|Home | About | Journals | Submit | Contact Us | Français|
The four proteins CDK8, cyclin C, Med12, and Med13 can associate with Mediator and are presumed to form a stable “CDK8 subcomplex” in cells. We describe here the isolation and enzymatic activity of the 600-kDa CDK8 subcomplex purified directly from human cells and also via recombinant expression in insect cells. Biochemical analysis of the recombinant CDK8 subcomplex identifies predicted (TFIIH and RNA polymerase II C-terminal domain [Pol II CTD]) and novel (histone H3, Med13, and CDK8 itself) substrates for the CDK8 kinase. Notably, these novel substrates appear to be metazoan-specific. Such diverse targets imply strict regulation of CDK8 kinase activity. Along these lines, we observe that Mediator itself enables CDK8 kinase activity on chromatin, and we identify Med12—but not Med13—to be essential for activating the CDK8 kinase. Moreover, mass spectrometry analysis of the endogenous CDK8 subcomplex reveals several associated factors, including GCN1L1 and the TRiC chaperonin, that may help control its biological function. In support of this, electron microscopy analysis suggests TRiC sequesters the CDK8 subcomplex and kinase assays reveal the endogenous CDK8 subcomplex—unlike the recombinant submodule—is unable to phosphorylate the Pol II CTD.
The Mediator complex is a general target of DNA-binding transcription factors and is required for expression of virtually all protein-coding genes (35). Four subunits—CDK8, cyclin C, Med12, and Med13—are known to reversibly associate with Mediator and modulate its biochemical function. In humans, CDK8, cyclin C, Med12, and Med13 are believed to associate as a subcomplex based in part upon biochemical and genetic experiments in yeast and lower metazoans. In fact, a stable “CDK8 subcomplex” containing CDK8, cyclin C, Med12, and Med13 has been isolated from yeast cells expressing TAP-tagged cyclin C (5). Genetic experiments in yeasts, Caenorhabditis elegans, and Drosophila melanogaster indicate that CDK8, cyclin C, Med12, and Med13 are each required for organism (but not cell) viability and function primarily as a unit (6, 20, 33, 53). That is, similar phenotypes result from mutation of any single subunit within the CDK8 subcomplex (7). Furthermore, the expression of CDK8, cyclin C, Med12, and Med13 is regulated differently with respect to other consensus Mediator subunits, at least in yeast (20). Genetic studies have also revealed that components of the CDK8 subcomplex play critical roles in development (60). For example, ablation of the CDK8 kinase in mice results in lethality at the preimplantation stage (57); deletion of Med12 causes severe defects in brain and neuronal development in C. elegans and zebrafish (21, 47, 56, 61), whereas the loss of either Med12 or Med13 causes identical defects in Drosophila eye and wing development (23, 52). Significantly, a number of disease-causing mutations in CDK8 subcomplex components are beginning to be uncovered in humans. For instance, a single point mutation in Med12 (R961W) has been linked to a rare syndrome affecting brain development and mutations within a different region in Med12—the Q-rich domain in its C terminus—can cause mental retardation in males (46, 48). Furthermore, congenital heart and neuronal defects can result from mutations within an isoform of Med13 (39), and CDK8 has been directly implicated in oncogenesis (12, 38).
Clearly, the subunits within the CDK8 subcomplex are important for proper control of developmental programs, and yet the biochemical mechanisms by which they regulate these processes are not fully defined. For one, it is unclear whether the CDK8 subcomplex functions only in the context of Mediator or whether it may operate in part as a separate, independent entity. Furthermore, no biochemical function has been attributed to Med12 or Med13, which together comprise a major portion (500 kDa) of the 600-kDa CDK8 subcomplex. Indeed, nearly all known regulatory functions of the CDK8 subcomplex have been attributed to its kinase activity (20, 40). For example, phosphorylation of different activators by yeast CDK8 (also called srb10) can alter their activity or cellular stability (8, 42, 54). Yeast CDK8 can also phosphorylate the RNA polymerase II C-terminal domain (Pol II CTD) prior to preinitiation complex assembly to inhibit transcription initiation (18), whereas human CDK8 appears to switch off transcription by phosphorylating cyclin H, a critical regulatory subunit within TFIIH (1). Thus, the kinase activity of CDK8 is a powerful regulator of gene expression. However, nothing is known about how CDK8 may be regulated, and few CDK8 substrates have been identified, particularly in humans.
We describe here the isolation and enzymatic activity of the CDK8 subcomplex purified both directly from human cells and also via recombinant expression of human CDK8, cyclin C, Med12, and Med13 in insect cells. Although our studies indicate the free CDK8 submodule can operate independently, it is clear that Mediator itself regulates CDK8 activity. Moreover, mass spectrometry (MS) and biochemical analyses suggest alternate factors work to control CDK8 submodule activity and stability apart from Mediator in cells. Significantly, we identify novel substrates for the CDK8 kinase (histone H3, Med13, and CDK8 itself) that were not anticipated based upon previous studies in yeast. Furthermore, our studies have uncovered a key biochemical distinction between Med12 and Med13 that also appears unique to higher organisms. Together, these results provide insight into CDK8 subcomplex regulation and suggest the subcomplex itself may regulate gene expression independently of Mediator in human cells.
