|Home | About | Journals | Submit | Contact Us | Français|
Cells devote considerable resources to nutrient homeostasis, involving nutrient surveillance, acquisition, and storage at physiologically relevant concentrations. Many Saccharomyces cerevisiae transcripts coding for proteins with nutrient uptake functions exhibit peak periodic accumulation during M phase, indicating that an important aspect of nutrient homeostasis involves transcriptional regulation. Inorganic phosphate is a central macronutrient that we have previously shown oscillates inversely with mitotic activation of PHO5. The mechanism of this periodic cell cycle expression remains unknown. To date, only two sequence-specific activators, Pho4 and Pho2, were known to induce PHO5 transcription. We provide here evidence that Mcm1, a MADS-box protein, is essential for PHO5 mitotic activation. In addition, we found that cells simultaneously lacking the forkhead proteins, Fkh1 and Fkh2, exhibited a 2.5-fold decrease in PHO5 expression. The Mcm1-Fkh2 complex, first shown to transactivate genes within the CLB2 cluster that drive G2/M progression, also associated directly at the PHO5 promoter in a cell cycle-dependent manner in chromatin immunoprecipitation assays. Sds3, a component specific to the Rpd3L histone deacetylase complex, was also recruited to PHO5 in G1. These findings provide (i) further mechanistic insight into PHO5 mitotic activation, (ii) demonstrate that Mcm1-Fkh2 can function combinatorially with other activators to yield late M/G1 induction, and (iii) couple the mitotic cell cycle progression machinery to cellular phosphate homeostasis.
Cellular growth and division are controlled by the temporal execution of programmed events that drive cell cycle progression. Several mechanisms that regulate the cell division cycle of S. cerevisiae are orchestrated by the cyclin-dependent kinase (CDK) Cdc28. While in large part this regulation occurs at the posttranscriptional level through targeted protein degradation (65), an important aspect of cell cycle regulation is also mediated at the level of transcription. More than 13% of S. cerevisiae genes are expressed in a cell cycle stage-specific fashion predominantly, yet not exclusively, via the action of one of three distinct classes of sequence-specific DNA-binding factors (16, 68, 81). These include SBF/MBF, Ace2/Swi5, and Mcm1, which respectively regulate G1/S-, M-, and M/G1-dependent transcription (81). These stage-specific roles are an approximation since some overlap occurs, most notably for Mcm1, which exhibits important functions throughout the cell cycle (23, 53).
Budding yeast Mcm1, along with Agamous and Deficiens in plants and mammalian serum response factor, is a founding member of a family of proteins containing the highly conserved 56-amino-acid MADS box (58, 63, 80, 95). Mcm1 is an essential gene product with diverse cellular roles in minichromosome maintenance, from which its name is derived, as well as cell cycle control, cell type determination, mating, arginine metabolism, and stress tolerance (14, 54, 78). Eighty amino acids near the N terminus of the 286-residue Mcm1 protein constitute the core fragment, which is sufficient for cell viability, minichromosome maintenance, and cell-type-specific transcription (17). This Mcm1 core fragment encompasses the MADS box, the N-terminal half of which makes sequence-specific contacts with DNA as a homodimeric binding complex. In addition, the C-terminal half of the MADS box plus an approximately 30-amino-acid extension specific to various MADS subfamilies, mediate heterotypic interactions with other DNA-binding factors, such as Arg80-Arg81, MATα1, MATα2, Yhp1, and Yox1 (54, 78).
In the cell cycle, Mcm1 functions with another factor(s) that occupies a closely apposed site (1, 48) to activate ~35 genes in G2/M (81). These genes are assigned to the CLB2 cluster (81) due to the prominent role of Clb2, a B-type cyclin, in their induction that includes positive-feedback control on CLB2 transcription by Clb2-Cdc28 (2). Subsequent studies elucidated that either of two distinct factors, Fkh1 or Fkh2, bind adjacent to Mcm1 to induce promoters of CLB2 cluster genes (9, 28, 34, 37, 67, 96). Fkh DNA-binding domains are highly homologous to those of the forkhead or winged-helix proteins in higher eukaryotes (11). Yeast cells lacking either Fkh1 or Fkh2 exhibit partial defects in the periodicity of mitotically induced genes, suggesting overlapping functions (28, 37, 67, 96). Interestingly, while a monomer of either Fkh1 or Fkh2 can bind its site in vitro, only Fkh2 efficiently binds the CLB2 promoter in vivo (29, 34, 37, 67, 92, 96). This is explained, at least in part, by the cooperative binding of Fkh2, but not Fkh1, with Mcm1 at promoters containing the bipartite Mcm1-Fkh site (29). The region that mediates direct interaction between Mcm1 and Fkh2, which is absent in Fkh1, is located N terminal to the Fkh2 winged-helix DNA-binding domain (9, 19). During G2/M, transcriptional activation by Mcm1-Fkh2 requires temporal recruitment of Ndd1, a coactivator that does not bind to DNA (19, 34, 66). Stable recruitment of Ndd1 to target genes is mediated by the forkhead-associated domain of Fkh2 that requires phosphorylation by Clb2-Cdc28 and the polo kinase Cdc5, whose gene is also a CLB2 cluster member (18, 19, 66, 70). This phosphorylation-dependent recruitment of Ndd1 is likely an underlying molecular event in the activation of G2/M-specific promoters upon CLB2 expression (81).
In contrast to our knowledge about CLB2 cluster gene regulation, relatively little is understood about activation of genes in the MCM cluster, which peak in late mitosis near the M/G1 boundary (81). The majority of MCM cluster genes contain Mcm1 binding sites of various quality, and only a subset of these Mcm1 sites lie adjacent to Fkh sites (81). The late-M-phase transcription of some of the genes in this cluster has been suggested to arise from a variant Mcm1 binding site, the early cell cycle box (53). Prominent MCM cluster members include genes under the control of the phosphate signaling (PHO) pathway, PHO5, PHO3, PHO11 and PHO12 (16, 81). Periodic expression was surprising, as PHO genes are induced by depletion of environmental phosphate (32, 43), and the rich medium used was thought to contain high phosphate (25). In addition, no previous study has shown direct binding of Mcm1 to the promoter of any PHO gene. With respect to the forkhead proteins, just Fkh2 was shown to bind to PHO5 and other genes regulated by the PHO pathway (see Fig. Fig.11),11), and only under conditions of severe oxidative stress (26, 49).
We have previously shown that mitotic induction of PHO5 occurs when inorganic phosphate (Pi) is at limiting concentrations in growth media (e.g., rich yeast-peptone-dextrose [YPD] medium) (57). PHO5 expression in YPD-grown cells is partially activated, expressing ~5 to 10% as much repressible acid phosphatase (rAPase) activity as is observed after overnight incubation in Pi-free medium (57). Populations of yeast cells growing synchronously in limiting Pi establish a four-stage oscillatory cycle of Pi starvation and replenishment (57). First, in G1-arrested cells where Pi uptake exceeds metabolic demands, excess Pi accumulates in the vacuole in the form of polyphosphate (polyP), a linear-chain phosphate polymer that buffers intracellular Pi concentration in yeast (35, 57, 84). Given this high cellular phosphate content, PHO genes are fully repressed because Pho4, a downstream DNA-binding activator (42), is phosphorylated by the Pho80-Pho85 cyclin-CDK and is exported to the cytoplasm (30, 33, 61). In high-Pi environments, residual phospho-Pho4 in the nucleus is unable to interact with its homeodomain-containing coactivator, Pho2 (33). Second, cells mobilize and exhaust vacuolar polyP reserves as they traverse S phase, presumably to meet higher cellular demands for Pi (57, 84). Third, intracellular Pi, which is not readily replenished by membrane-bound transporters under conditions of limiting external Pi, also declines. This leads to inactivation of Pho80-Pho85 by Pho81, an upstream-acting CDK inhibitor (CKI), which leads to increased nuclear retention of Pho4 (75). In the fourth and final stage, PHO genes encoding proteins with Pi-scavenging and -storage functions are induced, replenishing cellular levels of intracellular Pi and vacuolar polyP.
