Nearly 20% of the budding yeast genome is transcribed periodically during the cell division cycle. The precise temporal execution of this large transcriptional program is controlled by a large interacting network of transcriptional regulators, kinases, and ubiquitin ligases. Historically, this network has been viewed as a collection of four coregulated gene clusters that are associated with each phase of the cell cycle. Although the broad outlines of these gene clusters were described nearly 20 years ago, new technologies have enabled major advances in our understanding of the genes comprising those clusters, their regulation, and the complex regulatory interplay between clusters. More recently, advances are being made in understanding the roles of chromatin in the control of the transcriptional program. We are also beginning to discover important regulatory interactions between the cell-cycle transcriptional program and other cell-cycle regulatory mechanisms such as checkpoints and metabolic networks. Here we review recent advances and contemporary models of the transcriptional network and consider these models in the context of eukaryotic cell-cycle controls.
THE cell cycle is the sum of the processes by which cells replicate. That process, involving the replication and redistribution of all cellular components, requires the synthesis of many proteins, including constituents of cellular structures and organelles, enzymes that catalyze the many anabolic and catabolic processes required for their replication and distribution, as well as the regulatory proteins that govern those events. Although the genes encoding many of those proteins are expressed constitutively, a large number are expressed at or near the interval when they are needed. Consequently, progression through the cell cycle is accompanied by dramatic reorganization of gene expression that we refer to collectively as “cell-cycle-regulated transcription.” The nature and mechanism of that program of gene expression is the subject of this review.
The first cell-cycle-regulated genes were observed in yeast more than three decades ago (Hereford et al. 1981) when the mRNA encoding histone genes was observed to accumulate periodically during the course of the cell cycle in a synchronized population of cells. Since then, the list of cell-cycle-regulated genes has grown, slowly at first, one gene at a time, and then very rapidly, largely as a consequence of genome-wide approaches, to encompass between as much as 20% of the genome (Cho et al. 1998; Spellman et al. 1998; Pramila 2002; Orlando et al. 2008; Guo et al. 2013). Despite the relatively large number of individual genes that are periodically expressed, it has become clear that they fall into a relatively small number of gene families that are coregulated. Consequently, the entire program appears to be controlled by a relatively small set of specific transcriptional regulatory factors.
This general topic has been extensively reviewed (Bähler 2005; Wittenberg and Reed 2005; McInerny 2011) and in-depth reviews covering specific transcription factor families and cell-cycle-regulated gene clusters have been presented (Murakami et al. 2010; Cross et al. 2011; Eriksson et al. 2012). We will introduce the constituents and regulatory logic of the cell-cycle transcriptional circuitry with discussion weighted toward more recent contributions.
A general understanding of both the pattern of gene expression and the regulation of the cell-cycle transcriptional program is, in many cases, emerging. When viewed in its entirety, the program appears as a continuum of transcriptional activation and deactivation. However, we now appreciate that waves of gene expression are coupled to observable cell-cycle events, which, in most cases, depend on the activity of the cyclin-dependent protein kinase Cdk1 (Cdc28, see below). The transcriptional program guides the activity of Cdk1 by initiating the properly timed expression of specific cyclin genes. In turn, cyclin/Cdk1 complexes phosphorylate transcription factors and regulate their activity. Thus, there is a complex dynamic interplay between the transcriptional program, CDK activity, and cell-cycle progression (Figure 1). Waves of gene expression are associated with (i) cell-cycle initiation late in G1 phase prior to the initiation of S phase (G1/S transcription), (ii) S phase (S phase transcription), (iii) the transition from G2 phase into M phase (G2/M transcription), and (iv) the transition of cells from M phase back into G1 phase (M/G1 transcription). Genes within a coregulated cluster are not all activated at the same time but appear to be turned on in a precise order during an interval that can span 20% of the cell cycle (Eser et al. 2011; Guo et al. 2013). The consequence of this highly regulated pattern of transcription is the sequential periodic expression of upwards of 1000 genes.
The existence of this intricately orchestrated sequence of transcriptional activity raises a number of important questions that form the basis for contemporary studies. Which genes constitute the full cell-cycle-regulated set? How is the transcriptional activation within a coregulated cluster ordered, and how is the temporal order of clusters maintained? How is the cell-cycle-regulated transcriptional program influenced by regulatory pathways, including checkpoints and the yeast metabolic cycle (YMC)? Recent findings that suggest the transcriptional program can be uncoupled from cell-cycle progression leave in question the mechanisms that maintain synchrony between the transcriptional program and cell-cycle events. Finally, why is the temporal order of this transcriptional program important for execution of the yeast cell cycle and to what degree is this program conserved among eukaryotes? In the sections that follow, we will review the literature regarding these issues with a focus on the emerging models.
Evolution of Experimental Approaches and Models
Models of biological phenomena are necessarily constrained by the approaches we use to observe them. Views of cell-cycle-regulated transcription have certainly evolved as new technologies and approaches have been applied.
The single-gene approach
Most early studies identified cell-cycle-regulated genes one at a time. Typically, populations of cells were synchronized by arresting with mating pheromone, released from the arrest, and sampled over time. RNAs were isolated from the time series and subjected to Northern blot analysis using radioactively labeled probes from the gene of interest. Taken together, these early studies revealed that many of the genes identified by genetic methods as important cell-cycle regulators were themselves transcriptionally controlled during the cell cycle. Cell-cycle regulators, like the cyclins, were generally found to be transcribed in the phase in which they functioned (Wittenberg et al. 1990; Surana et al. 1991; Richardson et al. 1992; Kuhne and Linder 1993; Schwob and Nasmyth 1993). These findings pointed to a model in which the cell cycle is driven, at least in part, by successive waves of expression of regulatory proteins (Amon et al. 1993).
In 1998, the first global view of the cell-cycle-regulated transcription program was revealed when several groups examined synchronized populations of cells in time-series experiments using microarrays (Cho et al. 1998; Spellman et al. 1998). To date, five studies (using different types of microarrays and different means of synchronizing populations) have reported global transcriptional expression profiles in wild-type yeast cells as they progress through the cell cycle (Cho et al. 1998; Spellman et al. 1998; Pramila et al. 2006; Orlando et al. 2008; Granovskaia et al. 2010). These studies, and comparable studies in other systems (Rustici et al. 2004; Oliva et al. 2005; Peng et al. 2005), delineated the scope of the problem, identifying several hundred genes whose transcript levels oscillate during the cell cycle. The precise number of periodic genes is still a somewhat contentious issue, ranging from 600 to 1500 or between 10 and 20% of all yeast genes (discussed further below). This fraction is higher than expected, given the number of known cell-cycle regulators identified by genetics, and suggests that precise regulation of a large fraction of the genome might contribute to proper progression through the cell cycle.
The emerging picture from these studies also seemed to suggest genes are not transcribed in a small number of discrete intervals, but are instead turned on and off in a continuum as cells passed through the cycle. Nonetheless, the program could be divided into large blocks of genes (clusters) found to be transcriptionally active in specific cell-cycle phases. The lists of periodic genes enabled searches to identify promoter motifs governing cell-cycle-regulated transcription. This effort benefited greatly from the genome sequences of related Saccharomyces species, which allowed for comparisons that highlighted conserved promoter elements (Cherry et al. 2004; Teixeira et al. 2006). Bioinformatic analyses determined that genes within clusters shared common promoter elements and were likely coregulated by specific subsets of transcription factors (Spellman et al. 1998; Pramila et al. 2006; Orlando et al. 2008).
At the turn of the century, high-throughput chromatin immunoprecipitation-microarray chip (ChIP-chip) technologies enabled researchers to collectively identify the binding sites of all known transcription factors across the budding yeast genome (Ren et al. 2000; Simon et al. 2001; Lee et al. 2002; Harbison et al. 2004). These ChIP-chip experiments, like other early generation high-throughput technologies, likely produced false positives and false negatives at relatively high rates. Thus, functional experimental evidence is needed to confidently assign regulatory roles to specific transcription factor/gene combinations. Nonetheless, these studies provided physical evidence that clusters of genes identified by microarray experiments were indeed coregulated by specific transcriptional regulators expressed in discrete cell-cycle phases. It also became clear that the genes encoding transcriptional regulators were themselves located in these clusters (Simon et al. 2001). When the results of genome-wide transcription factor localization were combined with transcript dynamics from microarray experiments, it became apparent that the cell-cycle transcriptional program could be regulated by a relatively small network of serially activated transcription factors (Simon et al. 2001; Lee et al. 2002; Pramila et al. 2006). Furthermore, this limited network of factors can function independently or combinatorially to enhance the complexity of the pattern of transcriptional regulation (Kato et al. 2004). The idea that a transcription factor network could control the cell-cycle transcriptional program was a paradigm shift in the field (discussed below), and this shift was made possible only through the development and application of genomic technologies.
A qualitatively similar view of the cell-cycle transcriptional program emerged in Schizosaccharomyces pombe when microarray approaches were applied to synchronous populations of fission yeast (Rustici et al. 2004; Oliva et al. 2005; Peng et al. 2005). Cell-cycle-regulated transcripts in S. pombe are expressed in waves that can be subdivided into clusters of temporally coregulated genes. Many recognizable orthologs in Saccharomyces cerevisiae and in S. pombe exhibit similar transcript periodicity, suggesting that transcriptional regulatory mechanisms are conserved (Oliva et al. 2005; Orlando et al. 2007). Despite the fact that many of the transcription factors that mediate cell-cycle-regulated transcription and some specific genes regulated by those factors are conserved, there is considerable divergence among cell-cycle-regulated genes and programs, even among closely related yeasts (Eser et al. 2011; Guan et al. 2013). Furthermore, computational models indicate some altered behaviors that likely reflect unique aspects of cell-cycle progression in these highly diverged yeasts (Orlando et al. 2007). Comparison with other yeasts and more highly diverged metazoan systems has led to the concept of conservation of the general topology, but not the specifics, of the cell-cycle-regulated transcriptional circuitry (discussed below).
Caveats associated with population-based approaches
Taken together, the single-gene and genomic approaches have been phenomenally successful in providing a global view of how transcript levels are regulated during cell-cycle progression. However, it seems likely that some of the more subtle aspects of transcriptional regulation remain hidden, due to the nature of the experiments. The approaches described above rely on synchronous populations of cells to produce enough RNA to be interrogated by Northern blot, microarray, qPCR, or RNA-Seq. Because populations are used, measurements of mRNA levels necessarily represent the average level across the population. In an ideal case, all of the cells in the population would be in the exact same cell-cycle phase, and thus the measured values represent an average over cell-to-cell variance (biological noise). Although budding yeast is well known for the ease of synchronization using a variety of methods (mating pheromone, conditional mutants in cell-cycle genes, elutriation), synchronized populations are never fully synchronous, and they lose synchrony over time. Loss of synchrony is especially severe in budding yeast where the division is asymmetric (Hartwell and Unger 1977; Lord and Wheals 1980, 1981; Woldringh et al. 1993; Bean et al. 2006; Orlando et al. 2007). Thus, the measured values from time-series experiments using synchronized populations represent a convolution of values from cells distributed across cell-cycle phases. The result is that transcript oscillations appear to dampen over time. Convolution due to loss of synchrony can obscure the true amplitude of transcript oscillations as well as the precise timing of the appearance and disappearance of a specific transcript.