Med13 was isolated from a HeLa cDNA library and cloned into a modified pAC-GHLT vector (Pharmingen) in which the glutathione S-transferase (GST) and His tags were removed. Med12 cDNA was obtained from the lab of R. G. Roeder and cloned into a modified version of the pAC-GHLT vector that contained a C-terminal His tag. Cyclin C and Glu-tagged CDK8 were gifts from Emma Lees. High-titer viruses of each gene were coinfected into Sf9 cells at an multiplicity of infection ratio of 1:2:3:3 (K8/CC/12/13) at 27°C for 48 h.
Protease inhibitors (1 mM benzamidine, 1 mM sodium metabisulfite, 1 mM dithiothreitol [DTT], 1.1 μg of aprotinin/ml, and 25 μM phenylmethylsulfonyl fluoride) were added to all buffers immediately before use. Sf9 cells expressing recombinant CDK8 subcomplex components were lysed by Dounce homogenization in 50 mM Tris (pH 7.6), 150 mM NaCl, and 0.1% NP-40. Extracts were incubated with anti-Glu matrix (Abcam ab24587), washed with 400 column volumes (cv) 0.5 M KCl HEGN (0.1% NP-40, 50 mM HEPES [pH 7.6], 0.1 mM EDTA, 10% glycerol) and 80 cv 0.15 M KCl HEGN (0.02% NP-40, 50 mM HEPES [pH 7.6], 0.1 mM EDTA, 10% glycerol) and eluted with 0.15 M KCl HEGN supplemented with 1 mg of Glu-Glu peptide/ml. Elutions were loaded onto a glycerol gradient containing 300 μl of 80% glycerol, 850 μl of 40% glycerol, and 1,050 μl of 20% glycerol in 0.15 M KCl HEGN and spun at 200,000 × g for 6 h in a Beckman TLS-55 rotor. The final glycerol gradient size exclusion served to remove excess free CDK8/cyclin C from the full complex but did not affect kinase activity.
A typical preparation was started with HeLa nuclear extract from a 500-liter portion of cells; this was loaded on a Whatman P11 phosphocellulose column at 0.1 M KCl and eluted with 0.3 M KCl-1 mM MgCl2 HEG, dialyzed to 0.15 M KCl, and passed over a HiTrap Q column (GE Healthcare). Flowthrough fractions were precipitated by the addition of 50% (wt/vol) ammonium sulfate and further concentrated by using Amicon spin filters (Millipore). Residual Mediator was depleted by GST-SREBP affinity chromatography. Superose 6 fractionation isolated 400- to 800-kDa complexes. After preclearing with protein A/G resin (GE Healthcare), concentrated Superose 6 fractions were incubated with anti-CDK8 antibody-bound A/G resin, washed with 50 cv 0.5 M KCl HEGN and 10 cv 0.15 M KCl HEGN and eluted with 50 mM Tris (pH 8.0), 0.75 M ammonium sulfate, 40% ethylene glycol, and 0.1 mM EDTA. Eluates were desalted into 0.15 M KCl HEGN with a QuickSpin TE column (Roche) preblocked with bovine serum albumin; loaded onto a glycerol gradient containing 100 μl of 30% glycerol, 850 μl of 20% glycerol, and 1,150 μl of 15% glycerol in 0.15 M KCl HEGN; and spun as stated above. Endogenous CDK8 subcomplex was retained in the 30% glycerol fraction. (See Fig. Fig.66 for an overview of the purification protocol.)
A modified CTD-truncated GST-tagged Med12 [GST-Med12(aa1-1227)] was incorporated into the recombinant CDK8 kinase-dead complex by using the techniques outlined above. Having tags on two of the subunits [Glu-CDK8 and GST-Med12(aa1-1227)] allowed for the multistep affinity purifications summarized in Fig. Fig.2.2. Purifications included combinations of the following procedures: (i) Glu purification as described above; (ii) GSH-resin purification; (iii) glycerol gradient sedimentation (as outlined in above) using 20 to 80% glycerol; and (iv) anti-Med13 immunoprecipitation, followed by elution with 2% Sarkosyl. Since Sarkosyl disrupted protein-protein interactions, the anti-Med13 purification technique could be used only as a final step. For each affinity purification step (i.e., steps i to iv), bound material was washed extensively with 0.3 to 0.5 M KCl HEGN prior to elution from the resin. After tandem purifications, proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and stained with SYPRO-Ruby or Coomassie protein stains and quantitated by using densitometry.
Nuclear extract from 100L of HeLa cells was passed over a phosphocellulose column as described above. After the 0.3 M KCl wash Mediator was eluted 0.5 M KCl-1 mM MgCl2 HEG and loaded onto a HiTrap Q column (GE Healthcare). The flow through from the Q column was incubated with GST-SREBP or GST-VP16 bound GSH resin, washed with 250 cv 0.5 M HEGN, 50 cv 0.15 M KCl HEGN eluted with 30 mM GSH and run on a linear 15 to 40% glycerol gradient.
Recombinant CDK8 subcomplex or TCA precipitated endogenous CDK8 subcomplex samples were prepared for analysis by mass spectrometry as described previously (25). Briefly, the samples were separated by SDS-PAGE, SYPRO stained, reduced with DTT, alkylated with iodoacetamide, and digested overnight with trypsin (Promega). Recombinant protein identification was done by using an Agilent LC/MSD Trap XCT fed by an Agilent 1100 series high-pressure liquid chromatograph, while endogenous samples were run through a Waters nanoACUITY UPLC coupled to a Thermo Finnigan LTQ. The resulting tandem MS (MS/MS) data was searched by using the MASCOT search engine against the human IPI database.
