Mutations in a Partitioning Protein and Altered Chromatin Structure at ...

MOLECULAR AND CELLULAR BIOLOGY, June 2004, p. 5290–5303 0270-7306/04/$08.00⫹0 DOI: 10.1128/MCB.24.12.5290–5303.2004 Copyright © 2004, American Society for Microbiology. All Rights Reserved.

Vol. 24, No. 12

Mutations in a Partitioning Protein and Altered Chromatin Structure at the Partitioning Locus Prevent Cohesin Recruitment by the Saccharomyces cerevisiae Plasmid and Cause Plasmid Missegregation Xian-Mei Yang, Shwetal Mehta, Dina Uzri, Makkuni Jayaram,* and Soundarapandian Velmurugan* Molecular Genetics and Microbiology, University of Texas at Austin, Austin, Texas 78712 Received 22 September 2003/Returned for modification 19 December 2003/Accepted 18 March 2004

The 2␮m circle is a highly persistent “selfish” DNA element resident in the Saccharomyces cerevisiae nucleus whose stability approaches that of the chromosomes. The plasmid partitioning system, consisting of two plasmid-encoded proteins, Rep1p and Rep2p, and a cis-acting locus, STB, apparently feeds into the chromosome segregation pathway. The Rep proteins assist the recruitment of the yeast cohesin complex to STB during the S phase, presumably to apportion the replicated plasmid molecules equally to daughter cells. The DNAprotein and protein-protein interactions of the partitioning system, as well as the chromatin organization at STB, are important for cohesin recruitment. Rep1p variants that are incompetent in binding to Rep2p, STB, or both fail to assist the assembly of the cohesin complex at STB and are nonfunctional in plasmid maintenance. Preventing the cohesin-STB association without impeding Rep1p-Rep2p-STB interactions also causes plasmid missegregation. During the yeast cell cycle, the Rep1p and Rep2p proteins are expelled from STB during a short interval between the late G1 and early S phases. This dissociation and reassociation event ensures that cohesin loading at STB is replication dependent and is coordinated with chromosomal cohesin recruitment. In an rsc2⌬ yeast strain lacking a specific chromatin remodeling complex and exhibiting a high degree of plasmid loss, neither Rep1p nor the cohesin complex can be recruited to STB. The phenotypes of the Rep1p mutations and of the rsc2⌬ mutant are consistent with the role of cohesin in plasmid partitioning being analogous to that in chromosome partitioning. The 2␮m circle of Saccharomyces cerevisiae is a high-copynumber selfish extrachromosomal DNA element that resides in the nucleus and propagates itself stably in the cell population (2, 29). The plasmid does not seem to confer any selective advantage to its host under normal laboratory growth conditions, nor does it pose any noticeable disadvantage as long as the copy number does not rise significantly above the steadystate value of approximately 60 per cell. The stability of the plasmid approaches that of the yeast chromosomes, with the loss rate being as low as 10⫺4 to 10⫺5 per cell per generation. The structural organization of the plasmid and its genetic potential are devoted to two goals: (i) efficient plasmid segregation during cell division and (ii) the maintenance of the plasmid copy number with only modest deviations from the mean. The presence of a typical yeast replication origin in its sequence permits each plasmid molecule to be replicated once per cell cycle (35). A stability system consisting of two plasmidencoded proteins (Rep1p and Rep2p) and a cis-acting partitioning locus (STB) mediates equal or nearly equal distribution of the replicated molecules into daughter cells. The direct observation of reporter plasmids tagged with fluorescence indicates that the plasmid molecules are organized into a tightknit cluster in the nucleus and segregate as a cluster. Hence,

the copy number relevant for partitioning is effectively a unit. An amplification system consisting of a site-specific recombinase (Flp) and its target sites (FRT) arranged in head-to-head orientation provides a safeguard against an occasional missegregation event and a resultant drop in copy number in one of the two daughter cells. The amplification reaction is thought to be triggered by a recombination event that changes the direction of one of the two replication forks in a bidirectionally replicating molecule (8, 17, 31). There is credible suggestive evidence, based on genetic and rather sparse biochemical studies, that communication occurs between the partitioning and amplification systems. The control of plasmid gene expression appears to be directed toward maintaining an active partitioning system and a silent amplification system during steady-state growth. Apparently, this general regulatory scheme has built-in mechanisms for sensing a fall in copy number, and in response, rapidly triggering the amplification process (15, 23). The functional roles for the Rep proteins and the STB locus in plasmid partitioning are not understood at the molecular level. Previous studies have provided evidence, in vivo and in vitro, for self- and cross-interactions between the Rep1 and Rep2 proteins and for the binding of these proteins in vivo to the STB locus (1, 21, 22, 28, 30). These results are consistent with an earlier finding by Hadfield et al. (11) that Rep1p or Rep2p, in conjunction with a host factor or factors, is able to bind to STB in vitro. More recent results from Sengupta et al. (22) have demonstrated the ability of the carboxyl-terminal domain of Rep2p to bind DNA. The sum of these observations

* Corresponding author. Mailing address: Molecular Genetics and Microbiology, University of Texas at Austin, Austin, TX 78712. Phone: (512) 471-0966. Fax: (512) 471-5546. E-mail for M. Jayaram: jayaram @icmb.utexas.edu. E-mail for S. Velmurugan: velmurugan@mail .utexas.edu. 5290

VOL. 24, 2004

EFFECT OF Rep1p MUTATIONS ON PLASMID PARTITIONING

5291

TABLE 1. Strains and plasmids used for this study Strain or plasmid

EGY48 MJY311 MJY312 MJY313 MJY314 MJY315 MJY316 MJY317 MJY318 MJY319 MJY320 pSH2 pJG4-5 pGAD424 pSTB2 pTS408 pSV1 pBM272 pXY17 pXY97 pXY98 pXY99 pESC-TRP pSM41 pSV22

Genotype or salient features

Source or reference

MATa his3 trp1 ura3 LexAop6-LEU2 [cir⫹] MATa ura3-52 his3-200 ade2-101 lys2-801 leu2-3,112 trp1-901 tyr1-501 gal4-⌬512 gal80-⌬538 ade5::hisG::STBUAS-HIS3::HIS3[cir⫹] MATa ade2-101 ura3-1 leu1-3,112 trp1 his3-11::MCD1-3HA::KanMX [cir⫹] MATa ade2-101 ura3-1 leu1-3,112 trp1 his3-11::MCD1-3HA::KanMX [cir0] MATa ade2-101 ura3-1 leu1-3,112 trp1 his3-11::rsc2⌬::TRP1 MCD1-3HA::KanMX [cir⫹] MATa ura3-52 his3-200 ade2-101 lys2-801 leu2-3,112 trp1-901 tyr1-501 gal4-⌬512 gal80-⌬538 ade5::hisG::rsc2⌬::TRP1::STBUAS-HIS3::HIS3[cir⫹] MATa ura3-1 leu1-3,112 trp1 his3-11::GFP-LacI::URA3 smc1-2[cir⫹] MATa ura3-1 leu1-3,112 trp1 his3-11 smc1-2:: MCD1-3HA::KanMX [cir⫹] MATa ura3-1 leu1-3,112 trp1 his3-11 SMC1-6HA::HIS3 [cir⫹] MATa ura3-1 leu1-3,112 trp1 his3-11 smc1-2-3HA::KanMX [cir⫹] MATa ura3-1 leu1-3,112 trp1 his3-11 rsc1⌬::HIS3 MCD1-3HA::KanMX [cir⫹] LexA DNA binding domain vector B42 acidic patch activation domain vector GAL4- activation domain vector 2␮m circle-derived plasmid harboring REP2, ADE2, and LEU2 CEN-ARS vector for galactose-inducible expression of GFP fusion proteins 2␮m-derived Lac(O)-containing plasmid CEN-ARS plasmid carrying GAL1-GAL10 bidirectional promoter pBM272 expressing Rep1p (galactose inducible) pBM272 expressing rep1pY43A (galactose inducible) pBM272 expressing rep1pT32K (galactose inducible) pBM272 expressing rep1pK297Q (galactose inducible) 2␮m circle-derived plasmid carrying bidirectional GAL1-GAL10 promoter to overexpress epitope-tagged proteins pESC-Trp plasmid expressing Myc-Mcd1p (galactose inducible) pJG4-5 expressing Rep1p

