Cdk1 activity acts as a quantitative platform for coordinating cell cycle progression with periodic transcription

Cell proliferation is regulated by cyclin-dependent kinases (Cdks) and requires the periodic expression of particular gene clusters in different cell cycle phases. However, the interplay between the networks that generate these transcriptional oscillations and the core cell cycle machinery remains largely unexplored. In this work, we use a synthetic regulable Cdk1 module to demonstrate that periodic expression is governed by quantitative changes in Cdk1 activity, with different clusters directly responding to specific activity levels. We further establish that cell cycle events neither participate in nor interfere with the Cdk1-driven transcriptional program, provided that cells are exposed to the appropriate Cdk1 activities. These findings contrast with current models that propose self-sustained and Cdk1-independent transcriptional oscillations. Our work therefore supports a model in which Cdk1 activity serves as a quantitative platform for coordinating cell cycle transitions with the expression of critical genes to bring about proper cell cycle progression.

T he eukaryotic cell cycle is a highly regulated process that relies on intricate mechanisms to ensure faithful duplication and segregation of the genetic material. The succession of cell cycle phases coincides with and depends on the periodic transcription of specific sets of genes, a phenomenon that is conserved among all eukaryotes tested to date. Genome-wide studies in models ranging from yeast to human cells have led to the identification of a substantial pool of periodic genes that have been clustered according to their peak time of expression and associated with M (mitosis), G1, S (DNA replication) or G2 (refs 1-4). Although the overall list of genes belonging to this transcription program differs between species, a conserved core of periodic gene networks has now been revealed, highlighting its evolutionary importance 5 .
Initial insights into the control of cell cycle oscillations in gene expression were derived from studies in the budding yeast Saccharomyces cerevisiae (S. cerevisiae) 3,5,6 . In this system, the periodic transcriptional program was proposed to rely on the fact that each gene cluster contains the transcription factors (TFs) that control the expression of the following cluster. Moreover, the observation that oscillations in gene expression still occurred in G1-arrested cells suggested that the established circuit of serially induced TFs is sufficient to generate periodicity 7,8 . Even in the absence of all Cdk1 activities, B66% of cell cycle-regulated transcripts continued to be periodic 8 . From these results, a model emerged in which the kinetics of the transcriptional oscillations play a key role in bringing about the succession of cell cycle events. In this case, the major regulators of cell cycle progression, the cyclin-dependent kinases (Cdks), would act as effectors of the transcriptional oscillator: for example, entrainment of cyclin expression to this system could sequentially trigger the G1/S and G2/M transitions through the association of various cyclins with Cdks 8 . However, it remains unknown whether this architecture represents a conserved mode of regulation of periodic transcription among eukaryotes.
In the fission yeast Schizosaccharomyces pombe (S. pombe), cell cycle-associated periodic transcription involves B500 genes [9][10][11][12] and already appears to differ from what is observed in S. cerevisiae. The MCB-binding factor (MBF) complex regulates M/G1 and early G1 genes that encode for proteins involved in both the G1/S transition and the duplication of the genome during S phase. Later in G1, a set of periodic genes enriched in factors that are linked to cell separation after mitosis is regulated by the Ace2 transcription factor 9,13 , while expression of the histone genes in S phase depends on the GATA transcription factor Ams2 (ref. 14). Apart from those controlled by MBF, a number of genes induced at mitosis are regulated by the PCBbinding factor (PBF) complex 3 . Despite this apparently detailed knowledge of the regulation of periodic expression, little is known about what controls the transcription of the largest fractions of the G1, S and G2 clusters. In addition, the budding yeast TF network seems to be only partially present in S. pombe. For instance, although PBF regulates ace2 expression, which in turn activates the G1 genes, MBF targets are controlled independently of the PBF-dependent genes 3 . Furthermore, ams2 expression relies on MBF, but there is no evidence for a direct link between MBF-dependent transcription and the expression of histone genes. This suggests that the coupling of cell cycle progression with periodic transcription may be governed by a different logic in this system. Understanding these mechanisms may therefore lead to novel models for the coordination of the processes linked to cell proliferation in eukaryotes.
