Condensin Binding at Distinct and Specific Chromosomal Sites in the Saccharomyces cerevisiae Genome

Abstract

Mitotic chromosome condensation is chiefly driven by the condensin complex. The specific recognition (targeting) of chromosomal sites by condensin is an important component of its in vivo activity. We previously identified the rRNA gene cluster in Saccharomyces cerevisiae as an important condensin-binding site, but both genetic and cell biology data suggested that condensin also acts elsewhere. In order to characterize the genomic distribution of condensin-binding sites and to assess the specificity of condensin targeting, we analyzed condensin-bound sites using chromatin immunoprecipitation and hybridization to whole-genome microarrays. The genomic condensin-binding map shows preferential binding sites over the length of every chromosome. This analysis and quantitative PCR validation confirmed condensin-occupied sites across the genome and in the specialized chromatin regions: near centromeres and telomeres and in heterochromatic regions. Condensin sites were also enriched in the zones of converging DNA replication. Comparison of condensin binding in cells arrested in G₁ and mitosis revealed a cell cycle dependence of condensin binding at some sites. In mitotic cells, condensin was depleted at some sites while enriched at rRNA gene cluster, subtelomeric, and pericentromeric regions.

Condensin is a molecular machine of chromosome condensation and thus is one of the key players required for faithful chromosome segregation during mitosis. Vertebrates possess two distinct condensin complexes, with different expression patterns and chromosomal localizations (38, 39). Saccharomyces cerevisiae has a single condensin complex composed of five subunits: two SMC subunits (Smc2p and Smc4p) and three non-SMC subunits (Brn1p, Ycs4p, and Ycs5p/Ycg1p) that are highly conserved among eukaryotic organisms (8, 15, 22, 46, 52). All condensin subunits are essential for viability in yeast, and a loss-of-function mutation in any subunit impairs chromosome segregation during mitosis (8, 40, 50). Similar phenotypes are observed in metazoa when condensin subunits are inactivated (2, 13, 18). The purified condensin holocomplex can introduce positive supercoils into DNA in vitro in an ATP-dependent manner (1, 23), but the molecular nature of mitotic condensin activity in vivo remains largely unknown (1).

This limited understanding is, in part, due to a lack of data on the identity of the places where condensin acts in chromatin and on the molecular mechanisms directing condensin placement there. In vertebrates, condensin has been shown to be distributed in a reproducible pattern over the length of mitotic chromosomes, but only at the resolution of light microscopy (18, 34). The rRNA gene locus has been identified as the major binding site for condensin in S. cerevisiae (8). Condensin is also enriched in the nucleolar area in higher eukaryotes (5, 7), suggesting that its affinity for rRNA gene chromatin is universal.

In S. cerevisiae, condensin mutants display profound defects in nucleolar segregation, suggesting that a primary role for condensin in budding yeast is to promote resolution of rRNA gene cluster repeats (8, 30). However, several facts indicate that condensin might have some non-rRNA encoding locus binding sites in the genome. First, in strains where the episomal rRNA gene replaces the tandem chromosomal repeats (37), condensin remains essential, even though segregation of rRNA gene episomes is condensin independent (8). Second, smc2-8 mutant cells show a significant delay in mitotic segregation of long chromosomal arms, compared to wild-type cells (8). Third, yeast artificial chromosomes fail to segregate properly in condensin mutants under nonpermissive conditions (8). These results strongly indicate that condensin has non-rRNA gene targets in S. cerevisiae, as it does in higher eukaryotes (34, 38).

Identifying genomic sites of condensin binding should provide an important tool to study mechanisms of condensin-chromatin interaction and condensin activity in situ. Such an approach is difficult using the rRNA gene as a target, as the rRNA genes are repeated, multiple copies are essential, and individual repeats are functionally heterogeneous (41, 47). Finding condensin-binding sites using a chromatin immunoprecipitation (ChIP) approach in an arbitrary sample of nonrepeated DNA has proven to be difficult, presumably due to a relatively low efficiency of condensin cross-linking to chromatin (8, 49, 54). Thus, to verify the hypothesis that condensin binds at many places in the genome in addition to the rRNA gene, we conducted a genome microarray hybridization study with DNA extracted from immunoprecipitates containing the chromatin-bound condensin (ChIP-on-chip analysis) (27, 33). We identified and validated a number of condensin-bound sites in the yeast genome and showed their contribution to condensin-facilitated transmission of genetic information in mitosis. Condensin binding was investigated in relation to a variety of chromosomal features, including base composition, protein encoding, kinetochore formation, and heterochromatin and DNA replication. Finally, we detected two types of condensin binding: one that is present throughout the cell cycle and another that is mitosis dependent.