Reactions were carried out with 5 to 50 ng of purified CDK8 subcomplex and 125 to 500 ng of purified substrate in kinase buffer (25 mM Tris [pH 8.0], 100 mM KCl, 10 to 100 μM ATP, 10 mM MgCl2, and 2 mM DTT) with the addition of 2.5 μCi of [γ-32P]ATP at 30°C for 60 min. Reactions were separated by SDS-PAGE, silver stained, dried, exposed on a storage PhosphorScreen (GE Healthcare), and imaged with a Typhoon 9400 scanner (GE Healthcare). For kinase assays involving DNA-PK, 500 ng of linear DNA was added to activate DNA-PK, and 2 μM wortmannin was used to inhibit DNA-PK function. Purified DNA-PK was obtained from Promega (V5811).
Recombinant wild-type or kinase dead 4-protein CDK8 subcomplexes were incubated at 30°C for 60 min in kinase buffer (50 mM Tris [pH 7.9], 100 mM NaCl, 10 mM MgCl2, and 1 mM DTT) or in phosphatase buffer (supplemented with calf intestinal phosphatase) at 37°C for 60 min. Proteins were then separated by SDS-5% PAGE and visualized by silver staining.
Med12 (sc-5374), CDK8 (sc-1521), cyclin C (sc-1061), H3 (sc-8654), GST tag (sc-138), TCP-1 β (sc-1378), DNA-PK (sc-1552), and nucleolin (sc-8031) were obtained from Santa Cruz. H3 phospho-S10 (catalog no. 06-570), H3 phospho-S28 (catalog no. 07-145), and H3 phospho-T3 (catalog no. 07-424) were from Upstate. Med13 (A301-278A) and Med23 (A300-425A) were from Bethyl, Inc. Glu tag (ab1267) and His tag (ab9108) were from Abcam. Pol II (8WG16), Pol II S2P (MMS-129R), and Pol II S5P (MMS-134R) were from Covance. Med14 was a laboratory stock antibody.
Drosophila core histones were assembled with digested 5S G5E4 DNA into chromatin using the salt dialysis method as described previously (22). Assembled chromatin was visualized by ethidium bromide staining of phenol-chloroform-extracted micrococcal-nuclease-digested samples.
HCT116 and HEK293 cells were maintained in McCoy's 5A and Dulbecco modified Eagle medium (Gibco-Invitrogen), respectively, supplemented with 10% fetal bovine serum and an antibiotic-antimycotic mix.
Short-hairpin RNAs (shRNAs) targeting either CDK8 or MED12 were generated and expressed using the third-generation lentiviral delivery system previously described with minor modifications (10, 41). Briefly, complementary oligonucleotides encoding shRNAs were cloned into the lentiviral vector pLL3.7neo (a derivative of the original pLL3.7 vector with the coding region of green fluorescent protein replaced with the coding region of neomycin phosphotransferase). The shRNA/pLL3.7neo vectors were cotransfected into HEK293FT cells, along with the packaging vectors pMDLg/pRRE, pRSV-REV, and pMD.G. After 48 h, viral supernatants were harvested and used to transduce HCT116 cells. HCT116 clones that were neomycin resistant were grown and screened by immunoblotting for knockdown of the appropriate target. Cells were lysed in radioimmunoprecipitation assay buffer supplemented with a cocktail of protease and phosphatase inhibitors.
Endogenous CDK8 subcomplex samples were negatively stained with uranyl acetate, imaged at ×29,000 magnification on a Tecnai F20 FEG equipped with a Gatan 4k×4k charge-coupled device camera, and segregated into two-dimensional (2D) classes as described previously (36).
HeLa nuclear extracts were first concentrated by ammonium sulfate precipitation (50% [wt/vol]). Concentrated nuclear extract was passed through a Superose 6 10/300 GL column (GE Healthcare). Immunoblots of fractions collected were quantitated by using densitometry. Fractions larger than 1.4 MDa were considered to contain CDK8-Mediator, whereas fractions ranging from 1.4 MDa to 400 kDa were considered as eK8, and fractions with protein complexes smaller than 400 kDa were considered to consist of partial subcomplexes.
Quantitative Western analysis of Superose 6-fractionated HeLa nuclear extracts resulted in estimates of ca. 30% of the CDK8 subcomplex components present in the fractions between 1.4 MDa and 400 kDa—fractions consistent with the size of the endogenous CDK8 subcomplex. In agreement with this, when fractionated over a phosphocellulose column, ca. 30% of the total amounts of the CDK8 subcomplex components probed were retained in the P0.3M fraction (see Fig. Fig.6A).6A). Since the P0.3M fraction is greatly enriched for the endogenous CDK8 subcomplex, we view this as further evidence that as much as 30% of CDK8 subcomplexes may exist independently of Mediator in human cells. Proportionally more CDK8 (more than 50%) was found in the fractions from 400 kDa to 1.4 MDa, implying that other large complexes containing CDK8 may exist.