A central player in cell cycle progression is the conserved Cdk1 protein, the predominant member of the Cdk family. In the budding and fission yeasts, Cdk1 controls both the G1/S and G2/M transitions 15 . As mentioned previously, transcriptional oscillations still occur in S. cerevisiae cells that are arrested in their cycle, indicating that this periodicity is phase-independent and that Cdk activity acts in parallel or downstream of the transcriptional program. However, the ability to re-program the fission yeast cell cycle network and alter the sequence of cell cycle events simply by artificially changing Cdk1 activity 16 raises the possibility of a tight link between cell cycle phases and expression of critical periodic genes.
To address this, we take advantage of a recently described minimal cell cycle network in fission yeast in which Cdc2/Cdk1 is fused to the B-type cyclin Cdc13 (ref. 16). The level of this module oscillates through cycles of synthesis/degradation rather than strong cell cycle-regulated expression 9 . Importantly, its kinase activity can be finely regulated to alter the progression of the cell cycle at all phases of the process. This previously allowed us to demonstrate that cell cycle transitions are primarily driven by quantitative changes in the levels of a single qualitative Cdk activity (that is, the function of Cdk1 in association with a specific cyclin), a model that appears to be conserved in more complex eukaryotes [17][18][19] . Here we use this system to dissect the interplay between Cdk1 activity and periodic transcription. We first demonstrate that periodic gene expression in fission yeast does not show cell cycle-independent oscillations, regardless of the phase in which cells are arrested. We then uncover that the coupling between periodic transcription and cell cycle stages relies on a quantitative response to Cdk1 activity levels. We further show that cell cycle events neither participate in nor interfere with this transcriptional program, provided that cells are subjected to the appropriate Cdk1 activity levels. Our results challenge the widely accepted model of self-sustained, Cdk1-independent transcriptional oscillations that emerged from studies in budding yeast. We therefore propose that the regulation of transcription throughout the cell cycle is tightly linked to dynamic changes in the activity levels of the cyclindependent kinases. This may represent a conserved architecture in which Cdks act as a quantitative platform for coordinating critical cell cycle events with periodic transcription, providing new insight into the systems that control the proliferation of eukaryotic cells.

Results
Loss of periodic transcription in G1-arrested cells. To understand the interaction between cell cycle progression and periodic expression in the context of the fission yeast TF network, we assessed the transcriptional program upon inhibition of Cdk1 activity in G1, which prevents DNA replication. Indeed, cell cycle-independent transcriptional oscillations were initially revealed in budding yeast cells arrested in G1 by inhibition of mitotic cyclin function or using a temperature-sensitive allele of Cdk1/Cdc28 (refs 7,8). To control cell cycle progression in our model, we used minimal cells in which the activity of an engineered cyclin B-Cdk1 fusion can be altered by addition of the small molecule inhibitor 3-MBPP1 (ref. 16). Synchronous cultures were allowed to progress through mitosis and subsequently treated with high concentrations of 3-MBPP1 to arrest cell cycle progression in G1 (ref. 16) (Fig. 1a-c). Changes in transcript levels of genes representative of the different periodic clusters were then measured and compared with those in cycling cells.