Together, these results suggested that peak M-phase expression of PHO5 is driven primarily in response to cell cycle-dependent fluctuations in Pi. In support of this model, single deletions of PHO4 and PHO2, as well as the addition of exogenous Pi, eliminated PHO5 mitotic induction (57). Loss of Pho81, the upstream CKI of Pho80-Pho85, also severely impaired mitotic induction of PHO5 (57). We noted, however, that pho81Δ cells retained detectable levels of PHO5 mitotic cycling, suggesting at least one additional downstream regulatory input (57). Consistent with this notion, PHO5 expression was strongly induced by overexpression of CLB2 in cells arrested at M phase (81), implicating a cell cycle-dependent event(s) as the downstream input. It is unclear whether this effect of CLB2 overexpression was direct and, if so, how mitotic expression of PHO5 is controlled both through the PHO pathway via fluctuations in Pi level and activity of the master CDK Cdc28.
We have examined the role, in PHO5 mitotic induction, of sequence-specific cell cycle-dependent transactivators first shown to be involved in the cell cycle transition from G2 to M. We show an essential role for the MADS-box factor, Mcm1, and a partial requirement for the forkhead proteins, Fkh1 and Fkh2, in PHO5 mitotic expression. Induction by Mcm1 and Fkh proteins is direct as point mutations in a consensus Mcm1-Fkh site in the PHO5 promoter diminished mitotic expression. In addition, Mcm1-Fkh2 and, to a lesser extent, Fkh1, were found to associate directly with the PHO5 promoter by chromatin immunoprecipitation (ChIP) at specific cell cycle stages. These results elucidate a novel pathway in which Mcm1 and either of the forkhead proteins, Fkh1 or Fkh2, work in concert with Pho4 and Pho2 to establish peak expression of PHO5 in M/G1.
S. cerevisiae strains used in the present study (Table (Table1)1) were constructed by standard genetic methods (73) and are derivatives of S288C (10), except for strains DY2765, DY6669, DY12247, DY12872, and DY12878 that were derived from W303.
Systems for doxycycline (Dox)-regulated gene expression were derived in part from plasmids in reference (7) and will be described elsewhere (our unpublished data). To construct the “tet-on” strains, because MCM1 is essential for viability, one copy of the endogenous MCM1 promoter was first replaced with a Dox-responsive promoter (PtetO7) in a homozygous pho3Δ/pho3Δ background. Next, a single copy of YIpMR1337 was integrated in the PtetO7:MCM1/MCM1 and control MCM1/MCM1 strains at the HO locus using homologous targeting sequences (as described in reference 91). YIpMR1337 constitutively expresses both a repressor that dissociates from DNA (TetR-Ssn6) and an activator with a composite activation domain (AD) that binds to DNA in the presence of Dox (TetR′-VP16AD-P201AD). P201 is an 8-amino-acid AD that was selected for activation function from a library encoding random 8-mer peptides and is devoid of acidic residues (47). The “tet-off” MCM1 strain was constructed by first integrating plasmid YIpAAP1366, which constitutively expresses TetR-VP16AD that dissociates from DNA and TetR′-Ssn6 that binds to PtetO7 in the presence of Dox (i.e., opposite of tet-on), into a wild-type (WT) haploid strain. The endogenous MCM1 promoter was subsequently replaced with PtetO7.
Strains for most experiments were cultured in YPD containing 2% (wt/vol) Bacto yeast extract or yeast extract (BD Diagnostic Systems/Difco [DF0127-17-9] or U.S. Biological [Y2010 for Fig. Fig.6A6A only], respectively), 2% (wt/vol) Bacto peptone (BD Diagnostic Systems/Difco, catalog no. DF0118-17-0), and 2% (wt/vol) dextrose. All cultures were incubated at 30°C with shaking at 300 rpm, unless indicated otherwise. For the ChIP analyses, strains were grown in YPD plus 0.005% (wt/vol) adenine. For the Pi starvation experiment (see Fig. Fig.5),5), starter cultures were inoculated into rich yeast-peptone-2% (wt/vol) galactose. After overnight growth, the cells were washed and resuspended in defined Pi-free medium containing (per liter) 0.7 g of yeast nitrogen base without (NH4)2SO4, phosphate, or amino acids (Bio 101, catalog no. 4029-622); 0.74 g of complete synthetic medium supplement mixture; 3.9 g of MES [2-(N-morpholino)ethanesulfonic acid, pH 5.5]; and 13.4 mM Pi added back as described previously (57), with 2% (wt/vol) galactose as the sole carbon source. Next, the cells were washed and resuspended in defined Pi-free medium with 13.4 mM Pi added back and containing 2% (wt/vol) glucose for 3 h and then washed and resuspended in the same medium containing 13.4 or 0 mM Pi for repressive and activating conditions, respectively.
For most experiments, starter cultures of yeast were grown for 6 h or overnight in YPD to mid-logarithmic phase at 30°C and then diluted to an optical density at 600 nm (OD600) of 0.005 to 0.05 with fresh YPD and grown for an additional 6 h to overnight. Unless indicated otherwise, rAPase activity was assayed by a whole-cell liquid colorimetric method as previously described and reported as Miller units of activity, normalized to OD600 values for each culture (57). For tet-on and tet-off regulation of MCM1 (Fig. (Fig.22 and and3,3, respectively), strains were diluted into YPD plus the indicated concentrations of Dox.
Since the aberrant morphology of fkh1Δ, fkh2Δ, fkh1Δ fkh2Δ, and tet-off MCM1 strains (MCM1/mcm1Δ strains had normal morphology) precluded normalization to OD600, the total rAPase activity was assayed as follows. Overnight YPD cultures (10 ml) of WT, fkh1Δ, fkh2Δ, and fkh1Δ fkh2Δ strains (all in a pho3Δ background) were grown to a moderate density as gauged visually and then washed (10 ml) and resuspended (1 ml) in 0.1 M sodium acetate (pH 3.6) supplemented with protease inhibitors. Cells were then lysed by vortexing (5 to 10 times for 30 s each time) in the presence of 425- to 600-μm acid-washed glass beads, followed by vigorous agitation in a bead beater (twice for 60 s each time). Cell lysates were centrifuged 5 min at 14,000 × g, and the protein concentration was determined by using a bicinchoninic acid assay (Pierce). Approximately 0.5 ml of the cell lysate was used to assay for rAPase activity as described previously (57), except that the rAPase activity was normalized to the total cellular protein.
Since the activity of the pho4Δ mutants is below the linear range of spectrophotometric detection in the liquid rAPase assay, a color-forming rAPase plate assay was performed by staining the colonies with overlaid molten 1% soft agar (wt/vol) containing both 0.5 mg of α-naphthol phosphate and 5 mg of fast blue salt B per ml in 0.05 M acetate buffer (pH 4.2) (85). Cultures were grown to mid-logarithmic phase and adjusted to the same cell density (OD600 = 0.2), and then 3 μl was spotted on the plate and grown for 2 days at 30°C. Depending on the amount of rAPase activity of each strain, the colony color intensity varied from white (no detectable activity) to pink (weak activity) to deep red (strong activity) on the YPD plate.