To generate the recombinant CDK8 subcomplex, Sf9 cells were coinfected with high-titer viruses for CDK8, cyclin C, Med12, and Med13. The CDK8 subcomplex was purified from whole-cell extracts by immunoprecipitation with an anti-Glu resin, which targeted the Glu-tagged CDK8 subunit within the subcomplex. Fractionation of the extracts over ion-exchange and gel filtration columns prior to immunoprecipitation did not measurably improve the purity of the subcomplex (data not shown). A similar strategy was used to purify the kinase-dead CDK8 subcomplex, which contained a D173A point mutation in CDK8 that inactivates its kinase function (1). Silver-stained gels of the active (r 4wt) and kinase-dead (r 4kd) CDK8 subcomplexes are shown in Fig. Fig.1A.1A. Importantly, each subunit comigrated on a glycerol gradient, confirming their association in a large complex (data not shown). Immunoblots and electrospray MS/MS (ESMS) analysis confirmed the identity of each subunit as Med13, Med12, CDK8, and cyclin C (Fig. 1B and C). Furthermore, a 1:1:1:1 stoichiometry was established for the CDK8 subcomplex based upon densitometry measurements (see below). Thus, it appeared that human CDK8, cyclin C, Med12, and Med13 could assemble into a stable complex from recombinantly expressed subunits.
Comparison of silver-stained gels of the wild-type (4wt) and kinase-dead (4kd) CDK8 subcomplexes revealed a shift in Med13 and/or Med12 mobility, suggesting that these subunits may be phosphorylated by CDK8. To test this, we treated the recombinant wild-type (r 4wt) or kinase-dead (r 4kd) complexes with either ATP or phosphatase prior to gel electrophoresis. As shown in Fig. Fig.1D,1D, these experiments indicated the mobility shifts were due to differential phosphorylation; Western blots (not shown) and ESMS data confirmed that phosphorylation was occurring on Med13 (see below).
To explore a potential biochemical function for Med12 and Med13, we expressed and purified the CDK8/cyclin C pair alone and also together with either Med12 or Med13. Each three-subunit complex (CDK8/cyclin C/Med12 or CDK8/cyclin C/Med13), as well as the CDK8/cyclin C binary complex, was stable and could be purified by using standard techniques, indicating that Med12 and Med13 each independently interact with the CDK8/cyclin C pair (Fig. 1A and B). Subunits within these partial subcomplexes also comigrated on a glycerol gradient, confirming their assembly in a complex (data not shown). Together, these stable, recombinant subcomplexes offered a means to systematically examine the kinase activity of the CDK8 subcomplex.
Assessing the stoichiometry of the CDK8 submodule was complicated by several factors. First, Med13 is phosphorylated, which impacts its migration on a gel and causes it to overlap with the Med12 protein band. Specifically, Med13 in its unphosphorylated state migrates slightly below Med12, whereas in intermediate stages it overlaps with Med12; highly phosphorylated forms of Med13 will migrate higher than Med12 (Fig. (Fig.1D).1D). Because of its multiple phosphorylated states, Med13 is represented by multiple bands, which hinders an accurate quantitation. Thus, Med12, which shows no evidence of phosphorylation, represents a more reliable subunit for quantitation. However, because of overlapping Med12/Med13 bands, accurate quantitation of Med12 apart from Med13 was not possible. To remedy this problem, we expressed a CTD-truncated version of Med12(aa1-1227). This Med12 truncation does not impact its association with the subcomplex and has no apparent differences from full-length Med12 on impacting the CDK8 submodule kinase activity (data not shown). With respect to quantitation, this truncated form of Med12 enables a clear separation between Med12 and Med13 on a gel to accurately quantify Med12.
With respect to CDK8/cyclin C, we focused on cyclin C, because, like Med13, CDK8 presents multiple obstacles to accurate quantitation. First, CDK8, like Med13, is phosphorylated and therefore migrates as several bands. Second, the CDK8 band overlaps slightly with a nonspecific band that may represent the heavy chain from the antibody purification (Fig. (Fig.2).2). Fortunately, such issues do not accompany cyclin C. Thus, to ensure the most accurate results, we focused on Med12 and cyclin C for the assessment of stoichiometry within the CDK8 submodule. The results, which indicate a 1:1 stoichiometry for cyclin C and Med12, are summarized in Fig. Fig.2.2. Densitometry readings from Coomassie blue-stained gels were also consistent with a 1:1 stoichiometry between cyclin C and Med12. Despite the potential uncertainty in quantitation of CDK8 in these samples (noted above), the molar ratio of cyclin C to CDK8 ranged from 1:0.73 to 1:1.31 in our experiments, also suggesting a 1:1 overall stoichiometry. This could also be anticipated based upon predicted CDK8/cyclin C pairing within the complex.
Despite the problems associated with assessing the molar ratio of Med13 with densitometry (see above), it is worth noting that our measurements suggested a 0.5:1 ratio of Med13 relative to cyclin C. We are quite confident the true molar ratio is 1:1 because of the uncertainty due to its multiply phosphorylated states and also for additional reasons outlined below. First, in testing alternate purification protocols for the CDK8 subcomplex, we noted that Med13 could be selectively dissociated with 500 mM imidazole, suggesting that Med13 may be slightly labile within the subcomplex. Second, using titration gel densitometry techniques (plotting band intensity versus sample load), we noted that Med13 increased at the same ratio (i.e., the same slope) as cyclin C, suggesting a 1:1 ratio. Third, the molecular mass of a 1:1:1:1 subcomplex (600 kDa) is consistent with our electron microscopy (EM) data of the recombinant CDK8 subcomplex (M. T. Kneusel and D. J. Taatjes, unpublished results), whereas a 950-kDa subcomplex (required for a 2:2:2:1 ratio) is not. Taken together, these data suggest Med13 is present in a 1:1 molar ratio with cyclin C and the stoichiometry of each subunit within the CDK8 subcomplex is 1:1.
