Capping of nascent pre‐mRNAs is thought to be a prerequisite for productive elongation and associated serine 2 phosphorylation of the C‐terminal domain (CTD) of RNA polymerase II (PolII). The mechanism mediating this link is unknown, but is likely to include the capping machinery and P‐TEPb. We report that the fission yeast P‐TEFb (Cdk9‐Pch1) forms a complex with the cap‐methyltransferase Pcm1 and these proteins colocalise on chromatin. Ablation of Cdk9 function through chemical genetics causes growth arrest and abolishes serine 2 phosphorylation on the PolII CTD. Strikingly, depletion of Pcm1 also leads to a dramatic decrease of phospho‐serine 2. Chromatin immunoprecipitations show a severe decrease of chromatin‐bound Cdk9‐Pch1 when Pcm1 is depleted. On the contrary, Cdk9 is not required for association of Pcm1 with chromatin. Furthermore, compromising Cdk9 activity leads to a promoter‐proximal PolII stalling and sensitivity to 6‐azauracil, reflecting elongation defects. The in vivo data presented here strongly support the existence of a molecular mechanism where the cap‐methyltransferase recruits P‐TEFb to chromatin, thereby ensuring that only properly capped transcripts are elongated.
The C‐terminal domain (CTD) of the largest subunit of RNA polymerase II (PolII) plays a critical role in the integration of transcription with cotranscriptional events such as mRNA capping, splicing or polyadenylation (McCracken et al, 1997b; Hirose and Manley, 2000; Proudfoot et al, 2002). A variety of factors required for processing interact in a coordinated manner with the CTD, which consists of repetitions (27 in budding yeast, 29 in fission yeast and 52 in mammals) of the consensus YSPTSPS containing two main phospho‐acceptor sites: serine 2 (Ser2) and serine 5 (Ser5) (Hirose and Manley, 2000; Bentley, 2005). Among the CTD kinases identified, the most prominent ones belong to the CDK (cyclin‐dependent kinases) family. CDK7 is part of the general transcription factor TFIIH and phosphorylates the CTD on Ser5, whereas CDK9 associates with T‐type cyclin functions in P‐TEFb (positive transcription elongation factor) and favours Ser2, although several reports extended its normal activity to Ser5 and the HIV Tat protein shifts CDK9 preference from Ser2 to both Ser2 and Ser5 (Zhou et al, 2000). P‐TEFb was isolated by its ability to overcome arrest of PolII complexes during early elongation, a function that requires the CTD (Marshall and Price, 1995; Marshall et al, 1996). However, P‐TEFb had no detectable effect on purified PolII and was not required in a fractionated system (Peng et al, 1998), suggesting that its stimulatory effect occurs by alleviating the negative effect of factors present in the crude extract. These factors likely include the Spt4–Spt5 complex (Wada et al, 1998), which represses elongation and allows the recruitment of the capping machinery (Wen and Shatkin, 1999; Pei and Shuman, 2002; Yamada et al, 2006).
The ‘cap 0’ structure (m7GpppN1) is a specific modification of the 5′ end of eukaryotic mRNA and consists of a 7‐methylguanosine moiety attached through an unusual 5′–5′ link to the 5′ terminus of mRNA (Shuman, 2001). Three enzymatic activities are required to generate capped mRNA. RNA triphosphatase removes the γ‐phosphate group of the first nucleotide followed by the transfer of GMP to the remaining diphosphate by RNA guanylyltransferase and the addition of a methyl group at the N7 position on the guanine (Shuman and Schwer, 1995). Metazoans have a bifunctional RNA triphosphatase‐guanylyltransferase enzyme and a methyltransferase, whereas three enzymes are present in yeast: triphosphatase (Cet1 in Saccharomyces cerevisiae, Pct1 in Schizosaccharomyces pombe), guanylyltransferase (Ceg1 in S. cerevisiae, Pce1 in S. pombe) and methyltransferase (Abd1 in S. cerevisiae, Pcm1 in S. pombe). Capping is the earliest modification occurring when the transcript is about 30 nucleotides long (Jove and Manley, 1984; Rasmussen and Lis, 1993). In budding yeast, Ceg1 and Abd1 bind exclusively to Ser5 phosphorylated CTD (Cho et al, 1997; McCracken et al, 1997a, 1997b; Komarnitsky et al, 2000; Rodriguez et al, 2000; Schroeder et al, 2000), ensuring recruitment of capping enzymes to the promoter‐proximal‐arrested polymerase and selective capping of PolII transcripts. The phosphorylation on Ser5 is therefore likely to potentiate interactions with processing factors resulting in increased recruitment. Phosphorylation on Ser5 also dramatically enhances guanylylation (Ho and Shuman, 1999; Wen and Shatkin, 1999), suggesting that it allows allosteric activation. Removal of Ser5 phosphorylation by specific phosphatases early in elongation releases the guanylyltransferase from the CTD after PolII has transcribed a few hundred transcribed bases whereas the methyltransferase remains bound throughout the transcribed unit (Komarnitsky et al, 2000; Schroeder et al, 2000).