Analysis of the mitotic cycling of PHO5 transcript levels (Fig. (Fig.6B)6B) was performed with strains from which the highly homologous PHO3 gene had been deleted in order to avoid cross-hybridization exactly as described previously (57).
Yeast cells (10-25 ml) were grown in YPD without or with the indicated Dox concentration to an OD600 of ~1.5 and used to prepare protein extracts by a standard trichloroacetic acid precipitation method. The total protein was then quantified by using the bicinchoninic acid assay kit, and 10 to 30 μg of protein per lane was resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis. After electrotransfer to polyvinylidene difluoride membranes and blocking, the blots were incubated overnight with goat anti-Mcm1 antibody (Santa Cruz, catalog no. SC-12026) used at a 1:1,000 dilution and subsequently immunostained with horseradish peroxidase-conjugated anti-goat immunoglobulin G (IgG; Santa Cruz, catalog no. SC-2768) used at a 1:5,000 dilution. The blot was stripped and reprobed overnight with mouse monoclonal antibody specific to yeast 3-phosphoglycerate kinase (Pgk1; Molecular Probes, catalog no. A6457) used at a 1:1,000 dilution, followed by horseradish peroxidase-conjugated anti-mouse IgG (Santa Cruz, catalog no. SC-2055) used at a 1:10,000 dilution.
For ChIP of Mcm1 and both forkhead proteins, strain DY12872 (cdc28-13ts FKH1-6HA FKH2-18Myc) was grown to an OD660 of 0.4 at 25°C in YPD, a sample for asynchronous log phase was taken, and then the remainder was shifted to 37°C for 4 h to effect synchronous arrest in G1 phase (69). For synchronous release, cells were collected by filtration and then resuspended in fresh medium preincubated at 25°C. ChIP for Fig. Fig.99 was conducted exactly the same, except that the starter culture was divided into two equal aliquots and nocodazole (Noc; 100 μM) or dimethyl sulfoxide (Noc solvent) was added, followed by incubation for an additional 150 min. ChIP of Sds3 was performed with strain DY12247 (PGAL1:CDC20 SDS3-13Myc) after cell cycle synchronization by galactose withdrawal and readdition at 25°C in YP medium containing 2% (wt/vol) of both galactose and raffinose (8). For each ChIP experiment, at the designated times, 50 ml of culture was removed, and formaldehyde was added to 1% (vol/vol) for fixation overnight on ice. In parallel, ethanol at −70°C was added to 70% (vol/vol; final concentration) to 15 ml of culture for RNA purification. Also, ethanol at 4°C was added to 70% to 2 ml of culture, followed by staining with Sytox dye for flow cytometry. An asynchronous culture of DY12878 in logarithmic growth was used for the untagged control reactions.
Fragmentation of the chromatin by sonication, immunoprecipitation, and analysis for immunoprecipitated sequences by quantitative PCR were performed exactly as described in Voth et al. (92). The PHO5 promoter oligonucleotides used for quantitative PCR were AGCGGACGTCGTCTATAAACT and TATGCCTTGCCAAGTAAGGTG (59). The oligonucleotides for qRT-PCR analysis of PHO5 mRNA were TTATTCTCGTGGTGTGCATTT and CTTTAATAATTTGACTGAGGCATTG. They were determined to be specific for amplification of PHO5 and did not detect the highly homologous PHO3 sequences, as a PCR amplicon was only obtained when including DNA isolated from pho3Δ PHO5 cells (strain CCY577) but not from PHO3 pho5Δ (strain SPY4056) or pho3Δ pho5Δ (strain SPY4065) cells (data not shown). Moreover, the PHO5 qRT-PCR products quantified for Fig. Fig.88 (upper) were specific, as judged by direct sequencing.
We previously showed that depletion of vacuolar polyP stores in S/G2 precedes peak mitotic induction of PHO5 (57). Cell cycle oscillation of PHO5 transcript was not detected by blot hybridization of RNA isolated from strains containing single deletions of PHO4 and PHO2 or from cells grown in medium supplemented with Pi. However, significant oscillatory behavior persisted in cells inactivated for PHO81 (57), which encodes an upstream-acting CKI in the PHO signal transduction cascade (see Fig. Fig.11)11) (75). This implicated an additional cell cycle-dependent regulatory input at PHO5, perhaps at the level of transcription. Consistent with this, CLB2 overexpression also increased PHO5 mRNA levels in mitotically arrested cells (81). This suggests that PHO5 expression is upregulated by a transcriptional activator that responds to B-type CDK activity, possibly Mcm1-Fkh2-Ndd1, which, to our knowledge, has only been shown to induce transcription of CLB2 cluster genes in G2/M (18, 19, 37, 66, 67, 70).
Inspection of the PHO5 promoter region identified a strong candidate Mcm1-Fkh binding site (Fig. (Fig.1A).1A). This site aligned well with the Mcm1-Fkh consensus sequence and other sequences in the promoters of prototypical genes that are coregulated members of the MCM (KIN3 and SPO12) and CLB2 (CLB2 and SWI5) clusters that reach peak induction in M phase (48, 51, 81). In addition, the candidate Mcm1-Fkh site is located close to the core promoter of PHO5, from bp −47 to −25 and from bp −148 to −126 upstream of the TATA element and translational start codon, respectively (Fig. (Fig.1B).1B). This distance is similar to that of well-known mitotically induced promoters (81) and well within the range from which a single Mcm1 site has been shown to efficiently activate transcription (64).
Mcm1 is essential for viability and can function either as a transcriptional activator or as a repressor (22). Therefore, we assayed the effect of conditional MCM1 expression on rAPase activity in an isogenic set of diploid strains that differed only by the presence of a single Dox-inducible (tet-on) allele of MCM1 (PtetO7:MCM1) and the integrated plasmid YIpMR1337, which constitutively expresses the two regulators of PtetO7 (see Materials and Methods). One of these regulators, tetR-Ssn6, is a repressor that binds to DNA in the absence of Dox, whereas the other, tetR′-VP16AD-P201AD, is an activator that exhibits DNA binding in the presence of Dox. Therefore, addition of Dox dissociates the repressor and causes activator binding to PtetO7, which results in net induction of transcription, i.e., a tet-on phenotype. Each of these strains also contained a deletion of PHO3, encoding the so-called constitutive acid phosphatase, which allows measurements of secreted rAPase activity that essentially reflect PHO5 expression levels (20, 33, 57). Moreover, we have previously established that the majority of rAPase activity in pho3Δ cells grown asynchronously in YPD is attributable to oscillatory induction of PHO5 in mitosis (57).
