Activation of the Cyclin-dependent Kinase CTDK-I Requires the Heterodimerization of Two Unstable Subunits*

RNA polymerase II CTD kinases are key elements in the control of mRNA synthesis. They constitute a family of cyclin-dependent kinases activated by C-type cyclins. Unlike most cyclin-dependent kinase complexes, which are composed of a catalytic and a regulatory subunit, the yeast CTD kinase I complex contains three specific subunits: a kinase subunit (Ctk1), a cyclin subunit (Ctk2), and a third subunit (Ctk3) of unknown function that does not exhibit any similarity to known proteins. Like the Ctk2 cyclin that is regulated at the level of protein turnover, Ctk3 is an unstable protein processed through a ubiquitin-proteasome pathway. Interestingly, Ctk2 and Ctk3 physical interaction is required to protect both subunits from degradation, pointing to a new mechanism for cyclin turnover regulation. We also show that Ctk2 and Ctk3 can each interact independently with the kinase. However, despite the formation of CDK/cyclin complexes in vitro, the Ctk2 cyclin is unable to activate its CDK: both Ctk2 and Ctk3 are required for Ctk1 CTD kinase activation. The different specific features governing CTDK-I regulation probably reflect requirement for the transcriptional response to multiple growth conditions.

The C-terminal domain (CTD) 1 of the large subunit of RNA polymerase II contains a repeated heptapeptide that is highly phosphorylated in a portion of the molecules in the cell. This domain plays an essential role in the control of mRNA synthesis, as well as RNA processing, and its phosphorylation is a key feature of its function (1). The formation of initiation complexes on promoter DNA involves RNA polymerase II molecules with unphosphorylated CTD, whereas the CTD becomes phosphorylated during or after the transition to elongation. Three CTD kinases are involved in the regulation of RNA polymerase II transcription in the yeast Saccharomyces cerevisiae, all being cyclin-dependent kinases (CDKs).
The first yeast CTD kinase, Kin28 (2), is associated with cyclin Ccl1 (3) and is required for transcription of most genes (4). This complex, like its human counterpart the Cdk7 kinase complex (5), is a component of the general transcription factor TFIIH (6,7). Kin28 is an essential kinase that has been shown to phosphorylate the CTD only after the RNA polymerase II is associated with promoter DNA, thus promoting the elongation mode (8). Kin28 is also required for proper capping enzyme targeting in vivo (9). The second yeast CDK implicated in CTD phosphorylation, the nonessential Srb10/Ume5/Ssn3 kinase and its cyclin partner Srb11/Ume3/Ssn8, is a component of the holoenzyme (10). The Srb10/11 complex is structurally related to mammalian Cdk8 kinase and its associated cyclin C (11). Several studies established that the Srb10 kinase is implicated in the transcriptional repression of specific sets of genes (4,(12)(13)(14)(15). Hengartner et al. (8) showed that Srb10 is uniquely capable of phosphorylating the CTD prior to formation of the initiation complex on promoter DNA, with consequent inhibition of transcription. Srb10 has also been implicated in Gal4 phosphorylation, thus modulating its activity (16).
The CTDK-I complex has not been found in the holoenzyme. It was isolated by its ability to phosphorylate CTD-containing fusion proteins (17). It is composed of three nonessential subunits: the CTK1 and CTK2 genes encode a kinase and a cyclin respectively, whereas the third subunit, Ctk3, does not exhibit any similarity to known proteins (18,19). In vitro studies carried out in HeLa nuclear extracts have shown that CTDK-I can modulate the elongation efficiency of RNA polymerase II (20). Patturajan et al. (21) reported that deletion of the CTK1 gene results in an increase in phosphorylation of serine in position 5 of the CTD repeat, indicating that CTDK-I negatively regulates CTD Ser 5 phosphorylation. CTK1 deletion also eliminates the transient increase in CTD serine 2 phosphorylation observed during the diauxic shift, this latter defect being correlated with a defect in transcriptional induction during the diauxic shift, suggesting a role for CTDK-I in response to nutrient depletion. In humans, the closest kinase to Ctk1 is Cdk9 (for review, see Ref. 22), a CTD kinase that was first described as a positive transcription elongation factor (23), and that interacts specifically with the human immunodeficiency virus-1 transactivator protein Tat, which acts to enhance the processivity of RNA polymerase II (24,25).
How are these CTD kinases regulated? CDKs were first described as central regulators of the major transitions of the eukaryotic cell division cycle. Their activity is determined by cyclin binding, by both positive and negative regulatory phosphorylation and by CDK inhibitors (CKIs) (26,27). An important aspect of CDK regulation is the rapid destruction of their cyclin partners (28). Several studies both in yeast and in mammals have demonstrated that cell cycle cyclins are selectively degraded by ubiquitin-dependent proteasome pathways (29 -32). Although exceptions have been described (33,34), proteasome substrates are usually marked for degradation by conjugation with several molecules of ubiquitin, a highly conserved 76-amino acid-long polypeptide (33,35). The 26 S proteasome is a nuclear and cytosolic multicatalytic proteinase that breaks down targeted substrates to short peptides and recycles the ubiquitin molecules (36). CDKs involved in transcriptional control are activated by C-type cyclins that constitute a divergent group of cyclins (37,38). C-type cyclins exhibit sequence conservation for the cyclin box, i.e. the kinase binding domain (39,40), but their expression does not seem to fluctuate during the cell cycle (41,42), and little is known about their regulation. Cdk8⅐cyclin C phosphorylates cyclin H, resulting in down-regulation of TFIIH activities (43). In yeast, it has been shown that the C-type cyclin Srb11/Ume3 is regulated at the level of its turnover. Srb11/Ume3 is destroyed by multiple and independent pathways when cultures are subjected to diverse stresses (42,44). We previously showed that the Ctk2 cyclin subunit is phosphorylated and rapidly degraded, in a growthrelated fashion, by a ubiquitin-proteasome pathway (45). Interestingly, unlike what has been described for the G 1 cyclins, neither Ctk2 phosphorylation nor its destruction require its activated kinase.