Algorithmic approaches to population studies
To better identify more subtle aspects of the cell-cycle-regulated transcriptional program, several groups set out to use algorithmic approaches to deconvolve the measured values of transcript levels from microarray experiments (Bar-Joseph et al. 2004; Qiu et al. 2006; Rowicka et al. 2007; Guo et al. 2013). The goal of these approaches was to better approximate the dynamic behavior of transcripts in the average cell in the population. The groups made various assumptions about the underlying models that drive synchrony loss in time-series experiments, and each reported an increased performance in accurately inferring both the timing and amplitude of transcript levels during the cell cycle. Additionally, Guo et al. (2013) explicitly modeled asymmetric division using a branching process model that allowed them to infer differences in expression dynamics between mother and daughter cells. Thus, they were able to identify 82 genes that were expressed exclusively in daughter cells (see below). Taken together, these algorithmic approaches permit a more accurate view of the timing of transcript dynamics and suggest that the amplitude of transcriptional responses is, in many cases, substantially higher than previously recognized.
Although these algorithmic approaches likely provide a sharper view of transcript dynamics during the cell cycle, the values they return are still estimates rather than true measurements. Of course, the estimates are only as good as the underlying models for population asynchrony. Additionally, the algorithms report values for the average cell in the population and do not tell us much about the distribution of behaviors across cells in the population.
While there are computational methods to address the issues associated with approaches using synchronized populations, arguably the most direct way to address the issues is to measure transcript abundance in single cells. Ideally, one would be able to monitor the abundance of a variety of transcripts in living cells as they progress through the cell cycle. However, current protocols are limited to measuring only a few transcripts per experiment in fixed cells or to using the expression of destabilized proteins as a proxy for mRNA abundance in living cells.
Single-cell fluorescence in situ hybdrization (FISH) studies (Femino et al. 1998) can be used in yeast to measure the levels of a specific transcript. Because cells are fixed during the hybridization procedure, measurements of transcript fluctuations over time can only be made on a population of synchronized cells, and thus this approach would suffer from some of the same issues related to population synchrony. However, single-cell FISH was used effectively to demonstrate that individual yeast cells in an unsynchronized culture of slow growing yeast cells exhibit metabolic cycling of gene expression (Silverman et al. 2010). This approach, examining the coincidence of expression of several gene pairs, enabled investigators to establish positive correlation between genes expressed in the same phase of the metabolic oscillation and negative correlation between those expressed in opposite phases. Earlier studies relied on sampling synchronized populations of cells from a chemostat for microarray analysis (Klevecz et al. 2004; Tu et al. 2005). This approach also worked well when applied to cell-cycle-regulated genes by Silverman et al. (2010). Single-molecule fluorescence techniques, which have been applied successfully to monitor the initiation and elongation rates, as well as the stability of cell-cycle-regulated mRNAs in living cells, may be applicable in studies of the dynamics of cell-cycle-regulated transcription (Larson et al. 2011; Trcek et al. 2011).
The dynamics of periodic transcripts can also be inferred by time-lapse microscopy in live cells using strains with cell-cycle-regulated promoters fused to unstable variants of green or red fluorescent proteins (GFP/RFP) (Skotheim et al. 2008; Eser et al. 2011). Time-lapse microscopy on single cells does not suffer from population effects, and in cases where other markers are available, coherence between the expression of a fluorescent protein and other events can be determined. However, this approach does not directly measure transcript levels, and the kinetics of the appearance and disappearance of transcripts can only be estimated from the accumulation of the GFP protein. Precise estimation is blurred by variations in the maturation time of the GFP/RFP, and in the protein half-life, especially for short-lived transcripts. Furthermore, detection of GFP/RFP fluorescence is insufficiently sensitive to be useful for all but the most powerful promoters. Despite the low-throughput nature of the approach, transcriptional reporters with better signal to noise (e.g., luciferase) and advancements in high-resolution time-lapse microscopy should expand its utility (Howell et al. 2012).
Cell-Cycle-Regulated Gene Clusters: The Logic and Strategy for Expression
Despite the observation that transcripts appear and disappear in a continuum throughout the cycle, the cell-cycle transcription program has historically been viewed as four major waves of gene expression that define the dominant cell-cycle-regulated gene clusters including the G1/S cluster, the S phase cluster, the G2/M cluster, and the M/G1 cluster, based upon the coincidence of the timing of expression relative to the easily discernible cell-cycle events (reviewed by Bähler 2005; Wittenberg and Reed 2005; McInerny 2011). Despite their apparently coherent expression, the genes within a specific cluster are often coordinately regulated by distinct transcription factors or mechanisms. In addition to the prominent clusters, there are smaller gene clusters or single genes with distinct patterns of cell-cycle-regulated expression sometimes mediated by multiple transcription factors acting in concert. Our goal in this section is to discuss the major clusters and their modes of regulation.
The G1/S gene cluster
The G1/S gene cluster (also called simply the G1 gene cluster) comprises the largest and, arguably, the best-characterized family of cell-cycle-regulated genes in yeast. This family of upwards of 200 genes is expressed largely in G1 phase cells coincident with START and prior to the initiation of DNA replication (reviewed by bähler 2005; Wittenberg and Reed 2005; McInerny 2011). These genes are controlled by two heterodimeric transcription factors, SCB-binding factor (SBF) and MCB-binding factor (MBF), which bind to distinct but related DNA sequence motifs in the promoters of their target genes (Figure 2). SBF binds to SCB elements (SBF cell-cycle box) and MBF to MCB elements (MBF cell-cycle box), respectively (Lee et al. 2002; Teixeira et al. 2006). The distinct DNA-binding properties of those factors is attributable to their DNA-binding proteins, Swi4 for SBF and Mbp1 for MBF. Both factors bind DNA via a classical winged helix-turn-helix DNA-binding fold and associate with the common subunit Swi6. Although both DNA-binding proteins and Swi6 share a common domain structure, Swi4 and Mbp1 bind distinct DNA sequences, whereas Swi6 does not bind directly to DNA (McIntosh et al. 2000; Xu et al. 1997; Nair et al. 2003; Taylor et al. 1997, 2000, 2010).
SBF and MBF bind and regulate partially overlapping subsets of genes in the G1/S cluster because of the presence of both consensus sequence motifs in some of the promoters (Spellman et al. 1998; Iyer et al. 2001; Simon et al. 2001; Teixeira et al. 2006; Ferrezuelo et al. 2010). The extent to which the overlap is reflected in coordinate or differential regulation is only beginning to be understood (see below) (Bean et al. 2005; Eser et al. 2011; de Oliveira et al. 2012; Smolka et al. 2012; Travesa and Wittenberg 2012; Travesa et al. 2012). Finally, many potential and actual binding sites have been identified in promoters that do not appear to promote cell-cycle periodicity but may, in some cases, have functional consequences (Iyer et al. 2001; Horak et al. 2002; Bean et al. 2005; also see Haber 2012; Eriksson et al. 2012).
The G1/S gene cluster is rich in genes involved in processes associated with progression from G1 into S phase. Passage through START is particularly important for initiating the many events required for the new cell division cycle. Those include dramatic changes in morphogenesis leading to the formation of the bud, the nascent daughter cell with its associated septin ring marking the future site of cytokinesis (reviewed in Bi and Park 2012; Howell and Lew 2012). Additionally, the duplication of the spindle pole body, the first structures associated with assembly of the mitotic spindle (reviewed by Winey and Bloom 2012), and the initiation of DNA replication are triggered following START (reviewed by Remus and Diffley 2009). Although the separation of function is by no means absolute, in general, many genes involved in morphogenesis, including enzymes required for cell wall biosynthesis and septin ring components, are regulated by SBF, whereas genes for DNA replication and repair, including nucleotide biosynthetic enzymes and proteins acting at the replication fork, are regulated by MBF. The rationale for segregating SBF and MBF targets based upon function has, until recently, lacked explanation because the two transcription factors are coordinately regulated during the cell cycle. However, it has become apparent that segregation of targets into two groups can have dramatic consequences under specific physiological conditions (see below) (Doncic et al. 2011; Eser et al. 2011; Smolka et al. 2012; Travesa et al. 2012).
The hallmark of these two transcriptional regulators is their capacity to abruptly activate their target genes during G1 phase in response to the activation of G1 cyclin- associated Cdk1 (Figure 2) (Costanzo et al. 2004; Di Talia et al. 2007; Taberner et al. 2009). That regulation is disrupted by cdc28-ts mutants (Marini and Reed 1992). We now understand that the activation of G1/S genes is a consequence of the regulated accumulation of the G1 cyclin Cln3 (Tyers et al. 1993; Dirick et al. 1995; Stuart and Wittenberg 1995; Garí et al. 2001). The accumulation, activation, and localization of Cln3/CDK are critical determinants of the timing of START and are responsive to cell growth and cell size (Nash et al. 1988; Tyers et al. 1993; Baroni et al. 1994; Polymenis and Schmidt 1997; Miller and Cross 2000; Edgington and Futcher 2001; Jorgensen et al. 2002; Di Talia et al. 2007; Taberner et al. 2009; Ferrezuelo et al. 2012; Thorburn et al. 2013). Cln3, once accumulated to a critical level in the nucleus, activates Cdk1, leading to the phosphorylation of the SBF-bound transcriptional repressor, Whi5 (Costanzo et al. 2004; de Bruin et al. 2004; Schaefer and Breeden 2004). Upon phosphorylation, Whi5, which binds to the carboxy terminus of Swi6 through its GTB-containing domain, dissociates from promoter-bound SBF and is relocalized to the cytoplasm (Figure 2) (Costanzo et al. 2004; de Bruin et al. 2004; Di Talia et al. 2007; Skotheim et al. 2008; Travesa et al. 2013). Phosphorylation of Swi6 by Cln/CDK may also contribute to the regulation of Whi5 binding (Sidorova et al. 1995; Costanzo et al. 2004; Wagner et al. 2009). Dissociation and nuclear export of Whi5, in turn, leads to the activation of SBF and the expression of its target genes. Although Cln3/CDK-dependent Whi5 phosphorylation appears to promote both dissociation from promoters and export from the nucleus, those two phenomena have not been separated in terms of their effect on the activation of G1/S transcription (Costanzo et al. 2004; de Bruin et al. 2004; Di Talia et al. 2007; Skotheim et al. 2008; Travesa et al. 2013). A recent study has established that release of the Whi5 repressor is the critical determinant of the event historically referred to as START by Hartwell and colleagues (Doncic et al. 2011; Eser et al. 2011).
Activation of G1/S transcription factors
Our current understanding of SBF derives from its early identification (by the Nasmyth and Herskowitz laboratories) among the genes required for expression of the HO gene, which encodes the endonuclease that mediates switching of mating types via DNA rearrangements at the MAT locus (Breeden and Nasmyth 1987; Andrews and Herskowitz 1989; see Haber 2012). The known targets of SBF, which now number ∼100, include those encoding the G1 cyclins, Cln1 and Cln2, which, like Cln3, promote Cdk1-dependent activation of G1/S transcription (Cross 1988; Wittenberg et al. 1990; Nasmyth and Dirick 1991; Tyers et al. 1993; Wijnen et al. 2002). Although earlier studies supported the autonomy of Cln3/Cdk1 in the timely activation of G1/S genes (Dirick et al. 1995; Stuart and Wittenberg 1995), the application of single-cell techniques to the analysis of START led to the realization that the rapid accumulation of Cln1 and Cln2, and their potent activation of Cdk1, leads to a dramatic acceleration of nuclear export of Whi5. This positive feedback greatly enhances the coherency of activation of G1/S genes (Skotheim et al. 2008), confirming a model advanced many years prior (Cross and Tinkelenberg 1991; Dirick and Nasmyth 1991). Thus, the combined activity of the G1 cyclin-associated forms of Cdk1 is responsible for the switch-like behavior and irreversibility of the START event.