Some specific activators and the chromatin remodelling protein Brd4 are known to recruit P‐TEFb (Jang et al, 2005; Yang et al, 2005) and a two‐hybrid interaction between Cdk9 and the RNA triphosphatase has been reported (Pei and Shuman, 2002; Pei et al, 2003). However, it remains unclear how the recruitment of capping enzymes by the CTD phosphorylated on Ser5 is linked to subsequent Ser2 phosphorylation and elongation in vivo. Using chromatin immunoprecipitation (ChIP), we show that the P‐TEFb homolog of fission yeast (Cdk9–Pch1) forms a complex with and is recruited to chromatin by the cap‐methyltransferase in vivo.
Results and discussion
The fission yeast P‐TEFb forms a soluble complex with the cap‐methyltransferase Pcm1
In budding yeast, two CDKs (Bur1 and Ctk1) are equally similar in their sequence to mammalian Cdk9, the kinase subunit of P‐TEFb. The essential Bur1 kinase promotes transcription elongation, but does not seem to affect Ser2 phophorylation (Murray et al, 2001; Keogh et al, 2003), whereas the non‐essential Ctk1 associates with the elongating polymerase and is the main Ser2 kinase (Lee and Greenleaf, 1991; Ahn et al, 2004). Therefore, it appears that the metazoan Cdk9 function is shared between these two kinases in budding yeast. The fission yeast Pch1 cyclin belongs to the cyclin T family and has been isolated as a two‐hybrid partner of Cdc2 (Furnari et al, 1997). However, several lines of evidence indicate that Cdc2 is not its kinase subunit in vivo. Moreover, another Cdk (SpCdk9) related to metazoan Cdk9 was reported to interact with Pch1 in a two‐hybrid assay (Pei et al, 2003; Pei and Shuman, 2003). This prompted us to purify the proteins associated with Pch1 in vivo using the tandem affinity purification (TAP) method (Puig et al, 2001). A TAP‐tagged version of Pch1 was expressed from the endogenous locus and a soluble fraction was used for TAP purification. A silver‐stained gel of the final eluted product is shown in Figure 1A. Mass spectrometry analysis of the released peptides (Supplementary Figure 1) from the three bands excised from the gel revealed the presence of Cdk9, Pch1 and the mRNA cap‐methyltransferase Pcm1, suggesting that the fission yeast P‐TEFb Cdk9–Pch1 forms a soluble trimeric complex with Pcm1. The association of Pch1 with Pcm1 in a complex was confirmed by independent co‐IP presented in Figure 1B. These experiments confirm the data reported by Pei et al (2006) on the existence of the Cdk9–Pch1–Pcm1 complex.