CDK complexes are minimally composed of a catalytic and a regulatory subunit. The vertebrate TFIIH kinase contains an additional component, Mat1, that exhibits a zinc-binding motif typical of C3HC4 RING finger domains likely to mediate protein⅐protein interactions (46). Mat1 functions as an assembly factor, promoting a stable interaction between the Cdk7 kinase and the cyclin H that does not require Cdk7-activating phosphorylation (47)(48)(49). Also, association with Mat1 and core TFIIH modulates the activity of Cdk7 at the level of substrate specificity (50). Although the yeast TFIIH complex contains Tfb3 that is 32.5% identical to Mat1 (51, 52), a Kin28⅐Ccl1⅐Tfb3 complex has never been identified (53). However, Tfb3 has been implicated in transcription (51), and conditional alleles of TFB3 are severely defective in Kin28 T132A kinase activity, suggesting that it might be the Mat1 ortholog (54).
The function of the third component of the CTDK-I complex is unknown. The fact that deletion of each individual CTK gene generates cold-sensitive cells that display similar growth defects (19), suggests that Ctk3 is essential for the function of the complex. We studied Ctk3 expression in vivo, and we show that it is regulated at the protein level. Ctk3 contains a PEST motif involved in its degradation by the ubiquitin-proteasome pathway. Furthermore, binding of Ctk3 to the cyclin subunit is required for stabilization of both subunits. Ctk3 can directly interact with the kinase and is required, in addition to the cyclin, for its activation. It thus appears that, in contrast to other CDKs, the Ctk1 kinase is in fact regulated by a heterodimer composed of two unstable subunits.
Plasmids-For Ctk2 overexpression, the pFL44-CTK2 plasmid was used (URA3 2 m) (45). All the constructs were made via several PCR cloning steps. Briefly, 500-bp PCR fragments containing either the CTK1 or the CTK3 promoter were cloned into pFL36 (LEU2 CEN) (57). The different coding sequences were cloned downstream of sequences encoding two HA influenza virus hemagglutinin tags or a single c-Myc epitope tag prior to cloning downstream of their cognate promoter, yielding to the final constructs CTK1-HA, CTK1-Myc, CTK3-HA, and CTK3-Myc. The CTK2-HA construct (URA3 CEN) is described in Hautbergue and Goguel (45). The Myc-tagged CTK2 coding sequence was inserted into pFL38-PrCTK2 (45), yielding the CTK2-Myc construct.
The CTK3⌬PEST construct is identical to the wild type CTK3 construct except that 13 amino acids (see Fig. 1) are replaced by a leucine. CTK3⌬C50 contains a STOP codon in position 247 of the Ctk3 protein.
The ϩC50 construct encodes a 100-amino acid peptide composed of 25 amino acids encoding two HA-tags/Ctk3 24 N-terminal amino acids/one leucine/Ctk3 50 C-terminal amino acids. All constructs containing PCR fragments were sequenced (the Ctk3⌬PEST and Myc-Ctk1 proteins each contain a single mutation, His 58 3 Leu and Pro 35 3 Leu, respectively). For expression of the Ctk proteins from the T7 promoter, the different epitope-tagged coding sequences were cloned either into the pRSET5d vector (58) or the pET14b (Novagen) to express the (His) 6 -HA tagged Ctk1 protein. The plasmid pMT1089 (TRP1 2 m) encodes a (His) 6 -Myc-tagged ubiquitin under the control of the copper-inducible CUP1 promoter (32).
Northern Blots Analyses-Total RNA was extracted from 15 ml of cells by hot acid-phenol treatment as described (59). 30 g were separated on a 1% agarose gel containing 6% formaldehyde, blotted to a nylon membrane (Hybond Nϩ, Amersham Pharmacia Biotech), and rRNA was quantified by methylene blue staining. Hybridization to a PCR fragment containing the CTK3 coding sequence labeled with [␣-32 P]dATP by random priming (Appligene) was performed under standard methods.
Yeast Cell Extracts and Western Immunoblotting-Proteins were extracted in trichloroacetic acid by mechanical lysis with glass beads and analyzed by Western immunoblotting as previously described (45).
Detection of Ctk3 Ubiquitin Conjugates-Buffers described previously by Willems et al. (63) were supplemented with 15% (v/v) glycerol. Yeast pellets from 1.5 liters of cultures, with or without induction of the CUP1 promoter, were resuspended in G buffer and subjected to cellular lysis in an Eaton press. After ultracentrifugation at 100,000 ϫ g for 1 h, total ubiquitinated species from 10 mg of yeast proteins were loaded onto 50-l nickel columns (Amersham Pharmacia Biotech) packed into micro-tips with G buffer. After elution, ubiquitinated species were analyzed by Western immunoblotting.
