multi-chip Average (RMA) and analyzed using the Spotfire Decision Site.
ChIP was performed as previously described58. Briefly, serum starved HCT116 cells were either treated or not with complete media for 15 minutes, fixed with 1% (v/v) formaldehyde, harvested for whole cell lysate preparation and ChIP enriched DNA was analyzed by quantitative-PCR as described58. Antibody information can be found in Supplementary Table 1 and oligonucleotide sequences in Supplementary Tables 3–7.
Serum starved HCT116 cells were either untreated or treated with 10% (v/v) FBS McCoy’s 5A for ten minutes prior to isolation of nuclei. Nuclei from 107 cells were isolated in Run-on Lysis Buffer (10 mM Tris-HCl pH 7.5, 5 mM MgCl2, 10 mM NaCl and 0.5% (v/v) NP-40). Nuclei were washed once in 1 x Run-on Reaction Buffer without NTPs (20 mM Tris-HCl pH 7.5, 10 mM MgCl2, 150 mM KCl, 20% (v/v) glycerol), then the run-on was performed for the indicated times at 37 C in the presence of Run-on Reaction Buffer with NTPs (1 x Run-on Reaction Buffer, 0.06% (w/v) Sarkosyl, 0.5 mM ATP, CTP and GTP and either 0.2 mM UTP or 0.2 mM biotin-16-UTP). Run-on reactions were stopped by snap freezing in liquid nitrogen, RNA was purified using RNeasy Extraction Kit (Qiagen) and labeled RNA was further purified using streptavidin-coated Dynabeads (Invitrogen). cDNA was generated and quantified using methods described above for Q-RT-PCR.
Immunoprecipitation of Mediator complexes was performed by using anti-CDK8 (C-19 Santa Cruz Biotechnology) or anti-MED1 (M-255 Santa Cruz Biotechnology) immobilized on 50 μl Protein A/G Sepharose (Amersham Biosciences). After rocking approximately 2 hours at 4 C, beads were washed three times with 20 column volumes 0.5 M KCl HEGN (20mM Hepes, pH7.6; 0.1 mM EDTA; 10% (v/v) Glycerol; 0.1% (v/v) NP-40 alternative; 1 mM DTT; 1 mM benzamidine; 0.25 M PMSF; 2 μg ml−1 aprotinin) and twice with 20 column volumes 0.15 M KCl HEGN, 0.02% (v/v) NP-40 alternative and incubated with 1.5 mL HeLa nuclear extract rocking overnight at 4 C. Beads were then washed again three times with 20 column volumes 0.5 M KCl HEGN and twice with 20 column volumes 0.15 M KCl HEGN, 0.02% (v/v) NP-40 alternative. Bound material was eluted twice with 50 μl 0.1 M Glycine pH 2.7. Eluted proteins were separated by SDS-PAGE and resolved by silver staining or immunoblotting. Immunoblotting experiments analyzing serial dilutions of IP elutions were probed using antibodies against the indicated proteins. Antibody information is in Supplementary Table 1.
The CDK8-submodule and CDK8-Mediator complexes were purified as previously reported51.
This work was supported initially by a grant from the NIH (CA117907) and later by a grant from NSF (MCB-0842974) to the Espinosa lab. JME is a Howard Hughes Medical Institute Early Career Scientist. Work in the Taatjes lab is supported by grants from NIH (PO1 CA112131) and the Ellison Medical Foundation. AJD and CCE were supported in part by NIH training grant T32GM07135. We are thankful to members of the Espinosa lab for contributing ideas and support, especially Marybeth Sechler and Christopher Potts for technical assistance and Nathan Gomes for discussion.
Authors contribution statement.AJD performed most ChIP, Q-RT-PCR and western blots experiments, performed all nuclear run-ons and carried out microarray data analysis and wrote the manuscript. CE and DT generated all data regarding the CDK8-Mediator/P-TEFb interaction. JME performed ChIP, Q-RT-PCR and western blot experiments and carried out the microarray experiment.