|Home | About | Journals | Submit | Contact Us | Français|
Cohesin’s complex distribution on chromosomes and its implication in numerous cellular processes makes it an excellent paradigm for studying the relationship between the in vivo concentration of a protein and its in vivo function. Here, we report a method to generate systematic quantized reductions (QR) in the in vivo concentration of any yeast protein. Using QR, we generate strains with 13% and 30% of wild-type levels of the limiting subunit of cohesin, Mcd1p/Scc1p/Rad21p. Reducing cohesin levels reveals a preferential binding of cohesin to pericentric regions over cohesin-associated regions (CAR) on chromosome arms. Chromosome condensation, repetitive DNA stability, and DNA repair are compromised by decreasing cohesin levels to 30% of wild-type levels. In contrast, sister chromatid cohesion and chromosome segregation are unaffected even when cohesin levels are reduced to 13% of wild-type levels. The requirement for different in vivo cohesin concentrations to achieve distinct cohesin functions provides an explanation for how cohesin mutations can specifically lead to adult disorders such as Cornelia de Lange Syndrome and Roberts Syndrome without compromising the cell divisions needed for development and maturation. Our successful application of QR to cohesin suggests that QR is a powerful tool to study other proteins/pathways with multiple functions.
We were interested in understanding how the in vivo concentration of cohesin influences its ability to perform its cellular functions. In wild-type cells, enrichment of cohesin in the pericentric region of chromosomes generates sister chromatid cohesion proximal to the sister kinetochores, promoting bi-oriented attachment of the sister chromatids on the mitotic spindle and proper chromosome segregation. Cohesin also is enriched in 0.5–2 Kb cohesin associated regions (CARs) along chromosome arms and is induced to bind near the sites of DNA damage (reviewed in ). The localization of cohesin to these regions is thought to facilitate transcriptional regulation, chromosome condensation and DNA repair . Here, we address the importance of cohesin concentration in ensuring the occupancy of cohesin at these specific chromosomal sites and in executing cohesin’s diverse in vivo functions.
To address the significance of cohesin concentration (or any protein of interest) in vivo, we needed a system to systematically reduce the amount of a given protein in the cell. For this purpose, we exploited the inefficiency of tRNA nonsense suppressors in budding yeast. The SUP53 tRNA nonsense suppressor recognizes the UAG stop codon and inserts a leucine (Leu) into the growing peptide chain . However, it does so at approximately 30% of the efficiency of Leu tRNAs at the Leu codon CAG (Figure 1A). Thus, in a SUP53 haploid strain, substitution of the UAG stop codon for a Leu codon in a given gene should result in production of a wild-type protein at approximately 30% of the normal levels. The introduction of 2 or more UAG codons in a gene should generate isogenic strains with further quantized reduction (QR) in the synthesis. Importantly, this QR system generates a reduction in the in vivo concentration of a protein without changing the protein’s amino acid sequence or the timing of its synthesis. A similar QR system has been employed successfully to analyze bacteriophage T4 proteins  but has never been employed in eukaryotes until now.
To apply QR specifically to cohesin, we used QR to reduce the cellular concentration of Mcd1p (also known as Rad21p, Scc1p), a limiting subunit of the cohesin complex. We constructed mcd1 alleles with 1, 2, or 3 UAG stop codons in place of Leu codons encoding the 12th, 21st, and 26th residues of the Mcd1 polypeptide, which we refer to as 1 stop, 2 stop, and 3 stop alleles respectively. These mutant alleles were introduced into strains containing wild-type MCD1 on a URA3 containing plasmid as the sole source of MCD1 in cells. Cells were selected for loss of the wild-type MCD1 plasmid. We found that the 1 stop and 2 stop alleles, but not the 3 stop allele, support viability (Figure 1B, Figure S1). We next assessed the protein levels of Mcd1p in the 1 and 2 stop mutants using SDS-PAGE followed by western blotting. The 1 stop and 2 stop mutants have approximately 30%(+/− 7%) and 13% (+/− 2%) of wild-type Mcd1p levels, respectively (Figure 1C). Thus the QR system successfully modulates the levels of Mcd1p in vivo. Furthermore, this QR system revealed that the cellular concentration of Mcd1p in wild-type cells greatly exceeds that needed for viability.