The expression of MBF targets, like those regulated by SBF, occurs during late G1 phase and depends upon Cln3/Cdk1 activity (Tyers et al. 1993; Dirick et al. 1995; Stuart and Wittenberg 1995; Wittenberg and Reed 2005). MBF targets include genes encoding DNA replication and repair proteins and the cyclin Clb5, which is critical for timely activation of DNA replication. Interestingly, the SWI4 gene was recently shown to be a target of MBF, which is sufficient for much of its expression during G1 phase (Harris et al. 2013), suggesting there is crosstalk between the two major G1/S transcription factors. The mechanism of activation of MBF-regulated genes remains to be established. We do know that, unlike SBF, MBF acts, in part, as a repressor of its target genes outside of G1 phase, suggesting that Cdk1 acts to relieve its repressive activity (de Bruin et al. 2006). However, as of yet, no protein playing an equivalent role to Whi5 has been identified. This has led to the prevailing hypothesis that the direct phosphorylation of the transcription factor by Cdk1 is the critical event in transcriptional activation. Although both Swi6 and Mbp1 are known to be substrates of Cdk1 in vitro and are phosphoproteins in vivo, there is currently no evidence indicating that those phosphorylation events are required for activation of MBF targets (Sidorova et al. 1995; Siegmund and Nasmyth 1996; Ubersax et al. 2003; Geymonat et al. 2004; Holt et al. 2008).
Repression of G1/S transcription factors
Like SBF targets, MBF-regulated genes play important roles in the circuitry regulating G1/S transcription (Figure 2). One of its targets is the gene encoding the MBF-specific transcriptional repressor, Nrm1 (de Bruin et al. 2006). Unlike Whi5, which represses transcription from M phase through late G1 phase, Nrm1 acts only after G1/S transcription is strongly activated, leading to Nrm1 accumulation. Nrm1 binds to the carboxy terminus of Swi6 at MBF target promoters via interaction between its GTB-containing domain, a motif conserved with Whi5 and its fungal orthologs (Ofir et al. 2012; Travesa et al. 2013). This leads to the reestablishment of MBF as a transcriptional repressor such that expression of MBF target genes, including that of NRM1 itself, is terminated by negative feedback. Subsequently, as cells exit mitosis, the Nrm1 protein is targeted for degradation by the APCCdh1 ubiquitin ligase, preparing cells to reactivate MBF during the subsequent G1 phase (Ostapenko and Solomon 2011). In addition to Nrm1, MBF also promotes the expression of two B-type cyclins, Clb5 and Clb6. Whereas Clb5 ultimately activates Cdk1 to drive entry into S phase, Clb6/Cdk1 has been reported to phosphorylate Swi6, promoting its accumulation in the cytoplasm (Geymonat et al. 2004). It has been suggested that Swi6 is constitutively cycling between the cytoplasm and the nucleus such that the phosphorylation-dependent inhibition of nuclear uptake results in accumulation in the cytoplasm that is then reversed upon dephosphorylation subsequent to mitosis by the Cdc14 phosphatase (Queralt and Igual 2003; Geymonat et al. 2004). Interestingly, cycling of Swi6 between the cytoplasm and nucleus appears to be important for SBF-dependent transcriptional activity, although the molecular basis for that requirement has not been established (Queralt and Igual 2003). Like Swi6, nuclear localization of Whi5 is promoted by Cdc14-dependent dephosphorylation upon M phase exit. The capacity of Whi5 to relocalize to the nucleus if CDK activity is reduced during other cell cycle phases suggests that phosphatases other than Cdc14 must also be competent to dephosphorylate it (Charvin et al. 2010). Nuclear exclusion of Whi5 is also promoted by CDK-dependent phosphorylation, albeit earlier in the cell cycle than Swi6 (Costanzo et al. 2004; Di Talia et al. 2007; Taberner et al. 2009).
The repression of SBF does not seem to act through Nrm1-like negative feedback. The primary mechanism terminating expression of SBF targets is the accumulation of the B-type cyclin, Clb2 (Figure 2). Clb2/Cdk1 binds stably to Swi4 through its conserved ankyrin repeat and phosphorylates Swi4, leading to its dissociation from DNA and export to the cytoplasm, terminating SBF-dependent transcription (Amon et al. 1993; Koch et al. 1993; Siegmund and Nasmyth 1996). SBF-dependent expression of the genes encoding the Yox1 and Yhp1 transcriptional repressors may also contribute, albeit modestly, to the repression of SBF by repressing SWI4 expression (Pramila 2002). Although a similar mechanism has not been demonstrated for MBF, it is clear that in the absence of Nrm1, MBF can also be inactivated by a mechanism that depends upon B cyclin-Cdk1 (de Bruin et al. 2006).
Temporal regulation within the G1/S gene cluster
The application of algorithmic approaches to microarray analysis of the cell-cycle transcriptome has revealed an ordered timing in the activation of G1/S genes that had previously appeared to be expressed simultaneously (Eser et al. 2011; Guo et al. 2013). Although some hints of temporal resolution of gene expression within that cluster had been observed in prior analyses, the lack of reproducibility and the problems associated with differing amplitudes of expression had obscured those differences. Using high-throughput analysis of multiple array experiments coupled with single-cell fluorescence analyses, Eser and colleagues found that CLN1 and CLN2 were among the earliest expressed genes of the G1/S cluster, leading to the concept of “feedback first,” wherein the first targets to be expressed are those that promote further transcriptional activation of G1/S targets. That mechanism ensures the irreversibility of the process and is generally conserved in both closely related yeast species and in systems as divergent as humans. Their findings also revealed that NRM1, encoding the repressor of MBF-regulated transcription, is among the last genes of the cluster to be expressed, thereby ensuring that MBF targets are not repressed before they are adequately expressed. Different synchronization regimens can cause sequential waves of MBF-then-SBF or SBF-then-MBF activity such that genes containing both SBF- and MBF-binding sites are activated by whichever of the MBF or SBF factors is activated first and inactivated by whichever transcription factor is repressed first (Eser et al. 2011). Similar results were obtained using a mathematical deconvolution approach that increases the temporal resolution of microarray experiments (Guo et al. 2013). These findings suggests that cells can adapt their sequential program of gene activation in response to environmental or physiological conditions, perhaps resulting in increased fitness.
Coupling of G1/S transcription to the rest of the transcriptional program
The targets of the G1/S transcription factors are a large and diverse set of genes involved in a broad spectrum of events, many of which are initiated during the interval of G1/S gene expression. Of particular interest, in the context of understanding cell-cycle-regulated gene expression, are genes that act to limit G1/S transcription and those that induce subsequent waves of transcription. Nrm1 is perhaps the best example of a G1/S target that acts to limit G1/S transcription. Clb2 similarly limits SBF-regulated transcription, but it is not itself a G1/S target (see below). Hcm1 is the transcriptional activator of a large group of S phase genes (discussed below), and is expressed during G1/S from a promoter that is bound by both SBF and MBF (Iyer et al. 2001; Pramila 2002). Thus, Hcm1 targets are activated as a consequence of the burst of G1/S transcription, thereby, participating as a link in the cell-cycle-transcriptional circuit (Figure 1).
The S phase gene cluster
The G1/S transition, characterized by the initiation of DNA replication, is accompanied by the activation of two clusters of S phase genes. One cluster is composed of the genes encoding the histones, which make up the nucleosomes that package newly replicated DNA. Histone genes are regulated by an, as yet, poorly understood transcriptional regulatory mechanism involving SBF, Spt10, histone chaperones, and other factors. That S phase cluster has been recently reviewed by Eriksson et al. (2012) and will not be discussed further here. The second, much larger, cluster of genes expressed during S phase is that regulated by Hcm1 (Figure 3). Although less thoroughly studied than transcription factors controlling the other major bursts of cell-cycle-regulated gene expression, Hcm1 plays a central role in that regulatory circuitry. It is one of four members of the forkhead family of transcription factors found in budding yeast (Hcm1, Fkh1, Fkh2, and Fhl1 (Forkhead-like) (reviewed in Murakami et al. 2010). Like the other members of that family, which include mammalian Fox transcription factors, Hcm1 binds DNA directly via a winged helix DNA-binding motif. The presence of putative CDK phosphorylation sites in Hcm1 and its capacity to be phosphorylated by Clb/Cdk1 in vitro make it a likely target of the Cdk1 protein kinase (Ubersax et al. 2003).
The Hcm1 transcription factor appears to regulate ∼180 genes based upon the timing of their expression and the presence of the Hcm1-binding site in their promoters (Pramila et al. 2006). However, only a small fraction of those genes were found to bind Hcm1 by genome-wide chromatin immunoprecipitation (Horak et al. 2002), likely a consequence of the low reliability of the early studies using that technique. Among the targets are ∼10 genes involved in budding and morphogenesis and nearly 100 involved in some aspect of chromosome segregation. Unsurprisingly, these include protein components of chromosomes and elements of the chromosome segregation machinery, much of which is assembled during S and G2 phases. However, a role in the regulation of genes involved in budding was unexpected because bud morphogenesis is a well-established function of many genes in the G1/S gene cluster. Nevertheless, under most circumstances, budding occurs coincident with initiation of DNA replication so expression during S phase under the control of Hcm1 is timely. That said, Hcm1 is dispensable for budding but is required for the proper timing and execution of mitosis (Pramila et al. 2006). The synthetic lethality of hcm1 mutations with mutations in a number of genes involved in chromosome segregation is consistent with its importance in the orchestration of mitotic functions.
The relatively recent discovery of Hcm1 makes it something of a “missing link” in the transcriptional circuitry of the cell cycle. Its targets include regulators of all of the other major bursts of cell-cycle-regulated gene expression: the transcriptional repressors Whi5 (G1/S transcription) and Yhp1 (M/G1 transcription) as well as Fkh1, Fkh2, and Ndd1, the central regulators of the G2/M gene cluster (see below). Although these factors do not play a direct role in limiting G1/S transcription via feedback, Hcm1 indirectly activates Clb2, the feedback repressor of SBF genes, via its role in the expression of Fkh2 and Ndd1 activators of the G2/M gene cluster (Figure 1). Despite the central role of Hcm1 in the regulation of genes involved in cell-cycle-regulated transcription, hcm1 mutants are viable and retain detectable periodicity of S phase gene expression (Pramila et al. 2006; Simmons Kovacs et al. 2012). This suggests the involvement of additional, as yet unknown, transcription factors in the regulation of this gene family.
The G2/M gene cluster
Progression from S phase into G2 phase is accompanied by activation of a family of ∼35 genes falling into the G2/M cluster, also called the CLB2 cluster (Cho et al. 1998; Spellman et al. 1998). As the latter name implies, this cluster includes the genes encoding the B-type cyclin Clb2, along with its paralog Clb1 and other important cell cycle regulatory factors including Cdc5, the yeast polo-like kinase, Cdc20, a specificity subunit of the APC ubiquitin ligase, and the Ace2 and Swi5 transcription factors. The primary regulator of this gene cluster is the MADS box transcription factor Mcm1, in conjunction with the forkhead family member Fkh2 and the coactivator Ndd1, both members of the S phase gene cluster regulated by Hcm1 (Figure 3) (Loy et al. 1999; Kumar et al. 2000).