We first determined the behaviour of the mitotic cluster by analysing the PBF target gene ace2. As expected, its expression decreased as cycling cells exited mitosis and entered early G1 (Fig. 1d). Similar dynamics were observed in inhibitor-treated cells, consistent with the fact that Cdk1 activity is already strongly reduced at the time of inhibitor treatment due to APC-dependent degradation of the cyclin, an essential step in mitotic exit. Importantly, while an additional peak in ace2 expression was observed at the next mitosis in cycling cells (Fig. 1b,d), its levels remained low in G1-arrested cells throughout the time course. The MBF-target gene cdc22 (G1 cluster) peaked later but subsequently behaved similarly to ace2, with no second induction in inhibitor-treated cells (Fig. 1d). In contrast, their budding yeast orthologues, ACE2 and RNR1, are still periodically expressed in the absence of all Cdk1 activities 8 . Next, to assess the behaviour of the S-phase cluster, we focused on histone genes, which are among those with the highest induction amplitude 4,12 and are commonly used as markers for this cluster. hht1 was periodically expressed, peaking in cycling cells at the time of DNA replication (Fig. 1c,e). Intriguingly, hht1 expression was also induced in G1-arrested cells with similar kinetics and levels as in the control, but the transcript levels remained high for most of the time course before slightly decreasing towards the end of the experiment. This suggests that once cells have passed mitosis, commitment to DNA replication itself is not critical for induction of this cluster. However, its periodic behaviour is contingent on cell cycle progression through S and entry into G2. Finally, the G2 gene spd1 peaked during G2 in the cycling culture as anticipated (Fig. 1b,f) and remained low throughout the experiment in G1-arrested cells. Taken together, these results suggest that cell cycle-independent periodic oscillations in transcription are absent in G1-arrested fission yeast cells.
Loss of periodic transcription in G2-arrested cells. Autonomous transcriptional oscillations may require a specific trigger within the TF network that could be associated with discrete phases of the cell cycle. Interestingly, the organization of the budding yeast cell cycle is G1-driven, while fission yeast proliferation relies on strong regulation at the G2/M transition. This prompted us to assess if oscillations in transcription occur in G2-arrested fission Cultures were washed to allow cells to resume their cycle, progress through the metaphase to anaphase transition (15 min, Release), and finally treated with 10 mM 3-MBPP1 to induce a G1 arrest (DMSO was used as a control) 16 . Samples were then taken every 10 min for 120 min and gene expression changes were assessed. (b) Percentage of binucleated cells in a. Both control and inhibitor-treated cells initially undergo mitosis. However, cells subsequently treated with inhibitor at T0 remain blocked in G1 (see c) and do not progress to the next mitosis. n4100 at each time point. (c) DNA content analysis of cells in a. Although control cells progress through S phase (black profile) and cytokinesis, inhibitor-treated cells accumulate in G1 prior to S phase with a 1C DNA content (see Methods section for the interpretation of flow cytometry profiles in fission yeast). (d-f). Changes in gene expression for the indicated genes in a. C and A refer to cycling (DMSO-treated) and arrested (inhibitor-treated) cells, respectively. Fold changes are normalized to actin RNA levels and represented relative to the values at T0 (set to 1). See Supplementary Fig. 1 for additional representative genes. yeast cells. Synchronous cultures were allowed to proceed through mitosis, G1 and S, eventually entering the next G2 (Fig. 2a, T0). Half of the culture was then arrested in G2 using 1 mM 3-MBPP1, whereas the other half kept cycling (Fig. 2b,c). This allowed for a direct comparison of the transcription of periodic genes between G2-arrested and proliferating cells.
As seen in G1-arrested cells, periodic induction of ace2 and cdc22 in M and G1, respectively, was strongly dependent on cell cycle progression (Fig. 2d), as no activation was detected in inhibitor-treated cells. In contrast to the G1 experiment, hht1 (S) remained at low levels in arrested cells, consistent with a repression of its expression in G2 (Fig. 1e). We finally found that the G2 gene spd1 was not induced, despite cells being arrested in G2 (Fig. 1e). This suggests that the timing of inhibitor treatment occurred after spd1 levels had declined (compare Figs 1 and 2). Preventing another cycle through M, G1 and S into the next G2 therefore precluded another round of spd1 induction.
