Cell Cycle Dependent Nucleosome Occupancy At Cohesin Binding Sites In Yeast Chromosomes

Abstract

In the budding yeast, cohesin is loaded onto the chromosome during the late G1 phase, establishes sister chromatid cohesion concomitant with DNA replication, and dissociates by the telophase. Here, using oligonucleotide tiling arrays, we show that, at the anaphase, nearly all of the cohesin binding sites contain nucleosome-free regions. The majority of these sites remain nucleosome-free throughout the cell cycle, consistent with the suggestion of a DNA binding anchoring protein present at these sites, although such a region could also serve as part of a marker for the binding of cohesin in the next cell cycle. However, a third of these sites are remodeled in the G1 phase, being reoccupied by nucleosomes by the G1/S boundary, though their subsequent removal in the S phase appears to be independent of DNA replication. Whether this difference is a result of other functions of cohesin or of the chromatin remains to be elucidated.

Introduction

Cohesin is a multisubunit, ring-shaped protein complex that is the principal component of the structure holding sister chromatids together during mitosis and meiosis [1-3], though it has also been implicated in a number of other chromosomal functions [4]. In the budding yeast, the cohesin complex, consisting of 4 subunits, Scc1, Scc3, Smc1 and Smc3 [5], first becomes associated with the chromosome during the late G1 phase [6], establishes sister chromatid cohesion concomitant with DNA replication during the S phase [7, 8], and, following the cleavage of its subunit Scc1 at the anaphase, finally dissociates from the chromosomes by the telophase [9]. Even though this general order of events is now well characterized, it is not yet fully understood how cohesin mediates this robust physical connection between sister chromatids [2]. So far, several models have been proposed, which include a single cohesin ring encircling both sister chromatids [1], two interlocking rings with each ring also directly binding the DNA of one chromatid [3], and a single cohesin ring encircling one chromatid but attached to the other chromatid via an anchoring protein [10].

These different models, each based on well documented experimental observations [3, 10-12], would imply different predictions on the local chromatin structure at the cohesin binding sites, which has previously been suggested [13] to play a role in the specific localization of cohesin along the chromosome [6, 9, 13-15]. For example, since the ring structure formed by the cohesin complex is sufficiently large to encompass two nucleosome-packaged chromatin [5], a model of simple entrapment of the two sister chromatids would not necessarily require an alteration of the chromatin structure at the nucleosomal level. But on the other hand, for those models that include direct DNA binding, whether by the cohesin complex itself or through an anchoring protein, the nucleosomes at the cohesin binding sites may be removed in order to accommodate this interaction, similar to the now well understood case of transcription factor binding sites at which nucleosomes are generally depleted under most conditions [16-20]. To determine whether the local chromatin has been altered at the cohesin binding sites, we have investigated the cell cycle dependent nucleosome occupancy at each of the previously identified cohesin binding sites across more than three full length chromosomes of the budding yeast S. cerevisiae using high resolution oligonucleotide tiling arrays [6]. Strikingly, nearly all of the cohesin binding sites contain nucleosome-free regions at the anaphase. Among these regions, only about one third undergo chromatin remodeling in the G1 phase, while the remaining are nucleosome-free throughout the cell cycle. These findings support the notion that the local chromatin structure is an integral part of a cohesin binding site and could play an essential role in the demarcation and maintenance of chromosomal cohesion.

