G₁ cyclin–Cdk promotes cell cycle entry through localized phosphorylation of RNA polymerase II

Abstract

Cell division is thought to be initiated by cyclin-dependent kinases (Cdks) inactivating key transcriptional inhibitors. In budding yeast, the G₁ cyclin Cln3-Cdk1 complex is thought to directly phosphorylate the Whi5 protein, thereby releasing the transcription factor SBF and committing cells to division. We report that Whi5 is a poor substrate of Cln3-Cdk1, which instead phosphorylates the RNA polymerase II subunit Rpb1’s C-terminal domain on S₅ of its heptapeptide repeats. Cln3-Cdk1 binds SBF-regulated promoters and Cln3’s function can be performed by the canonical S₅ kinase Ccl1-Kin28 when synthetically recruited to SBF. Thus, we propose that Cln3-Cdk1 triggers cell division by phosphorylating Rpb1 at SBF-regulated promoters to promote transcription. Our findings blur the distinction between cell cycle and transcriptional Cdks to highlight the ancient relationship between these two processes.

The eukaryotic cell cycle is driven by a series of cyclin-dependent kinase (Cdk) complexes that promote cell cycle progression by phosphorylating key substrates. The first step of the cell cycle, from the G₁ to the S phase, is thought to require the inactivation of a transcriptional inhibitor. In human cells, cyclin D-Cdk4 and Cdk6 complexes phosphorylate the retinoblastoma protein Rb, and in budding yeast, Cln3-Cdk1 is thought to phosphorylate the transcriptional repressor Whi5. This results in the activation of the E2F and SBF transcription factors in animal and yeast cells, respectively, that commit cells to cell cycle entry through positive-feedback loops in which downstream cyclin-Cdk complexes hyperphosphorylate Whi5 and Rb (Fig. 1A) (1). However, in early to mid G₁, cyclin D-Cdk4,6 constitutively hypophosphorylates Rb (2), which is unlikely to account for its complete inactivation (3), and Whi5 phosphorylation by Cln3-Cdk1 has not been observed in vivo. Thus, key mechanistic aspects of the eukaryotic G₁-S transition model remain either unknown or untested.

Fig. 1. Whi5 is a poor substrate of Cln3-Cdk1.

(A) Budding yeast G₁-S network. (B) Immunoblots showing Whi5-3xFlag phosphorylation in G₁ cells collected by centrifugal elutriation treated with rapamycin to deplete Cln3-FRB from the nucleus or with dimethyl sulfoxide (see fig. S1, A to E). (C) Anti-V5 ChIP-seq signal of the indicated genotypes at the CLN2 locus. (D) Cell size distributions for the indicated genotypes. (E and F) Autoradiographs of in vitro phosphorylation of SBF-interacting proteins and histone H1 by the indicated cyclin-L-Cdk1 complexes.

To examine the prevailing model that Cln3-Cdk1 phosphorylates Whi5 (4, 5), we used Phos-tag–supplemented SDS–polyacrylamide gel electrophoresis (6) to separate distinct phospho-isoforms of Whi5, including multiple hypophosphorylated species not previously observed (Fig. 1B and fig. S1, B and F). We examined early G₁ cells collected by elutriation as well as cells released from G₁ pheromone arrest. Whi5 hypophosphorylation is slightly reduced upon release from pheromone but is otherwise constant throughout early and mid G₁ (Fig. 1B and fig. S1, A to F). Subsequently, at the G₁-S transition, Whi5 becomes fully hyperphosphorylated, presumably by the newly synthesized Cln1/2-Cdk1 complexes (4, 7). To determine whether the early G₁ hypophosphorylation of Whi5 is caused by Cln3-Cdk1, we conditionally depleted Cln3 from the nucleus in early G₁ using the rapamycin-dependent “anchor-away” technique (8). After Cln3 depletion, Whi5 hypophosphorylation was similar to that in untreated early and mid G₁ cells (Fig. 1B and fig. S1, A to E). We observed the same pattern in wild-type (WT) and cln3Δ cells after release from pheromone (fig. S1, F and G). Rapid Whi5 hyperphosphorylation in late G₁ is delayed by Cln3 depletion or cln3Δ, as reported previously (9, 10) (Fig. 1B and fig. S1, B, F, and G), likely because of Cln1/2-Cdk1 expression being delayed in the absence of upstream Cln3-Cdk1 activity. On the basis of these data and additional supporting experiments (fig. S1, H to L), we propose that Cln1/2-Cdk1 complexes complete Whi5 hyperphosphorylation in late G₁, but Whi5 hypophosphorylation in early G₁ is not predominantly caused by Cln3-Cdk1.