Bacterial Extracts-Escherichia coli BL21(DE3) cells (Novagen) were cotransformed with the pLysS vector (Novagen) and the plasmid coding for the protein of interest. Typically, cells were grown to an A 600 of 0.4 in 100 ml of LB medium. Expression of the T7 polymerase was then induced for 3 h with 250 M isopropyl-1-thio-␤-D-galactopyranoside. Cell pellets were resuspended in 2 ml of BB100 (25 mM HEPES, pH 7.5, 100 mM potassium acetate, 10 mM MgCl 2 , 1 mM dithiothreitol, 1 mM EDTA, 15% (v/v) glycerol) containing 1 mM phenylmethylsulfonyl fluoride and Complete protease inhibitor mixture (Roche Molecular Biochemicals). Cell disruption was completed by sonication. After ultracentrifugation at 100,000 ϫ g, levels of tagged proteins from the extracts were estimated by Western blot analyses.
Coimmunoprecipitation Experiments-Immunoprecipitations from yeast extracts were performed as described previously (45). Briefly, proteins extracted with glass beads from 10 ml of cells were subjected to immunoprecipitation with polyclonal antibody HA.11 (1/125, Babco) and protein G-immobilized on Sepharose beads (Amersham Pharmacia Biotech). Immunoprecipitations from bacterial or wheat germ extracts were performed in the BB100 buffer described above. Bacterial extract containing Myc-Ctk3 and HA-Ctk2 synthesized with [ 35 S]methionine in coupled transcription/translation wheat germ extract (Promega) were mixed overnight at 10°C. Immunoprecipitations were performed with HA-12CA5 antibody coupled to 20 l of magnetic beads according to the manufacturer (Dynal). After a 2-h incubation on a wheel, proteins were eluted with Laemmli buffer and separated by SDS-PAGE prior to Western immunoblotting.
Pulse-chase Experiment-50 ml of yeast cells transformed with the CTK3-HA construct were grown in SD medium to an A 600 of 0.4. Cell labeling was achieved for 10 min at 30°C with 1 mCi of Pro-Mix [ 35 S]methionine/cysteine (Amersham Pharmacia Biotech). After different times of chase by addition of 10ϫ chase medium (10 mM methionine, 2 mM cysteine, 4% yeast extract, 25% glucose), 10-ml aliquots were added to 60 l of STOP solution (0.5 M sodium azide, 0.5 M sodium fluoride) and centrifuged. Pellets were washed with water prior to freezing in liquid nitrogen. Protein extracts and immunoprecipitations were made following the in vivo immunoprecipitation procedure described above. Totality of eluted proteins were separated by SDS-PAGE before analysis by autoradiography.
Reconstitution and Immunopurification of Ctk Complexes-Equal amounts of different bacterial extracts each containing a tagged protein were mixed and incubated for an hour at 15°C. 15 g of purified monoclonal HA-12CA5 antibodies were coupled overnight at 10°C to 30 l of protein G-Sepharose beads in phosphate-buffered saline buffer containing 0.1% to 0.5% bovine serum albumin and 10% (v/v) glycerol.
Typically, 30 l of HA-coupled beads were packed into micro-tips. After equilibration with BB100 buffer, the different mixes of proteins were loaded three times consecutively on these mini-columns with one wash in between each. Native elutions of immunopurified complexes were performed in batch with 50 l of BB100 buffer supplemented with 0.1 g/l HA peptide for 20 min at 37°C. Extracts were either analyzed by SDS-PAGE or Western immunoblotting after addition of Laemmli buffer.
Gel Filtration Analysis-Bacterial extracts containing HA-Ctk2 and Myc-Ctk3, respectively, were mixed and immunopurified on an HA mini-column as described above. 5 l of bacterial extracts containing either HA-Ctk2 or Myc-Ctk3, or 50 l of Ctk2⅐Ctk3 immunopurified complex, were applied on Superdex 75 PC 3.2/30 columns (SMART system, Amersham Pharmacia Biotech). The flow rate during the chromatography was 40 l/min of BB100 buffer, and at 0.95 ml of elution, 30-l fractions were collected and analyzed by Western Immunoblotting.
Purification of Recombinant Ctk Complexes-Purification of high amounts of Ctk complexes was performed by ion exchange and affinity chromatographies on a fast protein liquid chromatography system (Amersham Pharmacia Biotech). Bacterial extract from 500 ml of cells expressing (His) 6 -HA-Ctk1 was loaded onto a 2-ml Bio-Rex 70 column (Bio-Rad) with PC␣(100) buffer (25 mM HEPES, pH 7.5, 10 mM MgCl 2 , 15% glycerol; supplemented with 100 mM potassium acetate). After 5 column-volumes of washing, an excess of recombinant HA-Ctk2, HA-Ctk3, or HA-Ctk2ϩHA-Ctk3 extracts, were loaded on the column. After washing with 10 column-volumes of PC␣(100) buffer, proteins eluted with a linear gradient (20 column-volumes) from 0.1 to 2 M potassium acetate. Fractions containing Ctk proteins were determined by Western immunoblotting with HA antibodies, pooled, and loaded onto 1-ml Hi-TRAP/Ni 2ϩ columns (Amersham Pharmacia Biotech) with PC␣(650) buffer. After 5 column-volumes of washing with PC␣(650) buffer, a linear gradient (5 column-volumes) from 650 to 100 mM potassium acetate was performed, and washing was achieved with 5 columnvolumes of PC␣(100) buffer. Proteins were eluted by step with BB100 buffer without dithiothreitol, supplemented with 0.3 M imidazole.