Results and Discussion

Nucleosome Occupancy at Cohesin Binding Sites at the Anaphase

To determine the nucleosome distribution along S. cerevisiae chromosomes at the anaphase, mononucleosomal DNA was purified from cells released from αfactor synchronization and hybridized to oligonucleotide tiling arrays. The quality of the synchronization, as well as the stage in the cell cycle, was determined using FACS and confirmed with optical microscopy by DNA staining before harvesting (see Supplementary Figure S1). DNA bound to mononucleosomes after micrococcal nuclease (MNase) digestion was purified, labeled, and hybridized to the tiling arrays, which cover chromosomes III, IV and V, and the right arm of chromosome VI at an average probe distance of 27 bp. After normalization to the reference with total genomic DNA, a chromosome wide map of nucleosome occupancy is produced. Since gene transcription is largely repressed at the anaphase [21], there should be minimal interferences from trans-binding factors on the nuclease digestion assay. Furthermore, since the Scc1 subunit of the cohesin complex is already cleaved at the anaphase [22], any steric occluding effect on the accessibility of the nuclease owing to the tight linkage between the sister chromatids should also be minimal. To improve the signal to noise ratio of the dataset, a Gaussian averaging (σ = 75 bp) is applied. Nucleosome-free regions are then identified from these data using one standard deviation as the threshold (see Supplementary Data). Owing to variations from array to array, the standard deviation was calculated separately for each experiment.

As shown with chromosome III (Supplementary Figure S2), there is a high degree of general agreement between these data and that published by Yuan et al. [18] in terms of nucleosome occupancy, although the precise positions of individual nucleosomes are not always discernable owing to residual cross hybridizations present with these arrays [23, 24]. Overall, more than 80% of the chromosomes are occupied by nucleosomes as expected [18]. Despite the somewhat lower signal to noise ratio of these short oligonucleotide tiling arrays, the results are sufficiently reliable to identify nucleosome-free regions along the chromosomes, albeit at a lower spatial resolution when compared with that by Yuan et al. [18].

For the budding yeast, Glynn et al. [13] identified a total of 1095 cohesin binding sites with a genome wide chromatin immuno precipitation (ChIP-on-chip) experiment. Similar results were also identified in [6], as a number of authors have also found [1, 2, 25-29]. Among these sites, a total of 222 are represented on our tiling array. Using these as a reference, we examined the nucleosomal occupation state centered at these coordinates within a window of ±500 bp. It is noted that in the ChIP-on-chip experiment [13], the average size of the chromatin fragments used in the immuno precipitation was on the order of 200-1000 bp and the hybridization was performed on an array of long probes, where the data were further processed with a larger rolling window (see Supplementary Data). Therefore, the criterion chosen here should be appropriate.

Unexpectedly, we find that virtually all of the identified cohesin binding sites (216 out of the total 222; >97%) are associated with at least one nucleosome-free region (Figure 1A). The spatial extent of these nucleosome-free regions varies somewhat in a range from 150 bp to 1 kb, equivalent to about 1 to 5 nucleosomes of packaged chromatin (including the linker DNA and 146 bp of DNA for the nucleosome core particle [30]). Using the maximum signal position in each nucleosome-free region at these sites as the coordinate to compare with the coordinate of the cohesin binding sites, the (absolute) mean difference in position is surprisingly small, only ~160 bp or about a single nucleosome. This is a much better overall accordance than what might be expected from a cursory examination of the raw data. However, since there are many more nucleosome-free regions than the number of the cohesin binding sites along the chromosomes (Figure 1B and Supplementary Figure S2), it is important to determine that this apparent association is statistically significant. Therefore, we estimated the probability with which these cohesin binding sites would coincide with nucleosome-free regions simply by chance. Using randomized 1kb segments along the chromosomes (see Supplementary Data), the occurrence of nucleosome-free regions present in these segments can be easily computed. With these calculations, we found that the p value is less than 10^-5 for the complete dataset. Therefore, the correlation of nucleosome-free regions with the cohesin binding sites is highly statistically significant.

Figure 1.

Nucleosome occupancy at the cohesin binding sites in yeast chromosomes at the anaphase. (A) The nucleosome occupancy at all 222 cohesin binding sites on chromosomes III, IV, V, and part of VI is shown. Each block spans 1 kb DNA in length, centered at the documented coordinates of the cohesin binding sites. The cohesin binding sites on the respective chromosomes are shown on the left (drawn in proportion). Regions not represented on the tiling array are coded in blue (ND) and the scale bar is in log2 with 1σ as the unit (σ: standard deviation of the respective dataset). Here, one standard deviation from the mean is used as the threshold to determine nucleosome occupancy. Clearly, most of these 1 kb segments contained well resolved nucleosome-free regions (green). (B) As a comparison, a continuous 31 kb region on Chromosome III is shown, where the majority of the chromosome is occupied by nucleosomes (yellow), consistent with previous findings [18]. Arrow heads point to the cohesin binding sites within this chromosome region. Color coding: occupied regions in shades of yellow and unoccupied regions in shades of green (see scale bar).

