In budding yeast, many genes are induced early in the cell cycle. Induction of these genes has been predominantly attributed to two transcription factors, Swi4-Swi6 (SBF) and Mbp1-Swi6 (MBF). Swi4 and Mbp1 are related DNA-binding proteins with dissimilar target sequences. For most G1/S-regulated genes that we tested in a cdc20 block-release protocol for cell-cycle synchronization, removal of both Swi4 and Mbp1 was necessary and sufficient to essentially eliminate cell-cycle-regulated expression. Detectable SBF or MBF binding sites (SCBs or MCBs) in the promoters or available genome-wide promoter occupancy data do not consistently explain this functional overlap. The overlapping ability of these transcription factors to regulate many promoters with very similar cell-cycle kinetics may provide robustness to the G1/S transcriptional response, but poses a puzzle with respect to promoter-transcription factor specificity. In addition, for some genes, deletion of Mbp1 or Swi4 enhances transcription, suggesting that these factors can also function as transcriptional repressors. Finally, we observe residual G1/S transcriptional regulation in the absence of Swi4 and Mbp1.
COMMITMENT to the cell cycle in eukaryotic cells occurs late in G1 at a point termed Start (Pringle and Hartwell 1981). Start coincides with a peak in transcription of >200 genes including the G1 cyclins CLN1 and CLN2, the B-type cyclins CLB5 and CLB6, and the mating-type switch endonuclease HO, along with many other genes involved in DNA synthesis, budding, and spindle pole body duplication (Cho et al. 1998; Spellman et al. 1998). Timely G1-specific transcription requires the G1 cyclin Cln3 (Tyers et al. 1993; Dirick et al. 1995; Stuart and Wittenberg 1995) and the cyclin-dependent kinase, Cdc28 (Koch et al. 1996). Cln3-Cdc28 promotes transcriptional activation by formation of the RNA polymerase II holoenzyme at the TATA boxes of the G1 cyclins CLN1 and CLN2, among others (Cosma et al. 2001). Cln3 activates cell-cycle initiation in a dosage-dependent manner (Cross 1988; Tyers et al. 1993). Regulation of cell size, pheromone sensitivity, and budding by Cln3 are dependent on the transcription factors MluI cell-cycle box binding factor (MBF) and Swi4/6 cell-cycle box binding factor (SBF) (Wijnen et al. 2002). Cln1 and Cln2 activation may represent the key event of Start. Cln1,2-Cdc28 complexes are involved in activating a number of pathways, including activation of B-type cyclins, bud emergence, and microtubule organizing center duplication (reviewed in Cross 1995).
Much of the Start transcriptional program depends on two related transcription factors, MBF and SBF (reviewed by Koch and Nasmyth 1994). MBF is composed of two proteins: Swi6, the trans-activating component, and Mbp1, the DNA-binding component. Mbp1 recognizes the MluI cell-cycle box (MCB), ACGCG (Koch et al. 1993). One study has found that Stb1, a Swi6-binding protein involved in Start transcription (Ho et al. 1999), is a specific regulator of MBF-dependent transcription (Costanzo et al. 2003). In addition to its role as a G1/S transcriptional activator, a role for MBF as a repressor during other stages of the cell cycle has been proposed (Koch et al. 1993).
SBF is composed of two proteins: Swi6 and Swi4, a homolog of Mbp1 that acts as the DNA-binding component of SBF. SBF recognizes the Swi4/6 cell-cycle box (SCB) CRCGAAA (Breeden and Nasmyth 1987; Andrews and Herskowitz 1989b; Taba et al. 1991). Consistent with the roles of SBF and MBF as DNA-binding factors that coordinately regulate the broad G1/S transcriptional regulatory program, three genome-wide location analyses show binding of MBF and SBF to a broad range of targets, in which SCB and MCB sites are enriched (Iyer et al. 2001; Simon et al. 2001; Harbison et al. 2004).
Whi5, a negative regulator of cell-cycle Start, was identified by small cell size in whi5 deletion strains (Jorgensen et al. 2002; Zhang et al. 2002). Whi5 binds to SBF and acts as a repressor of G1-specific transcription. Dissociation of Whi5 from SBF is dependent on Cln3, and in the absence of Whi5 the requirement for Cln3 in transcriptional activation is much reduced (Costanzo et al. 2004; De Bruin et al. 2004). A similar role for Whi5 as a repressor of MBF-dependent transcription has also been proposed (Costanzo et al. 2004).
The single deletion mutants swi4 and mbp1 are viable in Saccharomyces cerevisiae. However, the double deletion mutant mbp1 swi4 is inviable with arrest occurring in G1. The mbp1 swi4 lethality can be rescued by constitutive expression of CLN2, suggesting that the lethality stems primarily from failure to express G1 cyclins (Koch et al. 1993). A role for Cln3p as an activator of MBF and SBF has been proposed (Wijnen et al. 2002). However, cln3 deletion mutants are viable with delayed expression of MBF- and SBF-regulated genes whereas mbp1 swi4 strains are inviable, suggesting additional activators of MBF and SBF. BCK2 and CTR9 encode candidate SBF and MBF activators that are essential in the absence of Cln3 (Epstein and Cross 1994; Di Como et al. 1995; Koch et al. 1999). Furthermore, in cln3 deletion strains CLN1 and CLN2 are sufficient for transcriptional activation via a model of positive feedback (Cross and Tinkelenberg 1991; Dirick and Nasmyth 1991; Nasmyth and Dirick 1991; Tyers et al. 1993; Dirick et al. 1995; Stuart and Wittenberg 1995; Costanzo et al. 2004; De Bruin et al. 2004), in which Cln1,2 can inactivate Whi5 and/or directly activate SBF/MBF, thus driving their own transcription.
