|Home | About | Journals | Submit | Contact Us | Français|
The 2μm circle plasmid in Saccharomyces cerevisiae is a model for a stable, high-copy-number, extrachromosomal “selfish” DNA element. By combining a partitioning system and an amplification system, the plasmid ensures its stable propagation and copy number maintenance, even though it does not provide any selective advantage to its host. Recent evidence suggests that the partitioning system couples plasmid segregation to chromosome segregation. We now demonstrate an unexpected and unconventional role for the mitotic spindle in the plasmid-partitioning pathway. The spindle specifies the nuclear address of the 2μm circle and promotes recruitment of the cohesin complex to the plasmid-partitioning locus STB. Only the nuclear microtubules, and not the cytoplasmic ones, are required for loading cohesin at STB. In cells recovering from nocodazole-induced spindle depolymerization and G2/M arrest, cohesin-STB association can be established coincident with spindle restoration. This postreplication recruitment of cohesin is not functional in equipartitioning. However, normally acquired cohesin can be inactivated after replication without causing plasmid missegregation. In the mtw1-1 mutant yeast strain, the plasmid cosegregates with the spindle and the spindle-associated chromosomes; by contrast, a substantial number of the chromosomes are not associated with the spindle. These results are consistent with a model in which the spindle promotes plasmid segregation in a chromosome-linked fashion.
The 2μm plasmid is an extrachromosomal selfish DNA element in the Saccharomyces cerevisiae nucleus that harbors a partitioning system and a copy number control system for its stable high-copy-number propagation (2, 11, 21). Despite a copy number of approximately 60 per cell, the plasmid molecules exist as a close-knit cluster and segregate as one unit (18, 22). This effective reduction in copy number provides the rationale for the existence of an active segregation mechanism. The partitioning system has a beguiling simplicity. It consists of just two plasmid-encoded proteins, Rep1p and Rep2p, and a cis-acting locus, STB, which includes roughly six copies of a tandem array of a 65-bp consensus sequence. Yet the plasmid manifests a degree of stability comparable to that of the yeast chromosomes, the loss rate being as low as 10−4 to 10−5 per generation.
The remarkable efficiency of the partitioning system appears to stem from its ability to feed into the segregation mechanism established for the host chromosomes (14, 20, 22). The kinetics of chromosome segregation and 2μm circle segregation closely parallel each other during the yeast cell cycle (22). In several mutant yeast strains that missegregate chromosomes at the nonpermissive temperature (ipl1-2 and ndc10-2, for example), the plasmid shows a strong tendency to missegregate in tandem with the bulk of the chromosomes (14). The yeast cohesin complex, critical for the segregation of each pair of sister chromatids to opposite cell compartments, is recruited to the STB locus (8, 14) in a Rep1p- and Rep2p-dependent manner (14). Mutations in Rep1p that interrupt its interaction with Rep2p or STB or both prevent cohesin-STB association and lead to high plasmid loss rates (26). Similarly, altered chromatin structure at STB due to the rsc2Δ mutation blocks Rep1p (but not Rep2p) binding to STB, disrupts cohesin assembly at this locus, and causes plasmid missegregation (25, 26). The timing of cohesin association and the lifetime of the cohesin-associated state during the cell cycle are virtually the same for the plasmid and the chromosomes (1, 14, 19). When the dissociation of the cohesin complex during anaphase is blocked, the duplicated plasmid clusters fail to separate, as is the case for sister chromatids (14). Taken together, these observations are consistent with a general model in which cohesin association serves to hold together duplicated plasmid clusters, and cohesin dissociation triggers their movement to daughter cells.
In this report, we demonstrate a role for the nuclear microtubules in 2μm plasmid segregation that is distinct from their role in chromosome segregation. Upon disassembly of the mitotic spindle by nocodazole treatment or by inactivation of nuclear microtubules by a conditional mutation in TUB2, cohesin assembly at the STB locus is selectively abolished with no ill effect on cohesin acquisition by the chromosomes. The block in cohesin-STB association caused by spindle depolymerization is not due to cell cycle delay or arrest in G2/M per se and is not dependent on the activation of the spindle checkpoint. Consistent with previous observations, disruption of cohesin-STB association, caused in this case by the lack of spindle integrity, results in plasmid missegregation. These findings reveal a hitherto-unsuspected aspect of the molecular selfishness of the 2μm plasmid: not only does it poach an essential molecular component of chromosome segregation (namely, cohesin), but it carries out the act by enlisting the assistance of the principal infrastructure of the host mitotic apparatus. Evidence from a “spindle recovery” experiment and a programmed cohesin inactivation assay suggests that the timing of cohesin assembly at STB, likely concomitant with DNA replication, is critical for equal plasmid segregation. When a subset of the chromosomes are detached from the spindle by the mtw1-1 mutation, the plasmid still reveals a striking proclivity to localize to the same cell compartment as does the spindle, the bud being strongly preferred over the mother. Collectively, these observations reveal a hitherto-unsuspected role for the mitotic spindle in DNA segregation in yeast. In conjunction with earlier findings, they also suggest that the spindle-supported plasmid segregation is coupled to chromosome segregation.