Since chromosomes are still largely condensed at the anaphase [31], a potential concern is whether the lack of nucleosome occupancy in at least some of these regions could be a result of the higher order folding of the chromatin. Such foldings could limit the access of MNase to the linker DNA, especially under crosslinked conditions, resulting in a lower efficiency of digestion. Since only those DNA fragments with a length consistent with a mononucleosome were used in the hybridization (i.e., Figure 1), regions with higher order structures could then mistakenly appear as nucleosome-free, owing to the exclusion of longer DNA from hybridization. To assess the effect of this possibility, we purified the total DNA that remained after MNase digestion under identical conditions without further length selection, so that regions that have a poor MNase digestion efficiency should still produce a sufficient signal on the tiling array. The results from this latter experiment are in excellent agreement with the results using only mononucleosomal sized DNA, with the same nucleosome-free regions identified in the two cases (Supplementary Figure S3). These results, therefore, support the conclusion that these nucleosome-free regions are indeed correlated with the cohesin binding sites that were previously identified.

It should be noted that the cohesin complex appears to extend over 20-50 kb of pericentric sequences on the budding yeast chromosomes, showing up as higher and broader peaks compared to those on the arms in genome-wide ChIP-on-chip analysis [13, 32]. All of these peaks correlated well with nucleosome free regions in our dataset. However, the width of the nucleosome free regions did not have a clear correlation with the width of the cohesin peaks found in the ChIP-on-chip experiments [13]. The significance of this difference, if any, is not clear at present.

Loss of Nucleosome at Cohesin Binding Sites is Not Coupled to DNA Replication

It is now well documented that sister chromatid cohesion is established concomitant with DNA replication [7, 8, 33]. This direct coupling with DNA replication raises the important question of whether the nucleosome-free regions at the cohesin binding sites are generated when chromatid cohesion is established at the time of DNA replication, or whether these nucleosomes are removed much later, in conjunction with the cleavage of the cohesin subunit Scc1. Therefore, we further examined the nucleosome occupancy at the cohesin binding sites at three different time points in the cell cycle: early in the S phase where only a small fraction of the DNA completed replication and late in the S phase where the majority of the DNA was already replicated, as well as at the G2/M phase where DNA replication was complete (Figure 2A and Supplementary Figure S4). Somewhat surprisingly, the data clearly show that the overwhelming majority (> 90%) of the cohesin binding sites are nucleosome-free at each of these three cell cycle stages. Therefore, DNA replication is not a prerequisite for the formation of nucleosome-free regions at the cohesin binding sites and the cleavage of Scc1 that is known to trigger chromosome separation at the anaphase is likewise not correlated with the removal of nucleosomes at these sites. Of course, whether these sites could be occupied by other, as yet unidentified, DNA binding proteins [2], which could have a dependence on DNA replication, cannot be determined with the present assay and remains an open question for further examination.

Figure 2.

Cell cycle dependence of nucleosome occupancy at the cohesin binding sites. (A) Shown here is a side-by-side comparison of the nucleosome occupancy at all cohesin binding sites on chromosome IV at 4 different stages: S1 for early S phase, S2 for late S phase, M for G2/M phase and A for anaphase. A 1 kb region centered at each cohesin binding site is shown. Clearly, the overwhelming majority of the cohesin binding sites is nucleosome-free at each of these cell stages, indicating that at least for these sites, chromatin remodeling is not necessarily involved before or after DNA replication, nor after Scc1 cleavage. The full dataset is shown in Supplementary Figure S4. (B) A small fraction of the cohesin binding sites exhibits an S phase dependent nucleosome occupancy. Only18 of such sites are identified in our dataset and all are shown here, with chromosomal coordinates indicated on the right. In general, these regions exhibit a nucleosome loss from the S phase to the anaphase. Yet, even for these sites, the decrease in nucleosome occupancy is not clearly correlated with the timing of DNA replication. (C) A sizable fraction of the cohesin binding sites (74 out of 222) are reoccupied by nucleosomes during the G1 phase. Shown here is the nucleosome occupancy map of chromosome V obtained with cells arrested at the G1/S boundary with the α factor (right). As a comparison, the same map but obtained at the anaphase is shown on the left. The full dataset is shown in Supplementary Figure S5. These data suggest that chromatin remodeling must be involved in this change in the nucleosome occupancy at these sites. The color coding is the same as in Figure 1.