To address whether these results on representative genes reflect the global behaviour of the corresponding clusters, we performed genome-wide analyses in the same experimental setup, assessing periodic transcription of the M (20 min) and S (60 min) clusters (Fig. 2g). Genes within each cluster were grouped in three categories according to the amplitude of their induction 4,12 (Fig. 2, legend). Our results showed that 53% of the mitotic  Table 1 for the percentages of genes above the cutoffs for the ratios in expression between the cultures.
genes of the high expression group behaved similarly to ace2 when comparing cycling and arrested cells at 20 min (20_C/20_A in Fig. 2g and Supplementary Table 1, cutoff for the ratio cycling/ arrested cells is 1.5). However, other high-amplitude genes, such as ecm33, did not make the cutoff, suggesting that they might also be activated in the arrested cells. Nevertheless, when testing ecm33 by qPCR, we found no induction in inhibitor-treated cells (Fig. 2f). This prompted us to reduce the cutoff to 1.15, a value just below that measured for ecm33 (1.19). With this threshold, 82% of the high amplitude mitotic genes showed no periodic transcription upon cell cycle arrest (Supplementary Table 1). As expected by the timing of the cell cycle phases in these experiments and our qPCR data, expression of G1, S and G2 genes did not change in the 20_C/20_A analysis ( Fig. 2g; Supplementary Table 1). This again contrasts with the 66% of cell cycle-regulated transcripts that continue to oscillate in the absence of all Cdk1 activities in budding yeast 8 . At the 60 min time point, we observed that 85% of the entire set of S-phase genes showed significantly lower expression in the arrested cells (60_C/60_A, Fig. 2g and Supplementary Table 1), consistent with the hht1 qPCR results. We also noted that a significant percentage of M and G1 genes still made the cutoff (Supplementary Table 1), in contrast with the decrease in ace2 and cdc22 expression detected by qPCR. However, this likely results from a reduction in their basal transcript levels in the arrested cells during the experiment (see 20_A/60_A in Fig. 2g, Supplementary Table 1), a phenomenon that was less pronounced for the S and G2 clusters. Finally, a lower percentage of the G2 genes passed the threshold, even after 60 min, due to their later peak time (see spd1 in Fig. 2e).
Together with the data obtained with G1-blocked cultures, this suggests that the transcriptional program in fission yeast fundamentally differs from that in budding yeast, as periodic genes do not freely oscillate in cells arrested in the cycle. This could result from either the induced pause in cell cycle progression or a more direct link between changes in Cdk1 activity and transcriptional oscillations.
Periodic transcription is linked to cell cycle progression. Our results indicate that periodic transcription during the fission yeast cell cycle is linked to its different transitions. We next tested whether (1) the periodicity in expression of these genes only depends on the onsets of specific cell cycle events, with their subsequent decrease in transcription relying on a timer (triggertimer process), or (2) the induction and downregulation of a gene cluster are respectively coupled to entering and exiting associated cell cycle phases, as suggested by hht1 in G1-arrested cells (Fig. 1e). To address this, we assessed the transcription of representative genes when cells are temporarily blocked in mitosis, using a cold-sensitive mutant of the tubulin-encoding gene nda3. Switching the nda3-km311 mutant cells to 18°C activates the spindle assembly checkpoint, with cells arresting in early mitosis with high Cdk1 activity 20 . The culture was then shifted back to the permissive temperature, allowing for the resumption of cell cycle progression (Fig. 3a). Interestingly, we observed an accumulation of ace2 transcripts (M cluster) during the arrest that only returned to basal levels once cells were allowed to proceed through mitosis (Fig. 3b). A similar behaviour was observed for the MBF targets cdc18 and cdc22 (Fig. 3c,d), suggesting that their initial induction and subsequent downregulation are also linked to mitotic onset and exit, respectively. The S and G2-associated genes hht1 and spd1 were repressed upon mitotic entry and remained at low levels during the block, as anticipated (Fig. 3e,f); however, the expression of hht1 increased again 30 min after release, concomitant with cells progressing through G1 and S.
We therefore conclude that the periodicity in transcription of different clusters is dependent on sequential progression through cell cycle phases rather than operating on a trigger-timer model.

Periodic transcription is imposed by Cdk1 activity changes.