If Cln3 promotes cell cycle entry, but not through Whi5 phosphorylation, it is possible that Cln3 and Whi5 act as separate inputs regulating SBF. To test this, we performed chromatin immunoprecipitation sequencing (ChIP-seq) for Cln3 and the SBF components Swi4 and Swi6. Cln3 peaks were found at 68 gene promoters, 67 of which were also bound by SBF (table S3), including key SBF-binding sites in the CLN1 and CLN2 promoters (Fig. 1C and fig. S2, A, B, D, and E) (9). Cln3 localization to SBF-binding sites did not depend on Stb1 or Whi5, Cln3’s previously assumed targets (Fig. 1C and fig. S2D). Consistent with the separate input model, cln3Δwhi5Δ cells are larger than whi5Δ cells (4, 5), Cln3-dependent activation of the CLN2 promoter precedes loss of Whi5 protein from the promoter (9), and a hyper-active CLN3 allele (CLN3ΔC) reduces cell size more than whi5Δ alone (Fig. 1D).

That Whi5 is so poorly phosphorylated by Cln3-Cdk1 in vivo indicates that Cln3-Cdk1 may phosphorylate a different target involved in SBF-dependent transcription. To explore this, we examined Cln3-Cdk1 kinase activity toward SBF-interacting proteins in vitro. To purify Cln3-Cdk1, we fused CLN3 to CDK1 with a glycine-serine linker (CLN3-L-CDK1) because Cln3 is not as tightly bound to Cdk1 as other cyclins. When the endogenous CLN3 was replaced by this fusion, cells exhibited no cell cycle defects, and the removal of the linker after purification had little effect in vitro (figs. S3, A and H to J, and 4, A to C). Thus, this fusion complex appeared to function similarly to the WT complex. We note that kinase activity detected in a previously reported purification of Cln3-Cdk1 from insect cells (5), which we repeated, was not caused by Cln3 because activity was still present in a control lacking Cln3 (see the materials and methods; fig. S3, B and C). Consistent with our in vivo data, Cln3-L-Cdk1 poorly phosphorylated Whi5 in vitro, whereas Cln2-L-Cdk1 readily hyperphosphorylated Whi5 (Fig. 1E). Moreover, Cln3-L-Cdk1 was also a very poor kinase toward the SBF-associated proteins Swi6, Stb1, and Msa1 in vitro (Fig. 1F) (10, 11).

Cln3-Cdk1’s lack of in vitro activity against SBF-associated proteins and the model Cdk substrate H1 (Fig. 1F) suggested that Cln3 may promote the G₁–S transition independently of Cdk1 kinase activity. To test this, we fused CLN3 to a catalytically inactive (kinase-dead, KD) CDK1 allele (12) (CDK1^KD; fig. S3D). Although CLN3-L-CDK1^KD rescued the effects of deleting CLN3, immunoprecipitation revealed endogenous Cdk1 bound to Cln3-L-Cdk1 (fig. S3C). To prevent this, we introduced a cyclin box mutation (CBM) in CLN3 (13) (CLN3^CBM; Fig. 2A). Replacing CLN3 with CLN3^CBM-L-CDK1 resulted in WT-sized cells, indicating that the fusion allows Cln3^CBM to activate Cdk1 (Fig. 2B). CLN3^CBM-L-CDK1^KD cells were as large as cln3Δ cells, suggesting that Cln3-Cdk1 kinase activity is indeed required for Cln3 function (Fig. 2B). To confirm this, we examined Cln3-Cdk1 function in a simplified Cdk network driven exclusively by CLN3-L-CDK1, CLB5-L-CDK1, and CLB2-L-CDK1 fusions (Fig. 2, C and D). In this background, the addition of CLN3-L-CDK1^KD instead of CLN3-L-CDK1 was insufficient for proliferation, confirming that Cln3-Cdk1’s kinase activity promotes the G₁-S transition (Fig. 2, C and D, and fig. S3, E to G).