In Vitro CTD Kinase Assays-Extracts from High5 cells (Invitrogen) infected with baculoviruses expressing the GST-CTD fusion protein (a gift from J. Acker) were treated with the PP2A phosphatase prior to inhibition with okadaic acid and loaded onto a Hi-TRAP/glutathione column (Amersham Pharmacia Biotech). GST-CTD elution was achieved by competition with 20 mM reduced glutathione in BB100 buffer. Activities from 40 l of each purified complex (fraction 7, Fig. 7C), or Ctk1 alone, were assayed by incubation of 2 g of GST-CTD with 3 Ci of [␥-32 P]ATP in BB100 buffer. Enzymatic reactions took place for 1 h at 25°C, before concentration on Strataclean beads (Stratagene) and further elution with Laemmli buffer. The totality of the eluted proteins was separated by SDS-PAGE before analysis by autoradiography.

Ctk3
Is an Unstable Protein-The subunits of the yeast CTDK-I complex are encoded by three genes: CTK1, encoding a cyclin-dependent kinase subunit; CTK2, encoding a C-type cyclin subunit; and CTK3, encoding a 296-amino acid subunit of unknown function. We have previously shown that Ctk2 is processed through a ubiquitin-proteasome pathway (45). To analyze Ctk3 expression in vivo, its coding sequence was cloned on a single-copy plasmid under the control of its own promoter. The protein was fused at its N-terminal to either a c-Myc or two-HA epitope tags. The resulting constructs, CTK3-Myc and CTK3-HA, allowed expression of functional proteins, because each could restore ⌬ctk3 cell growth at the nonpermissive temperature (data not shown). Western blot analyses of ⌬ctk3 cell extracts expressing the Myc-tagged Ctk3 protein revealed a specific band of the expected size (Fig. 1A, lane 0Ј). To analyze Ctk3 stability, cycloheximide was added to cultures grown to an A 600 of 0.8 at 30°C to block cytoplasmic protein synthesis, and Ctk3 level was monitored by immunodetection at various times after cycloheximide addition (Fig. 1A). The half-life of Ctk3 was found to be relatively short (half-life, 30 min). Identical results were observed with the CTK3-HA construct (Fig.  1B). Because cycloheximide treatment might cause a stress affecting Ctk3 stability, we also performed a pulse-chase experiment. Cells were labeled with a [ 35 S]methionine/cysteine mix for 10 min prior to HA-Ctk3 immunoprecipitation after different times of chase. We observed a decrease in Ctk3 levels similar to that observed after cycloheximide addition (Fig. 1C). We conclude that, like the Ctk2 cyclin subunit, Ctk3 is an unstable protein in exponentially growing cells.
A number of proteins with a fast turnover contain PEST sequences (60). PEST regions are enriched in proline, glutamate, serine, threonine, and aspartate, these regions being uninterrupted by positively charged residues and flanked by lysine, arginine, or histidine residues (61). Several studies have shown that these peptide motifs are involved in protein turnover. Examination of Ctk3 sequence with the PESTfind algorithm revealed a strong potential PEST sequence (score of ϩ14.41) composed of a stretch of 18 amino acids located at the N terminus of the protein (Fig. 1D). We made a 13-amino acid deletion inside the potential PEST motif of the Myc-Ctk3 protein. The resulting protein, Ctk3⌬PEST, restored ⌬ctk3 cell growth similar to wild type, indicating that this region is not required for Ctk3 function (data not shown). Ctk3⌬PEST stability was determined after cycloheximide addition. We observed a substantial stabilization of the protein, because its levels were only slightly decreased 6 h after cycloheximide addition (Fig. 1E). Consistently, we also observed a higher level of Ctk3⌬PEST steady state as compared with wild type Ctk3 (data not shown). These results show that Ctk3 contains a functional PEST motif that is involved in its degradation.
Ctk3 Is Degraded by a Ubiquitin-dependent Proteasome Pathway-In eukaryotic cells, PEST sequences have been implicated in ubiquitination processes required for vacuolar degradation (62), as well as degradation by the proteasome (36). Biochemical and genetic evidence indicates that the ubiquitinproteasome pathway is involved in the degradation of abnormal proteins and regulatory proteins with naturally short halflives, such as cell cycle cyclins (35), and the C-type cyclin Ctk2 (45). To establish whether Ctk3 is a substrate for the proteasome degradation pathway, we examined its stability in the conditional pre1 pre2 mutant strain, which is defective in two of the proteasome catalytic subunits at 37°C (55). The CTK3-Myc construct was introduced into pre1 pre2 and wild type isogenic cells. Cell cultures were shifted to the nonpermissive temperature to block proteasome function and analyzed for Ctk3 levels after cycloheximide addition. In wild type cells, Ctk3 turnover was slightly different from that previously observed ( Fig. 2A; half-life, 15 min), indicating that, like Ctk2 turnover (45), it is dependent on genetic background and/or temperature. In pre1 pre2 cells, Ctk3 levels remained almost unchanged for 6 h after cycloheximide addition ( Fig. 2A). In conclusion, Ctk3 is stabilized in a mutant strain that exhibits defects in the activity of the proteasome, showing that its degradation is mediated by the 26 S proteasome pathway.