Despite the substantial amount of information that has been accumulated about SBF and MBF and the control of G1/S-regulated transcription, some mysteries remain. In some cases, it has been noted that removal of either Swi4 or Mbp1 has at best minor effects on the transcriptional activation of putative SBF or MBF target genes, respectively (Koch et al. 1993; Cross et al. 1994). In some circumstances Swi4 and Mbp1 may be functionally redundant, either because of the presence of both SCBs and MCBs in the promoters of some genes (e.g., CLN2; Stuart and Wittenberg 1994) or because of cross-binding of SBF and MBF to MCBs and SCBs, respectively (Koch et al. 1993; Partridge et al. 1997; Taylor et al. 2000). Functional redundancy of Swi4 and Mbp1 has never been tested directly, primarily because the lethality of swi4 mbp1 double mutants precludes simple analysis.
Here we directly analyze overlap of Swi4 and Mbp1 in the G1/S transcriptional regulation of multiple genes. We wished to distinguish between (i) Swi4- or Mbp1-specific regulation, so that the appropriate single deletion would abolish regulation; (ii) overlapping function of Swi4 and Mbp1, so that only the double deletion would abolish regulation; or (iii) Swi4- and Mbp1-independent regulation.
MATERIALS AND METHODS
Strains and plasmids:
All yeast strains used are described in Table 1. JB04-15D is a segregant of a cross between 2147-7C and K3294. JB03-19C is a segregant of a cross between 2147-7C and K2299. Plasmid pI411 was created by inserting ∼900 bp of the CLN3 promoter fused to 1.6 kb CLN2 coding sequence with three HA tags at the 3′ end between the SalI and SacII sites in the multiple cloning site (MCS) of pRS304. A SpeI digest of pI411 was used to transform 2147-7C by integration at the CLN2 locus to create JB13. Plasmid pKL035 was created by inserting ∼700 bp of the GAL-inducible promoter fused to 1.6 kb CLN2 coding sequence with three HA tags at the 3′ end between the SalI and SacII sites in the MCS of pRS404. A SpeI digest of pKL035 was used to transform 2147-7C by integration at the CLN2 locus to create JB14. JB21-2C was created in a similar way to JB13 by first transforming strain K2299 with SpeI-digested pI411 and then crossing the resulting transformed strain to JB04-15D and selecting segregants. All integrations were confirmed by an SspI Southern blot.
Growth and synchronization procedures:
YEP medium was used for all experiments, supplemented with the appropriate carbon source as indicated below. Yeast cultures were shaken at 250 rpm, in a volume no more than 20% of the container maximum at 30°. Cell-cycle synchronization was achieved by the cdc20 GALL-CDC20 block release in 2147-7C, JB13, JB03-19C, JB04-15D, and JB21-2C by growing cells to early log phase in YEP + galactose (3%) and then filtering them and growing them in YEP + glucose (2%) for 3 hr to arrest cells in metaphase. GALL is a truncated version of the GAL1 promoter that shows inducible but significantly lower expression than the full-length GAL1 promoter (Mumberg et al. 1994). Cells were released from the block by filtering back into YEP + galactose (3%).
Northern blot analysis:
Procedures for Northern blot analysis of mRNA were as described previously (McKinney et al. 1993). RNA was harvested from synchronized cultures every 15 min for 2–3 hr. Probes were generated by polymerase chain reaction using Research Genetics (Birmingham, AL) primer pairs for each gene tested. DNA fragments were radiolabeled by random-prime labeling using a Prime-It II kit (Stratagene, La Jolla, CA), and transcript levels were visualized and quantitated using a Molecular Dynamics (Sunnyvale, CA) STORM PhosphorImager system. Each blot contained a mutant time course and the relevant wild-type control time course run below it (2147-7C for JB03-19C and JB04-15D; JB13 for JB21-2C). All blots were probed with TCM1 as a loading control. TCM1 expression has been previously shown to be non-cell-cycle regulated and useful as a loading control for Northern blots (McKinney et al. 1993; Oehlen et al. 1996); lack of cell-cycle regulation of TCM1 was also observed in the genome-wide study of Spellman et al. (1998). To quantitate expression levels, each lane was background subtracted and normalized to the TCM1 signal. Duplicate, and in some cases triplicate, Northern blots for each gene gave highly reproducible expression profiles. For calculating expression levels, a single representative Northern blot was used for each gene. To calculate relative expression (Figure 1), normalized expression levels for each mutant were expressed relative to the peak expression for its particular control. For simplicity, in Figure 1 only one control time course [wild type (WT)] is shown for each gene. To provide a single quantitative measure of the degree of regulated expression in these experiments, we analyzed peak:trough ratios of expression, calculated as follows. For the two control strains (2147-7C and JB13) (Figure 2A), ratios of expression were defined as peak divided by trough value, where the peak is normalized to 1. The trough value is defined as the raw trough value plus an error (σ). σ is defined as the square root of the n − 1 average of squared differences in trough values in two