The yeast strains and plasmids employed in this study are listed in Table Table1.1. The strains used for chromatin immunoprecipitation (ChIP) assays contained an integrated copy of the MCD1 gene tagged with the hemagglutinin (HA) epitope (14). Their relevant genetic configurations are described in the text and indicated in the figures. The reporter plasmids containing Lac operator arrays and bound by the green fluorescent protein-Lac repressor or yellow fluorescent protein (YFP)-Lac repressor were visualized by fluorescence microscopy (14, 22). For some of the assays, the endogenous 2μm circles served as the reporter.
The method for localizing plasmids within the nucleus was adapted from a previously published study (7). Assays were carried out with a yeast strain housing a reporter plasmid tagged by the Lac operator-YFP-Lac repressor. The nuclear membrane was labeled using a fusion between the cyan fluorescent protein (CFP) and the nuclear pore protein Nic96p. To localize a reporter plasmid with respect to the spindle pole, the latter was labeled with a fusion between CFP and the Spc29 protein. The plasmid residence zone was determined by optical sectioning of the yeast nucleus along the vertical axis (Z-series sectioning). For each sample, 21 consecutive sections of 0.25-μm thickness each were examined, spanning 5 μm of the total thickness. Those sections simultaneously displaying fluorescence from the nuclear membrane (or the spindle pole) and that from the reporter plasmid were stacked, deconvolved, and projected as two-dimensional images. The plasmid signals were localized relative to the nuclear periphery by the use of concentric circles that define four zones of equal area. For zonal demarcations, the diameter of each nucleus was estimated as follows. The circumference of the nuclear membrane was traced along the midpoints between the outer and inner edges of the cyan fluorescence from the tagged Nic96p. A set of farthest end-to-end distances on this membrane contour were measured and averaged. The thickness of the membrane was exaggerated in the two-dimensional projections because multiple nuclear cross sections were layered over each other. Similar protocols of Z-series sectioning and deconvolution were employed in the mapping of plasmids with respect to the neighboring spindle pole. For a group of plasmid foci within a nucleus, the distance of each focus from the proximal spindle pole or the nearest point on the membrane was measured, summed, and averaged. Metamorph software (Universal Imaging Corporation, Downingtown, PA) was used to obtain the spherical reconstruction of the nucleus and to shrink the sphere to one-fourth, one-half, and three-fourths its volume.
DNA samples were digested with HinDIII, fractionated by electrophoresis in agarose, and blotted on to nylon membranes. They were hybridized to a radioactive probe prepared by randomly primed replication of a DNA fragment (obtained by PCR) containing the 2μm circle REP1 region and the yeast HIS3 marker. The ratio of the intensities of the REP1 and the HIS3 bands in the autoradiogram was estimated using Bio-Rad Image Analysis software (Molecular Imager).
Previous results have shown that the Rep1 and Rep2 proteins within the yeast nucleus tend to localize more strongly at or close to the spindle pole (22). Furthermore, depolymerization of microtubules using nocodazole causes the Rep proteins to become rather dispersed in the nucleus. At the same time, there is an increase in the width of the plasmid residence zone measured by confocal Z-series sectioning, indicating a loss of compactness of the plasmid clusters.
To precisely map reporter plasmids tagged with YFP fluorescence in the nucleus, we used a procedure adapted from Heun et al. (7). The nuclear envelope or the spindle pole, visualized by CFP fusions to Nic96p or Spc29p, respectively, provided the reference landmarks in this assay. Two-dimensional projections of deconvolved Z-series fluorescence images from 100 individual cells were used to allocate plasmids within the nucleus. The host strain was [cir+] and hence supplied an STB-containing reporter plasmid with the trans-acting components, Rep1p and Rep2p, of the partitioning system. As shown in Fig. 1A, a plasmid (shown in green) harboring STB was present as a tight cluster in close proximity to the spindle pole (shown in red). When the reporter plasmid lacked STB, the cluster was loosely organized, with a nearly twofold increase in the spacing between the spindle pole and plasmid foci. A similar increase was observed for the STB reporter in a [cir0] host, presumably due to lack of the Rep proteins (data not shown). Nocodazole treatment resulted in a roughly 2.5-fold increase in the distance between the spindle pole and the STB reporter, comparable to that caused by the lack of the Rep-STB system. By contrast, the effect of nocodazole on the reporter lacking STB was modest. The results were essentially identical for cells arrested in G1 with α-factor (shown here) and those in late G2/M in which replicated plasmid clusters had segregated into opposite cell compartments (data not shown).
In a corollary assay, we mapped plasmid positions to four arbitrarily defined zones of equal area comprising the nuclear cross section. According to this assignment, plasmid occupancy was confined primarily to zones 2 and 3, was independent of the partitioning system (presence or absence of STB on the plasmid), and was not affected by nocodazole treatment (see Fig. S1A in the supplemental material). Similar nonrandom distributions of plasmids were verified by reconstructing the nucleus in three dimensions and dividing it into four spherical shells of equal volume (see Fig. S1B in the supplemental material).
The above observations indicate that the global nuclear localization of plasmids to defined nuclear zones, with the center of the nucleus as the point of reference, is not dependent on the integrity of the mitotic spindle or that of the Rep-STB partitioning system. However, the picture changes when the spindle pole is the point of reference. Both the spindle and the partitioning system are essential for the tight clustering of plasmid foci, as well as their specific address proximal to the spindle pole. Perhaps nocodazole breaks the microtubule tether that holds the 2μm plasmid at this locale.