agrees with the generally accepted notion, albeit without direct experimental support, that the interactions among the components of the tripartite stability system (Rep1p-Rep2p-STB) are important for 2␮m circle partitioning. Sengupta et al. (22) have proposed that the Rep1 and Rep2 proteins may polymerize along the STB DNA to form a high-order structure involved in plasmid segregation. A recent observation suggests that the chromatin structure of the STB locus has a pronounced effect on its functionality in plasmid maintenance. In an rsc2⌬ strain lacking one of the RSC (remodel the structure of chromatin) complexes involved in chromatin remodeling, the 2␮m plasmid becomes highly unstable (34). The extreme simplicity of the plasmid partitioning system (just two proteins and a relatively short DNA stretch), contrasted with its impressively high efficiency—the ability to confer almost chromosome-like stability on the plasmid—has, until recently, remained somewhat of an enigma. For comparison, an elaborate mitotic apparatus and numerous protein factors are dedicated towards ensuring the fidelity of chromosome segregation. It now appears that the Rep-STB system may poach key molecular components of the chromosome segregation machinery and serve to couple plasmid partitioning to chromosome partitioning (14, 30). Seminal to this notion is the finding that the yeast cohesin complex is recruited specifically to the STB locus in a Rep1p- and Rep2p-assisted process (14). The timing of cohesin assembly and that of its disassembly during the cell cycle are remarkably similar for both the plasmid and chromosomes. Furthermore, a blockage of cohesin disassembly prevents the separation of replicated plasmid molecules into two distinct clusters, analogous to the

Roger Brent Our construct This This This This

study study study study

This study This study This study This study This study Roger Brent Roger Brent Clontech Our construct David Botstein through Clarence Chan Our lab 13a This study This study This study This study Stratagene Our construct Our construct

failure of sister chromatids to unpair under a similar constraint. These findings are consistent with a model in which plasmid partitioning requires the pairing of duplicated plasmid clusters during the S phase and their separation during anaphase via cohesin assembly and disassembly, respectively. With this study, we demonstrate that point mutants of Rep1p that disrupt its interaction with Rep2p, STB DNA, or both are unable to mediate plasmid-cohesin association. These mutants are also inactive in plasmid maintenance. Furthermore, the cycling of Rep1 and Rep2 proteins at the STB DNA as a function of the yeast cell cycle specifies the timing of cohesin recruitment to the plasmid and synchronizes the plasmid with chromosomes in cohesin assembly. In the absence of the RSC2 gene function, the chromatin structure at the STB DNA is altered (34) and Rep1p (but not Rep2p) fails to bind to this locus. Consequently, plasmid-cohesin association cannot occur in an rsc2⌬ background. Thus, the instability of the 2␮m plasmid caused by the Rep1p mutations described in this study and a similar effect resulting from a previously described mutation in RSC2 (35) are readily accommodated by a cohesin-mediated plasmid partitioning mechanism. Finally, even when the assembly of the Rep1p-Rep2p-STB ternary complex is normal, the prevention of cohesin recruitment to STB results in plasmid missegregation. MATERIALS AND METHODS Yeast strains and plasmids. The yeast strains and plasmids used for this study are listed in Table 1. The EGY48 strain, provided by Roger Brent’s laboratory, was used in dihybrid assays. Strains MJY311 and MJY315 were employed in monohybrid assays. Chromatin immunoprecipitation (ChIP) assays were per-

5292

YANG ET AL.

formed in strains 312-314 and 317-320. Plasmid and chromosome distributions were assayed by fluorescence microscopy of strain MJY316. Plasmids pSH2 and pJG4-5 were the basic vectors for cloning the genes for the bait and prey proteins for dihybrid assays (obtained from the Roger Brent laboratory). For monohybrid assays, pGAD424 provided the GAL4 activation domain vector into which DNA fragments encoding native Rep1p or Rep1p variants were introduced. Plasmid stability tests were carried out with the 2␮m circle derivative pSTB2 as the reporter. Green fluorescent protein (GFP) hybrids of Rep1p and its variants were constructed in pTS408. The plasmid pSV1 (30) was utilized for directly visualizing plasmid distribution in live yeast cells by fluorescence tagging. For some of the ChIP assays, Rep1p and Rep1p variants were expressed from the GAL10 promoter in pBM272, a CEN-ARS expression vector containing the bidirectional GAL1-GAL10 promoter. Plasmids pXY17 and pXY97-99 were pBM272 derivatives containing a REP1 insert and three rep1 (variant) inserts, respectively. A derivative of plasmid pSM41 was used to express epitope-tagged Mcd1p from the GAL1 promoter in G1-arrested cells. Further details about plasmid construction as well as plasmid maps are available upon request. Mutagenesis of REP1. Site-directed mutagenesis was performed with the REP1 gene cloned into pUC19 (pUC19-REP1). The plasmid template for the PCR-based mutagenesis procedure was prepared from a dam⫹ bacterial strain (Escherichia coli DH5␣). The oligonucleotide primers were designed to include sequence degeneracy at desired positions in order to obtain multiple substitutions at any chosen amino acid position in a single experiment. After amplification, the plasmid product was digested with DpnI to fragment the parental plasmid template which was then used to transform E. coli DH5␣. Plasmids isolated from the transformants were subjected to DNA sequencing to confirm the mutations. A subset of the mutants was then subcloned into appropriate yeast plasmid vectors for the studies reported here. Dihybrid and monohybrid assays. The interaction between wild-type Rep1p or its variants and Rep2p was tested by use of the dihybrid system (7, 28). The appropriate REP1 or rep1 fragments were cloned into pJG4-5 (Table 1) to obtain in-frame fusions with the acidic patch transcriptional activator sequence. The REP2 gene was cloned into the plasmid pSH2 (Table 1) in order to fuse it with the DNA binding domain of the LexA repressor. The hybrid proteins containing the activation domain were expressed from the GAL10 promoter, whereas the Rep2 hybrid containing the DNA binding domain was expressed constitutively from the ADH1 promoter. The criterion for an interaction between Rep2p and a partner Rep1p variant was the expression of the LEU2 reporter gene in the presence of galactose but not dextrose. A positive interaction for a protein pair by this test was confirmed by an alternative two-hybrid system (13). For monohybrid assays, a transcriptional cassette containing an approximately 375-bp STB fragment placed upstream of the basal HIS3 promoter was integrated at the chromosomal HIS3 locus (28). Translational fusions between Rep1p or its variants and the GAL4 activation domain were constructed in pGAD424 (Table 1) as described earlier (28). Growth of the tester strain in the presence of 40 mM 3-aminotriazole was considered evidence of a positive interaction. Plasmid stability assays. Wild-type Rep1p or Rep1p variants were fused to GFP by cloning the appropriate DNA fragments in pTS408, a CEN plasmid harboring URA3, as detailed earlier (1). The GFP-Rep1 hybrid proteins, expressed from the GAL10 promoter, were used to determine the cellular localization of the Rep1p variants as well as their activities in a plasmid stability assay. The test plasmid was pSTB2, a 2␮m circle derivative containing the ADE2 and LEU2 markers and the wild-type REP2 gene under its own promoter. Single colonies of the host strain (ade2) housing the Rep1p donor plasmid and the test plasmid were picked from galactose plates lacking leucine and uracil and spread out on yeast extract-peptone-dextrose or yeast extract-peptone-galactose plates. The loss of pSTB2 was indicated by the presence of red or red-sectored colonies when plates were scored after 3 days of growth at 30°C. The plasmid stability index was expressed as a ratio of the number of white colonies to the number of white colonies plus red colonies. A sectored colony was grouped with the white colonies when the sector(s) constituted less than one-fourth the colony size and with the red colonies when the sector size exceeded this value. ChIP assays. ChIP assays were performed as described by Mehta et al. (14). Antibodies to epitope-tagged Mcd1, Smc1, and native Rep1 and Rep2 proteins were used, as indicated in the relevant figures. The extract used in each ChIP assay was obtained from cells growing at an optical density at 600 nm of roughly 20. After processing, each sample yielded 500 ␮l of sheared chromatin, of which 10-␮l aliquots were employed as an input control and 400 ␮l was used for ChIP. The immunoprecipitated DNA was recovered in a final volume of 30 ␮l, and 1to 5-␮l aliquots were subjected to PCR amplification using primer pairs directed to the STB DNA or to a cohesin binding site on chromosome V. The results from

