Abstract
Proteins interact with genomic DNA to bring the genome to life; and these interactions also define many functional features of the genome. SBF and MBF are sequence-specific transcription factors that activate gene expression during the G1/S transition of the cell cycle in yeast1, 2. SBF is a heterodimer of Swi4 and Swi6, and MBF is a heterodimer of Mbp1 and Swi6 (refs 1, 3). The related Swi4 and Mbp1 proteins are the DNA-binding components of the respective factors, and Swi6 may have a regulatory function4, 5. A small number of SBF and MBF target genes have been identified3, 6, 7, 8, 9, 10. Here we define the genomic binding sites of the SBF and MBF transcription factors in vivo, by using DNA microarrays. In addition to the previously characterized targets, we have identified about 200 new putative targets. Our results support the hypothesis that SBF activated genes are predominantly involved in budding, and in membrane and cell-wall biosynthesis, whereas DNA replication and repair are the dominant functions among MBF activated genes6, 11. The functional specialization of these factors may provide a mechanism for independent regulation of distinct molecular processes that normally occur in synchrony during the mitotic cell cycle.
To identify the targets of SBF and MBF, we combined chromatin immunoprecipitation and microarray hybridization (Fig. 1). Proteins were crosslinked with formaldehyde to their target sites in vivo. DNA that was specifically crosslinked to either of the transcription factors was purified by immunoprecipitation using an antibody against either the native protein or an epitope tag that was fused to the protein. Polymerase chain reaction (PCR) analysis of immunoprecipitated DNA confirmed the specific association of Swi4, Swi6 and Mbp1 with several known target promoters, and other target promoters that are identified here (see Supplementary Information ). After reversal of the crosslinks, immunoprecipitated DNA was amplified and fluorescently labelled with the Cy5 fluorophore. We also fluorescently labelled a separate control DNA sample with the Cy3 fluorophore. Genomic target loci were identified by comparative hybridization of the immunoprecipitated and control DNA probes to a DNA microarray. The ratio of the Cy5 to Cy3 fluorescence intensities measured at each DNA element in the microarray provided a measure of the extent of binding of the transcription factor to the corresponding genomic locus. As the transcription factors were expected to interact primarily with promoters, we constructed DNA microarrays that represented all the intergenic intervals, as well as all the predicted coding sequences in the yeast genome. We plotted fluorescence ratios on a representation of the yeast genome to generate a map of the distribution of these proteins on the entire genome (see Supplementary Information).
Figure 1: Strategy for analysing genome-wide protein–DNA interactions.
![Figure 1 : Strategy for analysing genome-wide protein|[ndash]|DNA interactions.
Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, or to obtain a text description, please contact npg@nature.com](/nature/journal/v409/n6819/images/409533aa.0.jpg)
The reference probe can either consist of DNA generated in parallel from a strain bearing a deletion of the gene encoding the protein of interest (as depicted), or of unfractionated genomic DNA amplified and labelled in the same manner. Alternatively, an epitope-tagged version of the protein of interest can be immunoprecipitated with an antibody directed against the epitope. The DNA microarray includes all of the intergenic regions or promoters from the genome. The Cy5/Cy3 fluorescence ratio for each locus reflects its enrichment by immunoprecipitation (IP) and therefore, in general, its relative occupancy by the cognate protein.
High resolution image and legend (56K)To identify the most probable targets of SBF and MBF, we ranked genomic
loci according to their fluorescence ratios. We first determined the percentile
rank for each array element in each immunoprecipitation experiment, and then
the median percentile rank for each element across a set of experiments. The
distribution of the median percentile ranks for the 11 Swi4 immunoprecipitation
experiments was bimodal (Fig. 2a). Loci with median
percentile ranks above 93.8 formed a group that was distinct from the bulk
of the distribution. 163 loci (representing sequences upstream of 183 genes)
had median percentile ranks above this threshold. The average ranks of these
163 loci were significantly higher than would be expected if the rankings
in the individual immunoprecipitations were uncorrelated (P = 8.0 
10
-151). The genes associated with these loci therefore represent candidate
SBF targets (Fig. 3). When a putative binding site fell
between a pair of divergently transcribed genes, we considered both genes
as potential transcriptional targets. We use 'target' to mean
putative transcriptional target genes that are inferred from our immunoaffinity-microarray
results.