The remarkably similar effects of nocodazole treatment or the lack of the partitioning system on the specific nuclear address of the 2μm plasmid suggest a potential role for the mitotic spindle in plasmid segregation. Therefore, we tested whether nocodazole would interfere with plasmid-cohesin association, a key step in the partitioning process, as indicated by several lines of evidence (14, 26). The results are shown in Fig. Fig.1B1B.
The [cir+] host strain, in which ChIP assays were performed, contained an HA-tagged version of the MCD1 gene in its native chromosomal context. After 3 h of treatment with nocodazole, 80% or more of the cells were arrested in G2/M with 2C DNA content, as indicated by light microscopy and fluorescence-activated cell sorter (FACS) analysis (data not shown). Association of Mcd1p (the reporter for cohesin) with the STB DNA could not be detected in the drug treated cells; by contrast, its association with an arm binding site on chromosome V (19) was normal (Fig. (Fig.1B,1B, lane 2). Cohesin binding at centromeres was also unaffected by nocodazole (data not shown). As expected from previous work (14), the chromosomal and STB loci in the untreated cells were associated with Mcd1p (or the cohesin complex, by inference) (Fig. (Fig.1B,1B, lane 4).
We showed previously that the Rep1 and Rep2 proteins are essential for cohesin recruitment to STB (14). The observed nocodazole effect could result from either one or both of the Rep proteins being dislodged from STB. However, ChIP assays using antibodies to Rep1p or Rep2p showed that nocodazole did not affect Rep protein association with STB (Fig. (Fig.1B,1B, lanes 7 and 8). Consistent with prior results, the chromosomal cohesin binding site was not occupied by the Rep proteins (Fig. (Fig.1B,1B, lanes 7 and 8).
Is the disruption of cohesin-STB association upon nocodazole treatment caused by spindle depolymerization or rather by cell cycle delay or arrest in G2/M? To resolve this issue, ChIP analysis was repeated with the cdc20-1 yeast strain which, at the nonpermissive temperature, arrests at the metaphase-anaphase transition prior to cohesin disassembly (10). In the absence of nocodazole treatment, Mcd1p was detected at STB at the permissive (26°C) and at the nonpermissive (37°C) temperatures (Fig. (Fig.1C,1C, lanes 2 and 8). Nocodazole treatment eliminated this association at both temperatures (Fig. (Fig.1C,1C, lanes 5 and 11). In addition, ChIP results from a [cir+] mad2Δ strain revealed that triggering of the spindle checkpoint is not a prerequisite for the blockade of STB-cohesin association by nocodazole (Fig. (Fig.1D,1D, compare lanes 11 and 5).
The data from the nocodazole-ChIP assays suggest a role, either direct or indirect, for microtubules in the association of the cohesin complex with the 2μm plasmid. Strictly, our results can also be accommodated by the direct action of nocodazole on the cohesin-Rep-STB complex. However, the chances of a spindle independent action of nocodazole have been ruled out by further experiments (see below).
To verify that the nocodazole effects observed in the previous set of experiments (Fig. 1B to D) are mediated through the depolymerization of microtubules, the ChIP analysis was repeated using cold-sensitive tubulin mutants that affect cytoplasmic microtubules (tub2-104) or nuclear microtubules (tub2-402) or both (tub2-401) (9).
In the wild-type strain and the tub2-104 mutant strain (affected only in cytoplasmic microtubules), the association of Mcd1p with STB was normal at 30°C and 11°C, the permissive and nonpermissive temperatures, respectively, for the mutant (Fig. (Fig.2A,2A, lanes 2 and 5 of rows 1 and 2). In the tub2-401 (affected in nuclear and cytoplasmic microtubules) and tub2-402 (affected only in nuclear microtubules) strains, Mcd1p was present on STB at 30°C but absent at 11°C (Fig. (Fig.2A,2A, compare lanes 2 to lanes 5 of rows 3 and 4). At the same time, occupancy of the chromosomal locus by Mcd1p was not sensitive to any of the three mutations at either temperature (Fig. (Fig.2A,2A, lanes 8 and 11 of rows 2 to 4).
We have verified that the mutant strains arrest at 11°C as large budded cells with a 2C DNA content and exhibit the expected spindle phenotypes (Fig. (Fig.2B)2B) (9). Consistent with the lack of cytoplasmic microtubules that are required for nuclear elongation and spindle extension, nearly all of the tub2-104 cells showed brightly stained short spindles (Fig. (Fig.2B,2B, row 2). In roughly 70% of the tub2-402 cells, the tubulin fluorescence was localized at the periphery of the 4′,6′-diamidino-2-phenylindole (DAPI) staining region and towards the cytoplasm, with no indication of nuclear microtubules (Fig. (Fig.2B,2B, row 3). The remaining cells contained only fluorescent dots, likely specifying spindle poles, and no spindle structure was discernible (data not shown). In the tub2-401 mutant, >90% of the cells showed little or no sign of organized fluorescence (Fig. (Fig.2B,2B, row 4).
The analyses with the tubulin mutants demonstrate that the association of cohesin with the STB locus requires the integrity of only nuclear microtubules and is unaffected by the lack of cytoplasmic microtubules.