MOL. CELL. BIOL. ChIP experiments were verified by performing three independent experiments. The assays were consistent when tested over a fivefold range in the amount of input DNA, with either 20 or 30 cycles of PCR. Fluorescence microscopy. The 2␮m circle-derived reporter plasmid pSV1 was visualized under a Nikon inverted microscope as outlined previously (14), and chromosomes were visualized by standard 4⬘,6⬘-diamidino-2-phenylindole (DAPI) staining. Images were captured with a Photometrics Quantix camera from Roper Scientific and processed with MetaMorph software from Universal Imaging Corp. Miscellaneous procedures. Yeast transformation and standard genetic manipulations in yeast were performed according to published protocols (9, 18). Bacterial transformations, plasmid preparations, restriction enzyme digestion reactions, etc., were done as previously described (20). Supplemental information. The results of secondary experiments that are nevertheless directly relevant to the conclusions drawn from this study are summarized in supplemental materials at the following website: http://www.sbs .utexas.edu/jayaram/jayaramlab/mcb 2004.pdf.

RESULTS Distinct classes of mutations in REP1 that affect maintenance of the 2␮m plasmid. According to the current model for 2␮m circle partitioning, mutations that adversely affect the DNA-protein or protein-protein interactions within the RepSTB system must interfere with plasmid stability. We tested a subset of these predictions by site-directed mutagenesis of the Rep1 protein. The choice of Rep1p over Rep2p for this analysis was dictated by the higher degree of amino acid conservation for the former among yeast plasmids that resemble the 2␮m circle in structural and functional organization (27). Mutations were targeted to invariant or highly conserved amino acids. A subset of the generated Rep1p mutants was analyzed for interactions with Rep2p and STB. As described in previous work (1, 28), we employed a dihybrid assay to test the interaction between Rep1p mutants and Rep2p, with LEU2 used as the reporter gene. A positive interaction was declared for colony growth in medium lacking leucine in the presence of galactose but not dextrose (7). Similarly, we used a monohybrid test to examine the interaction between Rep1p mutants and STB DNA (28). A positive interaction was inferred from the induced expression of the HIS3 reporter gene and the consequent resistance of the tester strain to high levels of 3-aminotriazole, an inhibitor of the His3 protein. The data from the two types of interaction assays are assembled in Fig. 1. Of the 19 Rep1p mutants analyzed, 5 did not yield a detectable interaction with Rep2p (columns 1 to 3, 6 and 16 in Fig. 1A), whereas 4 had an extremely weak interaction (Fig. 1A, columns 10, 11, 13, and 17). Consistent with the results from previous deletion analyses (22, 28), four of the mutations responsible for the former phenotype were clustered within the amino-terminal 100 amino acids of Rep1p. Five of the Rep1p mutants did not yield a positive interaction with STB (Fig. 1B, columns 1, 8, 14, 18, and 19). One mutation, T32K, led to the loss of both Rep2p and STB interactions (Fig. 1A and B, columns 1). The expression levels of the mutants that tested negative in either of the two assays were comparable to that of wild-type Rep1p (Fig. 1A and B, bottom panels). The combined results from Fig. 1A and B, together with the performance of individual mutants in a plasmid stability assay, are summarized in Fig. 1C. Each of the Rep1p variants that did not bind to STB (class I), Rep2p (class II), or both STB and Rep2p (class III) failed to support the stable maintenance of a

FIG. 1. Interaction of Rep1p variants with Rep2p and STB and consequences in plasmid maintenance. (A) Dihybrid assays were performed with strain EGY48 (Table 1) and a LEU2 reporter as described previously (28). The wild-type Rep1 protein (C2) or point mutants derived from it were expressed as hybrids fused to the LexA protein from the inducible GAL10 promoter. In the negative control (C1), only the expression vector without any insert was present. The wild-type Rep2 protein fused to the B42 transcriptional activation patch was expressed from the constitutive ADH1 promoter. (B) In monohybrid assays, wild-type Rep1p (C2) or variants of it were fused to the GAL4 activation domain and expressed from the ADH promoter. The expression vector without any insert (C1) served as the negative control. A copy of the STB locus integrated into the chromosome served as the artificial UAS for the HIS3 reporter (28). The Western blots shown in the bottom panels were performed with antibodies to wild-type Rep1p or glucose-6-phosphate dehydrogenase (internal standard). (C) In a schematic linear representation of Rep1p, the amino acid substitutions in the variants are mapped. A summary of the interaction results from panels A and B and the activities of the four variant classes in plasmid maintenance are tabulated below. 5293

5294

YANG ET AL.

2␮m circle-derived reporter plasmid. In contrast, Rep1p variants that provided plasmid stabilities comparable to that of wild-type Rep1p (designated the “⫹” class) were positive for both Rep2p and STB interactions. Note that the Rep1 mutants were assayed for plasmid maintenance as GFP fusions and showed normal nuclear localization, as determined by fluorescence microscopy (data not shown). The results described above provide experimental support for two of the interactions, Rep1p-Rep2p and Rep1p-STB, being essential for normal plasmid segregation, as is implicit in the partitioning model. The possible functional relevance of the other interactions, Rep1p-Rep1p, Rep2p-Rep2p, and Rep2p-STB, in plasmid stability remains to be verified. It should be noted, though, that none of the Rep1p mutants used for the experiments described below were defective for interacting with wild-type Rep1p, as verified by in vivo dihybrid assays. For a subset of these mutants, the in vivo results were confirmed by in vitro GST pull-down assays (Fig. S1 in the supplemental material). Furthermore, the self-interaction of Rep2p was also unaffected, since it is not dependent on Rep1p (1, 21, 22). Rep1 mutations and recruitment of the cohesin complex to the STB locus. We demonstrated earlier that the yeast cohesin complex specifically associates with the STB locus in a Rep1pand Rep2p-dependent fashion (14). We also observed (in unpublished dihybrid experiments) that Rep1p or Rep2p can interact with the cohesin subunit Mcd1p. Since the tester strain contained an endogenous 2␮m circle ([cir]⫹), it is not clear whether, for either protein, this interaction is dependent on the presence of its partner. Furthermore, we do not know whether the interaction is direct or mediated through one of the other cohesin subunits or a bridging protein extraneous to the cohesin complex. Based on several pieces of circumstantial evidence, we have suggested that cohesin may play functionally similar roles in the chromosomal and 2␮m plasmid segregation pathways. If this is true, at least a subset of the Rep1 mutations that adversely affect plasmid partitioning is likely to abolish or interfere with cohesin recruitment to STB DNA. The association between cohesin and STB DNA was probed in exponentially growing yeast cells by ChIP assays using antibodies directed to a hemagglutinin (HA)-tagged Mcd1 protein (14) (Fig. 2). Similarly, the presence or absence at STB of Rep1p (or a mutant derived from it) and Rep2p was also monitored by ChIP assays employing antibodies raised against the wild-type Rep proteins. The results obtained with cells expressing native Rep1p, a representative Rep1p mutant belonging to class I (no STB interaction) or class II (no Rep2p interaction), and the only class III mutant (neither STB nor Rep2p interaction) are displayed in Fig. 2. Wild-type Rep1p and its variants were expressed in a [cir]0 host strain (cured of the 2␮m plasmid) from the inducible GAL10 promoter. A 2␮m circle-derived plasmid expressing Rep2p served as a reporter for the presence of cohesin at STB. As expected from prior observations, an Mcd1p-STB association was readily observed in the presence of native Rep1p (Fig. 2A, lane 2, top row). In contrast, in the presence of Rep1p(Y43A), Rep1p(K297Q), or Rep1p(T32K), this association was either extremely weak (Fig. 2B and D, top rows, lanes 2) or undetectable (Fig. 2C, top row, lane 2). Consistent with the interaction phenotypes of these mutants (Fig. 1), Rep1p(Y43A) (class II) was associated with

MOL. CELL. BIOL.