MATERIALS AND METHODS

Yeast culture and cell cycle methods.

Yeast cells were grown as previously described (4, 44). The yeast strains used are shown in Table 1. For the G₁ and mitotic arrest experiments, cells were synchronized by exposure to 5 μM α-factor (U.S. Biological). After incubation for 3 h, more than 90% of the cells were arrested in G₁. For mitotic arrest, α-factor was washed off with growth medium and the cells were released into medium containing 15 μg/ml of nocodazole (Sigma) for 2.5 h. Mitotic arrest, as determined by fluorescence-activated cell sorter and cell morphology analyses, was reached after 2.0 to 2.5 h of incubation.

TABLE 1.

S. cerevisiae strains used in this study

Strain	Genotype
W303	MATaade2 leu2 can1 his3 trp1 ura3
W303/pLF733	MATaade2 leu2 can1 his3 trp1 ura3 smc2-8::LEU2
W303/pAS532	MATaade2 leu2 can1 his3 trp1 ura3 SMC2:6His:3HA::LEU2
YPH499bp2	MATaura3 leu2 his3 trp1 lys2 ade2 bar1 pep4 SMC2:6His:3HA::LEU2
YPH499bp5	MATaura3 lys2 ade2 trp1 his3 leu2 bar1 pep4 YCS5:6His:3HA::URA3
YPH499bp6	MATaura3 lys2 ade2 trp1 his3 leu2 bar1 pep4 BRN1:6His:3HA::URA3
CH2524	MATaleu2 trp1 ura3 brn1-9::TRP1
ZW204	MATaade2 his3 leu2 lys2 trp1 ura3-52 ycs4-2:12MYC::HIS3
BLY07	MATaade2 his3 leu2 lys2 trp1 ura3 ycg1-2::KanMX (ycs5-2)
1aAS342	MATaade2 his3 leu2 lys2 trp1 smc4-1
1bAS330	MATaade2 his3 leu2 lys2 trp1 ura3 smc2-8
BUY549	MATα ade2 his3 leu2 trp1 ura3 ppr1Δ::HIS3 HMR::URA3
BUY549/pLF733	MATα ade2 his3 leu2 trp1 ura3 ppr1Δ::HIS3 HMR::URA3 smc2-8::LEU2
T1	MATα his7Δ

ChIP analysis.

ChIP was performed as previously described (54), except that 300-ml cultures were used. Chromatin-containing lysates were sheared by sonication to an average fragment size of ∼500 bp. Immunoprecipitations were done with anti-hemagglutinin (HA) antibody 12CA5 (Roche). Quantitative real-time PCRs (qPCRs) were performed using an MX3000P real-time PCR system (Stratagene). PCR mixtures (50 μl) contained 1 μl of template DNA (ChIP or input), 25 μl of 2× SYBR Green Master Mix (Stratagene), and 50 nM primers. PCR cycle parameters were 1 min at 95°C, 30 s at 55°C, and 1 min at 72°C. Primer dimerization signals were virtually eliminated by diluting primers to 20 nM and using the exponential (“linear”) range of the PCR for quantitative analysis. All primer sequences are available upon request. Forty-cycle (endpoint) products were analyzed on 1.5% agarose gels (see Fig. 3C and D and 5D and E), but quantitative values reported are from the qPCR analyses. The ratio of ChIP PCR product to the input PCR product (relative ChIP value, enrichment ratio) was calculated as previously described (54).

FIG. 3.