Our earlier work indicated that certain mutations, ipl1-2 and ndc10-2, for example, that affect specific steps of the chromosome segregation pathway also cause missegregation of the 2μm plasmid in a chromosome-coupled manner (14, 22). That is, the plasmid almost always stays in tandem with the bulk of the chromosomes. Since the plasmid and chromosomes respond differently to the lack of nuclear microtubules in cohesin loading, is it possible to uncouple the two in their segregation patterns in the tub2-402 strain at 11°C? Note, though, that chromosome missegregation is inevitable in the mutant at this temperature because of the absence of a functional spindle.
Normal segregation of the plasmid and of chromosomes (rough equality between the two cell compartments in the distribution of plasmid-associated fluorescence, as well as DAPI staining) seen in nearly all of the cells at 30°C was virtually absent at 11°C (Fig. (Fig.2C,2C, row 1). Instead, three classes of aberrant plasmid segregation (Fig. (Fig.2C,2C, rows 2 to 4) were observed. In the majority of cases, the entire set of plasmid foci was coincident with the chromosome mass indicated by the DAPI stain (Fig. (Fig.2C,2C, row 2) or clearly separated from it (Fig. (Fig.2C,2C, row 3). In the remainder, plasmid foci were split into two unequal groups, at least one of which was localized outside the DAPI staining area (Fig. (Fig.2C,2C, row 4). Caution is warranted, however, in interpreting the plasmid missegregation patterns because of the nuclear phenotypes of the tub2-402 mutant at 11°C (Fig. (Fig.2D).2D). As revealed by CFP-tagged Nic96p, the nucleus was contained entirely within one cell compartment in approximately 60% of the cells (Fig. (Fig.2D,2D, row 2). Interestingly, the primary missegregation class comprised approximately the same fraction of cells, with plasmid and chromosomes confined to the same cell compartment (62%) (Fig. (Fig.2C,2C, row 2). Hence, this phenotype could be accounted for simply by nonelongated nuclei at 11°C and is not particularly informative. Cells in which the whole set of plasmid foci had distinctly separated from the chromosomes accounted for ~24% of the population (Fig. (Fig.2C,2C, row 3); those in which a subset of plasmid foci had moved away from the chromosomes constituted ~12% of the population (Fig. (Fig.2C,2C, row 4). The sum of these two classes (36%) correlated well with the fraction of cells in which nuclei had elongated enough to straddle the two cell compartments (38%) (Fig. (Fig.2D,2D, row 3).
Thus, a mutation that specifically disrupts nuclear spindle assembly and abolishes cohesin association with the plasmid but not with chromosomes is able to uncouple, at least to a large extent, plasmid from chromosomes in their nuclear distribution patterns.
The Rep1 and Rep2 proteins and the associated 2μm circle molecules can be localized to a subregion of the DAPI staining zone in preparations of yeast chromosome spreads (14). A reporter plasmid that lacks the STB sequence was not detected in the spreads; nor is an STB-containing plasmid if Rep1p or Rep2p or both are missing in the host cell. Since nocodazole is known to cause dispersal of the Rep proteins in the yeast nucleus, expand the plasmid residence zone, and affect the plasmid address with respect to the spindle pole (Fig. (Fig.1A)1A) (22), the following questions become relevant. Does nocodazole affect the localization of the plasmid in chromosome spreads? If so, can the localization be reestablished when cells are allowed to recover in drug-free medium? Does the recovery lead to reassociation of cohesin with STB? Finally, and most significantly, how does the plasmid segregate following recovery? The answers are provided in the results shown in Fig. Fig.33 to to5.5. In these assays, cells blocked in G1 with α-factor were released into nocodazole-containing medium without the pheromone for 3 h to impose G2/M arrest and subsequently allowed to recover in nocodazole-free medium.
In contrast to chromosome spreads prepared from G1 cells (prior to nocodazole treatment), those from nocodazole-arrested cells (0 min) failed to display the Rep1 protein or the STB-containing reporter plasmid in roughly 80% of the cases (Fig. (Fig.3A,3A, left two panels). However, at 45 min into the recovery period, the spreads showed the presence of Rep1p as well as the plasmid in >85% of the cells (Fig. (Fig.3A,3A, right). At this time point, nearly all of the cells contained elongated spindles (Fig. (Fig.3C).3C). The pattern for Rep2p was the same as that for Rep1p (data not shown).
ChIP assays revealed that the Mcd1 protein, and hence cohesin by implication, was absent at STB after nocodazole treatment (0 min); it reappeared at STB at approximately 30 min from the start of recovery, was most prominent at 45 min, and was nearly gone by 60 min (Fig. (Fig.3B,3B, middle). By contrast, Mcd1p was present at the chromosome V binding site from 0 to 45 min, before disappearing almost completely at 60 min (Fig. (Fig.3B,3B, bottom). The marked reduction in Mcd1p at STB and the chromosomal site over the 45- to 60-min interval is consistent with its cleavage and consequent cohesin disassembly during anaphase. As shown in Fig. Fig.3C,3C, there was a strong correlation between the kinetics of spindle restoration and the reestablishment of STB-cohesin association during recovery from nocodazole treatment.
It should be pointed out that nocodazole treatment also resulted in the exclusion of the 2μm plasmid from chromosome spreads in G1-arrested cells (Fig. (Fig.4A).4A). When the drug was washed off without relieving G1 arrest, the plasmid regained its association with the spreads and its spindle pole proximal localization (Fig. (Fig.4B).4B). Upon subsequent release of the cells from pheromone arrest, the 2μm plasmid showed a normal segregation pattern in the ensuing cell cycle (Fig. (Fig.4C).4C). Furthermore, during nocodazole recovery of a strain harboring smc3-42(Ts), the plasmid cluster was detected at 45 min in chromosome spreads from cells maintained at the permissive temperature (26°C) or shifted to 37°C at the start of recovery (Fig. (Fig.4D4D).