Our QR alleles of MCD1 allowed us to address how the cellular concentration of Mcd1p affects its binding to different chromosomal regions/loci. The Mcd1p subunit only binds to chromosomes as part of the holo-cohesin complex . Furthermore, all the other subunits of cohesin absolutely require Mcd1p to bind to chromosomes . Because of this interdependency, Mcd1p binding to chromosomes is a direct measure of cohesin binding to chromosomes.
We began by analyzing the global chromosome binding of cohesin on spreads of mitotic chromosomes using indirect immunofluorescence with an antibody against Mcd1p. The levels of cohesin associated with chromosomes in 1 and 2 stop cells are reduced significantly compared to no stop cells (Figure 2A). To better assess the quantitative distribution of cohesin on chromosomes in the 1 and 2 stop strains, we analyzed cohesin binding to specific chromosome regions by chromatin immunoprecipitation using an antibody against Mcd1p. We monitored cohesin binding to three locations along chromosome (chr) III: the centromere, a pericentric CAR, and an arm CAR. Cohesin binding to most sequences within the arm CAR are reduced in the 1 and 2 stop mutants compared to wild-type cells (Figure 2B). In contrast, cohesin binding to sequences within the centromere (CEN) and pericentric CAR regions are largely unaffected (Figure 2B). To determine whether this pattern is also true for other chromosomes, we investigated cohesin binding at CEN XIV, an additional arm CAR site on chr XII, and a site within the rDNA repeats (Figure 2C, Figure S2). Similarly, cohesin binding is reduced in the 1 and 2 stop mutants compared to wild-type for almost all sequences within the arm CARs, but not at those within CEN XIV. The average fold reduction in binding relative to the no stop for all probes within centromere and pericentromeric regions is 1.3 +/−0.2 (1 stop) and 1.1 +/− 0.4 (2 stop) (Figure S2). These fold changes are not statistically significant. In contrast, the average reduction in cohesin binding to unique sequences within the ARM CARs are significantly reduced 2.2 +/−0.4 (1 stop) and 2.1 +/−0.3 (2 stop) fold (Figure S2). Thus the wild-type concentration of cohesin is necessary to ensure robust cohesin binding to arm sites, but is 8 fold greater than the concentration needed for robust binding to centromere and pericentromeric regions.
These results provide two important insights. First they reveal a previously unknown hierarchy of cohesin loading to chromosomes, with preferential binding of cohesin to centromere and pericentric regions over arm sites. This pericentric preference is unlikely to be mediated by the Scc2p/Scc4p loading complex, since the Scc2p/Scc4p loading complex is required for all cohesin loading (reviewed in ). Rather, this preference likely reflects the relative efficiency of specific modulators of pericentric and arm cohesin loading [6–11]. Second, the 1 and 2 stop mutants show indistinguishable levels and patterns of cohesin chromatin binding, despite the fact that the 1 stop mutant has almost 3 times as much cohesin in the cell. This implies that 13% of wild-type cohesin levels is sufficient to saturate the preferential CARs in the centromere and pericentric regions. Interestingly at 30% of wild-type cohesin levels, there is clearly an excess of cohesin that does not bind to the centromeric regions and also fails to bind to the low affinity arm CARs. This observation implies that these arm sites fail to recruit cohesin not because they cannot compete with the high affinity sites but rather because their occupancy requires a higher absolute cellular cohesin concentration.
Having established that the reduction in cohesin concentration leads to a non-uniform change in cohesin binding to chromosomes, we next examined the biological consequence of the reduction of cohesin levels. The fact that the 1 and 2 stop strains are viable suggests that proper chromosome segregation occurs in most cell divisions, and by inference sister chromatid cohesion must be generally intact. However, these strains may have a reduced fidelity of cohesion and chromosome segregation, particularly the 2 stop mutant which divides more slowly and exhibits an M phase delay (Figure S1). To test the amount of cohesion in these strains, we monitored sister chromatid separation in G2/M arrested wild-type, 1 and 2 stop strains using Flourescent In Situ Hydridization (FISH). We used a probe to regions 23–40 Kb or 400 Kb away from the chr 16 centromere. The levels of sister chromatid separation in G2/M arrested cells for the 1 and 2 stop strains at either locus are approximately 30%, levels indistinguishable from the spontaneous loss in the wild-type control (Figure 3A). This is in contrast to when cohesin is inactivated, where sister chromatid separation exceeds 50% . Thus the 1 and 2 stop mutants retain normal levels of cohesion even on the arms where the level of cohesin binding is reduced.