SpCdk9‐Pch1 is the functional homolog of P‐TEFb and the main Ser2 kinase in fission yeast
Both Spcdk9 (data not shown; Pei et al, 2006) and pch1 are essential genes (Furnari et al, 1997). In order to analyse the effect of SpCdk9 inactivation, we used chemical genetics, where a functionally silent mutation in the active site sensitises the target kinase to inhibition by small cell‐permeable ATP analogues that do not interfere with wild‐type (wt) kinases (Bishop et al, 2001; Shokat and Velleca, 2002). Based on previous work using the budding yeast CDK1 homologue (Bishop et al, 2000), we mutated threonine 120 in the active site (Supplementary Figure 2) of SpCdk9 to glycine and integrated the mutant at the locus under the control of its endogenous promoter. Growth of the Cdk9‐as (analogue sensitive) mutant was affected only in the presence of the Nm‐PP1 (4‐amino‐1‐tert‐butyl‐3‐(1′‐naphthylmethyl)pyrazolo(3,4‐d)pyrimidine) inhibitor and the expression levels of mutant and wt Cdk9 were similar (Figure 1C and D). After treatment with the inhibitor for 1 h, the phosphorylation of Ser2 in the CTD (detected using the H5 antibody) was strongly reduced when 1 μM was used and completely abolished at higher concentrations (10 μM) (Figure 1C). Phosphorylation of Ser5 (detected using the H14 antibody) was unaffected. These results suggest that SpCdk9 is the main, if not the only, Ser2 kinase in fission yeast, consistent with data from higher eukaryotes (Shim et al, 2002; Ni et al, 2004). This was confirmed by in vitro analysis performed on immunoprecipitated Pch1 or Mcs2, the cyclin subunit of TFIIH and S. pombe homolog of cyclin H. Both complexes phosphorylate the CTD (Figure 1E), but the Mcs2‐associated kinase (Mcs6) did so exclusively on Ser5, whereas the Pch1‐associated kinase (Cdk9) showed a marked preference for Ser2 as shown by phospho‐specific Western blot analysis (Figure 1F). To confirm that SpCdk9 is the genuine Cdk9 ortholog in fission yeast, we replaced its open reading frame with a cDNA of wt Human Cdk9 (a kind gift from O Bensaude; Garriga et al, 1996). To avoid interference from high‐level expression, integration was carried out at the cdk9 locus under the control of the endogenous promoter (Figure 1D and Supplementary Figure 3). Growth of the resulting strains was indistinguishable from controls. These data establish that SpCdk9 is the functional orthologue of metazoan Cdk9.
The SpCdk9, Pch1 and Pcm1 proteins colocalise on transcription units during constitutive or activated transcription
The existence of the soluble Cdk9–Pch1–Pcm1 complex led us to analyse the distribution of these proteins on chromatin. As no localisation data were available in fission yeast for either the general transcription factor TFIIH (responsible for Ser5 phosphorylation) or Tbp1 (TATA‐binding protein), we first determined the position of Mcs2 and of Tbp1 using ChIP (see Material and methods) with C‐terminal HA‐tagged proteins. The adh1 and act1 genes were first chosen for analysis because of their constitutive transcription rate, and four primer pairs spanning the promoter, ORF (two pairs) and termination regions (as indicated in Figure 2A) were selected for ChIP analysis. As shown in Figure 2A, HA‐tagged Tbp1 and Mcs2 crosslinked mostly upstream of the ORF as expected. When testing Rpb3, a subunit of PolII, we found even crosslinking across the open reading frames. Cdk9, Pch1 and Pcm1 have remarkably similar occupancy profiles with a maximum downstream of the TATA box (based on Tbp1 position) and then decreasing towards the 3′ region, suggesting a common loading and dissociation mechanism for these proteins. These data are consistent not only with the association of these proteins in complex, but also with their proposed function: Ser2 kinase for Cdk9‐Pch1 and methyltransferase for Pcm1. Our results are also in line with the recently reported occupancy of human (Gomes et al, 2006) and fly (Boehm et al, 2003) P‐TEFb and budding yeast cap‐methyltransferase (Komarnitsky et al, 2000; Schroeder et al, 2000). We next analysed the localisation of Cdk9, Pch1 and Pcm1 during transcriptional activation. The pyp2 gene encoding a phosphatase involved in osmotic stress (sorbitol 1 M) response was chosen both for the simplicity of the induction system and its more complex promoter structure recognised by the stress‐activated transcription factor Atf1 (Wilkinson et al, 1996). Figure 2B presents binding data before or after induction as in Figure 2A with additional primer pairs in the promoter region. Localisations in agreement with the results from constitutive genes (Figure 2A) were observed after induction. Nevertheless, the occupancy of Pcm1 and Cdk9–Pch1 does not precisely correlate, which might indicate that interactions with other proteins occur on chromatin. The binding of the Pcm1–Cdk9–Pch1 proteins also occurs in a more extended region upstream of Rpb3 recruitment, probably reflecting a more complex promoter structure as expected for an inducible stress control gene.