The link between cell cycle progression and periodic transcription may be brought about by different mechanisms. First, induction and downregulation of periodic genes could be strictly coupled to cellular events that are specific to each phase. Alternatively, oscillatory behaviour may directly derive from cell cycleassociated changes in Cdk1 activity levels rather than cell cycle progression per se. Indeed, Cdk1 activity is high at mitosis, strongly reduced upon mitotic exit, rises during G1 to pass the S-phase threshold and keeps increasing throughout G2 until the Wee1/Cdc25 feedback loop becomes activated 15 , at which point it reaches the mitotic threshold. These dynamic modulations in Cdk1 activity may provide an ideal quantitative framework for driving periodic transcription, regardless of cell cycle status.
To discriminate between these two possibilities, we uncoupled cell cycle progression from changes in Cdk1 activity. To this end, we used inhibitor-sensitive minimal cells carrying the nda3-km311 mutation, which allowed us to block cells in mitosis with high Cdk1 activity via temperature shift 20 and then repress Cdk1 using the 3-MBPP1 inhibitor while they remained arrested in M (Fig. 4a). In these conditions, cells experience a low Cdk1 activity without actually progressing through mitosis ( Supplementary Fig. 2a,b).
Consistent with our previous results, the PBF targets ace2 and slp1 (M phase) as well as the MBF targets cdc18, ams2 and cdc22 (M/G1 and G1) were upregulated in mitotically blocked cells, although with different dynamics. Strikingly, downregulation of Cdk1 activity even with low inhibitor concentration (1 mM) was sufficient to strongly repress the expression of these genes to basal levels while cells remained in mitosis (Fig. 4b,c). Similar results were observed with other high amplitude mitotic genes ( Supplementary Fig. 2c) and the Ace2 target eng1 (Fig. 4c). These data demonstrate that the periodicity of M and G1 genes is solely driven by changes in Cdk1 activity.
The S phase histone genes hht1 and hhf1 remained low in the control, arrested cells ( Fig. 4d and Supplementary Fig. 2d), indicating that G2/M levels of Cdk1 activity are sufficient to prevent the expression of this cluster. Interestingly, addition of 1 mM inhibitor had no effect on their expression, while 10 mM 3-MBPP1 resulted in rapid induction. Again, this was observed without resumption of mitosis. These data are in line with previous results showing that strong inhibition of Cdk1 activity in G2 through treatment with 10 mM inhibitor is necessary to allow entry into S without an intervening mitosis, while usage of 1 mM only maintains a G2 arrest 16 . Importantly, this difference in response between the two different inhibitor concentrations highlights the quantitative link between expression of periodic genes and Cdk1 activity levels.
For the G2 cluster, we did not observe striking changes in transcript levels of spd1 (Fig. 4e). This is consistent with its low expression during mitosis and the inhibition of its induction when G2 cells are treated with 3-MBPP1 (Figs 1 and 2).
We then tested the extent of this apparent lack of interplay between cell cycle events per se and periodic transcription by artificially altering the sequence of cell cycle phases, assessing the effect of bypassing mitosis on gene transcription. To this end, G2-blocked cells were treated with high concentrations of 3-MBPP1 (10 mM) and then released into lower concentrations (1 mM), allowing for G1 reset and re-replication of the genome without an intervening mitosis (Fig. 5a,b) 16 . Interestingly, we observed that compared with control cells maintained in 10 mM inhibitor, cells released in 1 mM 3-MBPP1 showed periodic induction of cdc18 and cdc22 (Fig. 5c), although with smaller fold changes than in our previous experiments (see Figs 1 and 2 and Discussion). Only a minor effect was detected for ace2. Consistent with the results in Fig. 4, elevated transcription of hht1 on the initial strong inhibition of Cdk1 activity was repressed by the switch to 1 mM inhibitor (Fig. 5c). These data reinforce our conclusions that altering Cdk1 activity, even while bypassing cell cycle phases, is sufficient to regulate periodic transcription.
Our results strongly point towards a new model in which autonomous transcriptional oscillations do not entrain cell cycle progression. Instead, periodic transcription occurs as a result of the operation of the cell cycle machinery through dynamic changes in Cdk1 activity.