Mcm1 plays diverse roles that derive from its capacity to multimerize with a diverse set of coregulators. This allows it to regulate distinct gene families including the M/G1 cluster genes (see below), the MAT locus genes (Tuch et al. 2008), and even to play a role in mating-type switching, which is apparently transcription independent (Coïc et al. 2006). The collaboration of Mcm1 with Fkh2 in the context of transcriptional regulation of the G2/M cluster is, in part, a consequence of the occurrence of adjacent Mcm1- and Fkh2-binding sites in G2/M cluster promoters (Lydall et al. 1991; Boros et al. 2003). The association of that heterodimer with the adjacent binding sites on those genes creates a platform for association of the coactivator Ndd1 to activate gene expression. Whereas other forkhead family members possess the capacity to bind to the DNA via their conserved winged helix motif, only Fkh2 can also associate directly with Mcm1 and thereby exert its effect on gene expression in that context. Interestingly, this binding interface has been conserved sufficiently for Fkh2 to bind to the human serum response factor (SRF), the homolog of Mcm1 (Boros et al. 2003). Fkh1, which lacks the Mcm1 interaction domain of Fkh2, binds to the same DNA motif as Fkh2 at CLB2 cluster promoters, but only in the absence of Mcm1 (Koranda et al. 2000; Hollenhorst et al. 2001).
The general view of regulation of the CLB2 cluster posits that Mcm1 and Fkh2 remain bound to DNA, and to each other, throughout the cell cycle. They repress transcription during G1 and S phases, but subsequent recruitment of Ndd1 activates those same genes as cells progress through G2 and into M phase. FKH2 and NDD1 are coexpressed as components of the S phase gene cluster. Ndd1 is destroyed during late M phase, presumably as a consequence of ubiquitylation by the APC (Loy et al. 1999). In contrast, Fkh2 is stable, persisting throughout the cell cycle (Koranda et al. 2000). Although association of Ndd1 with Fkh2 is the limiting event for activation of G2/M cluster expression, its accumulation alone is reported to be insufficient for association and transcriptional activation (Reynolds et al. 2003). Phosphorylation of Ndd1 by both Clb2/Cdk1 and Cdc5 polo-like kinase appears necessary for efficient recruitment to the Fkh2 FHA domain and robust activation of Clb2 cluster genes (Reynolds et al. 2003; Pic-Taylor et al. 2004; Darieva et al. 2006, 2012).
The requirement for Ndd1 phosphorylation constitutes a positive feedback loop, wherein both Clb2 and Cdc5, when expressed as G2/M cluster genes, feedback and activate their own gene expression. This posed a problem because, if these two protein kinases are required for activation of their own expression, it is unclear how they would be activated to initiate the feedback loop. This is particularly vexing in the case of Cdc5, which, unlike Clb2, has no paralogs in yeast and which appears to be the only kinase that phosphorylates S85 of Ndd1 (Darieva et al. 2006). However, this problem was recently addressed by the report that at least a subset of the Clb2 cluster genes (including CLB2) can be partially activated, even in the complete absence of B-type cyclins (Orlando et al. 2008). This finding suggests a mechanism by which the feedback loop can be activated in the absence of phosphorylation by Clb2.
In addition, phosphorylation of Ndd1 by Pkc1 (protein kinase C) has recently been shown to delay its association with Mcm1/Fkh2 and, thereby, prevent activation of Clb2 cluster genes by Ndd1 until conditions are appropriate (Darieva et al. 2012). Precisely what conditions are monitored via Pkc1 is unclear but it seems likely that it is related to cell wall synthesis or stress (reviewed in Levin 2011). Finally, it is believed that upon destruction of Ndd1 late in M phase, Mcm1/Fkh2-bound CLB2 cluster promoters are returned to their repressed state, prepared to initiate a new cell cycle (Loy et al. 1999). It remains unclear whether Ndd1 destruction alone is sufficient for repression of CLB2 cluster genes or whether that repression also involves a specific transcriptional repressor that acts either via negative feedback or some other mechanism.
This view of regulation of G2/M cluster genes, although sufficient to explain expression of CLB2 and other members, does not explain the regulation of the entire gene cluster. In fact, some genes within the cluster exhibit different behaviors in the mitotic cyclin mutants, suggesting that more than one regulatory mechanism exists for triggering transcription (Orlando et al. 2008). It has recently been recognized that other members of the family, including SPO12, despite being Mcm1/Fkh2 targets, are subject to an additional level of regulation by the homeodomain-containing repressor protein, Yox1 (Pramila 2002; Darieva et al. 2010). Yox1 was previously recognized as an important repressor in the regulation of the M/G1 gene cluster (see below). However, in addition to having a consensus DNA-binding site for Fkh2 downstream of that for Mcm1, some G2/M genes have a Yox1-binding site upstream, suggesting that Yox1 also plays a role in their transcriptional regulation (Darieva et al. 2010). In fact, occupancy of that Yox1-binding site precludes the binding of Fkh2. Consequently, Yox1 represses those genes having a Yox1-binding site in a manner similar to its effect on members of the M/G1 gene cluster. The basis for the mutual exclusivity of binding sites for Yox1 and Fkh2 remains unclear but presumably involves competition between Yox1 and Fkh2 for binding sites on Mcm1 and not the geometry of the DNA-binding sites themselves (Boros et al. 2003; Darieva et al. 2010). Regardless, this regulatory circuit is clearly effective in repressing gene expression at those loci and shifting the expression of those genes later in the cell cycle relative to the other members of the G2/M gene cluster (Pramila 2002). Importantly, this regulation does not affect Clb2 or Cdc5, which act early during the transcriptional burst to promote activation via positive feedback. This is similar to the “feedback first” phenomenon observed earlier in the cycle, wherein Cln1 and Cln2, encoded by the earliest expressed genes of the G1/S cluster, feedback to promote SBF- and MBF-regulated transcription.
The M/G1 gene cluster
As cells exit M phase and transition into G1 phase of a new cell cycle, cells activate four families of genes, distinguishable based upon either the function of their products or the transcription factors that regulate them. Together they comprise the M/G1 gene cluster. That cluster includes genes encoding the cyclin Cln3 and the Swi4 transcription factor, two proteins responsible for promoting G1/S transcription (SWI4 is also activated by MBF; see above). Components of the prereplication complex that prepare cells to initiate DNA replication during S phase, elements of the mating response pathway, and genes of the PHO regulon that are required for mobilization of phosphate are also members of that cluster (Haber 2012; Ljungdahl and Daignan-Fornier 2012).
The Mcm1 gene cluster
The largest family of genes in the M/G1 gene cluster, like the G2/M cluster, is regulated by Mcm1 but, in this case, without the involvement of Ndd1 (Zhu et al. 2000). Mcm1 is bound to the promoters of target genes, including Cln3, Swi4, and seven members of the prereplicative complex (Cdc6 and Mcm2–7) (Spellman et al. 1998), via an element referred to as an early cell-cycle box (ECB) (McInerny et al. 1997; MacKay et al. 2001). Mcm1 is regulated by one of two transcriptional repressors, the homeodomain proteins Yox1 or Yhp1, that hold expression of this cluster of genes in check outside of the M/G1 interval by binding directly to Mcm1 and a DNA element that flanks the ECB (Mai et al. 2002; Pramila 2002) (Figure 4). Their interaction with both Mcm1 and DNA is required to stabilize the repressive complex (Darieva et al. 2010). Consistent with the need to repress M/G1 cluster expression from G1 through M phase, YOX1 is expressed as a G1/S cluster gene during G1 and YHP1 somewhat later, perhaps as a consequence of the presence of Fkh2 sites in its promoter (Spellman et al. 1998; Pramila 2002). Expression of Mcm1 target genes does not occur until the Yox1 and Yhp1 repressor proteins are eliminated, presumably by APC-dependent ubiquitylation, during late M phase (Pramila 2002). Whereas Yox1 appears to repress gene expression at some G2/M cluster promoters by displacing Fkh2-Ndd1, no similar mechanism is known for M/G1 cluster genes (Darieva et al. 2010). Nevertheless, renewed accumulation of Yox1 and Yhp1 at the beginning of the next cell cycle certainly contributes to M/G1 gene repression.
The Sic1 gene cluster
The second family of M/G1 cluster genes, often referred to as the Sic1 cluster, is regulated by the paralogous transcription factors Ace2 and Swi5 (Dohrmann et al. 1992, 1996; McBride et al. 1999; Laabs et al. 2003) (Figure 4). These factors bind identical DNA sequences via zinc finger domains in vitro but regulate groups of genes in vivo that are only partially overlapping. CTS1 is a classical Ace2-regulated gene encoding chitinase, which is expressed only in daughter cells and mediates cell separation following mitosis (reviewed in Bi and Park 2012). In contrast, the best-studied Swi5 target, HO, is expressed only in mother cells (reviewed in Haber 2012). Both factors participate in the activation of SIC1, encoding a Clb-specific CDK inhibitor for which the cluster is named. Sic1 is expressed very late in M phase, where it contributes to the inhibition of Clb/CDK activity during exit from mitosis. Recent analyses suggest that the SIC1 transcription occurs primarily in daughter cells during G1 phase (Guo et al. 2013), where it prevents premature initiation of DNA replication (reviewed in Enserink and Kolodner 2010). Consequently, regulating the activity of Swi5 and Ace2 and their partitioning between mother and daughter cells is critical for proper cell-cycle regulation.
The differential regulation of Sic1 cluster genes by paralogous transcription factors that not only bind to the same DNA-binding site in vitro but, in many cases, bind to the same genes in vivo poses a conundrum (Dohrmann et al. 1992, 1996). By evaluating the requirement for Ace2 and Swi5 for expression of M/G1 genes and correlating that with the binding of Ace2 and Swi5 to the promoters of those genes, it became clear that the promoters of genes regulated solely by Swi5 are bound only by Swi5, whereas those regulated by both factors or by Ace2 alone bind both factors (Voth et al. 2007). Why, then, are some genes that bind both factors only dependent upon Ace2 for expression, whereas others can be activated by either one? This conundrum is, in part, resolved by the finding that promoters regulated only by Ace2 bind Swi5 along with Fkh1 and Fkh2, which act as “antiactivators” for Swi5 at those promoters (Figure 4). In contrast, genes that are activated by both factors or only by Swi5, including SIC1, CDC6, and HO, lack the Fkh1/2 binding site. Consequently, it appears that Swi5 binds to the promoters of all Sic1 cluster genes but only activates those at which Fkh1 or Fkh2 is not bound. Whether the Swi5/Fkh1/2 binds but fails to activate the same promoters in daughters, as well as in mothers, is unclear.