No significant binding of analysed proteins was observed under normal growth conditions, suggesting that in this case, the whole machinery, including TBP and general transcription factors, is not present on chromatin, but is rather recruited upon induction by stress. Taken together, these data show that the Cdk9–Pch1–Pcm1 complex first identified in soluble extracts (Figure 1A) is likely to also exist on chromatin, raising the hypothesis that the cap‐methyltransferase might recruit P‐TEFb. This would link the final step in cap assembly to Ser2 phosphorylation and elongation.
Cdk9 is recruited to chromatin by the Pcm1 cap‐methyltransferase
To test the above hypothesis, depletion mutants were generated by replacing the promoters of cdk9 and pcm1 at their endogenous locus with the thiamine‐repressible nmt81 promoter, which is widely used to switch off transcription (S/O) in fission yeast (Basi et al, 1993). Upon thiamine addition, both Cdk9 and Pcm1 were largely depleted from the cells within 4 h (Figure 3A). We assayed the phosphorylation level of the CTD on Ser2 and Ser5 with the H5 and H14 antibodies and in line with data obtained with the cdk9‐as mutant (see Figure 1D). A marked decrease in Ser2 phosphorylation was observed after Cdk9 depletion, whereas Ser5 was unaffected. Interestingly, depletion of Pcm1 also resulted in a decrease in Ser2 phosphorylation, whereas levels of Cdk9 under the same conditions remained unchanged (Figure 3A). These S/O strains allowed us to test the requirement of either Pcm1 or Cdk9 for the chromatin association of the other one. Pcm1 and Cdk9 occupancy was analysed for three genes: adh1 and act1 as described in Figure 2A and the housekeeping plasma membrane ATPase encoded by pma1, which has been widely used in budding yeast (Ahn et al, 2004). In each case, we probed the 5′ region where the abundance of Pcm1 and Cdk9 is maximal (see Figure 2A, data not shown for pma1). As a convenient method of quantitation, we calculated the ratio of ChIP signals normalised to input in the presence and absence of thiamine treatment for 6 h (Figure 3B). Therefore, a value of 1 would mean that thiamine treatment did not affect the level of occupancy. After addition of thiamine, when either Cdk9 or Pcm1 was depleted, the strong reduction in protein level observed in total extracts (Figure 3A) was reflected in the ChIP. Depletion of Cdk9 did not affect the level of chromatin association of Pcm1. In contrast, the binding of Cdk9 to chromatin was severely reduced in the absence of Pcm1 for all three genes tested. This effect could not be attributed to a global effect on Cdk9 stability as the steady‐state level of Cdk9 was unaffected (Figure 3A). We next examined the chromatin recruitment of the Pch1 cyclin in the absence of either Cdk9 or Pcm1. While Pcm1 was equally required for the recruitment of Cdk9 and Pch1 (Figure 3B), the depletion of Cdk9 only partially affected the chromatin binding of Pch1 with between 50 and 70% of Pch1 still present after Cdk9 depletion (Figure 3B). This suggests that a Pcm1–Pch1 complex is still formed in the absence of Cdk9 as confirmed by co‐IP (Figure 3C). However, we failed to detect an interaction between Pcm1 and Pch1 in the two‐hybrid system (Supplementary Figure 4).
The strong decrease in Cdk9 and Pch1 chromatin association when Pcm1 is depleted also explains why the Ser2 phosphorylation signal decreases after Pcm1 depletion (Figure 3A). We excluded an indirect effect of decreased PolII abundancy by performing Rpb3 ChIP analysis in the Pcm1‐depleted strain (Figure 3B and Supplementary Figure 5).
Taken together, these data show that the cap‐methyltransferase is recruited in the absence of Cdk9. This is likely to occur through interaction with the CTD phosphorylated on Ser5, which is not affected by Cdk9 depletion (Figure 3A). These observations also show that the presence of Pcm1 on chromatin is required to recruit or stabilise Cdk9 and Pch1. We favour the recruitment hypothesis because Cdk9 and Pcm1 are found in a soluble complex in the cell (Figure 1A).
A two‐hybrid interaction was reported between the mRNA triphosphatase Pct1 and Cdk9 (Pei et al, 2003), but could not be observed by affinity purifications (Pei et al, 2006 this work). ChIP analysis revealed that Pct1 occupancy unexpectedly covers the entire transcription unit closely following PolII occupancy (Supplementary Figure 6A). Depletion of Cdk9 did not affect Pct1 recruitment (Supplementary Figure 6B); so the physiological relevance of the two‐hybrid interaction between Pct1 and Cdk9 remains unclear.