The fact that plasmid localization in chromosome spreads and that the appearance of cohesin at STB in the recovering cells is coincident with or closely follows spindle reformation suggests possible spatial and/or functional regulation of plasmid-cohesin association by the spindle. Furthermore, the cohesin complex itself is not required for the specific nuclear location of the plasmid or its association with chromosomes, as indicated by the spreads.
If the chromosomal paradigm for the role of cohesin in segregation applies to the 2μm plasmid, cohesin loaded on the plasmid in a replication-independent manner is likely to be nonfunctional. We tested whether cohesin acquired subsequent to replication during recovery from nocodazole can mediate plasmid partitioning (Fig. (Fig.5A).5A). The plasmid segregation patterns observed in this experiment could be divided into four types: equal segregation (Fig. (Fig.5A,5A, row 1); unequal segregation (row 2), and complete missegregation (rows 3 and 4). Row 4, though, comprises a special class in which the chromosomes (stained with DAPI) failed to segregate (see below).
Equal segregation of the plasmid dropped from approximately 95% in large budded cells of the control group (not treated with nocodazole) to only about 30% in cells that had recovered from nocodazole but had not yet completed cell division (Fig. (Fig.5A,5A, row 1). Correspondingly, there was an increase in the class of cells missegregating the plasmid, with unequal segregation (approximately 45%) (Fig. (Fig.5A,5A, row 2) much more prevalent than nonsegregation (approximately 9%) (Fig. (Fig.5A,5A, row 3). In earlier experiments, it was noted that plasmids lacking a functional Rep-STB system showed an equal segregation frequency between 20 and 40% when cells were scored at 75 min after release from G1 arrest (22). The values are comparable to the 30% seen here after recovery from nocodazole, with the distinction that the partitioning machinery (Rep1p-Rep2p-STB) was intact in the present instance (Fig. (Fig.5A,5A, row 1). In other words, the absence of the spindle during the cell cycle window between G1 and G2/M or the lack of the Rep-STB system had the same functional consequence with respect to plasmid segregation. These results would be consistent with the concerted or sequential action of the mitotic spindle and the Rep/STB system in facilitating one or more steps, including cohesin recruitment, of the plasmid segregation pathway.
In three of the four cell types displayed in Fig. Fig.5A5A (rows 1 to 3), chromosomes segregated normally after recovery from nocodazole, as deduced from the equivalence of DAPI staining in the two cell compartments. Since nocodazole has no effect on the replication coupled recruitment of cohesin at the chromosomal locales, paired sister chromatids are expected to complete the subsequent steps in partitioning, once the spindle has been put in place. The fraction of cells in which both plasmid and chromosomes were restricted to the same cell compartment even after transfer to nocodazole-free medium (approximately 16%) (Fig. (Fig.5A,5A, row 4) likely indicates the preanaphase state of these cells or a failure on their part to resume the cell cycle.
There is no reason, a priori, to suspect that nocodazole has any adverse effect on 2μm circle replication. However, there is no experimental evidence that rules out this possibility. Missegregation caused by nocodazole can indeed be explained if the plasmid fails to replicate or grossly underreplicates in the absence of the spindle. The relative copy numbers of the 2μm circle in cells arrested in G1 (1C chromosome content) and those arrested in G2/M with nocodazole (2C chromosome content) were therefore estimated by Southern hybridization (Fig. (Fig.5B).5B). The intensity of the plasmid band was divided by that of the corresponding chromosome band for each of the lanes 1 to 4. The ratio of these values for lanes 3 and 1 was 0.93; that for lanes 2 and 4 was 1.32 (or close to 1). Thus, the molar ratio of plasmid to chromosome remained unchanged between the G1 and G2/M cells. Since chromosome duplication proceeds normally in nocodazole treated cells, so must plasmid replication.
In summary, despite the reemergence of the 2μm plasmid in chromosome spreads and the reassociation of cohesin with STB in response to spindle restoration in cells rescued from nocodazole arrest, the plasmid partitioning process is irreversibly damaged for the remainder of the cell cycle.
Our previous work indicated that the timely cleavage of Mcd1p is as essential for the separation of the duplicated plasmid clusters as it is for the separation of sister chromatids (22). The present work suggests that the late acquisition of cohesin by the plasmid in a postreplication fashion, as in the nocodazole recovery experiment, is ineffective in mediating equal partitioning (Fig. (Fig.33 and and4).4). In a separate experiment, it was noted that inactivating cohesin prior to DNA replication with the help of a temperature-sensitive (Ts) mutation in Smc1p leads to the missegregation of chromosomes and the plasmid (26). All of the above observations are consistent with a critical timing in the cell cycle when cohesin acquisition by the plasmid is functional.