Cohabitation of condensin with specialized chromatin. (A) Comparison of condensin peak densities in pericentric regions. The number of condensin sites was averaged for all chromosomes for regions defined as pericentromeric (CEN, 10-kb region centered at each CDEIII), CEN-L (10 kb to the left of the pericentromeric 10-kb region) and CEN-R (10 kb to the right of the pericentromeric 10-kb region) regions and compared to the average density of condensin peaks in the entire genome. Data were generated from the ChIP-chip experiment with W303/pAS532. (B) Smc2p-HA ChIP-qPCR scanning of the CEN4 region in an asynchronous cultures of W303/pAS532. Tiling PCR probes encompass 4 kb around CEN4 with 500-bp intervals. The TUB2 ChIP/input ratio in the same experiment was less than 0.01%. (C) Condensin sites are present in subtelomeric heterochromatin. Smc2p-HA chromatin enrichment was analyzed in W303/pAS532. PCR probes used for ChIP-qPCR measurements of condensin binding at the VI-R telomere region were the same as in reference 16. The agarose gel bands are the end products of ChIP-qPCR (ChIP and input [IN]), while numerical values correspond to linear-range qPCR data (tag + ChIP, %). Condensin is associated with the site located 0.6 kb from the telomeric repeats but not with other subtelomeric sites (0.35 or 1.4 kb from TEL). ChIP material from a W303 strain lacking the HA tag and PCRs using TUB2 primers were used as negative controls. (D) Condensin association at HML and HMR. PCR probes were as in reference 16. The gel shows the end products of ChIP-qPCR (ChIP and input [IN]). The linear-range qPCR data (tag + ChIP, %) show Smc2p-HA association with HMLα and HMRa in W303/pAS532. Negative controls were as in panel C.

FIG. 5.

Condensin targeting at different cell cycle phases. (A) Mitotic enrichment of condensin at RDN1 hot spots in mitosis. Data generated from Smc2p-HA ChIPs in W303/pAS532 (α-factor and nocodazole arrest). Variance bars generated as a result of averaging for four hot spots (54). (B) Comparison of condensin binding to selected sites in G₁ and M phases. Data were generated from Smc2p-HA ChIP (W303/pAS532) either after α-factor arrest or under nocodazole arrest, followed by qPCR. (5) and (3), 5′ and 3′ IGRs, respectively. TUB2 is a negative control. (C) Condensin binds the PTR2 IGR with higher affinity in G₁. Experimental conditions were as in panel B. x coordinates correspond to the centers of 300-bp PCR probes (500 bp apart). (D) Condensin binds HMLα and HMRa with higher affinity in G₁. Experimental conditions were as in panel B. The end products of ChIP-qPCR (gel) and linear-range qPCR data (bar graph) are shown. (E) Condensin is enriched at a subtelomeric site in mitosis. TEL VI, telomeric region VIR. The experiment and data analysis are as in panel D. Error bars in panels B, D, and E represent standard deviations. (F) A pericentric site has higher condensin occupancy in mitosis. Experimental conditions were as in panel B. Relative PCR probe positions around CEN4 are as in panel C.

Microarray analysis.

Two kinds of spotted microarrays were used. One contained all unique yeast open reading frames (ORFs), and the other contained all intergenic regions (IGRs). For both arrays, the primer pairs (Research Genetics) were designed to amplify genomic regions as described earlier (20). PCR products were validated by agarose gel electrophoresis. Amplified sequences (in 3× SSC [1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate]) were printed on polylysine-coated slides. All array elements are described in supplemental Tables S3 and S4.

For probe preparation, about 20 ng of the ChIP (Smc2p-HA precipitated) and total genomic DNA samples (reference control) were used in a two-step random PCR amplification as previously described (10). The amplified DNA fragments were in the range of 300 to 1,000 bp. These amplified DNA samples were used for Cy5-dUTP (DNA from immunoprecipitation) or Cy3-dUTP (genomic DNA) labeling (Amersham Pharmacia). Mixed Cy5- and Cy3-labeled DNA samples were purified with a QIAquick PCR purification kit (QIAGEN), concentrated through a YM-30 filter (Amicon), and used as probes for DNA microarray hybridization. Three independent ChIPs were performed for each array type (three ORF and three IGR arrays were analyzed).