STB, as was native Rep1p (Fig. 2A and B, top rows, compare lanes 5); Rep1p(K297Q) (class I) and Rep1p(T32K) (class III) were not (Fig. 2C and D, top rows, compare lanes 5). The Rep1p mutations had no effect on the association between Rep2p and STB (Fig. 2, top rows, lanes 8). The control assays showed that Mcd1p was associated with one of the cognate sites for cohesin on chromosome V regardless of the Rep1p status of the cells (Fig. 2, bottom rows, lanes 2). As expected, there was no binding of Rep1p (or its variants) or Rep2p to the chromosomal cohesin binding site (Fig. 2, bottom rows, lanes 5 and 8). The above results, in conjunction with previous observations (14), are consistent with the hypothesis that one critical role for the Rep proteins in plasmid partitioning is the recruitment of cohesin to STB. Furthermore, both Rep1p-Rep2p and Rep1pSTB interactions must be satisfied simultaneously for this recruitment to occur. An earlier analysis had shown that Rep1p by itself can associate with STB even in the absence of Rep2p, and vice versa (28). In principle, a class I Rep1p mutant, by virtue of its normal interaction with Rep2p, may be indirectly recruited to STB to establish the Rep1p-Rep2p-STB ternary complex. The outcome from the present experiments contrasts with this expectation and explains why the class II Rep1p mutants are unable to assist the loading of cohesin at STB. Cycling of Rep proteins at STB during the G1-S interval may ensure replication-dependent plasmid-cohesin association and coordination with chromosomal cohesin recruitment. A possible role for cohesin in the pairing of replicated plasmid clusters (similar to the pairing of sister chromatids) is suggested by the finding that, during the normal cell cycle, the timing of cohesin association and dissociation is the same for the 2␮m plasmid and the chromosomes (14). Furthermore, as is the situation with sister chromatids, a lack of cohesin disassembly due to a noncleavable form of Mcd1p causes the cells to arrest in a large budded state with a single plasmid cluster rather than two separated clusters. One plausible scenario suggested by these data, by analogy to chromosomes, is that the plasmid is primed for cohesin association only during its replication phase. However, when Mcd1p is expressed inappropriately in G1, it is bound to the STB locus but not to chromosomal sites, as inferred from chromosome spread assays and substantiated by ChIP analyses (14). Although cohesin binding to STB does not necessarily mean the establishment of cohesion between plasmids, the replication independence of Mcd1p binding raises some concern regarding its relevance to partitioning. An important question is whether, during each cell cycle, there is some mechanism for initiating a fresh round of replicationcoupled cohesin recruitment at STB. In this case, even if a fortuitous plasmid-cohesin association were to occur prior to replication, it would be subsequently overridden. Previous experiments showed that the binding of Mcd1p to STB in G1 reflects that of the whole cohesin complex (14). When either Smc1p or Smc3p harbors a T s mutation, the G1-expressed Mcd1p protein is not recruited to STB at the nonpermissive temperature. We wanted to know whether, like normal cohesin-STB association, that imposed in G1 is also mediated by the Rep proteins. The ChIP assays shown in Fig. 3A were performed in G1-arrested (and galactose-induced) cells expressing Myc-tagged Mcd1p from the GAL1 promoter. In the presence of wild-type Rep1 and Rep2 proteins, Mcd1p

VOL. 24, 2004


5295

FIG. 2. Association of Mcd1p with STB in the presence of wild-type Rep1p or its variants. ChIP assays were performed with a [cir]0 strain (lacking endogenous 2␮m circles) in which the MCD1 gene was tagged with three copies of the HA tag. The tag was present at the carboxyl terminus of the Mcd1 protein. Wild-type Rep1p or the indicated mutant was expressed from the GAL10 promoter present on a centromeric plasmid. The Rep2 protein was expressed from its native promoter in a multicopy plasmid. The antibody (Ab) used for each assay set is indicated above the respective panel. The PCRs were done with primer pairs specific to STB or to a chromosome V cohesin binding site (14). The whole-cell extract, the experimental sample, and the control without antibody treatment are referred to here and in subsequent figures as WCE, ChIP, and no Ab, respectively. The schematic to the right of each panel symbolizes the phenotype of Rep1p or the Rep1p variant used in the corresponding set of assays. (A) Normal interactions with Rep2p and STB; (B) no interaction with Rep2p; (C) no interaction with STB; (D) no interaction with Rep2p or STB.

was associated with STB, but not with the chromosome V binding site (Fig. 3A, lane 2). The Rep1 and Rep2 proteins were also associated with STB in these cells (Fig. 3A, lanes 5 and 8). In cells expressing a class I Rep1p mutant, Rep1p (K297Q), Mcd1p was not detected at STB (Fig. 3A, lane 11). As expected, the mutant Rep1p was not present at STB (Fig. 3A, lane 14), whereas Rep2p was (Fig. 3A, lane 17). Similar experiments showed that G1-expressed Mcd1p was not associated with STB when the Rep1 protein contained a class II (Y34A) or class III (T32K) mutation (Fig. S2 in the supplemental material). The mutant Rep1p protein was present at STB in the former case but not in the latter. Neither mutation had any effect on the Rep2p-STB association. Thus, even the untimely enlistment of cohesin by the plasmid during G1 is dependent on the functional interactions of Rep1p with Rep2p and STB. We next followed the patterns of occupancy of STB by Rep1p and Rep2p as a function of cell cycle progression in cells grown in dextrose. Simultaneously, the association of Mcd1p, expressed from its native promoter in an epitope-

tagged form, with STB as well as with its chromosome V binding site was also monitored. Both Rep1p and Rep2p were associated with STB at the time of release from alpha factor (Fig. 3B, 0 min). As expected, Mcd1p was absent from STB and chromosome V prior to the onset of S phase (Fig. 3B, 0 to 30 min). Strikingly, during the window between the late G1 and early S phases (Fig. 3B, 15 to 30 min), the Rep proteins were cleared from STB. They then reassociated with STB at the same time that an Mcd1p-STB or Mcd1p-chromosome V association was established (Fig. 3B, 30 min). The Rep proteins persisted at STB after cohesin disassembly from STB and chromosome V and into the G1 phase of the subsequent cell cycle (Fig. 3B, 90 to 120 min). The process of dissociation and reassociation was then repeated (Fig. 3B, 120 to 135 min). The choice of dextrose as the carbon source for the assays depicted in Fig. 3B made them roughly comparable to previously published cell cycle analyses on the stage-dependent acquisition of Mcd1p (or cohesin) by the 2␮m plasmid (14). Results from similar experiments performed in the presence of galactose (with Mcd1p overexpressed continuously during the

5296

YANG ET AL.

MOL. CELL. BIOL.

FIG. 3. Association of G1-expressed Mcd1p with STB: cell cycle dependence of Rep1p, Rep2p, Mcd1p, and Smc1p association with STB. (A) ChIP analyses were performed with [cir]0 cells arrested in G1 with ␣ factor in the presence of galactose. The tagged Mcd1p protein was expressed from the GAL1 promoter harbored by a 2␮m circle-derived plasmid (containing STB). Wild-type Rep1p or Rep1p(K297Q) was expressed under the control of the GAL1 promoter from a CEN-ARS plasmid. Rep2p was expressed in trans from a multicopy plasmid under its own promoter. (B) Assays were performed with a [cir]⫹ strain in which epitope-tagged Mcd1p was expressed from the normal chromosomal locale under the control of the native promoter for the MCD1 gene. After release from ␣ factor arrest, cells were sampled at the indicated time intervals and subjected to ChIP as well as FACS analyses. (C) The host strain contained a replacement of the native SMC1 gene by the epitope-tagged version. (D) The endogenous 2␮m circle or a derivative plasmid containing STB but lacking the REP1 and REP2 genes (pSTB) served as a reporter for Smc1p-STB association.