We have now followed plasmid segregation in a yeast strain carrying a single Ts allele of the MCD1 gene by releasing G1 cells from α-factor arrest and inactivating cohesin at different times during progression of the cell cycle (Fig. (Fig.5A).5A). The 0-min or 60-min temperature-shifted samples revealed high levels of plasmid and chromosome missegregation (unequal numbers of plasmid foci or unequal amounts of DAPI in the two cell compartments) (Fig. (Fig.6A,6A, bottom). The frequency of equal segregation improved markedly when the temperature shift was delayed until 120 min, tending towards that observed in the subpopulation of large budded cells maintained at 23°C (Fig. (Fig.6A,6A, top). The FACS data showed that chromosome doubling was completed in these cells at 120 min. Thus, blocking cohesin recruitment to STB entirely (0-min shift to 37°C) or inactivating cohesin in the midst of the S phase (60-min shift to 37°C) led to plasmid (and chromosome) missegregation. By contrast, the temperature shift imposed at the end of S phase (120 min) did not.
In the above experiment, there is a caveat that the Mcd1p subunit of cohesin could have potentially been cleaved at 120 min, so that the temperature shift was superfluous for cohesin disassembly. A variation of this assay was performed on a yeast strain containing a Ts allele of SMC3 (smc3-42), the native MCD1 under its own promoter, plus the MCD1-nc allele (encoding noncleavable Mcd1p) under the inducible GAL promoter (Fig. (Fig.6B).6B). The strain also harbored a fluorescence-tagged 2μm plasmid in one case and a fluorescence-tagged chromosome in the other. The cohesin complex assembled during growth in galactose, while being thermally unstable, would be immune to anaphase cleavage of Mcd1p. G1-arrested cells at 26°C were induced with galactose for 60 min, released into pheromone-free galactose medium, and subjected to temperature shift as in the experiment shown in Fig. Fig.6A.6A. FACS analysis of culture aliquots indicated that >90% of the cells completed S phase at 150 min. Cells shifted at this time point, and the control cells maintained at 26°C were examined for plasmid-chromosome segregation at 210 min.
The predominant fraction of cells maintained at 26°C revealed the fluorescence tagged reporter chromosome as a single dot (paired sister chromatids; 87%) and the reporter plasmid as a single cluster (unseparated sister clusters, presumably; 75%) (Fig. (Fig.6B).6B). The remaining cells contained two well-segregated chromosomal dots (13%) or plasmid clusters (25%). This result agrees with the previous observation that plasmids cannot be partitioned evenly into daughter cells in the absence of Mcd1p cleavage (22). In cells shifted to 37°C, normal segregation increased to 82% for the chromosome and 83% for the plasmid. Thus, once DNA replication has been completed (and presumably the duplicated plasmid clusters have been paired by cohesin), thermal inactivation of cohesin can yield equal segregation.
The Mtw1 complex, a component of the kinetochore, is required to establish the biorientation of sister chromatids that is monitored by the Ipl1/Aurora protein kinase (6, 17). In the mtw1-1 mutant yeast strain, the action of the Ipl1 kinase causes kinetochores to dissociate from the spindle at the nonpermissive temperature. Since the spindle checkpoint is activated in this mutant, cells are arrested in G2/M, and inhibition of the anaphase-promoting complex causes sister chromatids to remain paired by cohesin. Despite checkpoint activation, the chromosomes do segregate into two separate masses in a significant fraction of the arrested cells (6, 17). Strikingly, the short spindle, typical of the vast majority of the mtw1-1 cells, shows a strong tendency to preferentially migrate into the bud (17). These characteristic segregation phenotypes of the chromosomes and spindle in the mtw1-1 mutant prompted us to ask how the 2μm plasmid would segregate in this background. Would the plasmid follow the spindle-free chromosomes or the spindle or neither?
The chromosome segregation patterns in the mtw1-1 cell population arrested at 37°C, as revealed by DAPI, are depicted in Fig. Fig.7A7A (classes I through V). Roughly half the population consisted of cells with two well-separated chromosome masses, as indicated by DAPI staining (panel A, II to IV). The DAPI zones were roughly equal in class II, but unequal in classes III and IV. Note, however, that the DAPI equivalence in class II does not indicate normal chromosome segregation, since sister chromatids most often tend to cosegregate in the mtw1-1 mutant due to lack of cohesin cleavage. In the other two classes, the chromosomes existed as a single unresolved entity, either confined to one cell compartment (class I) or stretched along the bud neck (class V). In almost all the class I and class V cells, the chromosomes, the reporter plasmid, and the spindle were not separated from each other. Since we wished to know whether the 2μm plasmid preferentially associated with the spindle or with the chromosomes in the mtw1-1 background, these cells were not included in the analysis presented in Fig. 7B and C.
The class II to IV cells could be divided into five subtypes (Fig. (Fig.7B,7B, subtypes a through e) with respect to the segregation of the reporter plasmid they harbored: equal segregation (subtype a), complete missegregation (b and c), and unequal segregation (d and e). Equal segregation of the 2μm plasmid (harboring STB), which is the norm in wild-type cells, occurred only in approximately 6% of the mutant cells. Unequal segregation was observed in roughly 12% (subtypes d and e). In the largest fraction of cells (82%; subtypes b and c), the plasmid cluster was present only in one cell compartment. Of these, the bud residents (67%; subtype b) outnumbered the mother residents (17%; subtype c) nearly four to one. The strongly preferred localization of the 2μm plasmid in the bud correlates well with a similar bias noted by Pinsky et al. (17) for the migration of the short spindle in mtw1-1. By contrast to the 2μm plasmid, equal, unequal, and all-or-none segregation events for the ARS reporter plasmid (lacking STB) were comparable to those observed previously with wild-type yeast strains.