Figure 2: Defining criteria for identifying SBF and MBF targets.
a, Distribution of median percentile ranks of Cy5/Cy3 fluorescence
ratios from 11 independent microarray hybridizations analysing Swi4 binding.
The threshold for defining SBF targets is indicated by a dotted line.
b, Distribution of median ranks of Cy5/Cy3 fluorescence ratios from four
independent microarray hybridizations analysing Swi6 and Mbp1 binding.
c, Fraction of genes in a sliding 50-gene window whose expression profile
has a correlation coefficient of 
0.5 with that of a canonical
G1/S regulated gene, as a function of the median percentile ranks of Cy5/Cy3
fluorescence ratios from the MBF hybridizations. The threshold for defining
MBF targets is indicated by a dotted line.
Figure 3: Genomic targets of SBF and MBF.
Percentile ranks of intergenic fragments that meet selection thresholds
are inicated (blue–yellow colour scale). Loci with 
70%
overall nucleotide sequence identity to another yeast locus (potentially crosshybridizing)
are indicated (closed circles). The combination of Cy3 and Cy5 labelled probes,
the antibody used for IP (if used) and the culture conditions for each experiment
are summarized (left panel). Experiments 9, 10, 17 and 18 involved independent
crosslinking and IPs. DNA microarrays that included all yeast ORFs and other
features, in addition to the intergenic fragments, were used for experiments
3, 8, 13 and 14.
To identify MBF targets, we considered loci with high ranks in immunoprecipitation experiments targeting Mbp1 and Swi6. The median percentile ranks from the Mbp1 plus Swi6 immunoprecipitation experiments did not show a bimodal distribution (Fig. 2b); however, loci with median percentile ranks above 94 were highly enriched for genes with the G1/S expression profile, characteristic of known MBF targets (Fig. 2c). On the basis of this analysis, and on the distribution of known MBF target genes in the ranked list, we chose a median percentile rank of 94 as a threshold for considering loci to be MBF targets. Eighty-seven loci (upstream of 98 genes) ranked above this threshold (Fig. 3). Forty-three loci ranked above the selection thresholds for both SBF and MBF, and thus seemed to represent targets of both factors (see Methods).
Three lines of evidence support the conclusion that this survey has identified in vivo targets of SBF and MBF. First, the putative in vivo binding sites we identified are highly enriched for sequences matching the defined consensus binding sites. Second, most of the genes downstream of the putative binding sites are periodically expressed during the cell cycle with a peak in G1/S, a pattern typical of the previously identified targets of these factors. Third, putative target genes with known functions are highly enriched for functions related to DNA replication, budding and the cell cycle. We have, however, probably misidentified some genes as putative targets of these factors, and failed to identify some targets.
The putative SBF and MBF targets that we identified were highly enriched
for sequences matching the consensus binding sites for SBF and MBF12.
Roughly 49% of the unique intergenic fragments that were bound by Swi4 contained
a motif matching the defined SBF consensus binding site (CRCGAAA), whereas
only about 10% of all intergenic elements contain this site (P = 1.0 10
-30). Fifty-five per cent of the unique MBF intergenic targets contained
the MBF consensus binding site (ACGCGN), whereas roughly twenty per cent of
all intergenic fragments have such a site (P = 1.0 10
-9). Compared with the genome overall, there was also a greater tendency
for the SBF and MBF target promoters to have multiple binding sites for their
respective factors. The consensus site for SBF is found with roughly equal
frequency in open reading frames (ORFs) and intergenic regions in the yeast
genome. However, when we used DNA microarrays that contained nearly all of
the predicted ORFs, and the intergenic segments of the yeast genome, to map
Swi4 targets, we observed that in virtually every case in which an ORF had a
high percentile rank, its upstream intergenic fragment had a similarly high
rank (data not shown). Thus, despite the frequent occurrence of consensus
SBF binding sequences in coding sequences, SBF binds selectively to promoters
and not to coding sequences.