The heterozygous PtetO7:MCM1/MCM1 strain exhibited a haploinsufficiency phenotype, secreting less rAPase activity than WT cells (Fig. (Fig.2A,2A, compare bars 1 and 3 and bars 2 and 4). In other experiments (not shown), about half as much rAPase activity was secreted. Immunoblotting showed that the lower level of rAPase activity in PtetO7:MCM1/MCM1 cells is consistent with an ~2-fold reduction of the amount of Mcm1 protein (Fig. (Fig.2B,2B, compare lanes 1 and 3 and lanes 2 and 4). As a control, introduction of YIpMR1337, which expresses the Dox-dependent regulators, into otherwise WT cells reduced rAPase activity for an unknown reason (Fig. (Fig.1A,1A, compare bars 1 and 5 and bars 2 and 6). This decreased rAPase activity was not due to nonspecific reduction of the amount of Mcm1 (Fig. (Fig.2B,2B, compare lanes 1 and 5 and lanes 2 and 6) but nevertheless provides a reference level of activity for comparison to the tet-on strain. Untreated tet-on cells (containing both PtetO7:MCM1/MCM1 and YIpMR1337) did not appear to have reproducibly decreased rAPase activity compared to the reference strain containing only integrated YIpMR1337 (Fig. (Fig.2A,2A, compare bars 5 to 7). This is not due to failure to repress MCM1 expression in the tet-on strain in the absence of Dox, since much less Mcm1 protein accumulated compared to the reference strain (Fig. (Fig.2B,2B, compare lanes 5 and 7). Instead the reduction in rAPase activity upon integration of YIpMR1337 by itself (Fig. (Fig.2A,2A, bar 5) has probably obscured detection of further decreases in rAPase activity after replacing one copy of PMCM1 with PtetO7 (bar 7). In addition, PtetO7:MCM1 expression, as judged by Mcm1 levels, is not reduced further by YIpMR1337 integration in the absence of Dox, after normalization to the Pgk1 loading control (Fig. (Fig.2B,2B, compare lanes 3 to 7). A possible reason for this is competition of the tetR′-VP16AD-P201AD activator with tetR-Ssn6 repressor, raising basal expression of PtetO7 (7). In any case, the addition of 2 μg of Dox/ml for 16 h and overexpression of Mcm1 (Fig. (Fig.2B,2B, compare lanes 2 and 4 to lane 8) specifically increased activity in the tet-on MCM1 heterozygote by fivefold (Fig. (Fig.2A,2A, compare bars 7 and 8). These results suggest that Mcm1 transactivates PHO5 either directly or indirectly. The copy-number-dependent regulation also suggests that Mcm1 activity is rate limiting for PHO5 activation.
Glucose-mediated repression of MCM1 under the control of the GAL1 promoter (PGAL1:MCM1) in haploid cells was previously shown to abrogate transcription of CLB2 cluster genes and cause elongated budding morphology (1). To avoid a possible influence of carbon source on Mcm1 activity (3, 15, 51) and thereby on PHO5 expression, we used a Dox-repressible (tet-off) system to assess the effect of Mcm1 loss-of-function on mitotic expression of PHO5. Because MCM1 is an essential gene, we constructed a haploid strain in which the endogenous promoter of MCM1 (PMCM1) was replaced with PtetO7 in a strain containing YIpAAP1366. This integrating plasmid expresses TetR-VP16AD that dissociates from DNA and TetR′-Ssn6 that binds to DNA and represses transcription upon Dox addition. A tet-off haploid (PtetO7:MCM1 YIpAAP1366) was compared to a WT haploid (MCM1) strain after growth in YPD with or without 2 μg of Dox/ml. Cells were then analyzed for budding morphology by light microscopy (Fig. (Fig.3A)3A) and for Mcm1 protein level by immunoblotting (Fig. (Fig.3B3B).
Incubating WT cells with Dox did not alter their budding morphology (Fig. (Fig.3A,3A, top panels) or the amount of Mcm1 protein (Fig. (Fig.3B,3B, lanes 1 to 3). In contrast, even in the absence of Dox, an occasional cell in the tet-off MCM1 culture exhibited an elongated budding morphology typical of pseudohyphal or filamentous growth associated with delayed entry into M phase (1, 24, 36, 71) (Fig. (Fig.3A,3A, lower left panel). This is most likely due to diminished levels of Mcm1 in tet-off MCM1 compared to WT MCM1 cells, normalized for cytosolic Pgk1 protein in the immunoblot (Fig. (Fig.3B,3B, compare lanes 1 and 4). This could arise either because, in the absence of Dox, PtetO7 is transcribed more weakly than endogenous PMCM1, basal expression of PtetO7 may be repressed, or both. The majority of tet-off MCM1 cells became highly elongated and stopped dividing after Dox treatment (Fig. (Fig.3A,3A, lower right panel), as shown previously for cells with a conditionally repressed PGAL1:MCM1 allele (1). Consistent with this result, Mcm1 protein was not detectable in tet-off MCM1 cells treated with 2 or 5 μg of Dox/ml (Fig. (Fig.3B,3B, compare lanes 2 and 5 and lanes 3 and 6).
Having established a strong knockdown regulatory system for MCM1, we measured the rAPase activity in asynchronously growing YPD cultures to assess expression of PHO5 in M phase (Fig. (Fig.3C).3C). Compared to WT (MCM1 PHM4), rAPase activity in the tet-off MCM1 strain was reduced ~2-fold in the absence of Dox (compare bars 1 to 3) and by 14-fold in its presence (compare bars 2 to 4). These results parallel the decreased Mcm1 protein levels observed by immunoblotting (Fig. (Fig.3B)3B) and, in accord with the haploinsufficiency phenotype observed in Fig. Fig.2A,2A, suggest that Mcm1 is an essential, rate-limiting activator of mitotic PHO5 expression.
We previously noted that cells synchronized with α-factor pheromone at late G1 had elevated levels of polyP (57), most likely because Pi uptake exceeded usage in growth-arrested cells. Therefore, we considered the possibility that cells arrested in G2/M by depletion of Mcm1 also accumulated reserves of polyP and hence inhibited PHO5 transcription by buffering intracellular Pi concentration (57, 84). Indeed, polyP amounts increased by at least 15-fold in tet-off MCM1 cells arrested in G2/M by Dox addition (data not shown). To eliminate the potential repressive influence of this elevated polyP storage on mitotic PHO5 expression, we assayed the rAPase activity in WT and tet-off MCM1 cells with PHM4 deleted, whose gene product is required for polyP synthesis (57, 60, 84). As observed previously (57, 84), PHO5 expression was derepressed in MCM1 phm4Δ cells that lack detectable polyP compared to WT MCM1 PHM4 cells incubated with or without Dox (Fig. (Fig.3C,3C, compare bars 1 and 5 and bars 2 and 6). Despite this derepression of PHO5 upon eliminating polyP stores, depleting Mcm1 in tet-off MCM1 cells reduced rAPase activity by 5.4- and 19-fold in the absence and presence of Dox, respectively (Fig. (Fig.3C,3C, compare bars 5 and 7 and bars 6 and 8). Importantly, rAPase activity was reduced to similar absolute levels in both tet-off MCM1 strains, which are isogenic and differed only in their PHM4 or phm4Δ genotype (compare bars 3 and 7 and bars 4 and 8). This demonstrates that the role of Mcm1 in PHO5 activation is epistatic to the repression that polyP indirectly exerts on PHO5 transcription in sustaining intracellular Pi concentration. We conclude that Mcm1 is required for mitotic activation of PHO5 and that it acts downstream of the PHO signaling transduction cascade (see Fig. Fig.11),11), which responds to both Pi uptake across the plasma membrane and mobilization of vacuolar polyP reserves.
Mcm1 target sites are often located adjacent to sites that bind Fkh proteins at mitotically induced genes (29, 34, 37, 67, 92, 96). Moreover, we identified a strong consensus Fkh site in the PHO5 promoter (Fig. (Fig.1).1). An effect on mitotic induction of PHO5 in a double fkh1Δ fkh2Δ mutant was not detected in a previous study (96), possibly due to cross-hybridization of the highly homologous PHO5 and PHO3 transcripts to the cDNA probes affixed to the microarray (27).
To reexamine whether Fkh1 and Fkh2 regulate PHO5 mitotic expression, we constructed strains with single- or double-null mutations in the FKH genes in a pho3Δ background and assayed them for rAPase activity. In Fig. Fig.4A,4A, consistent with the known genetic redundancy of FKH1 and FKH2, only the double fkh1Δ fkh2Δ mutant showed the characteristic cell separation and morphology defects (28, 37, 67, 96). For rAPase activity, both strains with single fkh1Δ or fkh2Δ null alleles exhibited modest 25% reductions compared to WT FKH1 FKH2 cells dissected from the same tetrad (Fig. (Fig.4B,4B, compare bar 1 to bars 2 and 3). A fkh1Δ fkh2Δ double-null strain exhibited an additive loss in rAPase activity, at 60% of WT, again consistent with the redundancy of the two genes (Fig. (Fig.4B,4B, compare bars 1 and 4). These results suggest that Fkh1 and Fkh2 have redundant roles in PHO5 mitotic activation.