replicate Northern blots of strain 2147-7C for 25 genes. The ratio of two variables is obviously very large when one is small, but intuitively the denominator should not be allowed to get smaller than the measurement error. A precise formula can be derived for the average of the logarithm of two variables, each distributed with a mean and standard deviation. In the context of actual measurements this implies that intensities (after background subtraction) are positive, and that the standard deviation should be added to small quantities before taking the ratio and the logarithm. For the ratios of expression of 2147-7C the average of two replicate Northern blots for each gene was used. The ratios of expression found by microarray hybridization (Spellman et al., 1998; Figure 2A) were calculated by averaging expression ratios from four different methods of cell synchronization. To calculate ratios of expression in mutant strains relative to that in control strains (Figure 2B), the following was done: Mutant expression was defined as the normalized peak value of the mutant minus the average normalized trough value of the mutant. Control expression was defined as the normalized peak value of the control minus the average normalized trough value of the control. The ratio of expression was defined as mutant expression divided by control expression. We recognize that this metric, while convenient, does ignore some kinetic details of precise peak timing and peak width (see results). In all G1/S-regulated genes tested for which any regulated expression was detected in the various mutant strains, the peaks of expression remained at approximately the same point in the cell cycle as those in the wild-type strain. In none of the genes tested was there significant unregulated expression or peak expression at any point in the cell cycle other than G1/S. The raw quantitation of all the Northern blot data used is available upon request.
Analysis of published genome-wide binding data sets:
To compare the binding of MBF and SBF found in previous published genome-wide location analyses (Figure 2D; Figure 3) we used the supplementary data available from three data sets (Iyer et al. 2001, http://genome-www.stanford.edu/chromatinip/; Simon et al. 2001, http://web.wi.mit.edu/young/cellcycle; Harbison et al. 2004, http://web.wi.mit.edu/young/regulatory_code). For the first data set (Iyer et al. 2001), the figure3_data.xls file was used to identify MBF-bound, SBF-bound, and MBF + SBF-bound ORFs, including divergently transcribed ORFs, which satisfied the authors' criteria. For the second data set (Simon et al. 2001), the all_genes.xls file was used to sort out all ORFs with P-values <0.001 for binding of Mbp1 (or Mbp1 + Swi6) and Swi4. For the third data set (Harbison et al. 2004), the pvalbygene_forpaper_abbr.xls file was used to sort out all ORFs with P-values <0.001 for binding of Mbp1 (or Mbp1 + Swi6) and Swi4. For all lists any duplicate ORFs were removed and care was taken to correct any alternate ORF names. For all comparisons the systematic yeast gene names were used. A UNIX shell script was used to compare the MBF-only-, SBF-only-, and MBF + SBF-bound gene lists from the three data sets. To determine the conservation of MBF and SBF binding sites across sensu strictu Saccharomyces species, we used the yeast regulatory map available from the third data set (Harbison et al. 2004) at http://jura.wi.mit.edu/fraenkel/regcode/.
MCB and SCB consensus site counts:
To count the number of MCBs and/or SCBs in the promoters of the genes tested, a regular expression search of the upstream intergenic sequence of each gene was performed. For this analysis, we defined MCB as ACGCG and SCB as CRCGAAA (where R represents A or G).
ChIP was performed as described (Wilmes et al. 2004), with some modifications. Cells from an untagged strain (2147-7C), a Swi4-Myc-tagged strain (JB05-1B), and a Mbp1-Myc-tagged strain (JB06-1A) were synchronized by the cdc20 GALL-CDC20 block release at midlog phase. Cells were formaldehyde fixed 40 min after release, which is at approximately the peak of G1/S-regulated gene expression in this protocol. One-tenth of the whole-cell extract was set aside for input DNA. Mbp1-Myc and Swi4-Myc were immunoprecipitated overnight with a 1:250 dilution of 9E11 monoclonal antibody (Genetex). For amplification by PCR, one-fiftieth of the immunoprecipitated DNA and one-fiftieth of the input DNA were used as template. Multiplex PCR was performed using oligos for the non-G1/S-regulated URA3 as a control along with G1/S-regulated promoter-specific oligos. PCR was performed using PuReTaq Ready-To-Go PCR Beads (Amersham Biosciences, Arlington Heights, IL) in a RoboCycler 96 temperature cycler (Stratagene), separated on 1.5% agarose gels, and stained with 1× Sybr Green (Molecular Probes, Eugene, OR). Primer sequences are available upon request.
Mode cell volume measurements were made using a Beckman Coulter Z Series Z2 Coulter particle count and size analyzer. All cultures were grown to midlog phase in YPD and sonicated before analysis. For each strain, triplicate counts were obtained from the same culture and averaged. Mode cell volume for each strain was then obtained from those averaged size profiles.