Microarray images were acquired using a GenePix 4000B scanner (Axon Instruments). Intensities from Cy3 and Cy5 channels were normalized in the scanner. GenePix Pro 5.0 software (Axon Instruments) was used to filter usable data sets. BRB-Array Tools (http://linus.nci.nih.gov/BRB-ArrayTools.html) software was used to validate the reproducibility of hybridization replicas (within each array type). Data were compiled in Microsoft Excel and subjected to median normalization (see Tables S3 and S4 in the supplemental material). Normalized data were averaged for each microarray element, thus allowing us to merge the ORF and IGR data sets. The features with fewer than two measurements were excluded from all subsequent computations; however, some of them were validated by ChIP-qPCR. The merged whole-genome data set was analyzed using Peakfinder (11) to identify condensin-binding hot spots. Chromosome display of condensin peaks (see Fig. S1 in the supplemental material) was plotted using Promoter (55) (see Fig. S1 in the supplemental material). For the relative replication-timing correlation analysis (see Fig. 4C), only array features with three robust measurements were selected. Normalization and correlation analysis of condensin and ORC complex-binding sites at replication origins were done as previously described (28), using genome-wide measurements of condensin subunit Smc2p-HA (this work) and Orc1p or Orc1p-Orc6p enrichment (57).

FIG. 4.

Condensin-binding sites correlate with replication landmarks. (A) Negative correlation between Smc2p and ORC complex binding at chromosome (Chr.) VI replication origins. The resolution of this analysis is determined by the sizes of microarray elements (mean lengths: ORF, 1,404 nucleotides; IGR, 472 nucleotides). Correlation analysis after variance normalization was according to reference 28. (B) Calculation of the relative replication index. For each condensin-enriched locus, a relative replication time coordinate (index) was calculated as follows: the relative replication time for each condensin peak (t^x − t⁰) was divided by the time separating the nearest origin from the corresponding replication terminus (tⁿ − t⁰). Replication time values are from reference 42. Only loci that displayed condensin enrichment in three independent ChIP-chip experiments were included in this analysis. (C) Whole-genome distribution of condensin peaks relative to replication landmarks. The relative replication index for all euchromatic condensin peaks in the genome was calculated as illustrated for panel B, and the distribution of relative replication index values (2.5% bins, where each origin is 0% and each termination zone is 100%) was determined. Condensin depletion at the origin loci and enrichment at the termination zones are apparent. The mean number of condensin peaks per bin was 18.9 ± 14.3.

Minichromosome stability.

The stability of minichromosomes (circular centromere plasmids) was examined as previously described (51). Minichromosomes were transformed into wild-type (W303) and isogenic smc2-8 (W303/pLF733) strains. Transformed yeast cultures were incubated in synthetic dropout medium lacking uracil (for pRS416 or HMR-containing pRO3) at 23°C to an optical density at 600 nm of 0.3 and then shifted to the nonpermissive temperature (37°C) for 6 h or 12 h. Cells were plated on YPD medium, and the resulting colonies were replica plated onto synthetic dropout plates lacking uracil. Plasmid stability was calculated as the fraction of colonies that grew on selective medium.

RESULTS

Condensin is enriched at non-rRNA encoding loci located on all chromosomes.

We undertook a whole-genome screening for condensin-binding sites using a combination of ChIP and microarray analysis of the precipitated DNA (ChIP-chip). Genomic ChIP-chip experiments were performed using an asynchronous population of haploid cells with the fully functional HA-tagged SMC2 gene (SMC2:3HA). Microarrays containing S. cerevisiae ORFs and IGRs were used (see Materials and Methods). The data sets obtained with ORF and IGR arrays (see Tables S3 and S4 in the supplemental material) were merged after median normalization to give complete genome coverage. As a result of this ChIP-chip analysis, we established that condensin has reproducible peaks of enriched binding across the S. cerevisiae genome (see Fig. S1 in the supplemental material). Loci with at least a twofold enrichment in binding over the median value for the genome were defined as putative condensin-binding sites. The number of condensin-enriched sites per chromosome was a linear function of chromosome length, with an average spacing between the sites of 10.7 kb (Fig. 1A). Condensin enrichment was distributed between the coding and noncoding regions; however, condensin sites were not spread regularly: while the median distance between the neighboring peaks was 8.8 kb, many sites showed separation by as much as 20 to 50 kb (Fig. 1B). Since each 9-kb rRNA gene cluster repeat has at least four potential condensin-binding regions (8, 54), the relatively sparse distribution of condensin in the rest of the genome (Fig. 1B) could partially explain the visible enrichment of condensin-green fluorescent protein fusions in anaphase nucleoli, compared to the rest of the chromatin (8). Altogether, 1,121 condensin-enriched sites were found in the genome. Assuming that the rRNA gene cluster (not included in the arrays) contains at least 500 additional binding sites, we can estimate the number of condensin locations at less than 2,000 per genome. This number agrees well with the value of 1,800 condensin complexes per cell calculated from the mean concentration of the five condensin subunits (19).