A much more sensitive assay for the loss of cohesion is an increase in the rate of loss/missegregation of non-essential reporter chromosomes, either artificial minichromosomes in haploid strains or a homolog in a diploid strain. The rates of plasmid loss are not statistically different between wild-type and the 1 and 2 stop mutants, and the 1 and 2 stop mutants show only a very minor increase in chr I loss in diploids (Table 1). The proper cohesion and chromosome segregation in the 2 stop mutant demonstrate that only a small fraction of the total cellular cohesin is needed for these processes. This conclusion is consistent with our observation that cohesin preferentially binds to centromeric regions. Since sister chromatin cohesion around the centromere is effective for bipolar attachment and proper chromosome segregation, centromere-proximal preferential binding ensures that cohesion and accurate chromosome segregation persists even at low concentrations.
The slow growth and M phase delay of the 2 stop mutant must be due to another function of cohesin. In fact, it is known that cohesin is required for proper chromosome condensation [13–15] and that defects in condensation lead to partial M phase delays [16, 17]. To test whether chromosome condensation is affected by cohesin concentration, we monitored the structure of the RDN locus (~50–70 tandem rDNA repeats). Visualization of the RDN locus in wild-type cells by FISH reveals a transition from a diffuse ball in G1 phase to an easily recognizable line-like loop in mitosis . We measured the degree of condensation of the rDNA loops in M phase arrested cells by measuring the size of the loop as illustrated and the % of cells with no observable rDNA loops (Figure 3). Both the 1 and 2 stop mutants contain increasing numbers of cells with rDNA arrays that appear smaller and/or lack the loop like structure, suggesting that these mutants perturb the compaction process (Figure 3C). Thus, higher in vivo cohesin concentrations are required to execute proper chromosome condensation compared to sister chromatid cohesion.
Cohesin is recruited to the sites of DNA damage where it plays two roles in DNA repair. It promotes more efficient repair by holding a repair template, the sister chromatid, in close proximity[19, 20]. Second it holds the sister chromatids in register, such that repair of lesions does not lead to unequal sister chromatid exchange between repetitive DNA elements . To test what concentration of Mcd1p/cohesin was critical for efficient repair, we assessed the relative viability of wild-type, 1 and 2 stop strains when challenged with different types of DNA damaging agents. Indeed, the 1 and 2 stop mutants exhibit sensitivity to DNA lesions generated by camptothecin, while the 2 stop mutant shows increasing sensitivity to other DNA damaging agents, hydroxy-urea and phyleomycin (Figure 4A). The sensitivity of these mutants to these drugs could be caused by the inability of the mutants to efficiently repair the DNA lesions or their inability to retain cohesion upon induction of a cell cycle delay. However, at least for camptothecin we favor the former for two reasons. First, the 1 stop mutant shows no growth sensitivity to G2/M delay imposed by benomyl (Figure S3). Second, cohesion of the 1 and 2 stop mutants was assayed after prolonged arrest in G2/M and no defect in cohesion was observed (Figure 3A). The 2 stop mutant is sensitive to benomyl (Figure S3). Hence its sensitivity to phleomycin and hydroxy-urea may reflect a common sensitivity to cell cycle delay, although through a cohesion-independent mechanism. Taken together, these results suggest that the total concentration of cohesin in wild-type cells is close to the amount required for efficient DNA repair, especially for DNA lesions generated by camptothecin.
To address what amount of cohesin is critical to prevent unequal sister chromatid exchange, we looked at the excision of a URA3 reporter gene inserted into the RDN locus composed of ~50 to 70 rDNA repeats. Again both the 1 and 2 stop mutants exhibit an increasing rate of unequal exchange relative to wild-type, increasing to as much as 50 fold in the 2 stop mutant (Figure 4B). Corroborating this observation, we were unable to follow sister chromatid cohesion using the assay with lacO arrays, which consist of tandom arrays of a DNA sequence similar to the rDNA repeats, because these arrays cannot be maintained in the 2 stop mutant. Thus DNA repeat stability and DNA repair, like condensation, require greater levels of cohesin than the level needed for sister chromatid cohesion and chromosome segregation. In conclusion, the phenotypic similarities between the 2 stop mutant and condensin mutants suggest that the slow growth phenotype and G2/M delay may arise in the 2 stop mutant because of a defect in condensation. However, we cannot eliminate the possibility that these phenotypes may also be the consequence of another defect, such as in DNA repair.