To rule out possible effects of polyP reserves on PHO5 expression in strains deleted for FKH genes, we measured rAPase activity in WT, phm4Δ, fkh1Δ phm4Δ, fkh2Δ phm4Δ, and fkh1Δ fkh2Δ phm4Δ cells. Similar levels of rAPase were synthesized in each of these strains (data not shown), demonstrating genetic suppression of the PHO5 expression defects of fkh1Δ, fkh2Δ, and fkh1Δ fkh2Δ strains shown in Fig. Fig.4B.4B. We conclude that abolishing vacuolar polyP reserves and hence increasing intracellular starvation for Pi (57, 84) bypasses the requirement for Fkh1, Fkh2, or both forkheads in peak mitotic induction of PHO5. This is in contrast to the failure of loss of polyP to suppress the dramatic losses in rAPase activity observed in Mcm1-depleted cells (Fig. (Fig.3C3C).
Additional evidence argues that the substantial reduction in mitotic PHO5 expression in cells depleted for Mcm1 was not caused by the resulting G2/M arrest phenotype. First, after tet-off MCM1 cells were incubated with Dox overnight and then the antibiotic was removed by washing, the total protein content of cultures increased at a rate similar to that of an untreated culture (data not shown). This indicates that a substantial fraction of Mcm1-depleted cells retained viability and that the loss of rAPase activity was not caused by death of a large fraction of cells in culture. It is not possible to determine the percentage of viable cells in this experiment because of the filamentous phenotype that results from repression of MCM1 transcription. Second, rAPase activity was elevated 2.4-fold by metaphase arrest after glucose-mediated repression of PGAL1:CDC20, which encodes a mitotic activator of the anaphase-promoting complex (Fig. (Fig.5,5, compare bars 1 and 3). High Clb-Cdc28 activity in mitotically arrested cells has been shown to increase phosphorylation of both Fkh2 and the Ndd1 coactivator, which enhances Mcm1-Fkh2-dependent recruitment of Ndd1 and the expression of CLB2 cluster genes (19, 66, 70). Furthermore, PHO5 was strongly induced when the PGAL1:CDC20 strain was growth arrested in early M phase and then subjected to a subsequent period in Pi-free medium while growth arrest was maintained (Fig. (Fig.5,5, compare bars 3 and 4). The same experimental protocol yielded an identical result when a cdc15-1ts strain, which arrests in late M phase at the nonpermissive temperature, was used (25). These results indicate that the defects in mitotic activation of PHO5 in strains with loss-of-function mutations in MCM1 and FKH genes are not due to cell cycle arrest per se. To the contrary, arrest in early M phase by CDC20 shutoff partially derepressed PHO5 expression even in a PHM4+ background.
Although our data thus far implicate Mcm1 and Fkh proteins in mitotic induction of PHO5, it is unclear whether their role is direct or indirect. To address this, we made base substitutions (Fig. (Fig.1A)1A) in the candidate binding sites for Mcm1, Fkh or both factors in the PHO5 promoter at its native genomic location. The same mutations have been shown to disrupt essential protein-DNA contacts and thus abolish occupancy at CLB2 cluster targets in vitro and in vivo for Mcm1 and Fkh proteins (34, 37, 48, 50, 62, 67, 94). Strains bearing WT (PPHO5) and mutated promoters (PPHO5-mcm1, PPHO5-fkh, and PPHO5-mcm1 fkh) were assessed for PHO5 mitotic activation. Relative to the WT, rAPase activity was reduced ~2-fold in strains with mutations in either the Mcm1 or Fkh binding site (Fig. (Fig.6A,6A, compare bar 1 to bars 2 and 3) and 6-fold when both sites were mutated (Fig. (Fig.6A,6A, compare bars 1 and 4). Fkh2 can stabilize binding of Mcm1 to target genes bearing mutated or even unrecognizable Mcm1 sites (29, 48). Not unexpectedly, our data suggest that Mcm1 also stabilizes Fkh2 binding to weak sites, and thus it follows that mutation of sites for both factors is required to severely impair PHO5 mitotic activation.
We next determined whether mutations in the Mcm1-Fkh site affected the cell cycle-dependent oscillation of PHO5 transcript. YPD cultures of WT and PPHO5-mcm1 fkh strains were arrested in parallel in late G1 by α-factor and synchronously released from the block in fresh YPD lacking pheromone. Total RNA was isolated at 15-min intervals and assayed for PHO5 and TCM1 transcript levels via RNA blot hybridization (Fig. (Fig.6B).6B). Normalization of the level of PHO5 to TCM1 transcript, which is not subject to cell cycle regulation (16, 57, 81), revealed that the Mcm1-Fkh site mutations substantially reduced the amplitude of PHO5 mitotic induction. We conclude that the bipartite Mcm1-Fkh site in the PHO5 promoter is required for full rAPase activity in asynchronously growing cultures and for peak transcript accumulation in M/G1 in synchronized cultures.
The DNA-binding transactivators Pho4 and Pho2 bind cooperatively to the PHO5 promoter when cells are deprived of Pi, and both factors are essential for mitotic expression of rAPase activity (5, 6, 57). In addition, either reducing Mcm1 to a nondetectable level or simultaneous mutation of the Mcm1 and Fkh sites in the native PPHO5 severely impaired rAPase activity (Fig. (Fig.3C3C and and6A).6A). Moreover, these same site mutations substantially reduced the oscillatory amplitude of PHO5 mRNA (Fig. (Fig.6B).6B). To determine the relationship between the contributions of Mcm1, forkhead proteins, and Pho4 to mitotic activation of PHO5, we constructed all possible combinations of the PHO5 promoter mutants and a PHO4 deletion (Fig. (Fig.7).7). An equivalent number of cells of each original parent strain and three independent pho4Δ transformants derived from each parent strain were spotted onto a YPD plate. After overnight growth, the cells were assayed by a color-forming plate overlay assay for rAPase activity. The plate assay was used because it provides a more reliable, albeit qualitative, measurement in cells expressing low levels of rAPase activity (e.g., pho4Δ strains). The nonenzymatic background rate of hydrolysis of the phosphatase substrate used in the liquid assay is too high at the low levels of enzymatic activity assayed in the present experiment. In the plate assay, the darkness of each overlaid spot of cells is proportional to the level of enzyme-dependent substrate hydrolysis.
As expected, compared to the WT (PPHO5), cells with PHO4 deleted (PPHO5 pho4Δ) had vastly reduced levels of rAPase activity (Fig. (Fig.7,7, compare spot A1 to spots A2, A3, and A4 and E). Likewise, compared to the WT, point mutation of the Fkh binding site (PPHO5-fkh) substantially decreased rAPase activity levels (compare spots A1 and B1). Relative to the Fkh binding site mutation, rAPase activity was reduced even further by mutation of the Mcm1 binding site (PPHO5-mcm1; compare spots B1 and C1) by itself or in conjunction with the Fkh site (PPHO5-mcm1 fkh; compare spots B1 and D1). This is consistent with the findings above that Mcm1 plays a more prominent role in PHO5 mitotic induction than the forkhead proteins. Importantly, combining any of these promoter point mutations with a PHO4 deletion (PPHO5-fkh pho4Δ, PPHO5-mcm1 pho4Δ, and PPHO5-mcm1 fkh pho4Δ) resulted in further additive reductions in rAPase activity (compare spot B1 to spots B2, B3, and B4, spot C1 to spots C2, C3, and C4, and spot D1 to spots D2, D3, and D4). Taken together, this indicates that Mcm1, Mcm1-Fkh, and Pho4 activate PHO5 in M phase via separate, nonredundant pathways. Moreover, these data suggest that physiological levels of Mcm1 can activate PHO5 in mitosis independent of Pho4, and vice versa, albeit at a much lower level than when both transcription factors are present.