Another aspect of the differential regulation of genes by these transcription factors involves the asymmetric distribution of Ace2 to daughter nuclei following mitosis. That asymmetry is a consequence of the failure to retain Ace2 in mother cell nuclei following mitosis (Colman-Lerner et al. 2001; Weiss et al. 2002; Mazanka and Weiss 2010). Daughter-specific retention of Ace2 is regulated, at least in part, by the NDR/LATS family protein kinase Cbk1, a component of the so-called RAM complex (regulation of Ace2 and morphogenesis). A deconvolution algorithm developed by Guo and colleagues revealed the existence of a large group of genes subject to the daughter-specific transcription program (Guo et al. 2013). In addition to those genes that had been reported previously (Colman-Lerner et al. 2001; Di Talia et al. 2009), the high temporal resolution afforded by the deconvolution approach led to the identification of at least four temporally distinct waves of daughter-specific transcripts and identified several new daughter-specific genes. Transcription factor enrichment analyses suggest the earliest waves are regulated by Swi5 and Ace2, whereas the later waves appear to be regulated by Mcm1. These analyses also indicate a role for several other transcription factors in that regulation, including Sok2, Phd1, Ste12, Cin5, Yap6, and Tec1 (Guo et al. 2013).
One consequence of asymmetry is that chitinase, encoded by the Ace2 target CTS1, is daughter specific. This pattern of expression is responsible for the characteristic retention of the bud scar on mother cells, but not daughter cells, subsequent to cytokinesis. Although Ace2 and Swi5 can both bind to the CTS1 promoter, when Swi5 is bound, it fails to activate due to the binding of Fkh1/2 (Voth et al. 2007) (Figure 4). Only when Ace2 is bound, can it activate CTS1. Another consequence of the daughter-specific localization of Ace2 is the characteristic delay in budding in daughter cells relative to mother cells. This is, in part, a consequence of the Ace2-dependent delay in CLN3 expression and, thereby, delayed activation of G1/S transcription in daughter cells (Laabs et al. 2003; Di Talia et al. 2009). Although this impacts the range of cell size at which daughter cells bud under various nutrient conditions, it does not explain the phenomenon of cell size control (reviewed by Turner et al. 2012).
Finally, the Swi5 transcription factor exhibits specificity for binding to some cell-cycle-regulated promoters (e.g., CDC6 and HO), a property that is apparently not shared by Ace2 (Voth et al. 2007). Swi5 binding to the HO promoter, which occurs late in M phase, licenses that promoter for activation only in mother cells during the subsequent G1 phase. Similar activation does not occur in daughter cells due to the daughter-specific localization of the transcriptional repressor Ash1, which prevents Swi5 binding (Bobola et al. 1996; Sil and Herskowitz 1996; Cosma 2004; Shen et al. 2009). Thus, the selective binding of Swi5 at promoters in the absence of the Fkh1 and Fkh2 transcription factors defines genes that are uniquely regulated by Swi5.
Clb/CDK dependent phosphorylation is another critical contributor to Ace2 and Swi5 regulation during the cell cycle because it impacts their nuclear localization (Moll et al. 1992) (Figure 4). Phosphorylation within the nuclear localization sequences (NLSs) of Swi5 and Ace2 from S phase through M phase, catalyzed by increasingly active CDK, prevents transcriptional activation by preventing their entry into the nucleus. Only when B-type cyclins are destroyed during M phase and the Cdc14 phosphatase is activated upon exit from mitosis, are their NLS motifs dephosphorylated, allowing entry into the nucleus and promoter binding (Visintin et al. 1998). Thus, activation of Swi5 and Ace2 targets in early G1 is dependent on exit from mitosis. The asymmetric loss of Ace2 from mother cell nuclei depends upon its nuclear export sequence (NES), a motif that is not shared with Swi5 (Sbia et al. 2008). Inactivation of Ace2 by nuclear exclusion is necessary because it is stable throughout the cell cycle, whereas the bulk of Swi5 is apparently rapidly degraded upon localization to the nucleus (Tebb et al. 1993). Together, these mechanisms account for the selective inactivation of the M/G1 transcription factors in mother and daughter cells.
The MAT gene cluster
Another cluster of M/G1 genes that is expressed differentially based upon cell type is the so-called MAT cluster of haploid-specific genes regulated by the Ste12 transcription factor (Oehlen et al. 1996; Spellman et al. 1998). Although Ste12 participates in multiple transcription factor complexes, some as a heterodimer with Tec1 or Mcm1, regulation of the MAT gene cluster by mating pheromone occurs via binding to pheromone response elements (PREs) in the target promoters as a homodimer (Hwang-Shum et al. 1991). Genes in the MAT cluster encode many proteins involved in the response to mating pheromone, including the Fus3 MAP kinase and Far1, both of which are critical for induction of G1 phase cell-cycle arrest (Roberts et al. 2000). Whereas Ste12-regulated transcription is strongly induced in response to mating pheromone signaling, its target genes are expressed at a basal level in haploid cells of both mating types (Oehlen et al. 1996; Spellman et al. 1998).
Both basal and pheromone-induced expression of Ste12-regulated genes are restricted to the pre-START interval of G1 phase (reviewed by Dohlman and Thorner 2001; Bardwell 2005). Confining mating to G1 phase prevents inappropriate ploidy of the mating cells and proper organization of the apparatus for nuclear fusion. Transcriptional and post-transcriptional controls contribute to the G1 phase restriction. In particular, Far1, an inhibitor of G1 cyclin-associated CDK and regulator of cell polarity, is expressed as a target of both Mcm1 and Ste12 and, thereby, accumulates from M phase through early G1 phase (Oehlen et al. 1996). Fus3 activates both Ste12 and Far1 by phosphorylation, thereby, promoting gene expression and CDK inhibition. Conversely, pheromone signaling is inhibited and Far1 is targeted for destruction by ubiquitin-mediated proteolysis by G1 cyclin/CDK, which accumulates as cells pass START (McKinney et al. 1993; McKinney and Cross 1995; Henchoz et al. 1997). Cln/CDK also attenuates mating pheromone signal transduction by phosphorylating the MAPK cascade scaffold protein Ste5 (Oehlen and Cross 1994; Strickfaden et al. 2007; Bhaduri and Pryciak 2011). Because the substantial activation of Far1 by Fus3-dependent phosphorylation occurs only when the mating pheromone-signaling pathway is induced by mating factor, stable arrest resulting from Far1 inhibition of Cln/CDK does not occur in the absence of pheromone. Instead, under those conditions, Cln/CDK attenuates Ste12-dependent gene expression and promotes cell proliferation. Consequently, cycling cells are only sensitive to mating pheromone-induced cell-cycle arrest during early G1 phase.
The mechanism of periodic expression of Ste12-dependent genes in cycling cells is associated with basal signaling through the pheromone response pathway leading to basal activation of Ste12. Like pheromone-inducible expression, basal expression of the MAT cluster genes occurs only during G1 phase because the pathway is repressed by CDK, which remains active from START until the Clb/CDK inhibitor Sic1 is expressed as cells exit M phase. Although basal pheromone signaling is insufficient to cause cell cycle arrest, it is likely responsible for the enhanced rate of cell proliferation observed in pheromone signaling mutants even in the absence of pheromone (Lang et al. 2009).
The stability of both the START transition and pheromone arrest creates a tension between the mating pathway and cell proliferation. Both the stability and reversibility of those pathways is a consequence of the architecture of the cell-cycle- and pheromone-regulated transcriptional networks (Doncic et al. 2011; Doncic and Skotheim 2013). Pheromone arrest is stable due to pheromone-induced expression of Far1, Fus3, and other factors in the pathway. A drop in pheromone leads rapidly to mitotic growth due both to the rapid decay of signaling through the pathway and the associated activation of Cln/CDK activity resulting from the decreased expression of Far1. Cln/CDK activity promotes proliferation by attenuating the basal activity of the pheromone pathway and thereby Ste12-dependent transcription by promoting Far1 proteolysis and by activating G1/S transcription via positive feedback. Together, these responses flip the bistable switch governing START (reviewed by Ferrell 2011).
Emerging Views of Cell-Cycle-Regulated Transcription
The broad outlines of cell-cycle-regulated gene expression have been worked out for almost a decade. Yet, as can be seen from our discussion of the regulation of the cell-cycle gene clusters, substantial fleshing out and numerous revisions of that view have led to a much greater understanding. In the following vignettes, we introduce some of the recent developments in the general area of cell-cycle-regulated gene expression and a number of specific issues that are just beginning to influence our understanding of this important problem. Important topics impacting our understanding of cell-cycle-regulated gene expression, including the influence of nutrients, pheromones, stress responses, cell size, and other factors, are discussed in other Reviews and Yeastbook articles previously published and to come.
CDK’s role in cell-cycle-regulated transcriptional program
CDKs are widely believed to be the core mechanism controlling cell-cycle-regulated transcription. Support for the idea that periodic transcription occurs as a consequence of periodic cyclin/Cdk activity came from experiments in which specific groups of transcripts were found to be misregulated in cells mutated for specific cyclins (Dirick and Nasmyth 1991; Amon et al. 1993; Stuart and Wittenberg 1994; Koch et al. 1996). An attractive model was proposed by Nasmyth and colleagues in which successive waves of cyclin expression drive progression through the cell cycle as well as control periodic transcription (Amon et al. 1993; Koch et al. 1996). When cells commit to a new cell cycle, Cln3 activates a wave of transcription peaking in late G1. Clb2 then down-regulates a subset of those transcripts and simultaneously promotes the accumulation of a later wave of transcripts peaking in G2/M. A subsequent wave of transcripts peaking in late M/G1 is triggered by the destruction of Clb2 as cells exit mitosis.
Textbook models suggest that the temporal program of transcription is regulated by an interconnected regulatory network containing both CDKs and transcription factors. In these models, CDKs act as the key regulatory components that control the activity of cell-cycle transcription factors (Morgan 2007). There is good motivation for proposing this regulatory relationship, as many of the 10–15 transcription factors that are commonly placed in the cell-cycle network are phosphorylated by CDKs. CDK-mediated phosphorylation regulates both nuclear localization and the formation of transcription factor complexes at promoters (Ho et al. 1999; Costanzo et al. 2004; de Bruin et al. 2004). Consequently, they may serve as critical molecular switches for the interconnected network. In particular, G1-cyclin/Cdks activate SBF and MBF through phosphorylation of Whi5 (Costanzo et al. 2004; de Bruin et al. 2004) and, perhaps, Swi6 (Sidorova et al. 1995), triggering accumulation of many late G1 transcripts; SBF is then inhibited by the mitotic B-cyclin/Cdk complexes (Amon et al. 1993; Koch et al. 1996); B-cyclin/Cdks also phosphorylate and activate Ndd1 and Fkh2 to promote complex formation and trigger accumulation of many G2/M transcripts (Pic-Taylor et al. 2004); B-cyclins inhibit nuclear localization of Swi5 (Moll et al. 1991) and Ace2 (O’Conallain et al. 1999) during mitosis and, it is likely that B-cyclin destruction at the end of mitosis triggers late M/early G1 transcript accumulation (Breeden and Nasmyth 1987; Knapp et al. 1996; Kovacech et al. 1996; Toyn et al. 1997). Thus, successive waves of transcription could arise due to changing cyclin/CDK activities that act post-translationally to turn key factors on and off.
By contrast, in network models arising from genomic approaches, the temporal program of transcription can arise largely independent of CDK regulation (Simon et al. 2001; Lee et al. 2002; Pramila et al. 2006). In these models, successive waves of transcription emerge from oscillations in transcription factor abundance, rather than regulation of transcription factor activity. In a simplified model, transcription factors activate the expression of gene clusters that include genes encoding other transcription factors. Thus, the program could be controlled intrinsically by a network of serially activated transcription factors.