Although few substrates have been identified for human CDK8, this kinase clearly plays important roles in gene regulation (1) and has been linked to oncogenic transformation in colon cancer (12, 38). Known substrates of CDK8 include TFIIH and the CTD of the large subunit of Pol II (1, 31). As expected, the recombinant human subcomplex phosphorylated TFIIH (data not shown) and the Pol II CTD, whereas the subcomplex containing a point mutation in CDK8 did not (Fig. (Fig.3A,3A, lanes 4 and 5). Phosphorylated bands corresponding in size to CDK8 and Med13/Med12 were also evident with the active CDK8 subcomplex but not in assays with the kinase-dead mutant (Fig. (Fig.3A,3A, compare lanes 4 and 5). Subsequent ESMS analysis of each phosphorylated band confirmed their identities as CDK8 and Med13 (data not shown). Because Med13 and CDK8 were phosphorylated in the active subcomplex, but not the kinase-dead counterpart, we conclude that CDK8, and not a potential contaminating kinase activity, was phosphorylating these substrates. Phosphorylation of CDK8 and Med13 was also observed in the context of the entire CDK8-Mediator complex (Fig. (Fig.3A,3A, lane 6), indicating that subunit specificity is not altered upon subcomplex association with core Mediator. CDK8-Mediator was also observed to phosphorylate the Pol II CTD, as expected (data not shown). The absence of additional kinase substrates within Mediator itself suggests that the CDK8 subcomplex occupies a distinct, separable domain within CDK8-Mediator, a finding consistent with past structural studies (50).
During the course of this work, our lab isolated a derivative CDK8-Mediator complex that efficiently phosphorylated histone H3 in vitro (36). This result suggested that CDK8 is a histone H3 kinase, at least when associated with Mediator. To test whether the recombinant CDK8 subcomplex alone would display similar H3 kinase activity, we performed kinase assays against purified core histone octamers. The recombinant wild-type CDK8 subcomplex—but not the kinase-dead mutant—preferentially phosphorylated histone H3 within core histone octamers (compare Fig. Fig.3B,3B, lanes 5 and 6). Similar results were obtained with recombinant core histones (data not shown). As an additional control, we purified human TFIIH (Fig. (Fig.3C)3C) to test in the histone kinase assays. TFIIH is generally required to initiate transcription and contains a kinase, CDK7, that is critical for its biochemical function. Both CDK8 and CDK7 are capable of modifying the Pol II CTD in vitro (Fig. (Fig.3D);3D); however, TFIIH was unable to phosphorylate histone octamers in this assay (Fig. (Fig.3B,3B, lane 7), indicating a significant difference in CDK7 versus CDK8 substrate specificity. Given the emerging role of H3 phosphorylation in active gene expression (see below), this observation identifies a means by which CDK8 may positively regulate gene expression in human cells.
To identify the H3 residue(s) modified by the CDK8 subcomplex, we focused on the N-terminal tail of histone H3, which is a common site for posttranslational modifications. A series of kinase assays with purified, recombinant H3 were performed; antibodies against different phosphorylated forms of H3 were then used to probe for potential site(s) of CDK8 modification. As shown in Fig. Fig.3E,3E, these experiments revealed that H3S10, but not H3T3 or H3S28, is the major site for phosphorylation by the CDK8 subcomplex. Notably, whereas phosphorylation of H3T3 or H3S28 does not correlate with transcription, H3S10 phosphorylation is strongly tied to transcriptional activation (44). For example, H3S10 phosphorylation helps maintain a transcriptionally active state by inhibiting methylation of the adjacent H3K9 site, thereby preventing assembly of repressive HP1 complexes on chromatin (13, 19). Moreover, H3S10 phosphorylation strongly correlates with activation of genes regulated by such diverse activators as NF-κB, RARβ2, and myc in human cells (3, 28, 59, 64); furthermore, activation of heat shock genes in Drosophila occurs with a corresponding increase in H3S10 phosphorylation only at heat shock loci (43).
Recent work in our laboratory has shown that within the Mediator complex, CDK8 can efficiently modify H3 within histone octamers or chromatin templates (36). However, the activity of the free submodule may vary. Although the data in Fig. Fig.33 reveals the free CDK8 submodule can phosphorylate histone octamers, it does not address activity on chromatin. Therefore, we assembled core histone octamers into chromatin (Fig. (Fig.4A)4A) and tested whether the free CDK8 submodule would similarly modify H3 in this context. As shown in Fig. Fig.4B,4B, the free CDK8 submodule was not able to modify H3 in a chromatin context, even when doubling the amount of CDK8 submodule used (lane 8) or doubling the amount of chromatin added (lane 16). Only when incorporated into Mediator could CDK8 efficiently phosphorylate H3 within chromatin (Fig. (Fig.4B,4B, lanes 5 and 6 and lanes 11 and 12), and H3 phosphorylation was enhanced if CDK8-Mediator was recruited to chromatin by an activator (data not shown). These results indicate a major change in CDK8 function and demonstrate that Mediator itself acts to regulate CDK8 substrate specificity. Because the majority of H3 is tied up within chromatin in interphase cells, most CDK8-dependent phosphorylation of H3 likely occurs when associated with Mediator. However, it remains plausible that the free CDK8 subcomplex may cooperate with histone chaperones and/or chromatin disassembly factors to indirectly modify chromatin templates.