Previous ChIP and footprinting studies have shown that Mcm1, Fkh1, and Fkh2 associate with the promoters of genes in the CLB2 (CLB2, SWI5, and BUD4) and MCM (CDC6, CLN3, and SWI4) clusters in at least a fraction of the cells growing asynchronously (1, 29, 34, 37, 51). Consistent with these results, relatively modest fluctuations in Mcm1 binding to representative promoters during the cell cycle have been detected by ChIP between cells arrested in late G1 (α-factor) or M (Noc) or throughout the cell cycle after release from α-factor arrest (1, 34, 51). Fkh binding in synchronously growing cultures has only been studied at the CLB2 promoter (92). At CLB2, Fkh1 is bound at a constant, but significant, level above background, whereas Fkh2 association exhibited considerable cell cycle stage-specific oscillation. In contrast to these genes in the CLB2 and MCM clusters, significant Mcm1 and Fkh binding to the PHO5 promoter was not observed in nonsynchronized, logarithmically growing YPD cultures (Fig. (Fig.8C,8C, lower panel, and data not shown).
Therefore, ChIP analysis was performed on synchronized cultures to determine whether Mcm1 and the Fkh proteins directly associate with the PHO5 promoter in a cell cycle-dependent manner. We constructed a cdc28-13ts strain expressing C-terminally tagged versions of both Fkh proteins, Fkh1-6HA and Fkh2-18Myc, from their native genomic locations. Enlisting an anti-Mcm1 antibody as well allows all three factors to be immunoprecipitated individually from the same cross-linked samples for a direct comparison of binding in a single arrest-and-release time course. We previously used the same strategy with a strain in which both Fkh proteins were tagged to avoid stochastic effects and variation in synchrony between strains with different genotypes in separate cultures (92). Importantly, tagging the FKH genes did not affect the induction kinetics of the PHO5 promoter as assayed by rAPase activity (data not shown).
The cdc28-13ts strain was grown to early logarithmic phase in YPD, and then cells were synchronized at the nonpermissive temperature in G1 and released from the arrest point at 25°C (69). Separate aliquots of cells were removed at 10-min intervals for isolation of total RNA and for cross-linking chromatin in vivo for ChIP analysis of CLB2, CTS1 and PHO5 sequences associated with Mcm1, Fkh1-6HA, and Fkh2-18Myc. CTS1 is a member of the SIC1 cluster induced late in the cell cycle. Synchrony among the cell population was demonstrated by monitoring aliquots of ethanol-fixed cells for DNA content after Sytox staining and flow cytometry (Fig. (Fig.8A)8A) and for budding index (Fig. (Fig.8B),8B), the percentage of cells containing buds of various sizes. Both criteria demonstrated that a fraction of cells synchronously entered S phase by 40 min and that the majority were in S phase by 50 to 60 min and completed DNA synthesis by 80 min. The 2C DNA content for the remainder of the time course in the Sytox flow cytometry profiles arises from a postmitotic cell separation defect that is commonly observed in W303 strains bearing certain mutations after arrest and release (90). The budding indices in Fig. Fig.8B8B support this assertion since small buds had reemerged on 50% of the cells at 150 min after G1 release, that is, a substantial fraction of cells had begun a second S phase. Buds were not counted for the 70- to 130-min time points because they were not informative in that cells had elongated buds that just grew in size with no change in morphology.
Synchronous transition of the cell population through the cell cycle was further demonstrated by analysis of CLB2, PHO5, and CTS1 transcript levels by qRT-PCR (Fig. (Fig.8C,8C, top panel). In accordance with previous studies (2, 16, 57, 81, 83), CLB2 mRNA accumulated and peaked earliest, followed by PHO5 and then CTS1, consistent with their respective assignments to the CLB2, MCM, and SIC1 clusters (81).
In the ChIP analyses, CLB2 served as a positive control and showed strong occupancy by Mcm1 and Fkh2-18Myc throughout the cell cycle (92) (Fig. (Fig.8C,8C, middle panel). Interestingly, the level of binding of Fkh2-18Myc to CLB2 was highest at the arrest point and declined to a steady state at about 30 min after switching the culture to the permissive temperature. In contrast, a lower, but significant amount (≥2-fold) of Fkh1-6HA enrichment was observed at the CLB2 promoter at all time points. Fkh1-6HA enrichment increased to 7.4-fold at 100 and 110 min after release, preceding the decline in CLB2 mRNA.
At PHO5 (Fig. (Fig.8C,8C, bottom panel), the apparent level of Mcm1 binding was weaker than that observed at CLB2. Mcm1 occupancy of PHO5 was also highest at and immediately after release from the G1 cell cycle block. This binding declined as cells approached and passed through S phase and then exhibited an overall increase until the end of the cell cycle. Mcm1 binding to PHO5 was significant at all time points since it was more than twice the level of nonspecific enrichment of PHO5 sequences by preimmune IgG. Occupancy of PHO5 (and CLB2) was specific, since no significant binding of Mcm1 was detected at CTS1 (data not shown). Fkh2-18Myc exhibited a similar binding profile at the PHO5 promoter as Mcm1; however, the apparent binding was also substantially lower than at the CLB2 promoter. Although the Fkh2-18Myc ChIP signal is modest, it is clearly above the ChIP signal from the untagged control strain. Interaction of Fkh1-6HA with PHO5 sequences was the weakest, but a binding peak was observed from 100 to 130 min that was ~2-fold higher than the initial 30 to 90 min. The low, but steady, enrichment of Fkh1 binding over the same time period as when Mcm1 occupancy and PHO5 mRNA increased is consistent with the modest effects on rAPase activity of fkh mutants (Fig. (Fig.4B)4B) and mutation of the Fkh site alone (Fig. (Fig.6A6A and and7).7). We conclude that Mcm1 and the Fkh factors associate with the PHO5 promoter in a cell cycle-dependent manner.
The cdc28-13ts strain progressed synchronously through the cell cycle after release at 25°C (Fig. 8A and B). However, because Mcm1 binding at PHO5 was maximal at G1 arrest, we wanted to assess whether increased Mcm1 binding after S phase (Fig. (Fig.8C,8C, lower panel) was due to G2/M entry and/or a degree of asynchrony that yielded a fraction of cells that had entered G1. We repeated the same arrest-and-release experiment, except that the synchronously growing cells were divided into two aliquots and 100 μM Noc was added to one of them to subsequently block the cells in M phase. Binding of Mcm1 to the PHO5 promoter and open reading frame (ORF) of HCM1, a region negative for Mcm1 binding, was determined by ChIP at 0 and 150 min after release at 25°C and normalized to the signal of an asynchronous culture of the same strain. Figure Figure99 shows Mcm1 binding was again greatest at the G1 arrest point, when Cdc28 activity was inactivated (Fig. (Fig.9,9, compare bars 2 and 4), consistent with the results in Fig. Fig.8C8C (lower panel). Also, at 150 min after the cell cycle block was released, association of Mcm1 with the PHO5 promoter in the minus Noc culture declined because the cells had progressed into S phase (Fig. (Fig.9,9, compare bars 4 and 6). In contrast, the addition of Noc enriched PHO5 sequences in the anti-Mcm1 ChIP assay (Fig. (Fig.9,9, compare bars 6 and 8). This binding was specific since Noc addition did not increase Mcm1 at HCM1 sequences (compare bars 5 and 7). We conclude that Mcm1 binding increases at the PHO5 promoter in cells arrested in both G1 (by Cdc28 inactivation) and M (by Noc addition) phases.