Construction of a set of strains allowing cell-cycle synchronization in the absence of Swi4, Mbp1, or both:
To examine the roles of MBF and SBF as transcriptional regulators we have made use of single deletion strains of mbp1 and swi4. In addition, we have created double deletion strains of mbp1 swi4. To construct the normally inviable double mutant strain, we made use of the observation of Koch et al. (1993) that mbp1 swi4 mutants were viable when CLN2 was placed under control of a constitutive promoter. We observed reasonably efficient rescue of viable mbp1 swi4 mutant strains by inclusion of an integrated construct containing CLN2 under the control of the CLN3 promoter, PCLN3-CLN2 (Valdivieso et al. 1993; Levine et al. 1996).
To facilitate looking at mRNA levels throughout the cell cycle, all three deletion strains mentioned above along with all their respective controls were made synchronizable by the deletion of cdc20 and the addition of CDC20 under the control of the galactose-inducible promoter (GALL-CDC20) (Table 1). In the absence of Cdc20, cells arrest in metaphase due to inability to degrade the separase inhibitor Pds1 and the B-type cyclins Clb5 and Clb2 (Zachariae and Nasmyth 1999; Shirayama et al. 1999; Wasch and Cross 2002). Hence, galactose withdrawal from these strains results in a metaphase block, which is rapidly released upon readdition of galactose, giving a synchronous cell cycle. This M-phase block-release protocol may be preferable for the present purposes to the G1/S block-release protocols used previously to examine Swi4- and Mbp1-dependent expression (Koch et al. 1993; Cross et al. 1994; Stuart and Wittenberg 1994), since it allows a synchronized M/G1 transition in which regulated expression of SIC1 can be observed to confirm synchrony, and also because it separates the block release temporally from the G1/S transition where most Swi4- and Mbp1-dependent expression is thought to occur.
To assess the synchrony of the mbp1 swi4 PCLN3-CLN2 cdc20 GALL-CDC20 strain upon cdc20 block release, we examined expression of two cell-cycle-regulated genes expressed at different parts of the cell cycle, neither of which is thought to be under control of SBF/MBF. Expression of the cyclin-dependent kinase inhibitor SIC1 is induced in M/G1 by Swi5p (Toyn et al. 1996). SIC1 expression was regulated in a cell-cycle-dependent manner in the mbp1 swi4 double mutant similar to that in the respective control strain, indicating that the double mutant strain exits mitosis and activates Swi5 on schedule upon release of the cdc20 block. In addition, accumulation of SIC1 transcript in a second weaker peak after release in both wild type and the mbp1swi4 mutant indicates successful completion of a second mitosis following release of the block (Figure 1). Likewise, the SWI5 transcription factor itself, which is transcribed in S, G2, and M (Nasmyth et al. 1987) under control of Ndd1/Mcm1p/Fkh1,2 (Lydall et al. 1991; Althoefer et al. 1995; Koranda et al. 2000; Zhu et al. 2000), was regulated in a normal cell-cycle-dependent manner in the mbp1 swi4 double mutant (Figure 1). These controls indicate that the cell cycle proceeds with relatively normal kinetics in the mbp1 swi4 mutant following release of the mitotic cdc20 block through the next mitosis, as indicated by probing expression of genes controlled by transcription factors other than SBF/MBF, thus validating this assay for examination of genes thought to be under direct control of SBF/MBF.
Examination of a subset of genes in the G1/S program for Swi4/Mbp1 requirements:
To dissect the specificity and/or redundancy of SBF and MBF for transcriptional regulation of G1/S targets, we looked at expression profiles of cdc20 GALL-CDC20 synchronized cultures of mbp1, swi4, and mbp1 swi4 PCLN3-CLN2 strains (Table 1) by Northern blot analysis. We analyzed expression levels in the cdc20 block-release protocol of 23 G1/S transcripts in all these strains (Figure 2), as well as four additional transcripts analyzed only in WT and the single mbp1 and swi4 mutants (CLN1, CLN2, PCL1, and CLB6; data not shown). These transcripts were selected on the basis of the following criteria: A number of previously identified canonical SBF and MBF targets were selected to test along with a random sampling of transcripts found to be at least moderately cell-cycle regulated (Spellman et al. 1998) and weighted in favor of those with a consensus with respect to Mbp1 and/or Swi4 binding in the genome-wide binding data (Iyer et al. 2001; Simon et al. 2001; Harbison et al. 2004). Sample data for some of these transcripts are presented in Figure 1, and a compilation of the data for 23 G1/S transcripts and the two controls discussed above (SIC1 and SWI5) is shown in Figure 2B.
Because of the history of strain construction in this project, we used different controls for the single mbp1 and swi4 mutants than for the double mbp1 swi4 mutant. For the single mutants, the control strain was a simple MBP1 SWI4 wild type, while for the double swi4 mbp1 mutant, which was rescued by PCLN3-CLN2, the control strain was MBP1 SWI4 PCLN3-CLN2. To compare among all the experiments, it is important to establish that in the wild-type context, PCLN3-CLN2 did not significantly affect G1/S-regulated gene expression. Figure 2A demonstrates that this is the case: The peak/trough ratios for the 25 genes tested are generally quite comparable with or without PCLN3-CLN2 (CWP1 displays a significant effect of PCLN3-CLN2 on its peak/trough ratio relative to that of wild type and the effect of this on the ratios of expression in mutants is discussed below; a less dramatic effect of PCLN3-CLN2 is also seen in SWI5). These peak/trough ratios are also generally comparable to the genome-wide data of Spellman et al. (1998) (shown in Figure 2A), obtained using several different methods of cell-cycle synchronization.