Although some studies have linked CDK8 to transcriptional activation (9, 14, 32), the observation that CDK8 can mediate H3S10 phosphorylation provides a molecular mechanism by which CDK8-dependent activation might occur. Although other transcriptionally relevant kinases have been linked to H3S10 modification (e.g., PIM1 and IKKα), these appear to regulate only a subset of genes (3, 59, 64). In contrast, CDK8 (as well as core Mediator) appears to be recruited to promoter and enhancer regions genome-wide (2, 63), suggesting that H3S10 phosphorylation by CDK8 may be a general phenomenon. In support of this, shRNA knockdown of CDK8 causes a dramatic reduction in H3 containing the dual modification H3S10P/K14Ac, as reported previously (36). This tandem H3 modification results from cooperative activity between CDK8 and GCN5L within Mediator. Upon probing H3S10P alone, we observed that CDK8 knockdown did not significantly reduce the levels of this singly modified H3 in cells (data not shown). This result likely reflects high H3S10P levels from Aurora B kinase (a marker for mitosis) and also suggests CDK8 primarily functions within T/G-Mediator (i.e., together with GCN5L) when phosphorylating H3 (36).
Numerous biochemical roles for the CDK8/cyclin C pair have been identified (albeit mostly in yeast). No clear function, however, has been determined for Med12 or Med13. Previous biochemical and genetic data in yeast indicated that ablation of either Med12 or Med13 is phenotypically similar to mutations that inactivated CDK8 (7, 40). However, subtle developmental differences have been noted upon comparing null mutants in Drosophila: although CDK8, cyclin C, Med12, or Med13 mutant phenotypes are largely similar, Med12 or Med13 knockout cells did display differences in tarsal development compared to cells null for CDK8 or cyclin C (33). Interestingly, when the three-subunit partial CDK8 subcomplexes (CDK8/cyclin C/Med12 and CDK8/cyclin C/Med13 [Fig. [Fig.1A])1A]) were tested for kinase function, only the three-subunit complex containing Med12 was active, whereas the CDK8/cyclin C/Med13 complex showed only weak activity toward the human CDK8 substrates tested (Fig. (Fig.5).5). In fact, the three-subunit CDK8/cyclin C/Med12 complex possessed kinase activity comparable to the wild-type, four-subunit complex for all substrates tested. In addition, it is worth noting that the CDK8/cyclin C dimer was inactive in these assays, unless titrated to 10-fold higher concentration relative to the four-subunit CDK8 subcomplex (Fig. (Fig.5,5, lanes 7 and 15). We also used shRNA to knock down Med12 expression to further test the role of Med12 on CDK8 activity in human cells. However, this knockdown coordinately decreased CDK8 protein levels (data not shown) and therefore mimicked the CDK8 knockdown described above.
The data in Fig. Fig.55 reveal a previously unidentified biochemical function for human Med12 that distinguishes it from Med13. Future work will be necessary to identify the mechanism by which Med12 activates the CDK8 kinase. Notably, such a regulatory role for Med12 in yeast is not evident, as CDK8/cyclin C retains its kinase function independent of Med12 and Med13 in vitro (40). Whether Med12 is required for CDK8 activity in lower metazoans is not known; genetic experiments in worms and flies indicate that Med12- or Med13-null mutations are indistinguishable during embryonic development (6, 23, 52, 55, 60). However, Fraser et al. recently completed a comprehensive RNA interference screen in C. elegans which identified Med12 as one of only six “hub” genes that interact with a large number of genes from different signaling pathways (29). Whether this distinction is related to Med12 regulation of CDK8 kinase activity remains to be determined, but this C. elegans screen does suggest that Med12 may perform biochemical functions distinct from Med13 even in lower metazoans.
The purification of a stable, active human CDK8 subcomplex from recombinant subunits suggested that a similar subcomplex may exist in human cells. Although previous work has clearly demonstrated that CDK8 can reversibly associate with Mediator in human cells (37, 45), it has not been established whether free, endogenous CDK8 subcomplexes serve any potential regulatory role independent of Mediator. Given the kinase targets of the recombinant CDK8 subcomplex identified here (which include TFIIH, the Pol II CTD, and histone H3), it was important to establish whether the free CDK8 submodule may exist as a stable, biochemically active entity in cells. Therefore, we sought to isolate free CDK8 subcomplexes directly from human cells. Initial fractionation of HeLa nuclear extract over a phosphocellulose column allowed us to separate a potential CDK8 subcomplex from the majority of CDK8-Mediator: CDK8-Mediator was enriched in the P0.5M fraction, whereas CDK8 subcomplex components were also present in the P0.3M fraction (Fig. (Fig.6A).6A). Further purification from the P0.3M fraction was completed by tracking kinase activity and protein subunits over additional ion-exchange, gel filtration, and affinity resins (Fig. (Fig.6B).6B). This purification scheme resulted in isolation of a complex containing CDK8, cyclin C, Med12, and Med13 that was free of core Mediator subunits, as shown by immunoblot assays (Fig. (Fig.7A7A).