We used the pattern-finding programs MEME13 and Consensus14 to identify sequence motifs that were enriched in the high ranking (percentile rank > 98 for SBF and > 97 for MBF) target intergenic fragments. Both of these programs identified the consensus binding sites for SBF and MBF (CGCGAAAA and ACGCGN, respectively), based solely on their overabundance in the respective target loci. This suggests that these sequence motifs are indeed the primary determinants of binding of the respective factors to promoters (data not shown; and Supplementary Information). However, not every promoter bound by SBF or MBF in vivo contained a recognizable consensus binding site. Moreover, most of the coding sequences and many of the promoters that contain the consensus sequences show no evidence of binding to these factors in vivo.
SBF binding to DNA is cell-cycle regulated, occurring maximally at G1 (refs 15, 16). We saw only a modest
increase in fluorescence ratios measured for SBF target loci when cells were
treated with 
-factor, compared with the unsynchronized or benomyl-treated
cultures, perhaps because the fraction of G1 cells in unsynchronized cultures
or benomyl-treated cultures provided sufficient SBF-bound target sites for
enrichment by immunoprecipitation.
In a comparison of the global gene expression patterns in asynchronous
cultures of swi4
and wild-type cells, we found no significant
differences in expression of most of the putative SBF targets, including such
characterized targets as CLN1, PCL1, PCL2 and HO,
a result consistent with previous findings11 (See
Supplementary Information). Presumably, as the G1/S induction by SBF
is superimposed on a basal level of SBF-independent transcription17,
the effect would be more readily detected in synchronized cells. However,
the small set of genes whose transcript levels were significantly reduced
in the swi4 cells included several SBF targets identified by
chromatin immunoprecipitation, such as CWP1, CWP2, CIS3,
SVS1 and SRL1.
We examined the expression patterns of SBF and MBF target genes during
the mitotic cell cycle12, and sporulation18, using
published data (Fig. 4a). Whereas only 13% of all yeast
genes are transcriptionally cell-cycle regulated12, 66% of the
SBF targets and 68% of the MBF targets, as defined by the immunoprecipitation
results, were cell-cycle regulated (P = 1.0 10
-44 and 1.0 10-22, for SBF and
MBF, respectively). Genes with maximal transcript levels during G1 and S phase
showed the greatest enrichment, but there was also a less significant enrichment
for genes whose transcripts peaked during other stages of the cell cycle.
The putative target genes that were not cell-cycle regulated had slightly
lower immunoprecipitation-selection percentile ranks and a lower frequency
of consensus MBF or SBF sites than the cell-cycle regulated targets, suggesting
that at least some of these genes represent false positives in our immunoprecipitation
assay. This bias was more pronounced for the putative MBF targets. There was,
however, a significant enrichment for the presence of SBF consensus sites
in the promoters of SBF target genes that were not cell-cycle regulated. In
the set of MBF target genes that were not cell-cycle regulated, there was
a significant enrichment for genes with roles in DNA repair and replication,
as described below.
Figure 4: Expression profiles of SBF and MBF targets.
a, Expression patterns of SBF and MBF targets are indicated (red–green colour scale). Cell-cycle data are from ref. 12 and sporulation data are from ref. 18. The stages of the cell cycle are: M/G1, yellow; G1, green; S, blue; S/G2l, red; and G2/M, orange. Yellow boxes indicate the presence of consensus binding sites in the intergenic sequences upstream of each ORF (right), and the median percentile rank in IPs of the upstream sequences is also indicated (blue–yellow colour scale), as in Fig. 3. For each set of targets, the top panel contains cell-cycle regulated genes, the bottom panel contains genes that are members of divergently transcribed pairs in which the other member was cell-cycle regulated, and the middle panel contains the remainder of the non-cell-cycle regulated genes. b, Average expression profiles of the cell-cycle regulated targets of SBF and MBF, computed by averaging the log 2 (Cy5/Cy3) ratios. Note the specific induction of MBF targets during sporulation.