Although for the majority of the cohesin binding sites, the occupational pattern of nucleosomes does not significantly change from the early S phase through G2/M to the anaphase, there is a small (<10%), but significant, fraction of the cohesin binding sites that do exhibit a dynamic occupancy with a general trend of nucleosome loss with the progression from the S phase to G2/M (Figure 2B). A comparative analysis with the DNA replication timing profiles [34] at these 18 sites failed to identify a clear temporal correlation with the loss of the nucleosome in these regions though. Therefore, overall, these data are consistent with the conclusion that the absence of the nucleosome at the cohesin binding sites is not directly coupled to DNA replication.

A Significant Fraction of Cohesin Binding Sites are Reoccupied by Nucleosomes in the G1 Phase

Excluding the minor fraction of the cohesin binding sites at which nucleosome occupancy is dynamic during the S phase (which could be owing to other functions of the cohesin [4]), an attractive possibility is that a constitutively nucleosome-free region is required at the cohesin binding site throughout the cell cycle. To determine whether this is indeed the case, we further examined nucleosome occupancy with cells arrested at the G1/S boundary with α factor. Contrary to the above expectation, a third of the cohesin binding sites are fully occupied by nucleosomes at this cell cycle stage, while the rest remain nucleosome-free (Figure 2C and Supplementary Figure S5). Clearly, during the G1 phase, this subset of the cohesin binding sites has undergone chromatin remodeling to reload the histones back onto the DNA. It is also of interest to note that these G1 phase remodeled sites have included all of those sites identified to exhibit a dynamic occupancy during the progression of the S phase (i.e., Figure 2B). Like the sites with a dynamic occupancy during S phase, none of these G1 remodeled locations exhibited similar replication times or were correlated with any other identifiable functional traits (such as local histone modifications or proximity to tRNA).

Taken together, these results support the hypothesis that the substrate chromatin plays an important and integral role in cohesin-mediated functions. Since the majority of the cohesin binding sites are nucleosome-free throughout the cell cycle, an attractive hypothesis is that these regions alone or in conjunction with possible histone modifications [35] in the immediately adjacent chromatin could serve as a marker for the loading and/or the localization of the cohesin complex after the dissociation of the cohesin complex at the telophase. Alternatively, these regions could be occupied by a specific DNA binding factor to demark these regions specifically for sister chromatid cohesion. If such a protein does exist, its direct binding to DNA could exclude the formation of nucleosomes, similar to those sites that bind transcription factors [16]. Yet if this is indeed the case, then, at least for a third of the cohesin binding sites, such a protein would have to dissociate from the DNA to allow the reoccupation of these sites by the nucleosomes. What remodeling complexes are responsible for the loading and the subsequent removal of these nucleosomes are questions of great interest [10, 36-40]. The small fraction of cohesin binding sites that have an S phase dependence does not appear to correlate with other identifiable functional elements of the chromatin and are scattered throughout the chromosomes. Whether the sites that exhibit such a dynamic nucleosome occupancy are related to other functions of the cohesin than chromatid cohesion requires further examination. Clearly, the resolution of these questions is essential for an understanding of the mechanism that governs chromosome cohesion.