At the time when the first network models were proposed, the CDK and network models seemed incompatible. Both existing models had widely acknowledged gaps. The network models did not incorporate much of the post-translational regulation of transcription factors by CDK complexes that had been established by directed studies of specific transcription factors. However, although the CDK model was supported by rigorous experiments interrogating small subsets of target genes, this model had not been tested on a genome-wide scale.
The extent to which the cell-cycle-regulated transcriptional program was controlled by oscillations of specific cyclin/CDK complexes or by a network of serially activated transcription factors was examined directly by Orlando et al. (2008). Using budding yeast cells in which all six B-type cyclin genes had been inactivated (clb1,2,3,4,5,6∆), they showed that nearly 70% of the cell-cycle-regulated genes continued to be expressed at the proper time. In addition, the program of transcription repeated in a second cycle, despite the fact that the cells were arrested at the G1/S border by all conventional cell cycle measures. In a later study, similar results were observed in cdc28-4 cells at restrictive temperature, although the number and amplitude of the periodic genes was substantially reduced (Simmons Kovacs et al. 2012).
Taken together, these results suggested that interconnected networks of transcription factors could drive most of the periodic transcriptional program in the absence of Clb/CDK activity and function as an autonomous oscillator (Orlando et al. 2008; Simmons Kovacs et al. 2008, 2012). Nonetheless, while many transcripts continue to oscillate in the cells depleted for CDK activities, there were clear alterations in the behaviors of others. Thus, post-transcriptional regulation by CDKs may act more like a rheostat in the control of periodic transcription rather than simply functioning as an “on–off” switch. In general, as CDK activities decreased, the number of identifiable periodic genes also decreased (Simmons Kovacs et al. 2012). Furthermore, decreasing CDK activity was associated with diminished transcript amplitudes as well as increasing oscillatory periods. The regulatory mechanisms responsible for the loss of amplitude and increased period may be positive feedback loops in which cyclin expression is promoted by a transcription factor that is then phosphorylated by the cyclin/CDK complex, which increases the transcription factor’s activity (Simmons Kovacs et al. 2012). This regulatory network motif affects both G1 (SBF and Cln2; see section titled “Activation of G1/S transcription factors”) (Cross and Tinkelenberg 1991; Dirick and Nasmyth 1991; Skotheim et al. 2008) and G2/M gene clusters (Pic-Taylor et al. 2004).
Although CDKs may not be required as a core mechanism for generating the temporally ordered program of transcription during the cell cycle, they clearly play a role in maintaining the robust character of the program. Through positive feedback loops, CDKs provide the basis for important switch-like behavior observed at G1/S and G2/M transitions. Furthermore, the inhibition of Swi5/Ace2-regulated genes by mitotic cyclin/CDK ensures that the initiation of a new program of G1 transcription is dependent on exit from mitosis and the subsequent loss of Clb2/CDK activity. Thus, transcription factor networks and CDKs collaborate to coordinate the execution of the transcriptional program with the cell cycle (Simmons Kovacs et al. 2012).
Checkpoint control of cell-cycle-regulated transcription
Cell-cycle checkpoints maintain the appropriate order of cell-cycle progression by placing prerequisites on progression into a particular cell-cycle phase or on the execution of phase-specific events (Hartwell et al. 1994). For example, a checkpoint mechanism restricts progress into M phase when replication is incomplete, either due to ongoing replication or stalling of DNA replication forks. Similarly, another checkpoint mechanism restricts mitotic spindle elongation when chromosomes are not appropriately attached to the mitotic spindle. To elicit such responses, checkpoint pathways must impinge upon cell-cycle-regulatory mechanisms (Elledge 1996). Furthermore, they must allow the cell time to correct the problem yet retain the capacity to proceed once that problem is corrected.
The observation that the cell-cycle-regulated transcriptional program can proceed (albeit less robustly) when CDK activity is lost suggests that checkpoint arrest, which blocks CDK cycling, might allow the transcription program to be decoupled from cell-cycle events. This problem would be resolved if checkpoint pathways directly regulate the transcription network to maintain synchrony between transcription and cell-cycle progression. Some evidence suggests that checkpoints do, in fact, regulate transcription to enforce robust cell-cycle arrest (Gardner et al. 1999; Chu et al. 2007). Several global transcript-profiling studies indicate that the DNA replication and DNA damage checkpoint pathways trigger a checkpoint-dependent transcription program controlled by Dun1 kinase (Figure 5) (reviewed in Fu et al. 2008). The Dun1 kinase induces a well-described transcriptional response involving genes (such as RNR2 and RNR3) that mediate recovery from DNA damage or blocks to DNA replication (Elledge et al. 1993; Zhou and Elledge 1993; Allen et al. 1994; Gasch et al. 2001). Dun1 activity is stimulated by Rad53. Crt1, a repressor of DNA damage response genes, is phosphorylated by Dun1 upon checkpoint activation and released from the DNA, thereby, allowing transcriptional activators access to promoter regions of Crt1-target genes (Huang et al. 1998). In addition to its role in recovery from checkpoint stresses, genetic evidence suggests that Dun1 also plays a role in enforcing a robust checkpoint arrest (Gardner et al. 1999). Although some of these studies hint at regulation of cell-cycle genes (Gasch et al. 2001), a specific role for Dun1-regulated transcription in cell-cycle arrest has yet to be described.
In addition to activating the Dun1 pathway, Rad53 also directly phosphorylates Nrm1, the repressor of MBF-regulated transcription, preventing its association with MBF at target promoters (Figure 5) (de Oliveira et al. 2012; Travesa et al. 2012). Consequently, MBF targets, which would normally be repressed when DNA replication is underway, remain active, leading to the accumulation of their products. Because many of those targets are required for DNA replication and repair, this positions cells to correct the problems leading to induction of the checkpoint and to progress through S phase once that problem has been eliminated. A similar mechanism is observed in fission yeast where the Rad53 ortholog, Cds1, phosphorylates both Nrm1 and Cdc10, the S. pombe ortholog of Swi6. The result is persistent activation of transcription by MBF, the sole G1/S transcription factor in that organism (de Bruin et al. 2008b; Dutta et al. 2008; Aligianni et al. 2009; Gómez-Escoda et al. 2011). Finally, Rad53 also phosphorylates Swi6 in S. cerevisiae (Sidorova and Breeden 1997). Although the importance of that phosphorylation has not been established, it has been suggested that it delays expression of CLN1 and CLN2 in G1 cells responding to DNA damage.
The phosphorylation of Nrm1 by Rad53 specifically affects MBF, allowing SBF to be repressed, like it is in untreated cells (Travesa et al. 2012). Consequently, repression of SBF targets by Clb2/CDK occurs normally because Clb2 is stabilized in response to down-regulation of APCCdc20 activity by the checkpoint. Thus, inappropriate accumulation transcripts associated with morphogenesis and other processes promoted by targets of SBF is prevented. The differential regulation of the SBF and MBF transcription factors, and thereby their specific targets by the DNA replication checkpoint, provides one clear rationale for the existence of independent transcription factors regulating the two groups of G1/S cluster genes (de Oliveira et al. 2012; Smolka et al. 2012; Travesa et al. 2012). This dual regulation is dispensable in fission yeast where MBF, the sole G1/S transcription factor, is also subject to checkpoint regulation via phosphorylation of the transcriptional repressor Nrm1 (de Bruin et al. 2006, 2008b; Dutta et al. 2008). In that system, proliferation does not occur via budding and, consequently, morphogenesis is not regulated during G1 phase.
In addition to providing a mechanism by which the checkpoint can activate only a subset of G1/S genes, the differential regulation of SBF and MBF by the DNA replication checkpoint also poses a conundrum. SBF- and MBF-regulated genes can be repressed by the accumulation of Clb/CDK activity during the unperturbed cell cycle (Amon et al. 1993; de Bruin et al. 2006). Why, then, is SBF repressed normally during S phase in cells responding to the checkpoint signal, whereas MBF remains unaffected? This observation suggests that a mechanism invoked by the checkpoint signal allows MBF to evade the repressive activity of Clb/CDK. Although the mechanism of that evasion is unknown, it is possible that the checkpoint-induced phosphorylation of Swi6 by Rad53 makes MBF refractile to inhibition by Clb/CDK (Sidorova and Breeden 1997). Whatever the mechanism, this difference in their response to Clb/CDK provides another example of differential regulation of SBF and MBF during the cell cycle.
Although, the differential regulation of SBF and MBF by the DNA replication checkpoint provides a seemingly straightforward rationale for the involvement of two transcription factors in G1/S transcription, several observations suggest that the situation is really much more complex. First, many G1/S genes have multiple binding sites for both transcription factors that appear to be functional in vivo (Iyer et al. 2001; Horak et al. 2002; Bean et al. 2005; de Bruin et al. 2006). Second, the expression of many G1/S genes is impacted by inactivation of either of the two transcription factors, suggesting that both factors can contribute to their regulation (Bean et al. 2005; de Bruin et al. 2006). Finally, some genes undergo a switch from binding SBF to binding MBF during the course of the G1/S transcriptional burst as a consequence of the presence overlapping binding sites for the two transcription factors (de Oliveira et al. 2012). Those genes seem to be activated by SBF, dependent upon inactivation of Whi5, and repressed by MBF, via the activity of Nrm1. One rationale that has been proposed for this regulatory switch is to enable inducibility of those genes by the DNA replication checkpoint and to ensure that the genes are not expressed inappropriately if MBF is inactivated by mutation or other insult (reviewed by Smolka et al. 2012). Regardless of the reason, it is clear that the regulation of G1/S genes by the combined activities of SBF and MBF provides opportunities for more diverse modes of regulation than does the involvement of either transcription factor on its own.
We are just beginning to understand the complex mechanisms by which checkpoint effector pathways control cell-cycle-regulated transcription of the G1/S cluster. It also seems likely that checkpoints may regulate other clusters of periodic genes. Thus, it is likely that checkpoint control of cell-cycle-regulated transcription will be a rich area of future investigation.
Yeast metabolic cycle control of cell cycle and transcription
It has been recognized for some time that metabolic processes in yeast are periodically regulated (Futcher 2006 and references within) and may be coordinated with the cell cycle (e.g., Creanor 1978; Novak and Mitchison 1986, 1990). When budding yeast cells are grown in continuous culture conditions at appropriate densities and growth rates, the cells synchronize in robust metabolic cycles that can be monitored by periodic changes in dissolved oxygen. The reported periods of the YMC can vary substantially (Klevecz et al. 2004; Tu et al. 2005; Slavov and Botstein 2011), and seem to be linked to growth rate (Brauer et al. 2008; Slavov and Botstein 2011). The YMC regulates a large transcriptional program and appears to be coordinated with the cell cycle under slow growth conditions.
The YMC-regulated transcriptional program:
By sampling continuous cultures of metabolically synchronous populations of budding yeast over time, investigators have identified oscillations in gene expression that are coincident with the periodicity of the YMC. In one of the initial reports, cells were sampled over the course of three metabolic cycles and by microarray analyses found that the bulk of the genome exhibited oscillations in transcript abundance (Klevecz et al. 2004). The amplitudes of these oscillations were modest, with peak-to-trough ratios of ∼2. In a subsequent study, the McKnight group (Tu et al. 2005) also sampled metabolically synchronous cells growing in a chemostat over three metabolic cycles. Their findings suggested approximately half of the genes in the genome exhibited transcript oscillations coincident with the YMC. The amplitudes of these transcript oscillations were in the range of 10-fold (Tu et al. 2005), substantially higher than observed by Klevecz and colleagues.