We have demonstrated that Mcm1 and the forkheads are activators of PHO5 in mitosis (Fig. (Fig.2,2, ,3,3, ,4,4, ,6,6, and and7).7). However, PHO5 mRNA was at basal levels at 0 and 10 min after G1 arrest (Fig. (Fig.8C,8C, upper panel), points when there was high promoter occupancy by Mcm1-Fkh2 (Fig. (Fig.8C,8C, lower). Offering a potential explanation for this apparent paradox, previous work has shown that the Rpd3 histone deacetylase negatively regulates PHO5 expression and associates directly with PHO5 (39, 88, 89, 93). Recent work has also shown that Sin3 and Rpd3, most likely as components of the Rpd3L complex, are recruited to the CLB2 promoter in G1 and released as cells progress through START (12, 31, 87, 92). Therefore, we tested whether a similar temporal association of Rpd3L occurred with the PHO5 promoter by ChIP. The endogenous locus of SDS3, encoding an Rpd3L-specific subunit, was tagged with 13Myc in a PGAL1:CDC20 background. This strain was arrested in late M phase by removing galactose, and the association of Sds3-13Myc with PHO5 sequences was determined at various times after synchronous release into galactose-containing medium (Fig. (Fig.10).10). Peak association of Sds3-13Myc in the Rpd3L complex was detected at 35 min after removal of the cell cycle block. Rpd3L also showed periodic binding to the HO promoter that peaked at 35 min after release (S. Takahata and D. J. Stillman, unpublished data). This time corresponded to early G1, because expression of HO was detected at 40 min after release, which corresponds to late G1 (55).
We previously concluded that PHO5 mitotic activation in Pi-limiting environments is driven mostly independently of the cell cycle progression machinery (57). This conclusion was reached because PHO5 induction in M/G1 appeared to be abolished in cells lacking Pho4 and Pho2, which bind cooperatively to DNA, and when excess Pi was supplied (57). However, residual mitotic activation in cells lacking Pho4 (Fig. (Fig.7)7) or the upstream CDK inhibitor Pho81 (57) suggested one or more PHO-independent pathways of upregulation (Fig. (Fig.11).11). We have identified here a novel transcriptional input that includes the cell cycle regulators Mcm1, a MADS-box factor, and the winged-helix proteins Fkh1 and Fkh2. This is the first report of PHO5 regulation by sequence-specific DNA-binding factors other than Pho4 and Pho2. Strikingly, we found Mcm1 to be as essential for PHO5 mitotic activation (Fig. (Fig.3C)3C) as is Pho4-Pho2 (57) (Fig. (Fig.77).
In contrast to Mcm1, the forkhead proteins appear to play a significant, but less pronounced, role in PHO5 induction (Fig. (Fig.4B4B and and7).7). The need to delete both FKH1 and FKH2 in order to substantially reduce rAPase activity is consistent with their known partial redundancy (28, 37, 67, 96). Ndd1 may be weakly recruited to PHO5 by Mcm1-Fkh. This provides a possible explanation for the modest PHO5 activation defect of fkh1Δ fkh2Δ cells. Also, a smaller reduction in PHO5 mitotic activation in asynchronously growing fkh1Δ fkh2Δ cells due to lack of Ndd1 coactivator recruitment in G2/M may be offset by loss of the Rpd3L repressor, which is recruited in G1 (Fig. (Fig.10).10). In contrast, because Fkh2 cannot bind in vivo without Mcm1 (29, 34, 37, 67, 92, 96), loss of Mcm1 function may cause a more severe defect in PHO5 mitotic activation because recruitment of Ndd1 and an additional coactivator is disrupted. For example, Arg82/Ipk2, an inositol polyphosphate kinase, contributes to PHO5 activation (76, 82). Since direct association of Arg82/Ipk2 with PPHO5 has not been demonstrated, Mcm1 binding offers a potential mechanism for recruiting the kinase (21). Loss of both Arg82/Ipk2 and Ndd1 recruitment in cells lacking Mcm1, as opposed to only Ndd1 in Fkh-deficient cells, might explain the more severe PHO5 activation defect upon Mcm1 knockdown (Fig. (Fig.3C3C and and4B).4B). The more pronounced decline of PHO5 expression in Mcm1-depleted versus fkh1Δ fkh2Δ cells also agrees well with previous observations that Mcm1 binding is reduced, but not eliminated, in the absence of its interaction with Fkh2 (1, 9, 29, 37, 48, 50, 67). In contrast, these same studies showed an absolute requirement for Mcm1 in Fkh binding to promoters of CLB2 cluster genes. Furthermore, residual periodic expression of CLB2 and other CLB2 cluster genes was observed in fkh1Δ fkh2Δ mutants, whereas transcription was abrogated by Mcm1 depletion (1, 28, 37, 96).
Two lines of evidence indicate that Mcm1 and the two Fkh proteins regulate PHO5 directly. First, simultaneous mutation of candidate binding sites for Mcm1 and the forkheads in PPHO5 severely impaired rAPase activity and cell cycle periodicity of PHO5 mRNA (Fig. (Fig.66 and and7).7). Second, ChIP analysis showed specific, cell cycle-dependent association of Mcm1, Fkh2 and, to a lesser extent, Fkh1, with PHO5 over several time points in G2/M (Fig. (Fig.8C,8C, bottom panel). Since Fkh1 and Fkh2 bind the same sequence, they competed with each other for occupancy in vivo at five of five tested CLB2 cluster promoters (28). It is likely, therefore, that binding of each Fkh protein would increase at PPHO5 in the absence of the other (i.e., in FKH1-6HA fkh2Δ and FKH2-18Myc fkh1Δ strains). Consistent with Fkh2 possessing an Mcm1-interacting region that is absent from Fkh1, CLB2, SWI5, and other targets are also bound more strongly by Fkh2 than Fkh1 in vivo (9, 29, 34). Independent support for PHO5 being a bona fide Fkh target stems from detection of Fkh2 binding by genome-wide ChIP in H2O2-treated asynchronous cultures (26, 49). Under these conditions, Fkh2 possibly serves to coordinate Pi uptake and nucleotide synthesis for repair of oxidative lesions in DNA. Our results extend the role of Mcm1-Fkh2 beyond G2/M progression to include peak PHO5 expression in M/G1 in response to Pi levels that fluctuate during the cell cycle. Lower occupancy of PHO5 by Mcm1-Fkh2 (compared to CLB2) may in part explain the delay of PHO5 expression until M/G1.