To verify the suspected cosegregation of the 2μm plasmid and the spindle in the mtw1-1 mutant, the two were simultaneously localized by immunofluorescence (Fig. (Fig.7C):7C): the plasmid using antibodies to the LacO bound repressor and the spindle using antibodies to tubulin (Tub1p). In cells containing two equally segregated plasmid clusters (Fig. (Fig.7C7C--1),1), the spindle was almost always present in both cell compartments. When there was a single plasmid cluster confined to the bud compartment (C-2) or the mother compartment (C-3), the spindle almost invariably colocalized to the same compartment. In the case of unequal plasmid segregation (C-3 and C-4), the spindle most often stayed in tandem with the larger of the two clusters.
As a control, we followed fluorescence-tagged chromosome IV in a similar assay (see Fig. S2 in the supplemental material). Unlike the 2μm plasmid, chromosome IV showed a bias, albeit small, in the opposite direction. It was more often located away from the spindle (~60%) than associated with it (~40%) (see Fig. S2 in the supplemental material) (17). Unlike chromosome IV, chromosomes III and VII showed a strong tendency to remain associated with the short spindle in the mtw1-1 mutant (17). We also repeated this assay with the mtw1-1 mad2Δ double mutant, which is not blocked in metaphase and is competent for cohesin cleavage (17). The spindle phenotype in the large budded cells of the double mutant was the same as that in the metaphase-arrested single mutant. In addition, the 2μm plasmid remained tightly associated with the short spindle.
The plasmid, chromosome, and spindle patterns in the mtw1-1 mutant demonstrate that, during segregation, the 2μm plasmid maintains its coupling with the spindle and/or the subset of spindle-associated chromosomes.
The behavior of the 2μm plasmid in the mtw1-1 mutant is consistent with two models for plasmid segregation. In one, plasmid segregation is spindle dependent but chromosome independent. In the other, segregation is also dependent on a spindle-associated chromosome. If the plasmid is tethered to a chromosome, it can gain access to the spindle indirectly via kinetochore attachment to the spindle. This step may reinforce plasmid-chromosome association through additional spindle-mediated interactions. Furthermore, it can set the stage for critical events in plasmid segregation facilitated by the spindle, for example, cohesin recruitment. We therefore tested whether plasmid localization in chromosome spreads and cohesin association with STB is affected when chromosome-spindle association is blocked by the kinetochore mutation ndc10-2.
As shown in Fig. Fig.8A,8A, the plasmid cluster was detected in chromosome spreads at the permissive and nonpermissive temperatures in the ndc10-2 mutant. It disappeared from spreads upon nocodazole treatment and reappeared following nocodazole removal at both temperatures. Similarly, in ChIP analysis, cohesin association with STB was not affected by the mutation, and nocodazole blocked this association at 26°C and 37°C (Fig. (Fig.8B).8B). Furthermore, in a nocodazole recovery assay using a synchronized cell population (analogous to that shown in Fig. Fig.3),3), the mutation did not alter the temporal patterns of plasmid localization, spindle restoration and cohesin-STB association (data not shown).
Thus, plasmid localization as well as cohesin recruitment to the STB locus can occur normally in the presence of an intact spindle even when the chromosomes are blocked from spindle attachment.
The most significant and least-suspected outcome from the present study concerns the role of the mitotic spindle in the segregation of the 2μm plasmid. The integrity of the spindle is essential for the specific nuclear localization of the plasmid, its association with chromosome spreads, and the recruitment of the cohesin complex to its partitioning locus. Our current results, along with previous work (14, 26), suggest that the assembled cohesin complex plays similar functional roles in the partitioning of the plasmid and the yeast chromosomes. Yet, the plasmid utilizes a completely different mechanism from that of the chromosomes, involving the plasmid partitioning system and the host mitotic spindle to acquire the cohesin complex.
The displacement of the plasmid cluster from its normal nuclear locale or from chromsome spreads by nocodazole treatment, along with the accompanying loss of STB-cohesin association, suggests a spatial-structural contribution of the spindle to plasmid partitioning. At this time, the nature of the interaction between the plasmid and spindle is not understood. The interaction may be indirect, perhaps mediated through one or more spindle-associated proteins. A relevant question is whether, in addition to its role in channeling cohesin to STB, the spindle may also mediate the trafficking of plasmid clusters during segregation. In time-lapse assays, the 2μm plasmid and kinetochore proteins show almost identical localization and dynamics throughout the cell cycle (S. Velmurugan, unpublished data). The plasmid appears to be situated at the right place to potentially interact with a plus-end motor protein.
According to the prevalent models for cohesin assembly at chromosomal loci, an advancing replication fork pauses at a precohesin site that is already occupied by the Smc1 and Smc3 protein subunits of the complex (3, 23). The recruitment of the Mcd1 protein then requires the exchange of the resident DNA polymerase α within the replication complex for a novel polymerase σ (23). We do not know whether these steps hold true in the organization of cohesin at STB during a normal cell cycle. However, consistent with this mechanism, the Smc subunits of the complex are found associated with STB and not with other regions of the 2μm plasmid before the initiation of replication (26).