Although the 26 S proteasome degrades an unknown number of proteins that are recognized without undergoing ubiquitination, the ubiquitin system constitutes the major targeting process leading to selective degradation (33). Despite the fact that Ctk3 was readily observed by immunodetection, we did not observe any ubiquitinated forms. Indeed, due to deubiquitination and degradation by the proteasome, steady-state levels of ubiquitin species are often very low. In an attempt to detect these forms, wild type cells were cotransformed with the CTK3-HA construct and a multicopy plasmid expressing a (His) 6 -Myc-tagged variant of ubiquitin from the CUP1 inducible promoter (32). After elution of total ubiquitin species retained on a nickel column under denaturing conditions (to avoid the recovering of proteins interacting with Ctk3), proteins were analyzed by Western immunoblotting. Hybridization with Myc antibodies allowed detection of equal amounts of ubiquitin species when expression from the CUP1 promoter was induced (Fig. 2B, lower panel). Ctk3 species were further analyzed by hybridization with HA antibodies (Fig. 2B, upper  panel). We could observe high-molecular-weight bands corresponding to Ctk3 ubiquitin conjugates. Taken together, our results demonstrate that Ctk3 is a substrate for a ubiquitinproteasome pathway.
Ctk3 Turnover Is Influenced by the Ctk2 Cyclin-The turnover of cyclins involved in cell cycle progression has been described as being under the control of their associated CDKs (63,64). Because Ctk3 is an unstable protein that is part of a CDK complex, we wondered whether the Ctk1 kinase was implicated in the control of its turnover. The CTK3-Myc construct was introduced into ⌬ctk1 cells, and Ctk3 levels were followed after cycloheximide addition. In the absence of the kinase, Ctk3 stability was similar to that observed in wild type cells ( Fig. 3A; half-life, 30 min), indicating that Ctk3 turnover is a process that requires neither CTDK-I complex assembly nor its activity.
We next studied the stability of the Myc-tagged Ctk3 protein in ⌬ctk2 cells. Strikingly, the Ctk3 steady-state level was very low, because the protein could be detected only when the blots were overexposed (Fig. 3B). We further compared the steadystate level of the HA-tagged Ctk3 in these different strains in the same experiment (Fig. 3C). Indeed, Ctk3 levels were dramatically reduced in ⌬ctk2 cells as compared with ⌬ctk1 cells. Unexpectedly, Northern blot analyses revealed that ⌬ctk2 cells contained substantial amounts of CTK3 mRNA (Fig. 3D). Furthermore, the levels of CTK3 mRNA were similar in ⌬ctk2 and ⌬ctk1 cells (data not shown). Taken together, these results suggest that Ctk3 is synthesized and very quickly degraded in the absence of the cyclin.
Because Ctk2 levels seemed to influence Ctk3 turnover, we next examined the effect of Ctk2 overexpression on Ctk3 sta- bility. When the CTK2 gene was carried by a multicopy plasmid, Ctk3 was significantly stabilized (Fig. 3E; half-life, 120 min, compare with Fig. 1A). In conclusion, the data show that Ctk2 levels strongly influence Ctk3 turnover.
Ctk3 Influences the Ctk2 Cyclin Turnover-A striking feature of the CTDK-I complex is that, unlike that of cell cycle cyclins, the Ctk2 cyclin turnover is not under the control of the kinase that it activates (45). Ctk3 turnover does not require the Ctk1 kinase but is dependent on Ctk2 expression. These observations prompted us to study whether Ctk3 was involved in the control of the Ctk2 cyclin expression. The CTK2-HA construct that allows expression of a functional tagged-cyclin from a single-copy vector (45) was introduced into ⌬ctk1, ⌬ctk2, and ⌬ctk3 cells. As previously described (45), Ctk2 steady-state level was only slightly lower in ⌬ctk1 as compared with pseudo wild type cells (⌬ctk2 cells, Fig. 4A). In contrast, Ctk2 could only be observed in ⌬ctk3 cells when the blots were overexposed (Fig. 4A), suggesting that Ctk2 turnover was very fast in these cells. Also, in exponential cells, Ctk3 overexpression resulted in a significant stabilization of the cyclin (data not shown). We conclude that Ctk3 expression influences the Ctk2 cyclin degradation. To test whether this effect was specific, we analyzed the expression of the catalytic subunit of the CTDK-I complex, namely, the Ctk1 kinase. Unlike the two other subunits, Ctk1 appeared to be a stable protein in exponentially growing cells (half-life Ͼ 6 h, data not shown). Importantly, its steady-state level was not significantly affected in either ⌬ctk2 or ⌬ctk3 cells as compared with pseudo wild type cells (⌬ctk1, Fig. 4B). These results indicate that Ctk3 plays a specific role in the control of Ctk2 destruction.
In conclusion, Ctk2 and Ctk3 mutually control their respective turnover, a process that is independent of their associated kinase.