FIG. 1.

Condensin-binding sites are present on all chromosomes. (A) Correlation between chromosome length (x axis) and number of condensin-binding peaks (y axis). Condensin distribution peaks (twofold or greater above the whole-genome median value) were defined as described in Materials and Methods, using microarray analyses of Smc2p-HA ChIP (W303/pAS532 asynchronous culture). The regression line shown was established using the least-squares method. Loci with just one microarray data point available were excluded from this analysis. (B) Condensin-binding sites show a skewed distribution. The plot shows the distribution of distances between the two neighboring condensin-binding sites. Loci with just one microarray data point available are excluded. (C) ChIP-qPCR validation of condensin-enriched sites. Sites tested were either selected from condensin peaks associated with array elements (ORF or IGR) that were no longer than 500 bp or were selected from peaks that Peakfinder identified as residing at an IGR-ORF boundary (5′ or 3′). PCR primers were chosen that either included the whole locus or encompassed the IGR-ORF boundary (∼500 bp long with a center at the boundary), respectively. (5) and (3), 5′ and 3′ IGRs, respectively. TUB2 was used throughout as a negative control. (D) ChIP-qPCR validation of cold spots for condensin binding. Loci that Peakfinder identified as lacking condensin binding were validated by qPCR using criteria similar to those used for panel C. (E) Condensin subunits colocalize at the binding peaks. ChIP-qPCR analysis of sampled Smc2p-binding hot spots (identified from microarray analysis) in strains YPH499bp2 (Smc2-HA), YPH499bp5 (Ycs5-HA), and YPH499bp6 (Brn1-HA), all asynchronous cultures in an S288C background.

Several loci with elevated condensin binding in the ChIP-chip analysis were selected for validation by individual qPCR. As mammalian condensin binds only 200 bp of DNA (1), the validation sample (more than 30 sites) was selected from short (300- to 500-bp) microarray elements (ORFs or IGRs) that harbored condensin-enriched peaks to allow single-PCR locus verification (Fig. 1C). A separate sample included peaks located at boundaries between an ORF and an IGR. In this case, a 500-bp PCR probe was centered at the ORF-IGR boundary. Smc2p-HA showed a reproducible association with all the loci selected (Fig. 1C). In all cases, the TUB2 gene, previously identified by quantitative ChIP as negative for condensin binding (54), was used as a negative control and demonstrated only a background level signal. In addition, the absence of condensin binding was validated by quantitative ChIP for a sample of cold spots identified by microarray analysis (Fig. 1D). In a complementary validation approach, we compared binding of Smc2p-HA and the non-SMC condensin subunits Ycs5p-HA and Brn1p-HA to selected sites and found good agreement among the bindings of these three subunits (Fig. 1E).

Condensin binding shows no preference for DNA nucleotide composition.

In order to elucidate possible DNA determinants of condensin binding, we analyzed ChIP-chip data for correlation between the global condensin site distribution and genomic features. Unlike cohesin, which preferentially binds to AT-enriched and noncoding sequences (11, 29), condensin shows no significant preference for intergenic regions (60%). Moreover, unlike in the case of cohesin, only 15% of condensin-bound IGRs contain regions of converging transcription. We also compared the condensin enrichment distribution with nucleotide bias. Examples of this analysis are shown for chromosomes XII and III in Fig. 2A and B, respectively. We found no correlation between G+C content and condensin peaks. Moreover, the A+T content of condensin-enriched array elements is almost identical to the distribution of the A+T content of the entire array element set (Fig. 2C).