Earlier results (14, 26) and the present data demonstrate that lack of functional cohesin prior to or during DNA replication, as well as postreplication loading of cohesin on STB, leads to plasmid missegregation. In sum, the data suggest that functional cohesion is possible only in a replication-associated manner. The results agree with a cohesin-mediated binary mode of counting sister plasmid clusters, as is the established mechanism for sister chromatids.
The strong association of the 2μm plasmid with the spindle in the mtw1-1 mutant, while a substantial number of chromosomes are detached from it, places constraints on one of the currently entertained models for plasmid segregation in which replicated plasmid clusters hitchhike on sister chromatids (22). Not all chromosomes are uniformly affected by the mtw1-1 mutation. Whereas chromosome III is detached from the spindle quite infrequently, chromosome IV is detached in more than half the cell population (see Fig. S2 in the supplemental material) (17). In a revised hitchhiking model, the plasmid cluster can be tethered only to a member of the chromosome subset that remains stably attached to the spindle in the mtw-1 background.
As revealed by the ndc10-2 mutation, plasmid localization to chromosome spreads or cohesin association with STB can occur normally in the presence of a functional spindle even when chromosomes are not attached to the spindle. Yet, this mutation causes the 2μm plasmid to stay put with the bulk of the chromosomes in the mother, even though spindle elongation is normal (12, 14). This phenotype is not alleviated by combining the ndc10-2 mutation with mad2Δ to inactivate the spindle checkpoint and promote cohesin disassembly (Velmurugan, unpublished). Thus, spindle elongation and cohesin cleavage are not sufficient to promote plasmid segregation when chromosomes fail to do so. We have no reason to suspect that the ndc10-2 mutation itself may interfere with plasmid-spindle association. We failed to detect by ChIP assays any interaction between STB and representative member proteins of the inner, middle, and outer kinetochore complexes (S. Mehta, unpublished observations). Therefore, an alternative to the hitchhiking model that proposes spindle-dependent but chromosome-independent plasmid segregation (14) is also inadequate.
A tentative scheme for 2μm plasmid segregation that accommodates current as well as previous observations is summarized in Fig. Fig.8C.8C. In this model, the plasmid cluster interacts with the mitotic spindle as well as the chromosomes, although the nature of these interactions is as yet undefined. The three stages of plasmid segregation are (i) clustering, specific nuclear localization, and chromosome association of the plasmid; (ii) cohesin recruitment during replication and pairing of duplicated plasmid clusters; and (iii) cohesin dissociation and movement of the clusters away from each other.
The outcomes from this study underscore the importance of the spindle at all three stages of the plasmid segregation pathway. By contrast, sister chromatids that have been replicated and cohesed in the absence of the spindle can segregate normally when the spindle function is provided. Cohesin itself has no role during stage I, and stages I and II can be executed in the absence of chromosome-spindle attachment. Thus, plasmid interactions with a chromosome do not follow passively from the independent association of each entity with the spindle. Rather, the spindle appears to actively promote plasmid alliance with a chromosome, and the two then become coentities in the segregation process.
Two recent reports, based on a synthetic genetic array analysis using a ctf8 deletion strain and a microarray hybridization assay using a ctf4 deletion strain, have brought to light a previously unrecognized role for mitotic spindle integrity in sister chromatid cohesion (13, 24). Ctf8p is a component of the alternative replication factor C-like complex Ctf18-RFC, and Ctf4p is a constituent of the replication fork that binds DNA polymerase α. These factors contribute to sister chromatid cohesion in S. cerevisiae; however, the mechanisms are not understood. Among the synthetically lethal deletions uncovered using ctf8Δ were bim1Δ, kar3Δ, and vik1Δ (13), and each of the latter three deletions caused impaired cohesion at centromeric and arm sites. The kar3 deletion, but not bim1Δ or vik1Δ, was also identified in the screen with ctf4Δ (24).
KAR3 codes for a minus-end-directed microtubule motor protein (5) that is required for spindle integrity, karyogamy during mating (15), and proper mitotic spindle positioning (4). The data from Mayer et al. (13) suggest that the interaction of Kar3p with one of its accessory factors Vik1p is important in sister chromatid cohesion. An interaction between Kar3p and the cohesin subunit Smc1p had been reported earlier (16). Bim1p is a plus-end microtubule binding protein that is required for the correct orientation of the mitotic spindle and, as a result, for polarized cell growth. Its newly discovered role in cohesion is consistent with earlier results that link a lack of Bim1p to chromosome missegregation.
The above findings suggest that the spindle-cohesin connection is not exclusive to the 2μm plasmid, as we had originally believed. Nevertheless, the spindle effect on the plasmid is much more dramatic than that on the chromosomes. In a preliminary screen using the synthetic genetic array deletion strains, we noticed an elevated loss of a 2μm reporter plasmid in ctf8Δ; furthermore, ChIP assays have revealed the association of Kar3p with the STB locus (Mehta, unpublished). It is possible that the yeast plasmid has adapted and optimized one of the less prominent cellular mechanisms for establishing sister chromatid cohesion to serve a principal function in its own partitioning.
This work was supported by funds from the National Institutes of Health (GM64363). S.M. received partial support through the William Livingston Fellowship from the University of Texas at Austin.
We are grateful to David Botstein, Douglas Koshland, Sue Biggins, and Vincent Guacci for providing strains and reagents.
†Supplemental material for this article may be found at http://mcb.asm.org/.