Here we have shown that different in vivo concentrations of cohesin are required both to execute its distinct biological functions and to occupy cohesin-binding sites around the centromere and chromosome arms. The ability to generate robust cohesion with only 13% of the wild-type cohesin level indicates that only a small fraction of the total cohesin in wild-type cells may be dedicated to cohesion. If so, this small pool may have distinct biochemical and cell biological properties that would have been missed by studies of bulk cohesin. Condensation, DNA repair, and DNA repeat stability require high levels of cohesin (this study) likely because these levels ensure proper occupancy of arm sites (this study) and/or provide a dynamic pool that is not committed to cohesion and can move to the sites of DNA lesions. Finally, the requirement for different concentrations of cohesin for its distinct functions explains how mutations affecting this complex can specifically lead to cohesin disorders which do not compromise cell division [23–26].
In conclusion, we describe and validate a general method that allows quantitative reductions in the synthesis of cohesin in budding yeast. By analogy, this QR method may be applied to other proteins/pathways with multiple functions/targets. For example, this method could prove useful for identifying and understanding preferred substrates of kinases in the MAPK signal transduction pathway, preferred transcriptional targets of complex transcriptional responses like the Msn2p/Msn4p stress response, or preferred substrates of ubiquitin conjugating complexes like the anaphase promoting complex.
Yeast strains used in this study are listed in Table S1. All strains are derivatives of the JKM genetic background. Plasmid information is also given within the strain table, written as a part of the relevant strain containing the plasmid. Yeast strains were grown in SC or YEP media as described  supplemented with 2% glucose (EMD). Exponentially dividing cultures were arrested in G2/M using 15 μg/ml nocodazole (Sigma). Extra care was taken when growing the QR strains, since they have a tendency to generate suppressors that increase the amount of Mcd1p. This was particularly a problem when cells were outgrown for extended periods of time. For this reason, strains were streaked from a frozen stock, and every experiment was tightly assayed to make sure that Mcd1p levels were reduced.
Chromosome spreads were performed as described .
Chromatin Immunoprecipitation was performed as described .
Cells were spun down and washed with dH20 and frozen in liquid nitrogen. Pellets were resuspended in IPH150 buffer (50mM Tris pH 8, 150 mM NaCl, 5mM EDTA, 0.5% NP-40, 1mM DTT) containing Roche protease mini tablets. 200–450μM glass beads (Sigma) were added to the resuspended pellets and cells were broken by bead beating for 90 seconds using a BiospecTM mini bead beater. The soluble fraction was separated from the insoluble fraction by centrifugation at 14K rpm at 4ºC. An equal volume of 2x Laemmli buffer added to the supernatant and extracts were boiled. Standard procedures for Sodium-polyacrylamide gel electrophoresis (SDS-PAGE) and Western blotting were followed . A polyscreen PVDF membrane (PerkinElmer) was used to transfer proteins from polyacrylamide gels. Membranes were blotted with primary rabbit antibodies anti-Mcd1p (antibody 559, kindly provided by V. Guacci), or anti-Tub2p (antibody #43, Koshland lab). Antibodies were detected using SuperSignal West Dura extended duration substrate (Thermo Scientific).
Fluorescence was observed using a Zeiss Axioplans 2 microscope (100x objective, NA=1.40) with a Quantix CCD camera (Photometrics).
We thank Aaron Welch, Fred Tan, Margaret Hoang, Yixian Zheng and Chen-ming Fan for constructive comments on the manuscript; Ellen Cammon, and Patricia Cammon for excellent technical support; Cynthia Wagner, Judith Yanowitz, Jeff Han, and all members of our laboratory for advice and helpful discussions. Furthermore, we thank James Haber for strains. We apologize to our colleagues whose work we were unable to cite due to space constraints.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.