In Figure 2B, the data are condensed for each transcript as the relative cell-cycle-regulated induction (peak/trough) for each gene in the mbp1 swi4 double mutant compared to that in wild type (Figure 2B, top, green bars) or in the mbp1 or swi4 single mutants compared to that in wild type (Figure 2B, bottom, blue and red bars, respectively). Fifteen of the G1/S targets showed evidence of being under redundant positive control of MBF and SBF since their peak/trough expression ratios were not significantly reduced by single deletion of MBP1 or SWI4 but were significantly reduced by concurrent removal of both MBP1 and SWI4. Seven genes, including the G1 cyclins CLN1, CLN2, PCL1, and the B-type cyclin CLB6 (not shown) and YHP1 and YOX1 (Figure 1; Figure 2B) had significantly reduced peaks of expression after removal of SWI4 but were unaffected by removal of MBP1. Only one gene tested, CDC45, was found to have peak expression reduced significantly by a single deletion of MBP1 but to be unaffected by deletion of SWI4.
Interestingly, the genes tested exhibited varying degrees of residual control in the swi4 mbp1 double mutant. In Figure 2B, we indicate semi-arbitrary “normal,” “reduced,” and “off” cutoffs of >50%, <50%, and <25% of wild type as visual aids; 10 genes are in the off class, 8 are in the reduced class, and 5 are in the normal class. The G1/S-regulated genes in the normal class, PLM2, EXG1, DUN1, and ERP3, clearly demonstrate the existence of functional G1/S cell-cycle control by other factors beyond Mbp1 and Swi4, and a similar conclusion can be reached from the broad class of genes showing reduced but still clearly detectable regulation in the swi4 mbp1 double mutant (CLB2, though within the normal class, is a unique case; see results hereafter and discussion).
While the quantitative measure we chose to characterize expression was peak/trough ratio, for many genes there was an apparent delay in peak expression due to removal of SWI4 or MBP1 (see Figure 1 for examples). Since these differences are almost all single 15-min time-point differences in these time courses, we cannot accurately quantitate the actual peak displacement, but we note that these differences suggest a degree of independent function of Swi4 and Mbp1 at many promoters, at least at early times during the induction.
Mbp1 and Swi4 can act as transcriptional repressors at some promoters:
Three genes showed increases in peak expression following removal of Mbp1: CLB2, SPT21, (each about threefold elevated) and CWP1, showing a remarkable eightfold increase in peak expression. As shown in Figure 2A, the peak:trough ratio of CWP1 measured in the wild-type control is reproducibly and significantly lower than that of all other genes tested and this fact contributes to the eightfold increase in peak expression of CWP1 in the mbp1 strain. In all cases, an essentially normal pattern of cell-cycle regulation of these genes was observed; the main difference was the sharply increased peak value. This result suggests that Mbp1 is acting as a negative regulator of expression of these genes specifically at the normal peak expression time. Previously, Mbp1 was proposed to inhibit expression of its target genes during times when the genes were turned off (Koch et al. 1993). Our observations here suggest in addition that Mbp1 can also be an inhibitor during the G1/S interval, when it is thought to be activated for expression of its canonical targets (Koch and Nasmyth 1994). The three genes exhibiting elevated expression in response to MBP1 deletion differ in their response to removal of Swi4, with CLB2 showing very strong Swi4 dependence compared to SPT21 and CWP1. Perhaps related to this, Clb2 but not the other two shows a paradoxical increase in expression in the swi4 mbp1 double mutant compared to that in the swi4 single mutant (see discussion).
PLM2 peak/trough ratio was increased in the absence of either Swi4 or Mbp1 and was more strongly increased in the absence of both factors, suggesting that Swi4 and Mbp1 are jointly decreasing peak PLM2 expression. ERP3, ELO1, TOS4, and MSB2 all exhibit increases in expression in the absence of Swi4 and/or Mbp1 (Figure 2B), suggesting varying patterns of repressive or activating effects of the two factors at these promoters.
Previous work showed derepression of some G1/S-regulated gene expression (TMP1, POL1, and CLB5) in mbp1 G1 cells (Koch et al. 1993). Under the synchronization conditions of this study, we do not see derepression of TMP or CLB5 in G1 cells (Figure 1 and data not shown).
Site locations in promoters and published promoter occupancy data do not explain the pattern of Swi4/Mbp1 requirements:
Three genome-wide location analyses reported the binding of MBF and SBF across all yeast promoters (Iyer et al. 2001; Simon et al. 2001; Harbison et al. 2004, respectively). We attempted to use these genome-wide binding data sets to explain the results of our Northern blot analyses, with limited success. Out of the 23 G1/S-regulated genes and two control genes we assayed, 18 were assigned the same binding patterns in all three data sets (Figure 2D). Since the list of genes we assayed by Northern blot was heavily weighted in favor of genes that shared Mbp1 and Swi4 binding by the genome-wide binding data sets, the extent of shared binding in this list cannot be used to assess the overall similarity of these data sets (see Figure 3). Even among those genes with the same binding pattern in the three genome-wide binding data sets, there is little predictive power for the effect of deletion of either Mbp1 or Swi4 or for the presence or absence of a consensus binding site.