VOL. 24, 2004


cell cycle from the GAL promoter) corroborated the data in Fig. 3B regarding the cycling of Rep proteins at STB (Fig. S2 in the supplemental material). However, the timing of the DNA-protein association and dissociation steps of interest was different in the two analyses, as was expected because of the increased duration of the cell cycle and the inappropriate expression of Mcd1p during growth in galactose. As explained below, the ChIP and FACS data shown in Fig. 3B are mutually concordant, even though the FACS patterns from the two cell cycles are not as well matched as the ChIP patterns. One likely contributing factor to this discrepancy was the relaxation in the degree of synchrony as cells passaged from the first to the second cell cycle. Recovery from pheromone arrest might also introduce subtle timing differences between the two cycles, at least in their early phases. The absence of Mcd1p and the presence of the Rep proteins at STB at 105 min (with most cells displaying 2⫻ DNA content [G2/M] and only a minority having 1⫻ DNA content [G1]) and at 0 min (with almost the entire population being in G1) were not contradictory. Presumably, anaphase had been triggered in the 2⫻ class of cells, causing Mcd1p cleavage and the dissociation of the cohesin complex. Similarly, the 15-min population contained cells set to exit from G1 (note that S phase was established within 30 min), while the 120-min population consisted primarily of late G1 cells, with some entry into S phase indicated by the trailing shoulder of the G1 peak. Combining this information with the ChIP data, the transient exclusion of the Rep proteins from STB may be assigned to the time span straddling the end of G1 and the beginning of S phase. The association of STB with Rep1p, Rep2p, and Mcd1p in the 30-min and 135-min samples signifies that, despite quantitative differences, these populations were qualitatively similar in harboring mostly S phase-to-pre-anaphase cells. ChIP analyses analogous to those shown in Fig. 3B, performed by targeting HA-tagged Smc1p and native Rep proteins, showed that all three proteins occupied STB and exited from it at the same time points. The data for Rep1p and Smc1p are presented in Fig. 3C (rows 1 and 2); the pattern for Rep2p was the same as that for Rep1p (data not shown). Note also that the brief exit of Smc1p during the 0- to 30-min and 90- to 120-min intervals was specific to STB (Fig. 3C, row 2) and did not occur at the chromosome V site (Fig. 3C, row 3). Furthermore, Smc1p was detected at STB by the ChIP assay only in a [cir]⫹ strain, not in a [cir]0 strain (Fig. 3D, top row, compare lanes 2 and 5). The association between Smc1p and the 2␮m plasmid was specific to STB, and no interaction with DNA regions within the FLP, REP1, or REP2 genes was detected (results not shown). Together, these data suggest that the Rep proteins likely mediate the recruitment of Mcd1p via the Smc proteins. Note that Smc1p and Smc3p are present throughout the cell cycle, whereas the expression of Mcd1p only occurs close to the onset of S phase (24). In principle, the observed cycling of the Rep proteins at the STB locus guarantees that the 2␮m plasmid only acquires cohesin during each cell cycle in a replication-dependent manner and in coordination with the chromosomes, as would be consistent with the currently entertained partitioning models (14). In addition, the Rep1 and Rep2 proteins appear to constitute the cohesin loading complex for the plasmid, as inferred from the absence of Smc1p at the STB locus of a reporter in the

5297

[cir]0 background (hence lacking Rep1p and Rep2p) as well as the almost perfect match between the Rep proteins and Smc1p in their patterns of STB association during the cell cycle. In the case of chromosomes, the Smc1 and Smc3 proteins are associated only with cohesin binding sites, and this association is mediated by the Scc2p/Scc4p cohesin loading complex (5). The association between cohesin and STB is blocked in the rsc2⌬ background. A recent discovery by Wong et al. (34) revealed a strong correlation between the nucleosome architecture at STB and the stable propagation of the 2␮m plasmid. In the absence of the Rsc2 protein, a component of one of the nucleosome remodeling complexes in yeast, there is a high rate of loss of the 2␮m circle (34). Concomitantly, there is a finite alteration of the chromatin structure of the STB region. Is the plasmid instability in the RSC2 deletion background possibly brought on by a lack of cohesin recruitment to STB? The results from ChIP assays performed with an rsc2⌬ yeast strain and an isogenic wild-type counterpart, both containing a selectively maintained 2␮m plasmid derivative [cir]⫹, to monitor the Mcd1p association with STB are shown in Fig. 4A. Whereas Mcd1p was detected at STB and chromosome V in the wild-type strain (Fig. 4A, lane 2), it was absent from STB in the rsc2⌬ strain (Fig. 4A, top row, lane 5). In contrast, the deletion had no apparent effect on the association of Mcd1p with chromosome V (Fig. 4A, bottom row, lane 5). The ChIP results were then verified by a monohybrid assay performed with a chromosomally integrated copy of the STB DNA (Fig. 4B). In conformity with the data for Fig. 4A, Mcd1p interacted with STB in the wild-type strain (Fig. 4B, lane 2) but not with that in the rsc2⌬ strain (Fig. 4B, lane 4). In contrast to the deletion of RSC2, the deletion of RSC1 did not have any obvious deleterious effect on the association of Mcd1p with STB, as determined by a ChIP assay (Fig. 4C, lane 2). These data are consistent with the fact that, unlike the rsc2⌬ mutation, an rsc1⌬ mutation does not lead to an elevated loss of the 2␮m plasmid (34). The above findings suggest a model in which the Rsc2 protein facilitates 2␮m circle stability by preserving the chromatin structure of STB in a state that is conducive for recruiting cohesin. When this functional architecture of the locus is altered due to the lack of one of the RSC nucleosome remodeling complexes, STB is rendered incompetent in acquiring cohesin. As a result, cohesin-mediated equal partitioning of the plasmid becomes impossible. Absence of Rsc2 protein affects binding of Rep1p but not Rep2p to STB. In principle, the exclusion of cohesin from STB in the rsc2⌬ strain may result from the inability of Rep1p, Rep2p, or both to bind STB in its altered chromatin state. As shown in this study (Fig. 2), a mutation in Rep1p that eliminates STB binding also blocks STB-cohesin association. We therefore monitored the effect of the RSC2 deletion on the occupancy of STB by Rep1p and Rep2p by using ChIP assays (Fig. 4D) as well as monohybrid (Fig. 4E) assays. As expected, in the wild-type background, both the Rep1 and Rep2 proteins bound to STB (Fig. 4D, lanes 2 and 5, and 4E, lanes 2 and 3). In the absence of Rsc2p, the association between Rep1p and STB was abolished (Fig. 4D, lane 8, and 4E, lane 5). On the other hand, Rep2p binding to STB was unaffected by the rsc2⌬ mutant (Fig. 4D, lane 11, and 4E, lane 6). In the case of the rsc1⌬ mutant, both Rep1p and Rep2p

5298

YANG ET AL.

MOL. CELL. BIOL.

FIG. 4. Association of Mcd1p or the Rep proteins with STB in RSC2 and RSC1 deletion strains. (A, C, D, and F) Immunoprecipitations were done with the indicated antibodies in a [cir]⫹ wild-type strain or its rsc2⌬ and rsc1⌬ derivatives expressing HA-tagged Mcd1p from the native chromosomal locus. The [cir]⫹ designation for the rsc2⌬ strain refers to a 2␮m plasmid derivative that was maintained selectively and served as the source for the Rep1 and Rep2 proteins. For the rsc1⌬ strain, [cir]⫹ indicates the presence of the native 2␮m plasmid. (B and E) Monohybrid assays were performed as described in the legend for Fig. 1. Vector lanes, negative control assays in which the GAL4 activation domain was not fused to a test protein. The proteins tested for an interaction with STB are indicated above the corresponding lanes.