Using clustering approaches, both groups were able to identify sets of transcripts that appeared to be coregulated, and many of the coregulated genes shared related functions (Klevecz et al. 2004; Tu et al. 2005). Analysis of the data from Tu and colleagues revealed three “super clusters” of genes that were expressed in relatively discrete temporal intervals that correspond to three phases of the YMC: oxidative (Ox), reductive building (R/B), and reductive charging (R/C) (Tu et al. 2005). Within these clusters of oscillating transcripts were several cell-cycle-regulated genes, suggesting that the YMC transcriptional program and the cell-cycle transcriptional program are interconnected (Klevecz et al. 2004; Tu et al. 2005).
Botstein and colleagues approached YMC control of transcription from a different angle. They used continuous culture conditions to investigate the transcriptional response of yeast to a wide variety of growth limiting conditions in which cells were starved for a variety of different nutrients (Brauer et al. 2008). Under this broad range of growth conditions, they found that the expression of approximately one-fourth of the genes in the genome is linearly correlated (either positively or negatively) with growth rate. Surprisingly, this set of growth rate responsive (GRR) genes exhibited substantial overlap with a set of genes previously identified as environmental stress response (ESR) genes (Gasch et al. 2000), suggesting a functional relationship between the gene sets. Genes that exhibited positive correlation with growth rate tended to be genes inhibited by stress, while genes that were down-regulated in fast growth conditions were often activated by stress (Brauer et al. 2008). As well, there was considerable overlap between the GRR gene set and the YMC responsive gene sets identified by Tu et al. (2005) (Brauer et al. 2008). In a recent study using continuous culture conditions and a variety of nutritional limitations, Slavov and Botstein (2011) demonstrated that all of the GRR transcripts were, in fact, periodic during the YMC. Strikingly, negatively and positively correlated GRR transcripts oscillate in opposing YMC phases. Taken together, these findings suggest that the mechanisms controlling oscillation in transcript abundance may be integrating signals from stress, growth rate, YMC, and cell cycle.
YMC control of cell cycle:
The identification of cell-cycle-regulated genes as part of the YMC transcriptional program suggested that these genes were regulated by multiple conditions or that the YMC and the cell cycle may be coordinated (Klevecz et al. 2004; Tu et al. 2005). DNA content and budding analyses both suggest that the YMC and the cell cycle are indeed synchronized in continuous culture conditions (Klevecz et al. 2004; Tu et al. 2005). Both groups suggest that only a fraction of cells in the population [Tu et al. (2005) estimates 50%] enter the cell cycle during each metabolic cycle. In addition, it appeared that the YMC phases might be coherent with cell-cycle phases, with S phase being restricted from occurring during the Ox phase of the YMC (Klevecz et al. 2004; Tu et al. 2005).
Both groups proposed that the YMC gates the cell cycle so that the execution of S phase is restricted to the reductive phase and that this restriction is important for preventing oxidative damage of DNA. Supporting evidence for this hypothesis was reported in a subsequent study from the McKnight group (Chen et al. 2007). In mutant cells where S phase was allowed to proceed during the Ox phase of the YMC, they observed an increase in the spontaneous mutation rates. However, the mutant genes found to disrupt the phase coherence of the YMC and the cell cycle were themselves genes controlling the cell division cycle (Chen et al. 2007), suggesting the increase in mutation rate observed in these mutants could be related to improper execution of cell-cycle functions rather than replicating DNA during the oxidative phase of YMC.
In a recent study, the connection between the cell cycle and the YMC was directly tested (Slavov and Botstein 2011). By varying nutrient availability, and thereby altering the growth rate of continuous cultures, it was demonstrated that the restriction of S phase from the Ox phase of the YMC was dependent on the growth rate. Under specific growth rate conditions, synchrony between the cell cycle and YMC could be maintained, but execution of S phase occurred during the Ox phase of the YMC (Slavov and Botstein 2011). Interestingly, the cell cycle mutants that disrupted the coordination between S phase with the reductive phases of the YMC were also the mutants that exhibited the most substantial alterations in the growth rate and YMC period (Chen et al. 2007).
One interpretation of these data is that the duration of the cell division cycle influences the YMC period. Consistent with this possibility, it has been proposed that the coherence of cell-cycle phases with YMC phases may reflect the storage of carbohydrates under slow-growth conditions and a sudden “burn” of these carbohydrates during G1 phase (Futcher 2006). This rapid utilization of carbohydrates provides energy and drives rapid protein synthetic rates, providing a “finishing kick” for cells to pass through START and enter S phase (Futcher 2006). This coupling of cell-cycle and YMC programs seems to be restricted to conditions of slow growth, and oscillation of YMC or GRR transcripts has not been observed in microarray studies performed using rapidly growing cell populations synchronized in the cell cycle (Cho et al. 1998; Spellman et al. 1998; Pramila et al. 2006; Brauer et al. 2008; Orlando et al. 2008). This may be because the rapidly growing cells have abundant glucose and do not store substantial amounts of carbohydrates.
Many open questions remain regarding how and why budding yeast cells coordinate the YMC with the cell-cycle program. The interaction between the YMC and the cell cycle is likely to be more complex than a simple gating of the cell cycle by a metabolic oscillator. Although the mechanisms driving the YMC remain undiscovered, like the cell cycle, it is clearly capable of producing oscillations in an abundance of transcripts on a genome-level scale. The proposal that the cell-cycle period may be controlled by a transcriptional network oscillator suggests that the YMC and the cell cycle may be coupled via coordinated control of key transcriptional regulators (Orlando et al. 2008; Simmons Kovacs et al. 2008, 2012).
What is the role of chromatin in cell-cycle-regulated transcription?
Transcription takes place in the context of chromatin and, as a consequence, is influenced by it (Lenstra et al. 2011; Rando and Winston 2012). Cell-cycle-regulated genes have proven useful for understanding both general, and gene-specific contributions of chromatin to the regulation of expression. The specific roles of chromatin in the control of cell-cycle-regulated gene expression and the relationship between chromatin and the transcription factors that control the cell-cycle gene clusters are only beginning to be understood. In many cases, the specific vs. general roles of chromatin in the control of cell-cycle-regulated gene expression have yet to be established.
The earliest studies of the contribution of chromatin modification and remodeling to gene expression and, more specifically, cell-cycle-regulated gene expression were based, in large part, upon the pioneering studies of the G1/S gene, HO, by the Nasmyth and Herskowitz laboratories (reviewed by Stillman 2013; see Haber 2012). Most of the studies of the role of chromatin in the regulation of HO expression focused upon licensing for differential expression in mother cells, mediated by an upstream regulatory region known as URS1. However, the balance of the regulatory functions, mediated by URS2, are largely indistinguishable from expression of other SBF target genes (Peterson and Herskowitz 1992; Cosma et al. 1999, 2001). Consequently, most of the observations regarding URS2 can be generalized to CLN2 and other G1/S promoters.
Recent study of the CLN2 and HO promoters has focused, in part, upon the role for histone deacetylation by the Rpd3(L) HDAC complex in repression of SBF and MBF genes (de Bruin et al. 2008a; Takahata et al. 2009a,b, 2011; Wang et al. 2009). Rpd3(L) recruitment to SBF promoters depends upon the binding of the Sin3-associated protein Stb1 to promoters in a manner that depends upon the SBF-specific transcriptional repressor Whi5 (Wang et al. 2009; Takahata et al. 2011). At MBF genes, Stb1 binding occurs, but is, by necessity, Whi5 independent (Costanzo et al. 2003; de Bruin et al. 2008a; Takahata et al. 2009a). Stb1 binding is accompanied by the binding of Cdk1, which, via some combination of phosphorylation of Whi5, Stb1, and the transcription factors, promotes activation of the promoters. Cln/Cdk1-dependent phosphorylation of Whi5 and Stb1 leads to the dissociation and export of Whi5 from the nucleus and the dissociation of Rpd3(L) from promoters. Promoter activation occurs upon the loss of Rpd3(L) binding, presumably accompanied by the restoration of histone acetylation. Similar changes in histone acetylation during G1 and S phases are observed at both SBF and MBF target promoters. Although Rpd3(L) is an important factor in the regulation of histone acetylation state and involved in the repression of many genes, Stb1 may affect only a small subset of genes, including those in the G1/S gene cluster.
Activation of both CLN2 and HO expression, and likely the entire G1/S gene cluster, also involves the FACT complex, which promotes nucleosome eviction at those promoters (Takahata et al. 2009a,b; Wang et al. 2009). FACT acts just prior to, or at the time of, transcriptional activation and is required for G1/S expression. The binding of FACT to both CLN2 and HO promoters depends upon the Swi6 subunit of SBF along with Whi5 and Stb1. The G1/S genes provide the only example of a promoter-associated role for FACT, which is most often associated with transcriptional elongation (Rando and Winston 2012). FACT has also been implicated in CLN3 transcription, although whether that occurs via a direct or indirect mechanism remains obscure (Morillo-Huesca et al. 2010).
At least some of these factors play well-established roles at genes that are expressed independent of cell-cycle position, but the extent to which these transcription-factor-dependent mechanisms for chromatin modification and remodeling play specific roles at cell-cycle genes is not yet clear. However, those factors are clearly distinct from the more general role of nucleosome-depleted regions (NDRs) that encompass the SBF-binding sites at the CLN2 promoter and are required for the reliability of cell-cycle-dependent transcriptional activation (Bai et al. 2010, 2011). Consistent with that, the formation of NDRs at the CLN2 promoter does not require any of the known G1/S-specific transcriptional regulators. Additionally, histone H3 K79 methylation has also been associated with the function of SBF, Whi5, and Nrm1, although the relevance of that association to the regulation of G1/S promoters is not clear (Schulze et al. 2009).
There are many studies suggesting a role for chromatin in the regulation of other cell-cycle-regulated gene clusters, although most are less well studied. For example, Fkh2/Mcm1 promotes recruitment of Rpd3(L) leading to repression of G2/M cluster genes (Veis et al. 2007). Interestingly, in a two-step process, Rpd3(L) dissociates from G2/M gene promoters at the onset of S phase in a manner that depends upon Cln/CDK and, only then, does the combined activation of Clb2/CDK and Cdc5 promote eviction of a nucleosome from those promoters and binding of Ndd1 to Fkh2. Recent studies describing a role for histone chaperones and chromatin boundary proteins in chromatin remodeling at the histone gene cluster are reviewed elsewhere (Eriksson et al. 2012). Together, these studies suggest both direct and indirect roles for cluster-specific transcription factors in the recruitment and regulation of chromatin modifying and remodeling proteins. Nevertheless, the degree of uncertainty regarding the specificity of these processes makes this a fertile ground for future research.
Rationale for Cell-Cycle-Regulated Transcription: Need to Have or Nice to Have?