CLB5 is an example of the failure of the consensus motifs and binding data to predict function. Two genome-wide location analyses (Simon et al. 2001; Harbison et al. 2004) found that the intergenic region upstream of CLB5 is bound by MBF but not by SBF whereas a third study (Iyer et al. 2001) found neither MBF nor SBF binding to the CLB5 promoter. Northern blot analysis of yeast strains deleted for mbp1, swi4, or mbp1 swi4 shows that expression of CLB5 transcript is not affected by deletion of mbp1 or swi4 alone but is completely abrogated by a double deletion of mbp1 swi4 (Figure 1; Figure 2B). This suggests that in the absence of MBF, SBF is able to regulate CLB5 transcription normally despite the absence of any canonical SCB. The CLB5 promoter contains multiple MCBs but no canonical SCBs (Figure 2C). The regulatory element that SBF is binding in the CLB5 promoter could be the MCBs or an unknown site (for example, a highly degenerate SCB) or SBF could be acting through indirect regulation (Horak et al. 2002).
As an additional example, evidence from all three genome-wide location analyses (Iyer et al. 2001; Simon et al. 2001; Harbison et al. 2004) suggests that RAD27, a gene that encodes a protein involved in DNA replication or repair, is bound by MBF and not by SBF. Its promoter contains several MCBs and one SCB. Northern blot analysis of mbp1, swi4, and mbp1 swi4 strains shows that RAD27 expression is not perturbed by either mbp1 or swi4 single deletions but is strongly affected by the mbp1 swi4 double deletion (Figure 2B). This result suggests that RAD27, like CLB5, can be regulated by either MBF or SBF.
This difference in Mbp1 and Swi4 binding among the G1/S-regulated genes we assayed underestimates the extent of the difference between the data sets across the entire yeast genome, as we used overlap among the binding data sets in initial selection of the set of G1/S-regulated genes used in our Northern blot analysis. Across the genome, the three data sets shared 11–20% of MBF-only targets, 21–25% of SBF-only targets, and 5–7% of MBF + SBF targets (Figure 3). The evident high variability among the data sets may provide the simplest explanation for why these data sets were not highly predictive of which genes would be specifically vs. redundantly controlled by Swi4 or Mbp1.
We note further that simple inspection of the promoter sequences for canonical SCB or MCB sites was similarly only poorly predictive of the factor(s) required for appropriate gene regulation (Figure 2; some specific examples discussed above). In this site-counting measurement, we used only a single conventional definition of the SCB or MCB consensus; while other site definitions will naturally give different numbers of sites, no alternative definitions that we have tried (in an ad hoc, nonexhaustive search) have given a clearly better fit to the data.
Chromatin immunoprecipitation of Mbp1 and Swi4 shows dual binding of promoters:
Our data suggest that Mbp1 and Swi4 overlap functionally in the control of many genes, with little correlation to the presence or absence of recognizable SCB/MCB sites in the promoters of these genes. This could be explained by the idea of cross-binding of Swi4 and Mbp1 to MCBs or SCBs, respectively. Consistent with this idea, it was proposed that Swi4 regulates CLN1 by binding to MCB elements in the CLN1 promoter (Partridge et al. 1997). Alternatively, the effects of Swi4 or Mbp1 on regulation of some of these genes could be indirect, since other transcription factors are potential Swi4 targets (Horak et al. 2002). To begin to distinguish between these possibilities, we used ChIP analysis to examine the binding of Swi4 or Mbp1 to several promoters (Figure 4). We tested the specificity of binding by three criteria. First, we compared the amount of promoter-specific PCR product from anti-Myc immunoprecipitated enriched pools of DNA (IPs) from tagged vs. untagged strains and observed significantly higher amounts of product from the tagged strains (Figure 4, A and B). Second, we carried out the PCR reactions with inclusion of both promoter-specific oligonucleotides and oligonucleotides for amplifying a fragment of URA3, as a negative control (Figure 4B). No specific enrichment of URA3 amplification in the IPs from the Swi4-tagged strain could be observed. Slight enrichment of URA3 amplification was detected in the IPs from the Mbp1-tagged strain in a few reactions (Figure 4B). Third, we tested the IPs for enrichment of a fragment overlapping the CLN3 promoter and coding sequence (Figure 4A). The CLN3 promoter possesses no canonical MCBs or SCBs and is not a G1/S-regulated gene (Spellman et al. 1998; Mackay et al. 2001). No specific binding of Swi4 to the CLN3 fragment was detected; however, there was variable enrichment of CLN3 in IPs from the Mbp1-tagged strain. Thus, it appears that while Swi4 binding to all three G1/S promoters tested is specific, background apparently derived from nonspecific Mbp1-DNA interactions means that the data are suggestive but not conclusive of specific Mbp1 binding to the G1/S promoters. (We speculate that some “stickiness” of Mbp1 for nonspecific DNA could account for some of the variation in results in the genome-wide ChIP-chip experiments discussed above).