Cell cycle transitions do not affect Cdk1-driven periodicity. Our data demonstrate that periodic transcription is closely linked to changes in Cdk1 activity, to the extent that modulating Cdk1 function in cell cycle-arrested cells is sufficient to reprogram transcription. However, these experiments were performed in arrested cells and therefore do not address the potential interplay between the expression of a set of genes during a given cell cycle phase and the onset of the next phase. For instance, together with the APC-dependent reduction in Cdk1 activity at mitotic exit, entry into S phase may further downregulate mitotic and G1 genes. This may represent an additional mechanism for ensuring the proper sequence of cell cycle events.
To address this, we analysed periodic transcription in cells induced to simultaneously undergo M and S 16 . This is achieved through blocking cells in G1 with high concentrations of inhibitor, allowing for the accumulation of inactivated fusion protein, followed by removal of the inhibitor, which results in a rapid increase in Cdk1 activity. The abnormally high activity generated in this context is sufficient to trigger both S (low-activity threshold) and M (high-activity threshold) at the same time, giving rise to aberrant nuclei (Fig. 6a-c). We found that both the PBF target ace2 and the MBF target cdc22 were highly induced upon mitotic entry, in contrast to cells in which low Cdk1 activity was maintained (Fig. 6d). Furthermore, both the periodicity (30-40 min) and the difference in the timing of peak expression between these two genes (10 min) were maintained as in cycling cells. This is consistent with our model in which quantitative changes in Cdk1 activity, which still occur in these experiments through the function of the APC at mitotic exit, control periodic transcription. Importantly, all of these events were concomitant with DNA replication (Fig. 6c), suggesting that S-phase onset does not play a role in the observed downregulation of the M and G1 clusters during a normal cell cycle. We also noted that the capacity of these cells to undergo S phase appeared to be independent of prior activation of the MBFdependent program, which occurred during and peaked after bulk DNA replication. It therefore likely relied on either the basal expression of MBF targets during the extended G1 block in these experiments or the perdurance of essential factors. Finally, a significant increase in the expression of the S-phase histone gene hht1 was observed during the initial G1 block (Fig. 6e), as expected from its dependency on low Cdk1 activity (Fig. 4d). This was followed by a reduction in hht1 transcript levels upon   reaching high Cdk1 activity and eventually another increase, likely due to the degradation of the synthetic Cdk1 module by the APC at mitotic exit. Similar results were obtained when simultaneous S and M were induced from G2 cells that were first reset into G1 without an intervening mitosis 16 ( Supplementary  Fig. 3a-e). These data suggest that even when cell cycle progression is rewired to the extent that M and S overlap, a normal sequence of periodic gene expression is triggered upon experiencing mitotic Cdk1 activity. Undergoing DNA replication therefore has no direct impact on the Cdk1-driven oscillations in transcription.
To extend these conclusions, we performed RNA sequencing in the G1 arrest/release setup (Fig. 6a), with samples taken at 0, 20 and 40 min (Fig. 6f). Strikingly, as observed for ace2, a significant part of the mitotic cluster was strongly induced at 20 min and downregulated by 40 min (Fig. 6f and Supplementary Data 2). The latter is likely due to activation of the APC negative feedback loop by high Cdk1 activity. Moreover, a majority of G1 genes, including MBF and Ace2 targets, were induced at both time points. However, as anticipated from the cdc22 qPCR data (Fig. 6d), the expression ratios were on average lower at 40 min ( Supplementary Fig. 3f). When analysing the behaviour of S-phase genes, we noted that 44% made a standard 1.5 cutoff at 20 min and 69% at 40 min. A significant part of this cluster is therefore also induced in these experiments. Interestingly, changes in hht1 in the qPCR data and more generally of the histone genes in the RNA-seq data were limited and often below the cutoff. This suggests that distinct subsets of the S cluster may be subject to differential regulation. Finally, a significant fraction of the G2 genes were repressed in the T0 to 20 min comparison Interestingly, this indicates that rapid exposure to high Cdk1 activity can repress the G2 cluster during mitosis, while cells are expressing M genes. The repression of the G2 cluster was somewhat attenuated by 40 min (Fig. 6f), probably due to a reduction in Cdk1 activity at mitotic exit. This reflects the behaviour of spd1 and suggests that expression of G2 genes requires entry into G2 (Fig. 4e) but is repressed by high Cdk1 activity at mitotic onset. We conclude that the absence of strong interference between the onset of cell cycle transitions and transcriptional oscillations is an underlying feature of the periodic transcriptional program in fission yeast.