Association of Mcm1-Fkh2 with PHO5 in synchronized cdc28-13 cells was highest at and immediately after release from G1 arrest (Fig. (Fig.8C,8C, bottom panel). As Cln-associated kinase activity subsequently climbed (56), Mcm1-Fkh2 dissociated, but did not completely disappear, and then was rerecruited in G2. We previously showed that Fkh2 binding to CLB2 was low at the α-factor arrest point (92). Mcm1 binding to several M/G1-specific genes as determined by ChIP was also generally low in α-factor-arrested cells (51). The variation in initial promoter occupancies likely reflects differences between α-factor and cdc28-13 arrests. The seeming paradox of recruiting Mcm1-Fkh2 in G1 when PHO5 was transcriptionally silent is potentially resolved by recent reports that Fkh2 recruits the Rpd3 histone deacetylase, most likely as a component of the Rpd3L complex (12, 31, 87, 92). Rpd3 has previously been shown to associate directly with PHO5 in asynchronous cultures in log-phase growth (39). We have extended this observation by showing that Sds3, a subunit specific to Rpd3L, exhibits a peak of association with PPHO5 in G1 phase of the cell cycle prior to START (Fig. (Fig.10).10). It is possible that Rpd3 histone deacetylase activity helps establish and/or maintain the repressive chromatin configuration that silences PHO5 and CLB2 transcription in G1 phase of cycling cells (74, 77). Rpd3 is released from CLB2 by Cln-kinase activity as cells progress through START (87), much like the temporal binding profile for Mcm1-Fkh2 that we observed at PHO5. Thus, it is tempting to speculate that release of Rpd3L both at CLB2 and PHO5 may in part be attributable to phosphorylation-mediated dissociation of Mcm1, Fkh2 or both. Reassociation of Mcm1-Fkh2 in G2 is consistent with negative feedback of B-type cyclin-CDK activity on G1 cyclin activity (2). If Cln-Cdc28 phosphorylates Mcm1 and/or Fkh2, their affinity for DNA is likely to be decreased, but not abolished, because both factors associate significantly with CLB2 throughout the cell cycle (34, 87, 92) (Fig. (Fig.8C,8C, middle panel).
Other genome-wide binding studies using asynchronous cultures detected binding of Mcm1 and Fkh to CLB2 cluster promoters, but not to PPHO5 (41, 79, 86). A straightforward explanation for positive binding to CLB2 in these studies is localization to an extended nuclease-hypersensitive site of 4 to 5 Mcm1 sites and the single known Fkh site at this gene (1, 38, 46, 50, 77). Cooperative binding to this region would account for the high occupancy of Mcm1 and forkheads in both late G1 (α-factor arrest) and M phase (Noc arrest) (1, 34). We have built on these findings by showing that Mcm1 binding to CLB2 is abundant at all cell cycle stages and is not a consequence of growth arrest (92) (Fig. (Fig.8C,8C, middle panel). In contrast, association of Mcm1-Fkh2 with its single site at PHO5 oscillates considerably during the cell cycle (Fig. (Fig.8C,8C, bottom), and synchronization of the cell population is required for adequate ChIP sensitivity. Pho4 binding at PHO5 was also undetectable by ChIP in cells growing asynchronously in rich medium (13). Fusion of epitope tags to the C terminus of Mcm1 would further adversely affect the sensitivity of genome-wide binding studies using asynchronous cultures (26, 41, 49, 79), since such tagging reduces Mcm1 protein copy number (S. Pondugula and M. P. Kladde, unpublished observations). This is especially problematic because the level of WT untagged Mcm1 is already rate-limiting for PHO5 mitotic activation (Fig. (Fig.22 and 3B and C) and presumably for promoter occupancy.
Our genetic (Fig. (Fig.66 and and7)7) and ChIP (Fig. (Fig.8C8C and and9)9) results have demonstrated that Mcm1 directly upregulates PHO5 transcription through at least one additional pathway parallel to PHO signaling (Fig. (Fig.11).11). Specifically, we found that strains containing both a PHO4 deletion and mutations in Mcm1 and/or Fkh binding sites had further attenuated rAPase activity compared to cells bearing either the PHO4 or promoter mutation(s) alone (Fig. (Fig.7).7). This strongly suggests that Mcm1 and Pho4 induce PHO5 via independent and additive pathways. Interestingly, PHO5 mitotic expression was reduced by a factor of ~2 in the complete absence of both Fkh1 and Fkh2, i.e., fkh1Δ fkh2Δ haploid cells (Fig. (Fig.4),4), and when Mcm1 levels were approximately halved in diploid cells containing one repressed copy of MCM1 (Fig. (Fig.2).2). While this could merely be coincidental, it does suggest that Mcm1 plays a more prominent role in the mitotic induction of PHO5 transcription. Furthermore, whereas the loss of polyP does not even partially suppress the dramatic PHO5 mitotic defect upon Mcm1 depletion (Fig. (Fig.3C),3C), it does restore PHO5 expression to WT levels in the fkh1Δ fkh2Δ double mutant (data not shown). Since Fkh binding in vivo requires Mcm1 but not vice versa (1, 9, 29, 37, 48, 50, 67), this differential suppression strongly suggests that Mcm1 activates PHO5 in mitosis via distinct pathways; a minor one involving the forkheads (and perhaps Ndd1) plus an essential one where Mcm1 acts either alone or in conjunction with an unknown cofactor (Fig. (Fig.11).11). This situation at PHO5 may be entirely different than that at CLB2, whose cell cycle regulation is driven primarily, if not solely, by Mcm1-Fkh2, because PHO5 is also induced by Pho4-Pho2. Importantly, then, Mcm1 and Pho4-Pho2 are both essential in combinatorial control of PHO5 mitotic induction (57) (Fig. (Fig.3C3C and and77).
Why link PHO5 expression to the cell cycle? We previously found that PHO5 mRNA and vacuolar polyP stores oscillate inversely during the cell cycle (57). Maximal levels of polyP were present in late G1 and plunged to a minimum after the majority of cells had entered S phase. Once polyP reserves were exhausted and intracellular Pi concentration declined further, PHO5 transactivation was signaled via the PHO pathway (32, 43). The severity of Pi depletion influences the extent of Pho4 nuclear retention, occupancy at the PHO5 promoter, and rAPase expression (4, 20, 84, 85). Therefore, phm3Δ (= vtc4Δ) cells lacking detectable polyP (57, 61, 84) are apparently starved more severely for intracellular Pi compared to WT, and hence PHO5 activation occurred prematurely in the cell cycle and was substantially enhanced in magnitude (57). Consistent with this, Mcm1 increased occupancy of PPHO5 after S phase (Fig. (Fig.8C,8C, bottom panel), the cell cycle stage in which polyP is depleted (57), which preceded accumulation of PHO5 mRNA from G2 through M/G1 (Fig. (Fig.8C,8C, upper panel). Mcm1 binding also increases dramatically with prolonged Pi starvation (D. W. Neef, M. M. Reynolds, and M. P. Kladde, unpublished observations). Furthermore, PHO5 was strongly induced after shifting M-phase-arrested cells to Pi-free medium (Fig. (Fig.5).5). This convincingly demonstrates that yeast are able to sense and respond to low levels of Pi in phases besides G1, where nutrients and cell size are gauged in preparation for START (44). During activation in G2/M, Fkh2 and Ndd1 are phosphorylated by Clb-associated kinase (19, 66, 67, 70). Additional observations linking PHO signaling to the cell cycle include (i) the fact that phosphorylation of Pho2, possibly by Clb-Cdc28, is required to promote its direct interaction with Pho4 (45) and (ii) PHO4 mRNA peaks in late S or early G2 phase (68). Thus, in closing, phosphate homeostasis during the cell cycle is maintained through parallel contributions of Mcm1 and Mcm1-Fkh2 activity and the canonical PHO pathway (Fig. (Fig.11),11), essentially coupling Pho80-Pho85 and Clb-Cdc28 CDK activities to PHO5 mitotic activation.
We are grateful to C. Fox for providing FKH deletion strains, E. Herrero and EUROSCARF for constructs we adapted for Dox-dependent regulation of MCM1, and A. Pirrozo for help in constructing pAAP1366. We also thank M. Bryk for advice with ChIP analysis and S. Hoose, C. Pardo, and C. Ishida-Gutierrez for helpful discussions and critical reading of the manuscript.
This study was supported by Public Health Service grants CA95525 from the National Cancer Institute to M.P.K. and GM39067 and GM48624 to D.J.S.
Published ahead of print on 13 July 2009.