VOL. 24, 2004


were found to be present at STB, when analyzed by ChIP (Fig. 4F, lanes 2 and 5). This was expected since the deletion of RSC1 did not affect Mcd1p (or cohesin, by inference) recruitment to STB (Fig. 4C). The data from Fig. 4 suggest that the nucleosome organization at STB promoted by the Rsc2p-containing remodeling complex is specific to the recruitment of Rep1p. Rep2p can occupy STB even when the latter is in a nonfunctional chromatin state induced by the absence of Rsc2p. The RSC2 deletion and mutations of Rep1p that eliminate its interaction with STB (class I) (Fig. 1) are therefore functionally equivalent in destabilizing the 2␮m plasmid. According to our current thinking, their downstream effect of preventing plasmid-cohesin association is the likely cause for plasmid loss. At this time, we cannot rule out the possible alternative that the interaction between Rep1p and STB, which is abolished in the absence of Rsc2p, is what is essential for the nucleosome organization at STB. Blocking cohesin-STB association without affecting RepSTB interactions causes 2␮m circle missegregation. We previously showed that a temperature-sensitive mutation in a component of the cohesin complex (Smc1p or Smc3p) blocks its association with STB, as well as with chromosomal cohesin binding sites, at the nonpermissive temperature (14). Do the Rep1 and Rep2 proteins stay associated with STB under restrictive conditions? Furthermore, how is 2␮m circle segregation affected by the mutation? The answers to these questions, displayed in Fig. 5, have an important bearing on the suspected role of cohesin in plasmid partitioning. ChIP results for the smc1-2 (Ts) strain, obtained by using antibodies directed to the Mcd1 protein, confirmed our earlier finding that cohesin could not be detected at STB or at the chromosome V binding site in cells arrested at 37°C but was present at these locales in cells growing at 26°C (Fig. 5A, compare lanes 2 and 5). When ChIP was performed with the mutant Smc1p being tagged with Myc, occupancy of STB by the protein was detected at the permissive temperature (Fig. 5B, lane 2) but not at the nonpermissive temperature (Fig. 5B, lane 5). Assays with Rep protein antibodies showed Rep1p and Rep2p to be associated with STB at both temperatures (Fig. 5C and D, lanes 2 and 5). In one explanation of the above data, it is the whole cohesin complex, containing Mcd1p as well as Smc1p, that binds to STB. When preassembly of the complex is prevented, individual subunits can no longer bind to STB. In a second explanation, during the assembly of the cohesin complex at STB, the association of Smc1p with STB precedes that of Mcd1p. The latter explanation is favored by the data in Fig. 3C showing that Smc1p is associated with STB during G1, even though Mcd1p expression is not turned on until later in the cell cycle (5). In either case, the recruitment of cohesin to the 2␮m plasmid can be interrupted by a mutation in one of its subunits, while apparently normal interactions between the Rep proteins and the STB DNA are retained. We next examined the pattern of plasmid partitioning in the smc1-2 strain under permissive and restrictive conditions, as indicated schematically in Fig. 5E. Cells from a log-phase culture grown at 26°C were first arrested in G1 by use of alpha factor and were then released into a pheromone-free medium at 26 or 37°C. For the 26°C population, the plasmid partition-

5299

ing data were derived from large budded cells (constituting the major subset of cells) present 1 h after the cell cycle restart. For the 37°C population, the corresponding values were obtained from cells arrested in the large budded state (⬎80% of the population) after 2 h at the shifted temperature. In estimating the values for the two cell types depicted in Fig. 5E, the criterion for normal segregation (class I) was a rough equivalence between the two cell compartments in chromosome content as well as plasmid content, as revealed by DAPI staining and plasmid-associated fluorescence, respectively. For the plasmid, any inequity in partitioning was additionally displayed by the difference in the number of fluorescent foci between the clusters present in the two compartments. The most prominent missegregation phenotype at 37°C was one in which the plasmid and the chromosomes missegregated in a proportional manner. In this group of cells (class II), the two cell compartments clearly displayed nonuniformity in plasmid-associated fluorescence as well as DAPI fluorescence. However, the cell compartment with a higher chromosome content also had a higher plasmid content. The ratio of class II cells to the sum of class I and class II cells, called the aberrant segregation index, rose from ⬍2% at 26°C to approximately 70% at 37°C. The above analysis demonstrates that gross plasmid missegregation can be effected by obstructing cohesin recruitment to STB while leaving interactions between STB and the Rep proteins unaffected. Given this result, the incompetence of the class I to III Rep1p mutants in plasmid maintenance is most readily explained by their common inability to mediate the cohesin-STB association. DISCUSSION In this study, we have presented a molecular analysis of the 2␮m circle stability system, using three classes of mutations introduced into one of the partitioning proteins (Rep1p) as well as a host mutation (rsc2⌬) that affects the chromatin organization of the partitioning locus STB. We found that disrupting the interaction of Rep1p with STB (class I mutants), with its partner protein Rep2p (class II mutants), or with both Rep2p and STB (class III mutants) results in a high rate of plasmid loss. A similar plasmid instability phenotype has been documented for rsc2⌬ mutants as well (34). All three classes of Rep1p mutations abolished the normal association between the yeast cohesin complex and STB, as did the RSC2 deletion mutation. The effect of the latter was most likely manifested through the Rep1 protein, which, unlike the Rep2 protein, cannot bind to STB in the deletion background. These results corroborate several pieces of circumstantial evidence suggesting a possible role for cohesin during normal segregation of the 2␮m plasmid (14). In further support of such a role, we have now shown that plasmid missegregation can also be induced by directly preventing cohesin assembly at STB without interfering with the interactions between the Rep proteins or their binding to STB. In the following sections, we elaborate the implications of these findings in the context of the plasmid segregation model presented in Fig. 6. Relevance of DNA-protein and protein-protein interactions of the Rep-STB system in faithful plasmid segregation. Previous studies have shown that the Rep1 and Rep2 proteins can interact with each other in the absence of STB and that each

5300

YANG ET AL.

MOL. CELL. BIOL.

FIG. 5. Absence of Mcd1p-STB association and high level of plasmid missegregation in the smc1-2 strain at a nonpermissive temperature. (A to D) Association of Mcd1p, Smc1p, or the Rep proteins with STB as monitored by ChIP in exponentially growing cells at 26°C or in cells arrested in the large budded state after being shifted to 37°C for 2 h. (E) Plasmid distributions were examined in large budded cells from a growing culture maintained at 26°C or from a culture arrested at 37°C. The procedure for obtaining these cells is indicated schematically at the top. The STBcontaining reporter plasmid was tagged with green fluorescence by using the Lac operator/GFP-Lac repressor system as described previously (30). Chromosomal segregation was monitored by DAPI staining. Class I represents equal segregation patterns for both the plasmid and chromosomes. Class II represents the most prominent type of plasmid missegregation observed at 37°C. The less frequent types of missegregation were not scored in this assay.

VOL. 24, 2004


5301

FIG. 6. Stages in the pathway for 2␮m circle segregation. Rep1 and Rep2 proteins, recruited to STB in its functional chromatin state, interact with each other to form the stage I complex. This complex cannot be assembled in the rsc2⌬ background because Rep1 cannot associate with STB in its nonfunctional chromatin state. The class I and class III mutants of Rep1p cannot support stage I establishment because they cannot bind STB. Even though the class I Rep mutants can interact with Rep2p, this interaction is not sufficient to maintain them at STB. The class II Rep1p mutants fail to complete stage I because they are unable to interact with Rep2p. Competence to assemble stage I is a prerequisite for the appropriation of cohesin and progression to stage II. A blockage of stage I or its transition to stage II leads to plasmid missegregation. Stage III refers to the final steps in segregation culminating in the proteolytic cleavage of Mcd1p and the disassembly of cohesin. When the completion of stage III is prevented, as is the case when the assembled cohesin contains noncleavable Mcd1p, plasmid clusters fail to separate (14).

protein can interact with STB in the absence of the partner protein (11, 21, 22, 28). The present analysis with the Rep1p mutants revealed that both Rep1p-Rep2p and Rep1p-STB interactions must be simultaneously fulfilled in order for a Rep1p-Rep2p-STB ternary complex that is productive in plasmid partitioning to form (Fig. 6, stage I). A Rep1p variant that cannot interact with Rep2p is still recruited to the STB locus, but it cannot support subsequent steps in partitioning, even though Rep2p is also present at STB. Conversely, a Rep1p variant that cannot interact with STB is not rescued by wildtype Rep2p, with which it interacts. The variant protein is excluded from STB and thus fails to partake in plasmid maintenance. We propose that wild-type Rep1p and Rep2p are able to functionally interact with each other only after each protein has established its association with STB. The formation of this prepartitioning DNA-protein entity corresponds to stage I of the plasmid partitioning pathway (Fig. 6) and is required for the recruitment of cohesin to the plasmid. The effects of the Rep1p mutants from classes I to III on plasmid stability can be attributed to their failure to complete this stage. Role of chromatin organization at STB in completion of stage I. Our analysis demonstrates that the RSC2 deletion mimics the class I Rep mutations in its failure to support the

assembly of the stage I complex and provides a reasonable molecular explanation for the high level of plasmid loss observed by Wong et al. (34) in this genetic background. The RSC complex, of which there are at least two forms, is functionally similar to the SWI/SNF chromatin remodeling complex and is involved in repositioning nucleosomes in an energyconsuming reaction that is dependent on ATP hydrolysis (3, 10, 16, 19). The RSC complex appears to modulate the chromatin structure at centromeric regions and is required for normal kinetochore functioning and proper chromosome segregation (12). An analogous role for the complex in STB activity fits the observed effects of RSC2 deletion on 2␮m circle partitioning (34). The functional architecture of the STB DNA is dependent on only one of the RSC complexes, specifically the one containing Rsc2p but not the one containing its homolog, Rsc1p. In the absence of Rsc2p, a significant alteration in the nucleosome organization, as assayed by micrococcal nuclease digestion, occurs at STB, whereas other regions of the 2␮m plasmid are unaffected (34). We therefore suspected that the altered chromatin structure of STB is responsible for the poor maintenance of the 2␮m circle. Our current data suggest that a proper nucleosome arrangement at STB, induced by the Rsc2p-specific remodeling complex, is required for the association between Rep1p and STB.