Fig. 2. Cln3-Cdk1 kinase activity promotes G₁-S.

(A) Schematic of the different Cln3-Cdk1 complexes used. The CBM mutant cannot bind Cdk1 unless fused through the linker L. (B) Cell size distributions for the indicated genotypes. (C) Schematic of the experimental design for (D). The endogenous CDK1 promoter was replaced with a GALLpr. CLN3, CLB5, and CLB2 are fused to the indicated -L-CDK1. (D) Spot viability assays of the indicated genotypes on YPG (GALLpr ON) or YPD (GALLpr OFF).

Having established that Cln3-Cdk1’s kinase activity promotes the G₁-S transition, we next sought to identify its substrates through a candidate-based in vitro screen. We measured the specificity of purified Cln3-L-Cdk1 and other yeast cyclin-Cdk1 complexes toward >20 candidate targets (Fig. 3, A and B; fig. S4, A to D; and table S4). By far the most specific target for Cln3-L-Cdk1 was the RNA polymerase II subunit Rpb1 (Fig. 3C), which contains a C-terminal unstructured region (CTD) with multiple heptapeptide repeats (Y₁S₂P₃T₄S₅P₆S₇) (14). Truncations to isolate Rpb1’s unstructured C-terminal region, and then to remove the regions on either side of the CTD repeats, did not reduce phosphorylation. This implies that Cln3-Cdk1 targets residues inside the repeats that are also targeted by the canonical transcriptional kinases (except Bur1) (15) (fig. S4G).

Fig. 3. Cln3-Cdk1 phosphorylates S₅ in Rpb1’s CTD repeats.

(A) Autoradiographs of in vitro phosphorylation of candidate substrates by the indicated cyclin-L-Cdk1 complexes. (B) Cyclin-L-Cdk1 specificity is the ratio of activity toward a substrate relative to its activity toward histone H1. Each point corresponds to one substrate. (C) Autoradiographs of in vitro phosphorylation by Cln3-L-Cdk1 of Rpb1 truncations (WT denotes full length; ΔC denotes 1453 N-terminal amino acids; ΔN denotes 280 C-terminal amino acids). Note that the low purification yield of WT (~192 kD) results in reduced signal. (D to F) Autoradiographs of in vitro phosphorylation of indicated GST-CTD substrates by the indicated kinases. (G) Schematic showing S₅-specific phosphorylation by Cln3-Cdk1.

Phosphorylation of the different residues within these heptad repeats by the canonical transcriptional kinases regulates transcriptional initiation, elongation, and termination (14). We therefore sought to identify Cln3-Cdk1’s specific target residues by generating a series of model CTD substrates (Fig. 3D). Only mutation of the serine 5 residue prevented phosphorylation by Cln3-L-Cdk1. Conversely, the addition of S₅ to repeats lacking all serines restored phosphorylation (Fig. 3E). Phosphorylation was decreased by Y₁, P₃, and P₆ mutations but not by T₄ mutation (Fig. 3E and fig. S4E). The effect of the P₆ alanine substitution was expected because Cdk1 is a proline-directed kinase, whereas the P₃ requirement is similar to the −2P enhancement of Cln2-Cdk1 activity (16). Thus, Cln3-Cdk1, in contrast to other cell cycle cyclin-Cdk1 complexes, specifically phosphorylates S₅ in vitro, and this depends on the local amino acid sequence (Fig. 3, F and G).

That Cln3-Cdk1 colocalizes with SBF on the genome and functions as an S₅ CTD kinase in vitro suggests that Cln3-Cdk1 should be responsible for S₅ phosphorylation specifically at SBF-regulated promoters. To test this, we depleted Cln3 from the nucleus (8) and then performed ChIP-seq against Ceg1, a protein for which chromatin association depends on binding pS₅ (17) (Fig. 4A). This resulted in reduced Ceg1-FLAG signal specifically at SBF-regulated promoters (Fig. 4, B to D). Because the time scale required for nuclear depletion (>10 min) cannot distinguish between direct and indirect effects, we also performed a time course after chemical-genetic inhibition of CDK1^as (Fig. 4E). We observed an immediate and rapid decrease in the Ceg1-FLAG signal at SBF-regulated genes already within 1 min (Fig. 4, F to H, and fig. S7, A to D). We estimated SBF pS₅ turns over with a half-life of ~1 min (fig. S7D), comparable to that previously determined for other Cdk1 substrates (18). Although Rpb3 at SBF-regulated genes also decreases after 1NM-PP1 treatment, this occurred more slowly, suggesting that the loss of Ceg1 is unlikely to be caused by a loss of polymerase occupancy (fig. S7, E and F). Further supporting Cln3-Cdk1’s role as an S₅ kinase, cln3Δ reduced global S₅ phosphorylation in whole-cell extracts (fig. S5).