The CLN2 promoter shows specific binding of Swi4 and probably of Mbp1 (Figure 4). CLN2 transcription is at least partially Swi4 dependent (see Introduction; our data not shown). This promoter contains several MCBs and SCBs (Stuart and Wittenberg 1994), so Swi4 and Mbp1 could be binding to their cognate SCB/MCB sites or could occur through cross-binding. The finding that both Mbp1 and Swi4 are bound to the CLN2 promoter is in keeping with a study that showed a decreased level of dimethylsulfate protection of the CLN2 promoter in a strain lacking both Mbp1 and Swi4 compared to one lacking Swi4 alone (Koch et al. 1996).
TOS4 is under dual regulation by Mbp1 and Swi4 (Figure 2). Paradoxically, its promoter possesses neither MCBs nor SCBs on the basis of simple counting of consensus sites; despite this, we find specific binding of Swi4, and evidence suggestive of specific binding of Mbp1, to this promoter (Figure 4). YOX1 is Swi4 dependent (Figure 2) despite the absence of canonical SCBs from its promoter. We find specific binding of both Swi4 and probably Mbp1 to this promoter (Figure 4). It is possible that Swi4 is binding to the MCBs present in the YOX1 promoter and regulating expression in a manner analogous to Swi4 regulating CLN1 expression through MCBs in its promoter (Partridge et al. 1997). The YOX1 homolog YHP1 is similar to YOX1 in that it contains only canonical MCBs in its promoter but is controlled by Swi4 (Figure 2), and this promoter may show Swi4 and Mbp1 binding similar to YOX1 (preliminary data not shown).
In all three cases, CLN2, TOS4, and YOX1, the binding of Swi4 is not strikingly affected by the removal of Mbp1 and vice versa (data not shown). However, we note that the ChIP assay as we have carried it out is not a quantitative assay. Therefore, the question of quantitative regulation of the degree of Swi4 or Mbp1 binding to these promoters by the presence or absence of the other factor remains for future work.
Despite the caveats from the nonquantitative nature of these ChIP studies and the background binding problem for Mbp1, these results confirm the ability of Swi4 and probably Mbp1 to bind to promoters where simple sequence inspection makes it difficult to explain their binding. These findings could explain the functional overlap of Swi4 and Mbp1 in transcriptional regulation of these genes.
Evidence for function of Mbp1 in a SWI4 background:
The absence of SWI4 causes a notable phenotype, including slow growth, problems with morphogenesis, and large cell size (e.g., Ogas et al. 1991; Jorgensen et al. 2002; Igual et al. 1996). In contrast, little or no phenotypic consequence to deletion of MBP1 has ever been described to our knowledge. Perhaps consistent with this, of all the genes we tested, only CDC45 was found to be specifically Mbp1 dependent for its normal transcription. This could suggest the hypothesis that Mbp1 essentially functions solely in a backup pathway for Swi4. Since cell size is a sensitive indicator of cell-cycle progression overall (Jorgensen et al. 2002) we tested whether removal of Mbp1 altered cell size. We found a significant cell size phenotype for mbp1 strains (otherwise wild type) during exponential growth (Table 2). Loss of Mbp1 leads to a 20% increase in modal cell volume compared to that in controls. This volume increase is associated with a 5% increase in the proportion of budded cells, suggesting a possible delay in the replicative (budded) part of the cell cycle.
As noted above, many of the dually Swi4/Mbp1-regulated genes showed evidence for slight delays in peak expression in both the mbp1 and swi4 single mutants. The transcriptional delay in the mbp1 single mutants could be consistent with the moderate overall cell-cycle delay suggested by increased cell size and proportion of unbudded cells in the mutant (Table 2).
Do Mbp1 and Swi4 overlap in function?
This study was motivated by the clear importance of SBF and MBF for G1/S transcriptional regulation, combined with a number of observations showing puzzlingly small effects of deletion of the DNA-binding components of these factors on regulation of their putative targets (see Introduction). The standard view of SBF and MBF as direct and exclusive regulators of SCB-containing promoters and MCB-containing promoters, respectively, was clearly inadequate. The simple prediction that these factors might overlap in function at some promoters has been made (first by Koch et al. 1993), but was never tested directly by simultaneous deletion of the factors, due to inviability of the swi4 mbp1 double mutant. We exploited the observation that the swi4 mbp1 double mutant was viable upon ectopic expression of the G1 cyclin CLN2 (Koch et al. 1993) to directly test overlap of transcriptional regulation by SBF and MBF through the cell cycle.
Our results simplify understanding of G1/S-regulated expression in yeast: Swi4 and Mbp1 are substantially redundant with each other for control of the majority (19/23) of the G1/S-regulated genes that we tested (Figure 2). Only four genes [YOX1, YHP1, CDC45 (Figure 2B), and PCL1 (data not shown)] show specific Swi4 or Mbp1 dependence and lack of dependence on the other factor.
Swi4- and Mbp1-independent G1/S regulation:
Our data show clearly that G1/S activators of transcription that are independent of Swi4 and Mbp1 must exist. PLM2, DUN1, EXG1, and ERP3 are cell-cycle regulated at least as well in the absence as in the presence of Mbp1 and Swi4 (Figure 2B). We note in addition that many of the genes that we tested show reproducible (although significantly reduced) regulation of expression in the absence of Mbp1 and Swi4 (Figure 2B), suggesting Swi4- and Mbp1-independent regulation of G1/S accumulation of many transcripts, by an unknown mechanism.