FIG. 2.

Condensin-binding loci lack a base composition bias. (A and B) For chromosomes XII (A) and III (B), the condensin-binding profile produced by Peakfinder (see Materials and Methods) is shown in green (left y axis). A smoothed G+C content profile (5-kb sliding window) is shown in red (right y axis), aligned so that the x axis intercepts at the median G+C content value. Fifty-kilobase intervals are marked with gray lines. Asterisks mark condensin peaks located within 10 kb of centromeres, next to the rRNA gene cluster, and at the HM loci. (C) Comparison of the A+T content of condensin-enriched sites with the A+T content of the genome. A+T contents for each array element (ORF or IGR) were calculated and binned into 1% intervals (blue line, left y axis). A+T contents for each condensin-enriched site were also plotted in 1% intervals (green line, right y axis).

Functional condensin sites are present in specialized chromatin regions.

In order to elucidate genome features that might correlate with the distribution of condensin, we examined condensin localization at several specialized chromatin regions. Among the most distinct regions with specialized chromatin structure are centromeres (35) and pericentromeric regions (3). We found pericentromeric regions (defined as 10-kb spans centered at the CEN core) to be enriched for loci with elevated binding, with one condensin hot spot per 6 kb. This density is lower than that seen in the rRNA gene cluster but substantially greater than the average density in the genome as a whole (Fig. 1B) or the density seen in the 10-kb regions immediately flanking the pericentromeric 10-kb regions (Fig. 3A). Cohesin has been shown to be present at elevated levels in a broad (more-than-20-kb) region around centromeres (11, 29), where it facilitates establishment of bipolar kinetochores (53). In contrast, condensin association peaks do not include the centromere itself and are relatively narrow (Fig. 3B). In general, sites of condensin enrichment do not coincide (data not shown) with previously described patterns of cohesin association (11).

Heterochromatin blocks are other specialized chromosomal domains (45). In S. cerevisiae, the rRNA gene locus, the silent mating type loci (HML and HMR), and telomeric repeats are packaged into a chromatin structure that can silence polymerase II transcription (12, 21). We detected reproducible peaks of condensin binding at all subtelomeric regions (defined as unique DNA within 5 kb of telomeric repeats) that were present on the array (see Fig. S1 in the supplemental material). However, condensin targeting to these regions in vivo is unlikely to be functionally related to heterochromatin formation, as we did not detect condensin enrichment in the unique sequences closest to telomeric repeats (see example in Fig. 3C).

The silent mating type loci HMR and HML also contained condensin enrichment peaks (supplemental Fig. S1). Condensin binding was confirmed by ChIP-qPCR at both HMLα and HMRa. Condensin was bound to the silenced mating type genes rather than to the E silencer elements (HML-E and HMR-E) (Fig. 3D). However, this condensin enrichment does not appear to be directly related to gene silencing, i.e., heterochromatin formation. Quantitative mating assays showed that condensin mutants did not alter silencing of the mating type genes at HM loci (see Fig. S2A in the supplemental material), smc2-8 did not notably alter silencing of a URA3 gene inserted into the HMR locus (see Fig. S2B in the supplemental material), and the position and relative occupancy of condensin peaks at HML and HMR were not changed in a sir2Δ mutant, which is defective in HML and HMR silencing (data not shown). Minichromosome stability assays suggest, however, that condensin functions at HMR to promote the mitotic segregation of minichromosomes containing this locus (see Fig. S2C in the supplemental material).

Enrichment of condensin binding near replication origins and termination zones.

The replication fork termination regions are an example of specialized chromosomal domains that could potentially contain condensin-binding hot spots (26). While such sites are not precisely mapped in the S. cerevisiae genome, the well-defined replication fork barrier in the rRNA gene cluster (25) shows very strong condensin binding (8, 54). In order to address a possible correlation between DNA replication and condensin binding, first we compared condensin distributions with patterns of ORC binding (57) on chromosomes VI and III (not shown), where ORC ChIP has been validated by extensive origin mapping (9, 43, 58). Comparison between condensin-binding hot spots and ORC-binding sites (after variance normalization) (28) reveals a strong negative correlation (Fig. 4A). Thus, condensin binding is disfavored at ORC-binding sites.