5302

YANG ET AL.

It is also possible that the primary effect of an RSC2 deletion is a disruption of the Rep1p-STB interaction, which then leads to altered chromatin organization at STB. In contrast to Rep1p, Rep2p can associate with STB in the RSC2 deletion background. This is not surprising, since in monohybrid assays performed in vivo the interaction of one Rep protein with STB was not dependent on the other (28). As revealed by the class I Rep1p mutations, the stage I complex cannot be established in the absence of Rep1p at STB, even when the latter is occupied by Rep2p. Consequently, cohesin recruitment to STB is blocked. The high level of plasmid instability resulting from the lack of Rsc2p therefore fits into a cohesin-based partitioning scheme. Consistent with the dispensability of Rsc1p in plasmid stability, we found that an RSC1⌬ mutation had no effect on the association of Rep1p, Rep2p, and Mcd1p with STB. Timing of the assembly of the stage I complex during the cell cycle: preparation for stage II. The pattern of STB occupancy by the Rep1 and Rep2 proteins indicates that the stage I complex is assembled de novo during each cell cycle, just prior to or early in S phase. This timing corresponds closely to that of the expression of the Mcd1 protein and the loading of the cohesin complex at the cognate chromosomal locales (24). The Rep1 and Rep2 proteins continue to stay associated with STB beyond the disassembly of cohesin during anaphase and the subsequent completion of cytokinesis. As revealed by artificially expressing the Mcd1 protein during G1, the stage I complex from a given cell cycle persists through at least part of the G1 phase of the next cell cycle. It is then disassembled and reassembled, presumably to initiate a new round of cohesinmediated plasmid segregation. Why do the Rep proteins stay at STB for a certain length of time (approximately 15 min, as shown in Fig. 3B) after Mcd1p has already been cleaved? Perhaps the continued presence of these proteins is required for the process by which two plasmid clusters, separated from each other after Mcd1p cleavage, are distributed to potential daughter cells. According to plasmid segregation models that are currently under consideration (14), this may involve tethering of the plasmid clusters to sister chromatids, attachment of the clusters to the mitotic spindle or spindle-associated proteins, or an interaction between the plasmid and a cell component that is equally divided between daughters. The RSC2 complex also associates with STB in a stagespecific manner during the yeast cell cycle (B. Laurent, personal communication). This association precedes the cohesinSTB association by approximately 15 min, remarkably close to the time at which the Rep proteins exited from STB in our experiments (Fig. 3B). Perhaps the locus needs to be cleared of the Rep proteins before the RSC2 complex can gain access to it. Alternatively, the RSC2 complex may actively expel the Rep proteins as a prelude to or concomitant with the remodeling of nucleosomes at STB. It is possible that these events are linked in some way to DNA replication through the STB region. Rebinding of the Rep proteins to the remodeled STB would then set the stage for cohesin recruitment to the plasmid (Fig. 6, stage II). Chromosomes acquire cohesin coincident with DNA replication so that sister chromatids may be kept paired until it is time for them to be separated and pulled apart to opposite cell poles. Based on currently available evidence, it is believed that

MOL. CELL. BIOL.

the replication machinery pauses at chromosomal cohesin binding sites and exchanges the resident DNA polymerase with a new one, the sigma polymerase, which facilitates the deposition of cohesin at these sites (4, 6, 32, 33). Assuming that a replication-dependent pairing of two plasmid clusters is relevant to partitioning, the point in the cell cycle at which the plasmid becomes competent to acquire cohesin must be precisely timed. Despite the synchrony between the 2␮m plasmid and the chromosomes in the acquisition of cohesin, there are clear mechanistic differences between them regarding how this acquisition is mediated (14; S. Mehta and M. Jayaram, unpublished observations). We do not know whether a replication pause and polymerase replacement are required for cohesin loading on the plasmid. Furthermore, when it is expressed artificially in G1, the Mcd1 protein can associate with STB in a Rep protein-assisted manner, whereas it is excluded from a chromosomal cohesin binding site (Fig. 3A; also see Fig. S2 in the supplemental material). Nevertheless, the expulsion of Rep proteins from STB and the reassembly of the stage I complex at precise times in the cell cycle can potentially prepare the plasmid to receive a fresh batch of cohesin concomitant with DNA replication. Blocking the establishment of stage II results in 2␮m circle missegregation. When the assembly of the stage II complex is blocked with the help of a temperature-sensitive mutation in the cohesin component Smc1p (Fig. 6), gross missegregation of the 2␮m plasmid ensues. At the same time, the mutation does not affect the association of the Rep1 and Rep2 proteins with the STB locus, and by inference, the formation of the stage I complex. We suggest that the progression from stage I through stage II represents the natural sequence of events in the pathway for plasmid partitioning (Fig. 6). It should be noted that the order of events, the completion of stage I followed by the establishment of stage II, as depicted in Fig. 6, is strictly true only in a functional sense and does not necessarily portray a temporal sequence. It is formally possible that the association of the Rep proteins with STB (stage I) and cohesin recruitment (stage II) are concerted events. The relevant point is that the existence of stage II is predicted upon satisfying the interactions engendered by stage I. The present studies indicate that Mcd1p is not required for a Smc1p-STB association, whereas the Rep proteins appear to be essential (Fig. 3C and D). As a result, after the cleavage of Mcd1p, Smc1p (and likely Smc3p as well) can stay in contact with STB, but only for the same length of time that the Rep1 and Rep2 proteins remain at STB. These data suggest two possible models for cohesin assembly on the plasmid. In one, the preassembled cohesin complex associates with STB-bound Rep proteins by means of interactions mediated through one or both of the Smc proteins. In the second model, cohesin is recruited to STB in a hierarchical fashion. The Smc proteins interact first with the Rep proteins, to be joined later by Mcd1p. Completion of plasmid segregation: stage III. Our present experiments did not address the events in plasmid partitioning beyond stage II. However, based upon previous observations (14), it seems likely that the proteolytic cleavage of Mcd1p (25, 26) and the dissolution of the cohesin bridge to separate the sister clusters are some of the critical events during stage III of plasmid segregation. How the separated clusters find their final destination in the daughter cells remains an open question.