Fig. 4. Cln3-Cdk1 drives the G₁-S transition by phosphorylating S₅ on Rpb1’s CTD at SBF-regulated genes.

(A) Cln3-FRB is conditionally anchored away from the nucleus by rapamycin. Ceg1-5xFLAG ChIP-seq measures Rpb1 pS₅. (B to D) Ceg1-5xFLAG ChIP-seq signal at the CLN2 promoter (B) and average coverage at SBF-regulated genes (C) and non–G₁/S genes (D). (E to H) As in (A) to (D) except CDK1^as is inhibited by 1NM-PP1. (I and J) Cell size distributions for the indicated genotype. Ccl1-L-Kin28 fusion proteins were expressed from a CLN3 promoter and recruited to SBF through Swi6 by rapamycin-induced binding. (K) Summary model.

We reasoned that if Cln3-Cdk1 functions as an Rpb1-CTD-S₅ kinase at SBF-regulated genes, then we should be able to bypass the requirement for Cln3 by providing an alternative S₅ kinase at SBF-regulated promoters. To test this, we used a rapamycin-dependent binding system (8) (fig. S6, A and B) to conditionally recruit a fusion protein of the canonical CTD S₅ kinase, Ccl1-L-Kin28, to SBF (Fig. 4I). Recruitment of Ccl1-L-Kin28 to SBF fully rescued the size and cell cycle phenotypes of cln3Δ cells, and this rescue depended on Kin28 kinase activity (Fig. 4J and fig. S6, C to I). We therefore propose that Cln3-Cdk1 promotes the first step in the budding yeast cell cycle by directly phosphorylating S₅ on Rpb1’s CTD at SBF-regulated genes (Fig. 4K).

Our findings here add Cln3-Cdk1 to the list of transcriptional kinases that promote gene expression through CTD phosphorylation (14). Unexpectedly, Whi5 and the other SBF-associated factors that we tested were very poor substrates of Cln3-Cdk1, and our in vitro screen suggested that Cln3-Cdk1 has few, if any, targets other than the Rpb1. Such extreme cyclin specificity may help to order the cell cycle (16). If, as we suspect, Cln3 has no targets driving another cell cycle event, then the level of Cln3 can be used to exclusively modulate the G₁-S transition without prematurely triggering downstream cell cycle events.

Although the cell cycle Cdks and transcriptional Cdks are all part of the same kinase family, their functions are generally thought to have diverged (19). Our findings break this dichotomy by showing that cell cycle Cdks can directly activate transcription at specific genes. Unexpectedly, we observed that Kin28 also binds SBF promoters (fig. S2, C and D), indicating that both Kin28 and Cln3-Cdk1 cooperate to phosphorylate Rpb1, consistent with the reported genetic interactions between KIN28 and CDK1 (20). The functional overlap of cell cycle and transcriptional Cdks has also been suggested by the dual function of the Kin28 orthologs in other eukaryotes that both activate the cell cycle Cdks and function as global S₅ CTD kinases (21). In addition, Cdk1 and Cdk2 were identified as the first RNA polymerase II CTD kinases in vitro, but whether they function as such in vivo is unknown (22). Regardless, it is now apparent that the two branches of Cdks that regulate cell division and transcription have overlapping functions, which suggests the possibility that their primordial ancestor regulated both processes. Such an ancient link may be reflected here in our discovery that Cln3-Cdk1 drives cell cycle progression by directly activating transcription.

Data and materials availability:

ChIP-seq data are available at the Gene Expression Omnibus (GEO) repository under no. GSE169271. Other materials are available upon request.