Mbp1 and Swi4 as repressors:
A number of genes (SPT21, CWP1, CLB2, PLM2, ERP3, ELO1, TOS4, and MSB2) show increased peak expression in the absence of Swi4 and/or Mbp1 in various patterns, suggesting that Swi4 and/or Mbp1 may act as repressors for many genes. As noted above, this phenomenon is distinct from the previously characterized repressive effect of Mbp1 on some of its targets during G1, when the targets are normally not expressed (Koch et al. 1993).
CLB2 is a unique case in that its peak expression level is higher in a double mbp1 swi4 mutant than in the single swi4 mutant, in which expression is greatly reduced (Figure 2B). The Swi4 dependence of CLB2 expression, combined with the ability of Clb2 to antagonize Swi4-dependent gene expression (Amon et al. 1993), suggests the possibility of a negative feedback loop, whereby Swi4 would activate CLB2, and Clb2 would subsequently inactivate Swi4. This would prevent CLB2 expression from activating prematurely and then preventing Swi4-dependent expression from occurring at all. Swi4 dependence of CLB2 expression may be due to the repressive effects of Mbp1 on CLB2 transcription. Swi4 does bind to the CLB2 promoter by ChIP (in both the published genome-wide studies and in our unpublished data); we lack clear data as to whether Mbp1 binds the CLB2 promoter (data not shown). The near-normal cell-cycle regulation of CLB2 in the mbp1 swi4 double mutant (Figure 1) is presumably due to the previously characterized Ndd1/Mcm1/Fkh1,2 regulators (Lydall et al. 1991; Althoefer et al. 1995; Koranda et al. 2000; Zhu et al. 2000; Reynolds et al. 2003).
A remaining puzzle for specific site recognition:
For the large class of genes for which Swi4 and Mbp1 overlap in function, the mechanistic basis for the overlap is still unclear. The simplest explanation, which doubtless applies in some cases, is independent binding of Swi4 and Mbp1 to canonical binding sequences in some promoters (as may be the case for CLN2; Stuart and Wittenberg 1994). Another possibility is cross-binding of Swi4 and Mpb1 to canonical sequences for the other factor. There is biochemically detectable cross-binding of Swi4 to MCBs and of Mbp1 to SCBs (Dirick et al. 1992; Primig et al. 1992; Partridge et al. 1997; Taylor et al. 2000). Expression from simple SCB-element reporter plasmids is specifically dependent on Swi4 (Andrews and Herskowitz 1989a); thus, cross-binding of Mbp1 to SCBs may not be an adequate explanation in general. (The reciprocal measurement of the degree of exclusive dependence on Mbp1 for expression of MCB element reporter expression has not been reported to our knowledge.)
It is also possible that many promoters contain noncanonical binding sequences for Swi4 and/or Mbp1. The canonical sequences for binding Mbp1 or Swi4, ACGCG and CRCGAAA, respectively, are not very similar by inspection beyond the CRCG core (although the proteins use a different method of inspection than the human eye does!) In addition, both of these are rather short sequences, which occur many times in the genome presumably without effectively regulating transcription; it seems likely that any proposed hybrid sequence that could be recognized by both factors would have even less information content. This leaves us with little understanding of how these sequence-specific binding factors bind effectively to the “right” sites and not the “wrong” sites. Binding of other factors to SBF/MBF-regulated promoters may provide an environment that might encourage Swi4 or Mbp1 binding even to apparently nonconsensus sites.
The technique of measuring genome-wide promoter occupancy mapping by “Chip-ChIP” methods is a powerful and very useful method, and initially we expected that it would help to resolve many of these mysteries. Unfortunately, in the present specific experimental context we have not found the available public data sets to be useful in helping us to predict what genes will be regulated by Swi4 or Mbp1. It is also notable that multiple carefully performed studies using very similar reagents and procedures give binding occupancy data that are substantially nonoverlapping (Figure 3). We do not have any concrete suggestions to make here, but wish only to provide a caveat for the use of these very valuable data sets.
Gene duplication followed by partial divergence of function is a well-known theme in evolution and is frequently observed in transcription factors. This process provides fascinating challenges with respect to coevolution of binding specificities and binding sites and either divergence or conservation of functions of the related factors. Swi4 and Mbp1 present an excellent case study for working out some of these issues. Here we have shown that these factors have substantially conserved the ability to regulate the majority of their target genes. We document an apparent reversal of specificity in the case of several targets, including the main mitotic cyclin CLB2, for which Swi4 appears to be an activator and Mbp1 a repressor. Last, we pose a paradox with respect to specificity of DNA binding and promoter occupancy.
We thank Michael Mwangi, Rahul Siddharthan, and Jonathan Widom for helpful discussions and Kim Nasmyth and Richard Young for providing strains. We also thank Vincent Archambault for helpful advice regarding ChIP. This work was supported by a Howard Hughes Medical Institute Predoctoral Fellowship (J.M.B.), National Institutes of Health grant PHS GM047238 (F.R.C), and the Burroughs Wellcome Fund (F.R.C, E.D.S.).
Communicating editor: A. P. Mitchell
- Received April 18, 2005.
- Accepted June 7, 2005.
- Copyright © 2005 by the Genetics Society of America