Ctk2⅐Ctk3 Heterodimerization-To study potential physical interaction between Ctk2 and Ctk3, we first performed coimmunoprecipitation experiments from yeast extracts. ⌬ctk1 cells were cotransformed either with the CTK2-Myc and CTK3-HA constructs or with the CTK2-HA and CTK3-Myc constructs. HA-tagged proteins were immunoprecipitated. As revealed by hybridization with Myc antibodies, Ctk2 was coimmunoprecipitated with Ctk3 (Fig. 5A, left panel); and reciprocally, Ctk3 was coimmunoprecipitated with Ctk2 (Fig. 5A, right panel). These results demonstrate that, even in the absence of the kinase subunit, Ctk2 and Ctk3 are part of the same complex in vivo.
To test a direct interaction between these two proteins, constructs were made to express tagged versions of the Ctk2 and Ctk3 proteins from the T7 promoter. HA-tagged Ctk2 was synthesized in wheat germ extracts, whereas Myc-tagged Ctk3 was bacterially expressed. Extracts were incubated together prior to subjecting the HA-tagged proteins to immunoprecipitation. We observed that Ctk2 immunoprecipitation resulted in coimmunoprecipitation of the Ctk3 protein (Fig. 5B), demon-strating a direct interaction between these two subunits.
To further examine Ctk2⅐Ctk3 complexes, gel filtration experiments were performed. Bacterial extracts containing either the HA-Ctk2 or the Myc-Ctk3 protein were subjected to gel filtration chromatography on a Superdex 75 column. HA-Ctk2 and Myc-Ctk3 were eluted in fractions whose apparent molecular mass corresponded to 44 and 33 kDa, respectively, consistent with their theoretical size (41 and 37 kDa, respectively) (Fig. 5C, upper and middle panels). In addition, Ctk3 aggregates were found in the void volume. When both extracts were preincubated and HA-Ctk2 was further immunopurified prior to loading on the gel filtration column, elution resulted in the presence of both factors in the same fractions corresponding to an apparent molecular mass of 78 kDa (Fig. 5C, lower panel). These data demonstrate that, in vitro, Ctk2 and Ctk3 interact to form heterodimers.
Ctk2⅐Ctk3 Interaction Is Required for Their Mutual Stabilization-Taken together, our data suggest that Ctk2 and Ctk3 turnover are controlled by their heterodimerization. To investigate further the role of Ctk2⅐Ctk3 interaction, we constructed several deletion mutants of the Ctk3 protein. 2 We concentrated our study on a C-terminal 50 amino acid truncation. The resulting protein, Ctk3⌬C50, was unable to restore ⌬ctk3 cell growth at the nonpermissive temperature (data not shown), indicating that this region is essential for Ctk3 function. The CTK3⌬C50-HA and CTK3-HA genes carried on single-copy plasmids were separately introduced into wild type and ⌬ctk2 cells, and the resulting extracts were analyzed. In wild type cells, Ctk3⌬C50 steady-state levels were extremely low as compared with wild type Ctk3, whereas, in ⌬ctk2 cells, both Ctk3 and Ctk3⌬C50 amounts were extremely low (Fig. 6A). Importantly, unlike wild type Ctk3, Ctk3⌬C50 levels were identical in both strains, indicating that the mutant protein was very rapidly degraded, regardless of the presence or the absence of the cyclin encoding gene. We next analyzed Ctk2 expression by introducing the CTK2-Myc construct in ⌬ctk3 cells expressing Ctk3 and ⌬ctk3 cells expressing Ctk3⌬C50, respectively. In the presence of wild type Ctk3, we could easily detect Ctk2 species (Fig. 6B, Ctk3), whereas, in marked contrast, Ctk2 was difficult to detect in the sole presence of the Ctk3⌬C50 mutant protein, even when blots were overexposed (Fig. 6B, ⌬50). Altogether, these results show that both Ctk2 and Ctk3⌬C50 are very unstable proteins. A possible explanation to account for this observation was the absence of interaction between Ctk2 and Ctk3⌬C50. To test Ctk2⅐Ctk3⌬C50 interaction directly, the tagged proteins were expressed in E. coli and incubated overnight prior to loading on an HA column and subsequent HA peptide elution. Remarkably, although the Ctk3⌬C50 protein was immunopurified as efficiently as the wild type Ctk3 (Fig.  6C, compare lane 2 to lane 3, lower panel), Ctk2 did not copurify with the mutant protein (Fig. 7C, lane 3). In contrast, Ctk1 was efficiently copurified with Ctk3⌬C50 (Fig. 6D, lane 3), indicating that the mutant protein is still correctly folded. These results show that the Ctk3 50 C-terminal amino acids are specifically required for Ctk2 binding. In the absence of this interaction, both Ctk2 and Ctk3 become very unstable.
Activation of the Ctk1 Kinase Requires Both Ctk2 and Ctk3 Subunits-In higher eukaryotes, the third component of the TFIIH kinase complex has been described as an assembly factor that promotes in vitro the association of the CDK with its cyclin. To analyze whether Ctk3 is required for Ctk1⅐Ctk2 complex formation, constructs were made to express tagged versions of the Ctk1 protein from the T7 promoter. Different combinations of proteins expressed in E. coli were mixed to allow protein⅐protein interaction to occur, prior to immunopurification on HA columns. We observed that the Ctk1 kinase copurified with the Ctk2 cyclin, regardless of Ctk3 presence (Fig. 7A, compare lane 3 to lane 2). The addition of the third subunit is not required for this interaction, suggesting that Ctk3 is not an assembly factor in vitro. Interestingly, we also observed that Ctk1 copurified with Ctk3 (Fig. 7A, lane 4), showing a direct interaction between the kinase and the third subunit.