High resolution image and legend (138K)The known functions of many of the putative targets that we have identified raised interesting questions about the regulation of the cell cycle (Fig. 3). The G1 cyclins CLN1 and CLN2, as well as PCL1 and PCL2, were among the SBF targets that we expected to find. Unexpectedly, the promoters of the G2/M-specific B-type cyclins CLB1 and CLB2 also seemed to be bound by Swi4 in our microarray analysis, and in a confirmatory PCR analysis (data not shown). The significance of this result is unclear; SBF may be involved in repressing the transcription of these genes during G1/S or, as the mitotic B-type cyclins are required to shut off SBF-induced transcription during G2 and M phase15, this observation may point to a negative feedback loop in SBF regulation.
Why are two different transcription factors used to mediate nearly identical
transcriptional programmes during the cell-division cycle in yeast? A possible
answer is suggested by differences in the functions of the genes that they
regulate. Many of the known transcriptional targets of SBF have roles in cell-wall
biogenesis and budding11. Indeed, a remarkably large fraction
of the putative SBF targets identified in this study encode proteins that
are involved in various aspects of bud formation, and synthesis of membrane
and cell-wall components, based on functional definitions in the MIPS
(http://www.mips.biochem.mpg.de/proj/yeast/) and YPD (http://www.proteome.com)
databases. Twenty-seven per cent of the known genes in the SBF target set
fell in these functional categories, which is a significant enrichment (
P = 1.0 10-5) over their representation
in the genome as a whole (11%). Twenty-five per cent of the putative MBF target
genes have known roles in DNA replication, recombination and repair (on the
basis of MIPS and YPD definitions), which is a significant enrichment (
P = 1.0 10-5) over their representation
in the genome overall (6%)2. There was no enrichment of DNA
metabolism-related genes among SBF targets, nor were cell-wall morphogenesis
genes over-represented in the MBF targets.
The functional specialization of the putative targets of SBF and MBF was accompanied by differences in their overall expression programmes that suggest a possible regulatory logic to this division (Fig. 4b). The molecular programmes required for cellular growth and for the characteristic events at successive stages of the cell cycle occur with stereotyped synchronicity during simple mitotic proliferation. In different cell types, or during different physiological or developmental processes, however, the same molecular programmes may need to be orchestrated in different ways. In mitotically proliferating yeast, budding and DNA replication occur simultaneously; however, during the meiotic S phase, DNA replication occurs without budding. The converse is seen during mating (shmoo formation) and during pseudohyphal or invasive growth, when synthesis of membrane and cell-wall components occurs without concomitant DNA replication. Yeast cells must therefore have a mechanism to regulate genes devoted to DNA replication separately from those with major roles related to budding. As noted previously19, we observed that many of the putative MBF targets were genes that are strongly induced during sporulation (Fig. 4; data from ref. 18). The product of the B-type cyclin CLB6, a prominent MBF target, is also required for initiation of S phase during meiosis20, and several other MBF targets involved in DNA replication and repair have crucial roles in meiosis.
Several of the putative SBF targets are implicated in mating and pseudohyphal growth. RSR1 and WSC2 have a positive role in pseudohyphal growth; CAP1, GIC1 and GIC2 are involved in shmoo formation; SRL1, the highest-ranking target of SBF in our immunoprecipitations, is induced by TEC1, a positive regulator of pseudohyphal growth; and CLN1 is required for pseudohyphal growth21. The expression of TEC1 and PHD1, another positive regulator of pseudohyphal growth, was reduced in the swi4 deletion strain (see Supplementary Information), which suggests a positive function for SBF in the regulation of these genes.