Materials and Methods

Cell culture

S. cerevisiae (derived from W303, MATa ho ade2-1 trp1-1 can1-100 leu2-3, 112 his3-11, 15 ssd1, clb1–4^tsura::URA3/GPD-TK [41]) was grown in YPD at 25ºC and arrested at the G1/S boundary with 10 μg/ml of α factor (GenScript, Piscataway, NJ) for 3 hr prior to release. To release the cells from the arrest, the cells were first washed to remove the α factor from the medium and then resuspended in fresh medium supplemented with 50 μg/ml of pronase (Sigma, St Louis, MO). The cells were then processed at 15 min (early S phase), 40 min (late S phase), 65 min (G2/M) and 95 min (anaphase) after release (see below). The cell cycle stage of the culture was monitored by FACS and optical microscopy by DNA staining (Supplementary Figure S1). The final concentration of the cell culture was kept below 1.0 at OD₆₀₀.

DNA purification

Mononucleosome preparation was carried out according to the published protocols [18]. Briefly, cells were first fixed with 2% formaldehyde directly in the medium for 15 min at 25ºC under constant agitation. The fixation was terminated with the addition of glycine to the final concentration of 125 mM. The cells were then washed twice with 20 mM Tris-HCl pH 7.5, 150 mM NaCl and resuspended in 1 M sorbitol, 50 mM Tris pH 7.4, 10 mM β-mercaptoethanol containing 0.25 mg/ml lyticase (Sigma, St Louis, MO) to remove the cell wall (30 min at 30ºC). Spheroplasts were pelleted and resuspended in 1 M sorbitol, 50 mM NaCl, 10 mM Tris pH 7.4, 5 mM MgCl₂, 1 mM CaCl₂, and 0.075% NP-40 (for permibilization of the cell membrane; Sigma, St Louis, MO), and micrococcal nuclease (MNase; Worthington, Lakewood, NJ) was added (to the final concentration of 0.27 unit/μl) and incubated at 37ºC for 45 min. To recover the undigested DNA, proteins were digested with 0.1 mg/ml of proteinase K (MP Biomedicals, Solon, OH) and the formaldehyde crosslink was reversed by incubation at 65ºC overnight. After phenol extraction and ethanol precipitation, RNA was digested with an RNase cocktail (25 U/ml of RNase A and 1 U/μl of RNase T1; Ambion, Austin, TX). Mononucleosomal DNA was purified by excising the 150 bp band from a 1.5% agarose gel.

To evaluate the effect of higher order chromatin folding, total DNA after MNase digestion under the above conditions was collected without gel purification. The purified DNA was amplified with random priming (BioPrime DNA labeling system; Invitrogen, Carlsbad, CA) and labeled with 50 μM of biotin-ddUTP by terminal transferase (20 U/μl; Roche, Indianapolis, IN). Total genomic DNA was also purified as the reference. To minimize possible bias owing to the amplification procedure, the genomic DNA was first sheared to 100 bp-1 kb fragments with a Branson Sonifier 450, followed by the same amplification procedure by random priming and biotin labeling. Two independent experiments were performed for each condition, including that with total genomic DNA.

Array hybridization and Data analysis

15 μg of biotin labeled DNA was hybridized to Affymetrix yeast tiling array (SC3456a520015F; Affymetrix, Santa Clara, CA), following the protocol in Lengronne et al. [6]. The tiling array used in these experiments contained 92812 probe pairs covering a total of 2.55 Mb yeast DNA sequence. The data was acquired using GeneChip Scanner 3000 (Affymetrix, Santa Clara, CA). The raw data were first normalized with dCHIP [42]. To improve the quality of the data, 8251 probes (<9% of the total) were excluded based on the following criteria: (1) they have multiple hits in the S. cerevisiae genome; (2) the signal from the genomic DNA hybridization was too weak (a threshold of (PM–MM)/ MM ≤ 0.25 was used; PM = perfect match; MM = mismatch). The signal intensity (PM–MM) along each chromosome was averaged with a Gaussian rolling window (σ = 75 bp). Each dataset was normalized by that of the genomic DNA (two replicates) and log2 ratios were calculated. A region was assigned as nucleosome free if the log2 ratio was more than one standard deviation below the mean. The data discussed in this article are deposited in NCBIs Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/) and are accessible through GEO Series accession number GSE8130.

Supplementary Material

01

Appendix A. Supplementary Data

Supplementary data and additional information associated with this article can be found in the online version at Genomics online.