Discussion
Progression through the eukaryotic cell cycle requires the coordination of a host of pathways that operate in an orderly manner to robustly sustain the sequence of cell cycle events. In particular, temporal regulation of the expression of cell cycle components, including phase-specific Cdk1 substrates, plays a central role in this process. Although the periodic expression of distinct gene clusters at each cell cycle stage has been established, little is known about the networks that link cell cycle progression with transcriptional oscillations. Interestingly, the possibility to re-wire the fission yeast cell cycle, bypassing phases by externally altering Cdc2/Cdk1 activity 16 , suggests that critical Cdk1 substrates are either constantly present, allowing for rapid response to Cdk1 activity changes, or expressed in coordination with the different Cdk1-dependent cell cycle transitions. In this work, we propose a novel model for the orchestration of cell cycle processes in which both the transitions through the major phases and the periodic expression of cell cycle genes are directly dependent on the same profile of quantitative changes in Cdk1 activity.
In contrast to budding yeast, where periodic gene expression still occurs in arrested cells [6][7][8] , our data demonstrate that the transcriptional network that operates in fission yeast is associated with cell cycle progression. It has been suggested that the Rad53dependent S-phase checkpoint can alter periodic transcription in budding yeast 21 . However, in the minimal system, it was previously shown that the DNA replication checkpoint is not activated in G1 or G2-arrested cells, excluding a role for this pathway in the observed inhibition of transcriptional oscillations 16 . Importantly, our study reveals that periodic expression is in fact not linked to the completion of cell cycle phases per se: quantitative changes in a single qualitative Cdk1 activity are sufficient to control the transcriptional oscillations even when cells are prevented from undergoing cell cycle transitions.
We observe that the mitotic genes, including PBF targets, are activated only when cells are exposed to high Cdk1 activity. Interestingly, this was also true for the expression of MBF targets (for example, cdc18 and cdc22). This suggests that full activation of the periodic transcription program associated with DNA replication occurs prior to G1 and S in normal cells, when they commit to mitosis with elevated levels of Cdk1 activity. Mitosis per se is not essential for this, as induction of MBF targets is still observed when this event is bypassed, although to a lesser extent. This attenuated response likely results from an intermediate Cdk1 activity (between the S and M thresholds) in cells released in lower inhibitor concentrations (Fig. 5). In contrast, induction of the S phase histone genes is dependent on low Cdk1 activity. Again, their expression can be induced even in mitotic cells by simply inhibiting Cdk1. This may optimize the coupling of periodic transcription with genome duplication, as Cdk activity must first be strongly reduced for replication origin licensing and then increased to pass the S phase threshold. Finally, we also uncovered that cell cycle transitions do not interfere with the transcriptional program imposed by Cdk1 activity. These findings therefore support a model in which periodic oscillations in gene expression can be generated from any point in the cell cycle, provided that cells are exposed to the proper levels of Cdk1 activity.
Our results also imply that key aspects of periodic transcription in fission yeast are set by the drastic and unique changes in Cdk1 activity that occur between mitotic onset (high activity, induction of PBF and MBF targets) and mitotic exit (low activity resulting from cyclin B degradation, induction of S genes). The delay between PBF and MBF target expression could be a consequence of a difference in sensitivity to Cdk1 activity. PBF may be more sensitive to phosphorylation by Cdk1 and therefore activated earlier than MBF, with a direct impact on the timing of expression of their respective targets. Alternatively, this may rely on a timer activated at high Cdk1, although the potential mechanisms involved remain unknown.    Table 2 for the percentages of genes above the ratio cutoff.