Our results show that the CTDK-I complex contains two unstable subunits that can both interact independently with the kinase. We thus wondered whether the cyclin alone, or the third subunit alone, was able to activate Ctk1. This question was difficult to address in vivo because when Ctk2 (or Ctk3) is not expressed, as a consequence, Ctk3 (or Ctk2) is very quickly degraded. To analyze the CTD kinase activity of different complexes, we reconstituted and purified Ctk complexes by ion exchange and affinity chromatography (Fig. 7B). The coelution of the HA-tagged subunits was evidenced by Western immuno-2 V. Goguel, unpublished data. blotting (Fig. 7C). The Ctk proteins contain identical epitope tags, thereby showing identical reactivity to anti-HA antibodies. They coelute in nearly stoichiometric amounts within the different complexes. CTD kinase activity was assayed using as substrate a purified GST/human-CTD fusion protein expressed in insect cells. No contaminant CTD kinase activity was detected with GST-CTD alone (Fig. 7D, lane 1) or by addition of protein fractions eluted after purification from an E. coli/empty vector extract (Fig. 7D, lane 2). When GST-CTD was incubated with the kinase subunit alone, no labeling was detected (Fig.  7D, lane 4). On the other hand, CTD kinase activity was recovered when the three subunits were present (Fig. 7D, lane 3). Interestingly, neither Ctk2 nor Ctk3 alone could activate Ctk1 (Fig. 7D, lanes 5 and 6). In conclusion, Ctk2 and Ctk3 can each interact independently with the kinase, however, both proteins are required for its activation. DISCUSSION In principle, CDK activation is a process achieved by cyclin binding and CDK activating phosphorylation. We show in this report that activation of Ctk1, a CDK involved in the control of RNA polymerase II transcription, requires its association with two unstable subunits: a C-type cyclin and a subunit displaying no homology to known proteins. Interestingly, the turnover of both subunits is controlled by their direct interaction, pointing to a divergent mechanism for CDK regulation.
Activation of the Ctk1 CDK Requires Its Association with Two Distinct Subunits-CDKs are flexible proteins that are regulated in several different ways. In addition to cyclin binding, their complete activation is generally achieved by phosphorylation of the CDK by a CDK-activating kinase. We do not know whether the Ctk1 CDK is phosphorylated in vivo. However, Ctk1 expressed in E. coli can support CTD phosphorylation, suggesting that, like the two other yeast CTD kinases (54,65), this CDK can be active without itself becoming phosphorylated.
So far, of the three yeast CDKs implicated in RNA polymerase II transcription control, only CTDK-I has been described as a heterotrimeric complex. We show that Ctk3, the third component of the CTDK-I complex, is not required for the association of the unphosphorylated Ctk1 kinase with its cyclin subunit in vitro. These results suggest that, unlike Mat1, Ctk3 is not an assembly factor. Also, we observed a direct interaction between Ctk3 and the CDK. In terms of protein interaction, Ctk3 thus behaves like CKIs that have been shown to bind both to cyclin and CDK subunits (for review see Ref. 66). However, Ctk3 is a specific factor that does not share any structural similarity with CKI proteins. The Ctk3 sequence does not appear to contain a cyclin box, and as yet we do not know the Ctk1⅐Ctk3 binding domains.
Based on sequence similarity and functional properties, Cdk9 has been proposed as a potential human ortholog of Ctk1 (24). The transactivator Tat recruits the Cdk9 CTD kinase by binding to the RNA TAR element (for review see Ref. 22) and modifies the substrate specificity of the Cdk9 complex (67). It has been suggested that genes regulated by CTDK-I may share common cis-acting RNA elements that, similar to the TAR element, might attract the CTDK-I complex (21). It is interesting to note that Ctk3 shares with the Tat protein the ability to bind to both the CDK and the cyclin subunit, raising the possibility that Ctk3 and Tat might be functionally related.
Strikingly, despite the formation of Ctk1⅐Ctk2 complexes in vitro, the Ctk2 cyclin is unable to activate its CDK. Indeed, both Ctk2 and Ctk3 subunits are required for Ctk1 CTD kinase activation. Because Ctk2 and Ctk3 expression are mutually dependent on each other, it is difficult to study in vivo, when the CDK might be phosphorylated, whether a single subunit could activate Ctk1. When Ctk2 was overexpressed from the strong PGK1 promoter, we observed a high steady state of the cyclin in ⌬ctk3 cells (data not shown). However, ⌬ctk3 cell growth was not rescued at the nonpermissive temperature. This observation supports the in vitro data and suggests that the cyclin cannot by itself activate its CDK. It thus appears that, within the CTDK-I complex, the cyclin function requires two distinct factors, pointing to a divergent mechanism for activation of a CDK implicated in RNA polymerase II CTD phosphorylation.