In order to address the possibility that condensin peaks in euchromatin are correlated with replication termination regions, as observed in the rRNA gene cluster (8, 54), we examined condensin binding in the vicinity of origins and in the regions between the origins (likely containing regions of replication termination) for all 16 chromosomes. First, we utilized replication-time data (42) to divide chromosomes in two equal parts: origin-proximal and termination-proximal regions (not shown). Condensin-binding sites were 1.5-fold more frequent in replication termination-proximal regions than in origin-proximal regions. In order to exclude the possibility that condensin enrichment detected at the termination zones was driven simply by the observed negative correlation with ORC-binding sites, we conducted a similar analysis at a much higher resolution. Using replication timing data (42), we calculated a relative time coordinate (index) for condensin peaks, expressed as a fraction of the distance between the nearest replication origin and the nearest replication terminus (Fig. 4B). This relative-scale approach confirmed our original assessment of condensin enrichment at the replication termination zones and a strong negative correlation between condensin binding and replication origins themselves (Fig. 4C). However, this analysis also revealed an unexpected enrichment of condensin in sequences adjacent to, but not at, replication origins. Compared to other sequences, condensin-enriched sites were 5 to 6 times more frequent in replication termination zones and about 2.5 times more frequent in sequences adjacent to replication origins (Fig. 4C).

Dynamic changes in condensin association with chromosomes during the cell cycle.

Previous studies have determined that condensin is bound to chromatin throughout the cell cycle. Furthermore, the cellular level of condensin increases only twofold prior to each S phase (8), as do most “structural” chromosomal proteins needed to assemble newly synthesized chromatin. Thus, the mitosis-specific increase in condensin association seen at the nucleoli (8, 54) should involve condensin relocalization, with some interphase condensin-bound sites becoming less occupied during mitosis. This relocalization has been hypothesized (48) to be required for mitotic chromosome condensation activity (8, 24, 31).

To compare the condensin association with chromosomes in the G₁ and M phases, we performed quantitative Smc2p-HA ChIP using cultures arrested with α-factor or nocodazole, respectively. The average condensin occupancy at the four identified hot spots (54) in the rRNA gene region was about twofold greater in mitosis-arrested cells than in G₁ cells (Fig. 5A). Most non-rRNA encoding loci showed little difference in condensin binding (Fig. 5B), indicating that many sites have a constitutive level of condensin occupancy. However, at some loci, such as the PTR2 IGR, HML, and HMR, a significant decrease in condensin binding was observed in mitosis compared to G₁ cells (Fig. 5C and D). In contrast, condensin binding at a subtelomeric site on chromosome VI and at a pericentromeric site on chromosomes IV increased in mitotic cells (Fig. 5E and F). Thus, three distinct modes of condensin binding to chromatin are evident. First, at many sites on chromosome arms, condensin occupancy does not change from G₁ to mitosis. Second, in a subset of loci (e.g., PTR2 and HM loci), condensin binds in G₁ but is depleted by the time cells are in mitosis. Third, some special loci show condensin-binding enrichment throughout the cell cycle that is further increased in mitosis (rRNA genes, subtelomeric regions, and CEN4).

DISCUSSION

While condensin complexes have been shown to be present on all chromosomes in higher eukaryotes (14), in S. cerevisiae both genetic and cytological evidence previously suggested that the essential function of condensin is to facilitate segregation of rRNA gene-containing chromosomes (8). This contradiction was partially resolved by the demonstration of unstable yeast artificial chromosome maintenance in condensin mutants (8, 17), indicating that yeast condensin has a potential to affect segregation of chromosomes lacking the rRNA genes and hence to bind in vivo some non-rRNA gene sites. Furthermore, the diffuse intranuclear localization of condensin in interphase yeast cells, the delayed anaphase segregation of chromosomal arms in condensin mutants (8), and the continued condensin requirement in strains with an episomal rRNA gene repeat suggest that S. cerevisiae condensin does act at additional sites dispersed in the genome.