To understand the rationale for regulating the expression of genes during the cell cycle, it is helpful to examine which genes are controlled. Deciding whether or not a gene is cell-cycle-regulated is not as trivial as it might seem. Many definitions of “cell-cycle regulated” can and have been applied, yielding a variety of estimates of the number and identities of cell-cycle-regulated genes in S. cerevisiae (Cho et al. 1998; Spellman et al. 1998; de Lichtenberg et al. 2005; Pramila et al. 2006; Orlando et al. 2008; Granovskaia et al. 2010; Guo et al. 2013). The number of transcripts reported as cell-cycle regulated has varied from 416 to 1270. A consensus set of 440 genes (Orlando et al. 2008) was identified by each of three distinct studies (Spellman et al. 1998; Pramila et al. 2006; Orlando et al. 2008). The remaining genes, which are unique to each of the studies, could result from differences in strain backgrounds or experimental conditions including synchrony procedures, growth medium, microarray platform, or the approach used to identify periodic genes.
So which genes are cell-cycle regulated? The best answer is: There is no single, identifiable set of cell-cycle-regulated genes. Because there is no precise and universally accepted definition of cell-cycle regulated, there is no single algorithmic method available for identifying genes with periodic behaviors over the cell cycle. In addition, each of the algorithms returns a rank-ordered list of genes with no obvious gap in the distribution that distinguishes periodic from nonperiodic genes. Thus, all cut-offs chosen by investigators are somewhat arbitrary, and it is likely more accurate to say that there is a continuum of cell-cycle-regulated genes.
Despite the difficulty in precisely defining a single set of cell-cycle-regulated genes, it is likely that ∼20% of genes in the yeast genome are expressed periodically during the cell cycle at a substantial energy cost to the cell. This observation suggests that coordination of gene expression with cell-cycle events is important for the well being of cells. Several distinct though potentially overlapping rationales have been advanced as reasons for phase-specific expression of various genes.
First, some genes are expressed at the time they are required during the cell cycle. This is clearly true for a large number of genes whose function is required for, or is related to, morphogenesis, DNA replication, and chromosome segregation. The logic for this pattern of expression is referred to as “just in time.” Whether genes are actually expressed just in time is not always so clear. For example, because cytokinesis occurs subsequent to M phase, one might expect the expression of genes encoding the structural elements of the bud neck, including septins, to occur in M phase rather than G1. However, those genes are, in fact, expressed just in time because the structures at the bud neck, which play roles from G1 through M phases, are assembled during G1 phase prior to budding, precisely when G1/S gene expression is maximal (reviewed in Bi and Park 2012; Howell and Lew 2012). Similar scenarios may explain the periodic expression of some of the genes for which no rationale is currently apparent.
Many genes that fall into the just-in-time category, despite being expressed at the time of their essential function, need not be expressed precisely at that time and, in some cases, need not be expressed periodically at all. In fact, several genome-wide studies have revealed the vast majority of yeast genes cause no significant growth disadvantage, at least when constitutively overexpressed in cycling cells from the inducible GAL1 promoter (Liu et al. 1992; Sopko et al. 2006; Douglas et al. 2012).
In addition to those genes that are expressed just in time, there are a large number of genes expressed at a specific cell-cycle interval during which they have no apparent function. In some cases, the apparent lack of phase-specific function may be a consequence of our lack of knowledge, but in other cases the genes may just “tag along” with other genes expressed during the same phase. That is, the pattern of expression may be adequate but not necessarily advantageous. However, some of those genes might be better described as just in case, because their expression may prepare the cell to deal with a sporadic event that does not occur in all cell cycles. For example, whereas many DNA repair proteins are expressed prior to every S phase, damage may occur only in a subset of cells during any round of DNA replication.
Finally, some genes that are expressed just in time are deleterious when expressed outside of the interval during which they function. Those genes often encode proteins with regulatory functions that are temporally coupled to their accumulation. For example, the genes encoding the B-type cyclins Clb2 and Clb5 and others encoding the anaphase regulators Cdc20 and Pds1 perturb cell-cycle progression when they are expressed continuously throughout the cycle, with the caveat that the studies in question involve overexpression (Nasmyth 1993; Sopko et al. 2006). However, even in the context of overexpression, such genes are surprisingly rare.
So, why express a gene just in time when that pattern of expression is unnecessary? One possibility is that just-in-time expression is simply a matter of economy. Often, proteins expressed outside of the interval during which they function are disposed of by proteolysis or other mechanisms (e.g., those encoding the G1 cyclins Cln1–3 and the CDK inhibitor Sic1) (Lanker et al. 1996; Willems et al. 1996; Verma et al. 1997). Consequently, there is a large energy cost to the cells of expressing those genes outside of their functional interval. Although expressing those proteins throughout the cycle may not produce a noticeable or debilitating phenotype, it may result in a slight competitive disadvantage under optimal or suboptimal conditions. Consequently, such changes would not be noticed in typical laboratory growth conditions. A definitive answer to these questions is lacking because, except for studies involving overexpression, the impact of continuous expression of cell-cycle-regulated genes on overall fitness has not been systematically investigated.
The Topology of Cell-Cycle-Regulated Transcriptional Circuitry is Conserved
One of the overarching questions arising in the study of model systems is the degree to which their regulatory circuitry is conserved with humans. Comparative analysis of cell-cycle transcriptional programs across the Eukaryota has provided one window through which to examine that question. The general organization and structure of the cell-cycle apparatus is conserved among the organisms so far examined. Finally, all of those organisms express a portion of their genes in a cell-cycle-dependent manner. Consequently, it has been anticipated that the conservation of the cell-cycle-dependent transcriptional machinery will be substantial. This supposition has been met with both confirmation and contradiction.
The cell-cycle transcription program has now been examined on a genomic scale in plants, a variety of yeasts including budding and fission yeast, and in humans (Spellman et al. 1998; Whitfield et al. 2002; Menges et al. 2003; Bar-Joseph et al. 2004; Rustici et al. 2004; Pramila et al. 2006; Gauthier et al. 2008). Although it is often difficult to assess in light of the evolution of both sequence and function, comparative analysis of cell-cycle-regulated gene expression has led to the conclusion that the ontology of genes under cell-cycle control is often conserved (CDK regulators, chromatin proteins, transcription factors, DNA replication and repair proteins, mitotic proteins, etc.), but the expression pattern of orthologs is not always conserved (Lu et al. 2007; Orlando et al. 2007). The degree of divergence seems to increase with the evolutionary distance between the species and is even obvious in closely related Saccharomyces species, cerevisiae and bayanus (Guan et al. 2010, 2013; Eser et al. 2011). This seems surprising, considering the similarities in cell-cycle organization between those organisms. However, the observation that the timing of expression of cell-cycle-regulated genes in yeast is flexible, and, in some cases, completely dispensable (see above) suggests that such changes in expression might be easily accommodated. Of course, the differences in periodicity at the transcription level do not account for the loss or acquisition of the broad array of post-transcriptional mechanisms that might restore periodic regulation of many of those gene products. In fact, differences in the periodicity of transcription that have coevolved with changes in post-transcriptional regulation have been documented (de Lichtenberg et al. 2005; Jensen et al. 2006). Another indication of the evolutionary flexibility of cell-cycle-regulatory phenomenon is the finding that within protein complexes where one or more subunits is periodically expressed, the identity of the periodically expressed subunits may differ between species (Jensen et al. 2006). In fact, it appears that just-in-time assembly may be important and conserved, whereas just-in-time expression may be dispensable (de Lichtenberg et al. 2005). This notion is consistent with the finding that within closely related yeast species, there is extensive repurposing of conserved transcription factor subunits to form novel complexes with other factors and produce entirely new transcriptional outcomes (Tuch et al. 2008).
Like cell-cycle-regulated transcripts, comparison of the regulators of periodic gene expression reveals strong conservation of the general topology of the systems and dramatic divergence of the details. This is well illustrated by analysis of the transcriptional machinery regulating the G1/S gene clusters of fungi and mammals (reviewed by Wittenberg and Reed 2005; Cross et al. 2011; Bertoli et al. 2013b). The activating members of the E2F family of transcriptional regulators, E2F1 being the best studied, exhibit many parallels with SBF (Figure 6) (Dimova and Dyson 2005). E2F1/DP1 binds to promoters in quiescent cells in complex with a transcription repressor, in this case Rb, the retinoblastoma tumor suppressor protein, a functional homolog of Whi5. Upon entry into the cell cycle, Rb is phosphorylated by a G1 cyclin/CDK, in this case CycD/Cdk4/6, functional analogs of Cln3/Cdk1, leading to activation of the E2F1/DP1 targets. Finally, the G1/S genes are repressed by the binding of the E2F6, a repressive form of E2F, as cells progress into S phase, leading to the dissociation of E2F1/DP1 from promoters (Bertoli et al. 2013a). This precisely parallels the regulation of SBF in budding yeast. However, examination of E2F1, DP1, and Rb, reveals no recognizable protein sequence conservation, or even structural conservation, with Swi4, Swi6, or Whi5 (Cross et al. 2011).
Whether this is an example of rapid evolution between fungi and humans or convergent evolution directed at a redundant function remains to be established. However, it is clear that the E2F/Rb family and not the SBF/Whi5 family of regulators exists in the plant kingdom, which is evolutionarily distant from both fungi and mammals (Cross et al. 2011). Furthermore, despite the lack of relationship between the transcriptional regulators, all of these organisms use a conserved, albeit diversified, family of CDK kinases to coordinate gene expression with the cell cycle (Enserink and Kolodner 2010).
In contrast to the lack of conservation of SBF, E2F, and their regulators, other cell-cycle-associated transcriptional regulators from yeast exhibit strong sequence and structural conservation with other eukaryotes. For example, both the MADS box transcription factor, Mcm1, and Forkhead transcription factors, Fkh1 and Fkh2, are relatively well conserved among eukaryotes. However, at least in the case of the MADS box transcription factors, their role in cell-cycle regulation appears not to be conserved. SRF is a well-established regulator of growth-associated genes, but not those expressed periodically during the cell cycle (Shore and Sharrocks 1995). In plants, MADS box transcription factors are prominently associated with developmentally regulated gene expression (reviewed by Masiero et al. 2011). Interestingly, although yeast Mcm1 plays a prominent role in the regulation of G2/M and M/G1 regulated genes, it is also involved in a number of functions that are not related to the cell cycle.
In contrast, the human Forkhead protein, FoxM1, shares many regulatory properties with Fkh2. That regulation appears not to involve either a MADS box transcription factor or an Ndd1 ortholog, as it does in yeast. However, like in yeast, the activity of FoxM1 is required for the proper expression of a number of mitotic genes during G2/M and, like the Mcm1/Fkh2/Ndd1 complex, it is subject to regulation by both Cdk1 and the human Polo kinase, Plk1 (reviewed by Murakami et al. 2010).
Together these observations suggest a complex evolutionary relationship between the factors mediating cell-cycle-regulated gene expression and their targets. Perhaps unsurprisingly, it is the final product, the overall execution of cell-cycle events, that is best conserved rather than the details of the regulatory pathway that controls that outcome.
The authors thank Anna Travesa, Danny Lew, Nick Buchler, Sara Bristow, Adam Leman, Christina Kelliher, Mark Chee, and anonymous reviewers for helpful discussion and comments on the manuscript. S.B.H. is supported by grants from the Defense Advanced Research Projects Agency and the National Institutes of Health (NIH) P50-GM081883. C.W. is supported in part by grants R01 GM059441 and R01 GM100354 from the National Institutes of Health.
Communicating editor: M. Tyers
- Received May 9, 2013.
- Accepted September 16, 2013.
- Copyright © 2014 by the Genetics Society of America