Our observation that mitosis serves as the central node for establishing periodic transcription also raises the question of feedback regulation from the expression of specific clusters. Indeed, mutations in transcription factors controlling M genes [22][23][24] or alteration of the RNA polymerase II Mediator complex 25 have an impact on mitotic entry 26,27 . This may provide additional layers of regulation to ensure the proper sequence of cell cycle events, together with the quantitative dependence of the G1/S and G2/M transitions on Cdk1 activity levels. Interestingly, the Mediator complex is specifically recruited to both PBF and MBF target genes in early mitosis 27 , and it may thus play a role in linking periodic transcription with the control of Cdk1 activity.
Our findings provide a new understanding of the regulation of cell cycle progression and periodic transcription. Consistent with previous work in fission yeast 16,28 and the accumulating evidence in other systems for an important quantitative component in the control of cell proliferation [17][18][19] , the conclusions presented here strongly argue that qualitative differences in Cdk activity (that is, activities mediated by distinct cyclins) are not essential for core cell cycle control. Not only can cell cycle transitions be triggered by changes in the levels of a single Cdk activity, but we now show that the underlying organization of the expression landscape that is essential for cell proliferation is also dependent on such changes. This suggests that the events associated with cell proliferation in eukaryotes may function at two levels. Basic cell cycle progression may rely on simple, quantitative inputs from Cdks, without a significant contribution from cyclin specificity. On the contrary, responding to challenging environments may require Cdk1 to associate with different cyclins. Thus, the complexity observed in wild-type cells may have evolved as a response to changing conditions, when a purely quantitative operation of cell cycle control and periodic transcription may not have been sufficient to sustain efficient cell proliferation. Taken together, our results give rise to a novel model for how eukaryotic cells may orchestrate gene expression with the control of the major cell cycle regulators during growth and division, which may provide a new perspective for future investigation of these processes in higher eukaryotes.
Assessment of cell cycle progression. For determining the percentage of binucleated cells in the population (mitotic index), cells were heat-fixed for 2 min at 70°C on glass slides and stained with a 1:1 DAPI (1 mg ml À 1 ): Blankophor (MP Biochemicals) solution. This stained the DNA as well as the cell wall and division septum, allowing the scoring of both binucleated cells and aberrant nuclei ('cut' cells). Imaging was performed using a Zeiss Axio Observer (Carl Zeiss Inc.) equipped with a Lumencor Spectra X illumination system and a Hamamatsu Orca Flash 4.0V2 sCMOS camera.
DNA content analyses were performed by flow cytometry. Cells were fixed in 70% ethanol, washed in 50 mM sodium citrate, treated with RNase A (0.1 mg ml À 1 ) and stained with propidium iodide (2 mg ml À 1 ). DNA content was determined using a BD Accuri C6 flow cytometer. The fission yeast cell cycle has a short G1, with S phase occurring prior to cytokinesis. This has important consequences on the interpretation of flow cytometry profiles. As a result of the short G1, cells have a 2C DNA content for most of their cell cycle. In synchronized populations, a transient 4C peak is observed when S phase occurs (cells are then binucleated), which is then resolved when cells undergo cytokinesis. This 4C peak is almost undetectable in asynchronous wild-type cells as it only represents a small percentage of the entire population. A larger 4C fraction can result from cytokinesis defects or re-replication without an intervening mitosis. Conversely, appearance of a 1C peak reflects an elongation of G1 or a delay in S phase, leading to cytokinesis occurring prior to DNA replication. Finally, cell size can have a significant effect on the cytometry profiles, as non-nuclear staining by propidium iodide increases with size. The profiles become slightly shifted to the right in larger cells and to the left in smaller cells, despite identical nuclear DNA contents.