These data strongly suggest the existence of a checkpoint that ensures that Cdk9 is not recruited to chromatin before capping has occurred, thus making sure that uncapped transcripts are not elongated. We therefore tested if the absence of Pcm1 or Cdk9 activity impeded elongation by PolII.
The polII stalls at promoter‐proximal 5′ regions when Cdk9 activity is compromised
The ratio of Rpb3 crosslinking signal in the absence or presence of either 1 or 10 μM of the Cdk9‐as inhibitor NmPP1 (see Figure 1C and D) was calculated along three genes (Figure 4A). The inhibitor had no effect on a wt strain (data not shown; Figure 1C and D). At these inhibitor concentration, a reduced association of PolII at the 3′ end of genes is observed together with significantly increased PolII abundance, up to 60% for act1, in the 5′ regions close to the promoter. For pma1, the effect is less marked, likely owing to the fact that, for technical reasons, the first probe is located further downstream in the open reading frame. This behaviour is reminiscent of elongation mutants (Keogh et al, 2003), and shows that PolII stalls when Cdk9 kinase activity is inhibited and the elongation pause cannot be overcome. In line with these data, depletion mutants of Cdk9 or Pcm1 and the cdk9‐as mutant all have increased sensitivity to 6‐azauracil (6‐AU), a drug that reduces GTP and UTP levels (Exinger and Lacroute, 1992; Mason and Struhl, 2005), thereby inhibiting elongation rate and processivity (Figure 4B). Although polymerase stalling and elongation defects are already observed at low inhibitor concentrations (1 μM), growth is affected only when a higher inhibitor concentration (10 μM) is used and serine 2 phosphorylation is completely abolished (Figures 1C, D, 4A and 4B). This suggests that other defects associated with the loss of phosphorylated serine 2 are the primary cause of cell death.
Similar to cell‐cycle progression, where the completion of one step enables the onset of the following one, key steps during transcription are interconnected to avoid the wasteful production of incomplete or misprocessed mRNA. A ‘checkpoint’ model has been proposed for the coupling of pre‐mRNA capping and early transcription elongation, where following the addition of a cap to the nascent RNA, P‐TEFb binds to and phosphorylates PolII to allow productive elongation (Wen and Shatkin, 1999; Orphanides and Reinberg, 2002; Pei et al, 2003).
Here we described the molecular basis of this checkpoint mechanism (Figure 5) by showing that the Pcm1 cap‐methyltransferase forms a tight soluble complex with Cdk9–Pch1, the fission yeast P‐TEFb homologue. When this connection is disrupted, Cdk9 is not properly recruited to chromatin and PolII elongation is severely affected. The operative replacement of fission yeast Cdk9 by its human ortholog expressed from the endogenous locus suggests that this mechanism is likely to be conserved in metazoans.
Materials and methods
Yeast strains, media and mutant construction
All yeast strains used in this study are presented as Supplementary Table 1. All experiments were performed in rich YES medium except for thiamine treatment, where minimal EMM medium was used as reported by Moreno et al (1991). HA‐ and TAP‐tagged strains and nmt81 promoter replacement were generated as described (Bahler et al, 1998; Van Driessche et al, 2005). The cdk9 locus was reluctant to recombination using short stretches of homology and in that case, upstream and downstream flanking regions of about 1 kb were amplified by PCR and cloned in adequate vectors. Restriction fragments were purified and used for transformation. Human Cdk9 cDNA (Garriga et al, 1996) was transferred in adequate vectors for integration at the locus. The S. pombe cdk9‐as mutant was generated in vitro using the ‘Quickchange’ kit (Stratagen) and integrated at the locus using a selection marker. Two inhibitors (Toronto Research Chemicals Inc.), Nm‐PP1 (4‐amino‐1‐tert‐butyl‐3‐(1′‐naphthylmethyl)pyrazolo(3,4‐d)pyrimidine) and Na‐PP1 (4‐amino‐1‐tert‐butyl‐3‐(1′‐naphthyl)pyrazolo(3,4‐d)pyrimidine), were tested. Nm‐PP1 was chosen because it gave stronger inhibition. Cloning details are available upon request. All integrations were verified by PCR and sequencing for mutant integration.