Most convincingly, MS analysis confirmed the presence of all four CDK8 subcomplex components in this sample (Fig. (Fig.7B),7B), whereas core Mediator subunits were not represented (however, see Fig. Fig.7B).7B). Notably, CDK8-like (CDK8L; also called CDC2L6 or CDK11), a paralog of CDK8, was observed to copurify with the endogenous CDK8 subcomplex. Immunoblots of the purified endogenous subcomplex showed no evidence of the CDK8 antibody cross-reacting with CDK8L. For this reason, and the fact that the CDK8 antibody used in the final affinity purification step is directed against a region of CDK8 (amino acids 424 to 464) not conserved within CDK8L, we conclude that CDK8L most likely associates together with CDK8 in a single subcomplex. Confirmation of this will ultimately require reagents specific for CDK8L.
MS analysis also identified other submodule-associated factors, including DNA-PK, GCN1L1, and the TRiC chaperonin complex (Fig. (Fig.7B).7B). The presence of DNA-PK and TRiC was further substantiated with immunoblotting experiments (Fig. (Fig.7C;7C; antibodies against CDK8L and GCN1L1 are not available). In addition, CDK8 subcomplex components were observed to coimmunoprecipitate with DNA-PK and the TRiC subunit TCP-1β (Fig. (Fig.7D).7D). Given the rigorous purification protocol, including a final CDK8 antibody binding/elution step followed by separation based upon size (glycerol gradient step), it is likely that TRiC, DNA-PK and GCN1L1 are in fact associated with the CDK8 submodule and do not copurify as individual entities. However, we cannot rule out the possibility that TRiC, DNA-PK, or GCN1L1 may bind the CDK8 submodule in a mutually exclusive fashion. Collectively, the data in Fig. Fig.77 demonstrate that the free CDK8 subcomplex can exist as a stable entity in human cells. Factors associated with the endogenous CDK8 subcomplex, however, distinguish it from the recombinant, four-subunit CDK8 subcomplex and suggest a means to regulate its activity and stability within cells. Given the critical role of CDK8 in gene regulation and oncogenesis (1, 12, 38), it is likely that multiple mechanisms have evolved to control its biological function.
The biological relevance of DNA-PK, GCN1L1, and TRiC association with the CDK8 submodule awaits elucidation in future studies. DNA-PK is well established in DNA repair mechanisms and is peripherally implicated in transcriptional regulation (49); much less is known about GCN1L1, although its yeast ortholog has been shown to bind and regulate the GCN2 kinase (15, 26). The copurification of the chaperonin TRiC with the endogenous CDK8 submodule is particularly intriguing. TRiC does not appear to be a general chaperonin. Instead, it displays specificity for a subset of regulatory complexes, including elongin BC and SMRT-HDAC3 (11, 16). The results shown here identify the CDK8 submodule as another potential TRiC-regulated complex. Notably, TRiC is critical for the formation of an active human cyclin E/CDK2 complex (58), and therefore its association with the CDK8 submodule may reflect similar roles in the formation of a stable, active subcomplex. Also noteworthy is the ability of TRiC to modulate aggregation of poly-Q tracts in the Huntington protein (4, 51). Med12 contains a Q-rich region at its C terminus and TRiC may help regulate interactions from this domain.
That TRiC might regulate CDK8 submodule interactions is further supported by EM analysis of the endogenous CDK8 submodule. Alignment and 2D classification of single-particle images revealed the presence of complexes of size and shape consistent with the free CDK8 submodule, as expected (data not shown). In addition, 2D averages of single-particle images clearly showed TRiC complexes whose large, central cavity was occupied with protein density (Fig. (Fig.7E).7E). Mammalian TRiC has a distinct barrel shape consisting of stacked eight-membered rings with an outer diameter of ~150 Å and a length of 160 Å. TRiC utilizes ATP hydrolysis to aid in folding cargo proteins located within its 80 Å-diameter interior cavity. EM analysis and 3D reconstruction of the recombinant CDK8 subcomplex (M. T. Knuesel, K. D. Meyer, and D. J. Taatjes, unpublished results) indicates that its overall dimensions are slightly larger than the TRiC cavity could accommodate. However, conformational shifts within the CDK8 subcomplex may facilitate its sequestration within the TRiC central cavity, and/or the CDK8 submodule may exist as a partially assembled entity within TRiC. However, since both CDK8 and Med13 were found to coimmunoprecipitate with an antibody against TRiC (Fig. (Fig.7D),7D), this suggests the entire CDK8 subcomplex might associate with the TRiC chaperonin. Although additional experiments are necessary to delineate the means by which TRiC might regulate the CDK8 submodule, the EM data suggest that TRiC may sequester the submodule as a simple means of regulating its interactions with other proteins such as core Mediator or its kinase substrates.