VOL. 24, 2004


Does plasmid segregation occur in a state in which the chromosome is tethered so that each cluster can “hitchhike” in opposite directions with each one of a pair of sister chromatids? Or does this directed movement take place independent of chromosome attachment? Perhaps mutations or experimental conditions that cause differential loading of cohesin on the plasmid and chromosomes or differential unloading of cohesin from them may help us to tackle these currently unresolved issues. ACKNOWLEDGMENTS We are grateful to Douglas Koshland and Vincent Guacci for providing yeast strains. We thank Brehon Laurent for communicating results prior to publication. This work was supported by funds from the National Institutes of Health (GM064363). S.M. is a recipient of the University Continuing Fellowship (2002) as well as the William S. Livingston Fellowship (2003) from the University of Texas at Austin. REFERENCES 1. Ahn, Y. T., X. L. Wu, S. Biswal, S. Velmurugan, F. C. Volkert, and M. Jayaram. 1997. The 2micron-plasmid-encoded Rep1 and Rep2 proteins interact with each other and colocalize to the Saccharomyces cerevisiae nucleus. J. Bacteriol. 179:7497–7506. 2. Broach, J. R., and F. C. Volkert. 1991. Circular DNA plasmids of yeasts, p. 297–331. In J. R. Broach, J. R. Pringle, and E. W. Jones (ed.), The molecular biology of the yeast Saccharomyces. Genome dynamics, protein synthesis and energetics. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y. 3. Cairns, B. R., A. Schlichter, H. Erdjument-Bromage, P. Tempst, R. D. Kornberg, and F. Winston. 1999. Two functionally distinct forms of the RSC nucleosome-remodeling complex, containing essential AT hook, BAH, and bromodomains. Mol. Cell 4:715–723. 4. Castano, I. B., P. M. Brzoska, B. U. Sadoff, H. Chen, and M. F. Christman. 1996. Mitotic chromosome condensation in the rDNA requires TRF4 and DNA topoisomerase I in Saccharomyces cerevisiae. Genes Dev. 10:2564– 2576. 5. Ciosk, R., M. Shirayama, A. Shevchenko, T. Tanaka, A. Toth, and K. Nasmyth. 2000. Cohesin’s binding to chromosomes depends on a separate complex consisting of Scc2 and Scc4 proteins. Mol. Cell 5:243–254. 6. Edwards, S., C. M. Li, D. L. Levy, J. Brown, P. M. Snow, and J. L. Campbell. 2003. Saccharomyces cerevisiae DNA polymerase epsilon and polymerase sigma interact physically and functionally, suggesting a role for polymerase epsilon in sister chromatid cohesion. Mol. Cell. Biol. 23:2733–2748. 7. Finley, R. L., and R. Brent. 1986. Interaction trap cloning with yeast. Oxford University Press, Oxford, United Kingdom. 8. Futcher, A. B. 1986. Copy number amplification of the 2 micron circle plasmid of Saccharomyces cerevisiae. J. Theor. Biol. 119:197–204. 9. Gietz, R. D., and A. Sugino. 1988. New yeast-Escherichia coli shuttle vectors constructed with in vitro mutagenized yeast genes lacking six-base pair restriction sites. Gene 74:527–534. 10. Goodwin, G. H., and R. H. Nicolas. 2001. The BAH domain, polybromo and the RSC chromatin remodelling complex. Gene 268:1–7. 11. Hadfield, C., R. C. Mount, and A. M. Cashmore. 1995. Protein binding interactions at the STB locus of the yeast 2 micron plasmid. Nucleic Acids Res. 23:995–1002. 12. Hsu, J. M., J. Huang, P. B. Meluh, and B. C. Laurent. 2003. The yeast RSC chromatin-remodeling complex is required for kinetochore function in chromosome segregation. Mol. Cell. Biol. 23:3202–3215. 13. James, P., J. Halladay, and E. A. Craig. 1996. Genomic libraries and a host strain designed for highly efficient two-hybrid selection in yeast. Genetics 144:1425–1436. 13a.Johnston, M., and R. W. Davis. 1984. Sequences that regulate the divergent GAL1-GAL10 promoter in Saccharomyces cerevisiae. Mol. Cell. Biol. 4:1440–1448.

5303

14. Mehta, S., X. M. Yang, C. S. Chan, M. J. Dobson, M. Jayaram, and S. Velmurugan. 2002. The 2 micron plasmid purloins the yeast cohesin complex: a mechanism for coupling plasmid partitioning and chromosome segregation? J. Cell Biol. 158:625–637. 14a.Mehta, S., M. Jayaram, and S. Velmurugan. 2003. Yeast plasmids: models for the evolutionary persistence of selfish extrachromosomal elements, p. 249–271. In S. G. Pandalai (ed.), Recent research developments in molecular and cellular biology, vol. 4. Research Signpost Press, Trivandrum, India. 15. Murray, J. A., M. Scarpa, N. Rossi, and G. Cesareni. 1987. Antagonistic controls regulate copy number of the yeast 2 micron plasmid. EMBO J. 6:4205–4212. 16. Ng, H. H., F. Robert, R. A. Young, and K. Struhl. 2002. Genome-wide location and regulated recruitment of the RSC nucleosome-remodeling complex. Genes Dev. 16:806–819. 17. Reynolds, A. E., A. W. Murray, and J. W. Szostak. 1987. Roles of the 2 micron gene products in stable maintenance of the 2 micron plasmid of Saccharomyces cerevisiae. Mol. Cell. Biol. 7:3566–3573. 18. Rose, M. D., F. Winston, and P. Hieter. 1990. Methods in yeast genetics. CSHL Press, New York, N.Y. 19. Saha, A., J. Wittmeyer, and B. R. Cairns. 2002. Chromatin remodeling by RSC involves ATP-dependent DNA translocation. Genes Dev. 16:2120– 2134. 20. Sambrook, J., and D. W. Russell. 2001. Molecular cloning: a laboratory manual, 3rd ed. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y. 21. Scott-Drew, S., and J. A. Murray. 1998. Localisation and interaction of the protein components of the yeast 2 micron circle plasmid partitioning system suggest a mechanism for plasmid inheritance. J. Cell Sci. 111:1779–1789. 22. Sengupta, A., K. Blomqvist, A. J. Pickett, Y. Zhang, J. S. Chew, and M. J. Dobson. 2001. Functional domains of yeast plasmid-encoded Rep proteins. J. Bacteriol. 183:2306–2315. 23. Som, T., K. A. Armstrong, F. C. Volkert, and J. R. Broach. 1988. Autoregulation of 2 micron circle gene expression provides a model for maintenance of stable plasmid copy levels. Cell 52:27–37. 24. Tanaka, T., M. P. Cosma, K. Wirth, and K. Nasmyth. 1999. Identification of cohesin association sites at centromeres and along chromosome arms. Cell 98:847–858. 25. Uhlmann, F., F. Lottspeich, and K. Nasmyth. 1999. Sister-chromatid separation at anaphase onset is promoted by cleavage of the cohesin subunit Scc1. Nature 400:37–42. 26. Uhlmann, F., D. Wernic, M. A. Poupart, E. V. Koonin, and K. Nasmyth. 2000. Cleavage of cohesin by the CD clan protease separin triggers anaphase in yeast. Cell 103:375–386. 27. Utatsu, I., S. Sakamoto, T. Imura, and A. Toh-e. 1987. Yeast plasmids resembling 2 micron DNA: regional similarities and diversities at the molecular level. J. Bacteriol. 169:5537–5545. 28. Velmurugan, S., Y. T. Ahn, X. M. Yang, X. L. Wu, and M. Jayaram. 1998. The 2 micrometer plasmid stability system: analyses of the interactions among plasmid- and host-encoded components. Mol. Cell. Biol. 18:7466– 7477. 29. Velmurugan, S., S. Mehta, and M. Jayaram. 2003. Selfishness in moderation: evolutionary success of the yeast plasmid. Curr. Topics Dev. Biol. 56:1–24. 30. Velmurugan, S., X. M. Yang, C. S. Chan, M. Dobson, and M. Jayaram. 2000. Partitioning of the 2-micron circle plasmid of Saccharomyces cerevisiae. Functional coordination with chromosome segregation and plasmid-encoded Rep protein distribution. J. Cell Biol. 149:553–566. 31. Volkert, F. C., and J. R. Broach. 1986. Site-specific recombination promotes plasmid amplification in yeast. Cell 46:541–550. 32. Wang, Z., I. B. Castano, A. De Las Penas, C. Adams, and M. F. Christman. 2000. Pol kappa: a DNA polymerase required for sister chromatid cohesion. Science 289:774–779. 33. Wang, Z., and M. F. Christman. 2001. Replication-related activities establish cohesion between sister chromatids. Cell Biochem. Biophys. 35:289–301. 34. Wong, M. C., S. R. Scott-Drew, M. J. Hayes, P. J. Howard, and J. A. Murray. 2002. RSC2, encoding a component of the RSC nucleosome remodeling complex, is essential for 2 micron plasmid maintenance in Saccharomyces cerevisiae. Mol. Cell. Biol. 22:4218–4229. 35. Zakian, V. A., B. J. Brewer, and W. L. Fangman. 1979. Replication of each copy of the yeast 2 micron DNA plasmid occurs during the S phase. Cell 17:923–934.