Two‐hybrid assay and yeast expression vectors
Two‐hybrid assay was performed following the manufacturer's instructions (Clonetech Matchmaker II). For cloning in the two‐hybrid vectors, cDNA was amplified by PCR using Phusion Taq polymerase (Finzyme). PCR products were digested and cloned in either pAS‐1 or pACT‐II as indicated in the figure. The fission yeast vector expressing HA‐Mcs2 or HA‐Pch1 was previously described (Hermand et al, 1998).
Tandem affinity purification
TAP purification was performed essentially as described (Tasto et al, 2001). Shortly, 12 l of the Pch1‐TAP strain were grown to an optical density (OD) of 0.5 (5.5 × 106 cells per ml), pelleted and frozen in liquid nitrogen by passing the pellet through a 50‐ml syringe to produce ‘noodles’. The extract was obtained using an electrical mortar/pestle (Reich RM100) under liquid nitrogen. The following steps were as described (Tasto et al, 2001). Aliquots of the elution product were separated on Novex gradient gels and silver stained (Silver Quest, Invitrogen). The protein bands were cut from the gel, destained and extracted, followed by dialysis into ammonium bicarbonate. After trypsination, peptides were analysed by MS‐MS and resulting data were compared with the fission yeast proteome. For the co‐IP in Figure 1B, extracts from a 100 ml culture volume were prepared as described above.
Western blotting, co‐IP and kinase assay
Total protein extracts were prepared as described (Moreno et al, 1991) and separated on 10% (7.5% when Rpb1‐CTD was analysed) polyacrylamide gels. After transfer on PVDF membranes, the following antibodies were used: anti‐HA, H5 (recognizing phospho‐Ser2) and H14 (recognizing phospho‐Ser5), all from Covance‐Eurogentec and PAP (Sigma). For detection, Fuji films and ECL+ (GE Healthcare) were used. Co‐IPs and kinase assays were performed as decribed (Hermand et al, 1998).
Chromatin immunoprecipitation assay
Strains were grown in rich medium (YES) at 32°C to OD 0.4. For pyp2 induction, cells were grown to an OD of 0.8 and an equal volume of prewarmed YES+2 M sorbitol was added to a final concentration of 1 M sorbitol and incubated for an additional 15 min. Proper induction was verified by Northern blot (data not shown).
ChIP were performed essentially as described (Kuras, 2004) with anti‐HA antibodies (F7; Santa Cruz Biotechnology) bound to protein‐A–Sepharose or IgG–agarose for TAP‐tagged proteins. Input and immunoprecipitated DNA samples were used as template in real‐time PCR reactions. Primers located along the adh1, act1, pma1, tef3 and pyp2 loci were designed with Primer Express 1.0 (Applied Biosystems; see Supplementary Table 2 for amplicon sizes and primer localisation and sequences). PCR products ranged from 80 to 100 bp. PCR reactions were performed with SYBR Green Mix (Applied Biosystems). Relative quantitation using a standard curve method was performed for each primer and 96‐well Optical Reaction plates (Applied Biosystems) were used in an Applied Biosystems 7900HT real‐time PCR instrument (absolute quantification method). Input DNA values were used to normalise ChIP, which are presented as a percentage of precipitated DNA (IP)/total DNA (IN). All experiments were performed in triplicate and mean values with standard deviations are presented.
Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).
Supplementary Figure 1
Supplementary Figure 2
Supplementary Figure 3
Supplementary Figure 4
Supplementary Figure 5
Supplementary Figure 6
Supplementary Table 1
Supplementary Table 2
We are grateful to Olivier Bensaude for reagents and Edouard Delaive for help with Mass Spectrometry analysis. We thank Thomas Westerling, Tomi Mäkelä and Damien Coudreuse for critical reading of the manuscript. LT was supported by a FRIA fellowship. AG was supported by an FEMS fellowship and an FRIA fellowship. DH was supported by the Human Frontier Science Program (LT00665/2003). This work was supported by FNRS grant 1.5.076.02 (to DH) and FRFC grant 2.4504.00/1999 (to JVDH and DH).
- Copyright © 2007 European Molecular Biology Organization