Ctk3 Is a Substrate for a Ubiquitin-Proteasome Pathway-Our study of Ctk3 expression revealed that, in exponentially growing cells, it is a relatively unstable protein (half-life, 30 min). Ctk3 turnover is markedly affected in a mutant strain deficient for two of the proteasome catalytic subunits (pre1 pre2) (55), indicating that its destruction is mediated by the proteasome pathway. We identified a nonessential PEST-region located at the N-terminal of the protein involved in Ctk3 turnover. PEST regions are usually composed of phosphorylation sites involved in protein targeting to the ubiquitination machinery. Several studies indicated that phosphorylation of G 1 cyclin PEST regions by their associated CDKs was the signal that triggered their degradation (63,68,69). As yet, we do not know the phosphorylation state of Ctk3 nor its potential role in the regulation of its turnover. However, it is clear that the Ctk3 PEST motif is not a target for the CDK of the complex, because Ctk3 turnover does not require CTDK-I complex assembly nor its activity. This feature is different from what has been described for G 1 cyclins, consistent with the fact that the CTDK-I kinase is subject to a distinct type of regulation.
Ctk3 joins the growing list of naturally short-lived proteins targeted to the proteasome. Interestingly, of the three subunits composing the CTDK-I complex, the two activating subunits Ctk2 and Ctk3 are regulated at the level of protein turnover, whereas the catalytic subunit is a stable protein.
Ctk2⅐Ctk3 Heterodimerization Controls Their Respective Turnover-The results presented here demonstrate that the Ctk2 cyclin plays an essential role in the control of Ctk3 turnover. In cells lacking Ctk2, Ctk3 turnover is dramatically increased, whereas in contrast, Ctk2 overexpression generates a substantial decrease in Ctk3 turnover. Strikingly, Ctk3 steadystate levels also influence Ctk2 turnover. In other words, changing the relative levels of Ctk2 (and Ctk3) leads to important changes in the degradation kinetics of Ctk3 (and Ctk2). Furthermore, Ctk2 and Ctk3 coimmunoprecipitate in the absence of the kinase subunit, and in vitro, they form heterodimers. In cells expressing a Ctk3 mutant protein lacking the cyclin binding domain, both Ctk3 and Ctk2 are very unstable. The fact that both proteins are quickly degraded when they cannot bind to each other strongly suggests that their heterodimerization protects both subunits from degradation by the proteasome.
Ubiquitination of the p27 CKI requires trimeric complex formation with the cyclin and CDK subunits (70). It has also been reported that polyubiquitination of cyclin B is likely to target the complex to the proteasome where Cdc2 is released from cyclin B (71). Because phosphorylated G 1 cyclin bound to Cdc28 has been observed, 3 it is probable that yeast G 1 cyclin is recognized by the ubiquitination machinery while complexed with the CDK. Thus, high affinity association of a CKI or a cell cycle cyclin with its CDK would necessitate regulated proteolytic degradation to break down the complex. In contrast, our data indicate that neither Ctk2 nor Ctk3 destruction occurs within the CTDK-I complex.
Protein interactions have often been described as a means to 3 E. Ceccarelli and C. Mann, submitted for publication. regulate transcription factor turnover, but this is the first description of the regulation of cyclin turnover by heterodimerization with a specific factor. Interestingly, Johnson et al. (72) also reported mutual stabilization of yeast proteins: the MAT transcription factors. Coexpression of a1 and ␣2 which occurs in a/␣ diploids results in a dramatic stabilization of both proteins. The authors further demonstrated that a1 and ␣2 turnover is controlled by their direct physical interaction that masks proteolytic signals. Ctk2 and Ctk3 are both constitutively expressed in the cell, implying that the degradation of these two unstable regulatory factors requires either prevention of their interaction or promotion of the dissociation of Ctk2⅐Ctk3 complexes. What are the regulatory signals that could induce Ctk2⅐Ctk3 dissociation? Interaction with another factor (such as a component of the ubiquitination machinery for example) could bring conformational changes leading to Ctk2⅐Ctk3 dissociation. Alternatively, Ctk2 and/or Ctk3 protein modification could trigger this dissociation. Several lines of evidence suggest that Ctk2 phosphorylation is a signal for its rapid destruction (45). In vivo, Ctk3 immunoprecipitation from wild type as well as ⌬ctk1 cells allowed recovery of both phosphorylated and unphosphorylated species of the cyclin, indicating that Ctk2 phosphorylation is not a signal for Ctk2⅐Ctk3 dissociation. Because Ctk2⅐Ctk3 binding prevents their respective degradation, it is tempting to speculate that, similar to the case of a1 and ␣2, target sites for the ubiquitination machinery are not exposed when the two proteins are associated, and that only their dissociation would expose, at least transiently, these degradation sites. It will be important to determine whether degradation determinants in Ctk2 and Ctk3 overlap (or are closely juxtaposed to) the interaction surfaces of these two molecules.
The CTDK-I Complex: A New Mechanism for CDK Regulation?-Most CDKs interact sequentially with multiple cyclin subunits (27), cyclin binding thus regulating kinase temporal activity, as well as substrate specificity. In contrast, only one cyclin has been characterized per yeast CTD kinase. Interestingly, Ctk1 activation supposes a balanced concentration of two specific regulatory subunits. Control of protein degradation by regulating protein⅐protein interaction could establish modulations in CTDK-I activity rather than the classic on/off switch observed for other CDKs. In addition, Ctk2 and Ctk3 might respond to distinct signals thus allowing a larger spectrum of signals to be integrated. The different specific features governing CTDK-I regulation probably reflect requirement for the transcriptional response involved in cell adaptation to multiple growth conditions.