Our results support a model in which SBF is the principal controller of membrane and cell-wall formation11, whereas MBF primarily controls DNA replication6. The need for both DNA replication and membrane and cell-wall biogenesis during G1/S in dividing cells accounts for the parallel functions of SBF and MBF in the mitotic cell cycle. It is possible that SBF has an independent and distinct role in mating and pseudohyphal growth, whereas MBF has a role in meiosis. In fission yeast, an Mbp1-like factor, p73pct1, is involved in regulating both mitosis and meiosis22, 23. In budding yeast, it is thought that Swi6 is involved in meiosis, but to our knowledge, the role of Mbp1 in meiosis has not been examined directly24. Experiments that directly examine the function of SBF and MBF in mating and meiosis, respectively, may further clarify this system.
Our results indicate that although the sequences that define canonical binding sites for SBF or MBF are common in the yeast genome, only a fraction of these sequences are actually bound by the corresponding factors in vivo. Moreover, most of the in vivo binding sites are adjacent to genes that show a characteristic cell-cycle transcription pattern. The use of DNA microarrays, together with protein-affinity isolation methods, should be generally applicable to the investigation of molecular systems that involve physical interactions of proteins with the genome.
Methods
DNA microarrays
We synthesized roughly 6,700
PCR primer pairs to amplify nearly every intergenic region in the yeast genome.
Primers were designed to amplify genomic DNA immediately adjacent to the coding
sequence of every ORF or non-ORF feature, up to the end of the neighbouring
ORF or non-ORF feature. Intergenic elements were designated by prefixing the
SGD (Saccharomyces Genome Database, http://genome-www.stanford.edu/saccharomyces)
designated name of the ORF or non-ORF feature that was immediately to its
left with an 'i'. The primers and their sequences can be obtained
from Research Genetics (Huntsville, Alabama). We prepared DNA microarrays
as described25.
Crosslinking and immunoprecipitation
For chromatin
immunoprecipitations, 5 108 cells were
used with the conditions as described26. Briefly, cells were
treated with formaldehyde to crosslink proteins and nucleic acids, and then
lysed. Extracts were sonicated with a Branson 350 Sonifier with a microtip,
at a power setting of 7 and a 60% duty cycle. Samples were pulsed eight times
for 12 s to shear chromatin to a final average size of about 0.5–2
kilobases. We used antibodies at a 1:100 dilution for immunoprecipitation.
Anti-haemmagglutinin antibodies and anti-Myc antibodies were from BabCo (Berkeley
Antibody Company). Affinity-purified anti-Swi6 polyclonal antibody was a gift
from K. Baetz and B. Andrews. The chromatin–antibody complexes were
precipitated with Protein A Sepharose beads (Pierce) and washed twice with
lysis buffer and twice with 1 TBS (150 mM NaCl, 20 mM Tris–HCl,
pH 7.6) for 5 min each. Chromatin was eluted from the beads
with 1% SDS, 1 TE (50 mM Tris–HCl, 10 mM EDTA)
and washed with 0.67% SDS, 1 TE. Crosslinks in the immunoprecipitated
chromatin were reversed by heating at 65 °C for at least 10 h
and the DNA was purified as described27, and resuspended in
10 
l of 1 TE.
Microarray hybridizations and analysis
A detailed
protocol for PCR amplification and fluorescent labelling of immunoprecipitated
DNA is available at http://www.microarrays.org. Microarrays were
hybridized at 65 °C for 6 h and washed as described previously,
then scanned with a GenePix 4000A scanner (Axon Instruments). We quantified
fluorescence intensities using GenePix Pro software (version 3.0), and uploaded
data to a custom database for analysis. Data were filtered to exclude spots
with obvious defects, or where the absolute signal intensity in both channels
was below an empirically determined threshold. In considering loci whose percentile
ranks were above the selected thresholds as putative targets, we excluded
loci with similarly high ranks in control experiments (within 2 s.d.
for the SBF targets, and within 1 s.d. for the MBF targets).
The control experiments were immunoprecipitates performed using antibodies
specific to the HA epitope tag, carried out with a strain lacking any of the
HA fusion proteins (Fig. 3, columns 1, 2 and 3).