Following purification of the endogenous CDK8 subcomplex, we tested its kinase activity against known human substrates. Notably, the endogenous subcomplex displayed similar—but not identical—kinase activity relative to the recombinant subcomplex. The endogenous sample resembled the recombinant CDK8 subcomplex in its phosphorylation of histone H3, Med13, and CDK8 (Fig. (Fig.8A8A and data not shown). Although these results may reflect the activity of associated factors—particularly CDK8L—the similarity relative to the recombinant subcomplex implicates CDK8 in these modifications. We cannot, however, rule out phosphorylation by CDK8L in these experiments. CDK8L is a paralog of CDK8 and may possess similar substrate specificity. In contrast, DNA-PK is unlikely to contribute to the kinase activity observed in Fig. Fig.8A8A because DNA-PK requires association with DNA to become an active kinase (49) and the endogenous subcomplex is not active in assays containing DNA (i.e., chromatin substrates [Fig. [Fig.8A]).8A]). Moreover, addition of the DNA-PK inhibitor wortmannin had no effect on kinase activity of the endogenous CDK8 subcomplex under the conditions tested here (Fig. (Fig.8B8B).
Strikingly, the endogenous subcomplex was unable to phosphorylate the Pol II CTD. When normalizing for kinase activity toward histone H3, the endogenous CDK8 subcomplex phosphorylated the CTD with 2% efficiency compared to the recombinant subcomplex (Fig. (Fig.8C).8C). Given this clear difference in substrate specificity relative to the recombinant subcomplex—and relative to CDK8-Mediator, which can also phosphorylate the Pol II CTD—we sought to further test this result in human cells. Therefore, we examined whether knockdown of CDK8 would impact the cellular levels of phosphorylated Pol II CTD and observed that CDK8 knockdown did not appreciably impact global CTD phosphorylation patterns (serine 2 or serine 5 [data not shown]). These results are consistent with studies of yeast CDK8 (srb10), in which CDK8 kinase inhibition did not notably impact global Pol II CTD phosphorylation (32). Because human (and yeast) cells contain multiple kinases capable of Pol II CTD phosphorylation, the minimal impact of human CDK8 knockdown on global CTD phosphorylation was not unexpected. It remains possible that CDK8-dependent phosphorylation of the Pol II CTD might regulate a subset of human genes.
Given the functionally diverse targets for the CDK8 kinase, including histone H3, TFIIH, Med13, and Pol II CTD, precise temporal or context-dependent regulation of CDK8 substrate specificity may be a critical factor controlling gene expression in humans. We have identified several means by which CDK8 kinase activity might be regulated in cells. First, the kinase activity of the CDK8 submodule is dependent upon Med12. Second, incorporation into Mediator regulates CDK8 substrate specificity toward H3S10 within chromatin. Third, MS analysis of the endogenous CDK8 subcomplex identified CDK8L, the TRiC chaperonin, GCN1L1, and DNA-PK as potential modifiers of CDK8 activity within the free subcomplex, and evidence that TRiC might help regulate CDK8 submodule interactions was obtained by EM analysis of endogenous subcomplexes. It is also plausible that phosphorylation or other posttranslational modifications within the subcomplex itself may impact its kinase activity. Since the CDK8 subcomplex appears to regulate transcription genome-wide (e.g., via regulation of Mediator), a variety of mechanisms may exist to control CDK8 function.
That the CDK8 subcomplex is stable and biochemically active apart from Mediator in cells suggests the subcomplex may function autonomously to regulate gene expression. However, how might the free subcomplex be targeted to its substrates? Notably, Med12 and CDK8 are known to interact directly with a number of DNA-binding transcription factors, which would provide a means for targeting the free CDK8 subcomplex to specific genomic locations (14, 17, 24, 47, 56, 62). Alternately, the CDK8 subcomplex may contain chromatin-targeting domains (e.g., cryptic bromodomains or chromodomains) that remain unidentified. Recruitment may also be mediated through one of the subcomplex-associated factors, such as DNA-PK or the TRiC chaperonin, identified in the MS analysis. It is important to note, however, that the Mediator complex is more generally targeted by DNA-binding transcription factors; consequently, association with Mediator may enable more widespread targeting of the CDK8 subcomplex to appropriate regulatory sites. Although the majority of CDK8 subcomplexes appear to be Mediator-associated in cells, we estimate that a significant fraction (up to 30%) of CDK8 submodule components may exist independent of Mediator in HeLa cells (see Materials and Methods and Fig. Fig.99).
It is notable that histone H3, Med13, and CDK8 itself are not modified by the CDK8 subcomplex in yeast, nor is the yeast CDK8 subcomplex dependent upon Med12 for activity (18, 40). This is perhaps to be expected given the extensive sequence divergence within Mediator subunits over evolutionary time. Indeed, compared to the rest of the general transcription machinery (e.g., TFIID, TFIIH, and Pol II), Mediator has diverged most significantly (30). This substantial sequence divergence may be reflected in the distinct activity and substrate specificity of the human CDK8 subcomplex.
We thank Bob Roeder for providing the Med12 cDNA, Emma Lees for providing cDNAs for CDK8 and cyclin C, and Karolin Luger for cDNAs encoding core histones. The 5S G5E4 template was a gift from Jerry Workman.
This study was supported by the NCI P01 CA112181 (D.J.T.) and R01 CA117907 (J.M.E.) and in part by NIH grants T32 GM07135 and T32 GM065103 (M.T.K. and K.D.M.).
Published ahead of print on 1 December 2008.