Using ChIP-chip analysis of the whole S. cerevisiae genome, we established that condensin is indeed distributed over all chromosomes, with distinct binding sites spaced 2 to 45 kb apart (Fig. 1A and B; see Fig. S1 and Tables S3 and S4 in the supplemental material). We validated more than 30 such sites using quantitative ChIP and demonstrated that condensin-binding sites likely represent a short stretch of chromosome (less than 500 bp, see Fig. 5C and F). This size is in good agreement with in vitro data on the extent of condensin-DNA interaction (about 200 bp) (1). The size of these euchromatin condensin-binding sites also agrees well with the size of condensin-binding hot spots in the rRNA gene cluster (54). The absence of a nucleotide composition bias in condensin binding (Fig. 2) may indicate that condensin binding to particular sites is determined by features of chromatin or higher-order chromosome structure, and not the DNA sequence per se.

The distribution of condensin-binding sites appears to be specific for this SMC complex and is very different from the recently published distribution of cohesin, another SMC-based complex (11, 32). These two complexes show nonoverlapping whole-genome distributions (data not shown). Moreover, the absence of a strong condensin preference for intergenic regions, extended pericentromeric flanks, and AT-enriched DNA (all preferential cohesin occupancy sites) suggests that two complexes have different mechanisms of loading or of relocalization after loading.

While our data indicate that condensin is unlikely to be a kinetochore component (see also reference 36), it should be noted that the temperature-sensitive phenotype of an smc2-6 mutation is suppressed by overexpression of Ndc80p/Hec1p, a kinetochore component (59). In addition, condensin involvement in centromere function in metazoa (6, 38, 56) has been reported. Thus, we cannot exclude the possibility that the pericentromeric enrichment of condensin we observe (Fig. 3A) is functionally significant. All available data suggest that condensin binding does not interfere with or reinforce gene silencing by heterochromatin (see Fig. S2 in the supplemental material), although it may play a role in the mitotic segregation of heterochromatin-containing regions (see Fig. S2C in the supplemental material).

Comparisons of the interphase and mitotic occupancy of condensin sites supports a chromatin-to-chromatin relocalization mechanism as a source of the condensin found in the nucleolus during mitosis (8). Some euchromatic sites display elevated condensin binding during interphase but are depleted of condensin during mitosis (Fig. 5C and D). Such sites may provide condensin for the regions (including the rRNA gene cluster, pericentromeric, and subtelomeric regions) with increased condensin occupancy in mitosis. It remains to be determined if this subpopulation of condensin complexes is a passive storage pool or if condensin is active at these sites during interphase.

One intriguing possibility to emerge from this study involves the suggestion that both the genomic distribution of condensin and its mitotic relocalization are driven, at least in part, by DNA replication dynamics. First, two of the latest replication regions in the genome, the rRNA gene cluster and subtelomeric regions (42), are both condensin enriched in mitosis. Second, the whole-genome distribution of condensin-enriched loci shows strong correlations with replication landmarks: a strong negative correlation with replication origins themselves, a mild enrichment in sequences adjacent to replication origins, and a marked enrichment in zones of replication termination (Fig. 4).

While the negative correlation with replication origins can probably be explained by constitutive and tight ORC at these sites, the condensin enrichment in sequences next to origins is unexplained. One possibility is that this mild enrichment results from the spillover of uniformly loaded condensin that the tightly bound ORC complex excludes from origins. Another possibility is that condensin binding preferentially occurs at the sites where polymerase runs terminate. As origin boundaries have two converging runs of lagging DNA polymerase, this could account for the observed twofold condensin enrichment. The significant enrichment of condensin at the fork termination zones, respectively, correlates with converging leading-strand polymerases and collision of the forks themselves. One feature of unreplicated DNA ahead of the fork, its positive overwinding, could be potentially attractive for condensin binding. This, in turn, suggests a possible mechanism for constitutive condensin targeting—recognition of sites where replication creates a favorable topological environment for condensin binding.