Identification of a Mid-anaphase Checkpoint in Budding Yeast

Yeast rich medium, YPGal and YPD, contained 2% galactose and glucose, respectively, 2% peptone, and 1% yeast extract. Yeast minimal media (SD-URA) contained 0.67% yeast nitrogen base, 2% glucose or galactose supplemented with 0.5% casamino acids, and adenine (16.5 μg/ml).

All strains used in this study are isogenic derivatives of J178-1D (Mat a ade1 met14 ura3-52 leu2-3,112 his3). The 45-kb dicentric chromosome III strain J178#24 (his4Δ::URA3 GALCEN3) was constructed by fragment-mediated transformation of the conditional centromere at the HIS4 locus (Hill and Bloom, 1989). The rad9Δ dicentric strain was constructed by fragment-mediated transformation into J178#7 (his4Δ::URA3 GALCEN3:: his4), as previously described by Brock and Bloom (1994). The 3-kb dicentric plasmid yC² (Hill and Bloom, 1987) was introduced into J178-1D by standard transformation. All dicentric strains were grown in galactosecontaining medium to maintain chromosome III or the plasmid yC² as monocentric.

Immunofluorescent staining of yeast cells was performed essentially as described by Pringle et al. (1989). Cells were grown to early to mid-logarithmic phase growth in galactose-containing medium. Cells were pelleted, washed once with sterile water, resuspended in glucose-containing medium, and returned to 32°C for 1–2 h. Cells were fixed in 3.7% formaldehyde for 2 h at 25°C before processing for immunofluorescence. Microtubules were visualized with the rat anti–α-tubulin antibody YOL1/34 (Accurate Chemical and Scientific Corp., Westbury, NY). Rhodamine-conjugated goat anti– rat (Cappel Labs, Malvern, PA) was used as the secondary antibody. Cells were incubated with DAPI (1 μg/ml) for 5 min before the addition of mounting medium. Images were acquired and stored for further analysis, as described for FISH. Fluorescent digital images were analyzed with distance measuring tools within Metamorph software (Salmon et al., 1994).

To visualize the nucleus, cells were fixed in 3.7% formaldehyde for 1 h at 25°C. Fixed cells were washed once with water and resuspended in 1 μg/ ml DAPI for 5 min. The cells were washed once with water and mounted in mounting medium.

J178#24 was grown to an A₆₆₀ = 0.2 density in galactose-containing medium before switching to YPD for 1 h. Cells were fixed and stained with DAPI (1 μg/ml for 5 min) or subjected to fluorescent in situ hybridization (FISH) as described by Guacci et al. (1994). Images were visualized with a microscope (Microphot FXA; Nikon) using a 60X/1.4NA PlanApo objective. Metamorph software (Universal Imaging Corp., West Chester, PA) was used to control a Metaltek filter wheel (Salmon et al., 1994). A shuttered mercury lamp was used for fluorescence excitation (490 nm). Images were collected with a cooled CCD camera (C4880; Hamamatsu) and stored on an optical drive (Pinnacle Micro, Irvine, CA). 100–200 nuclei were counted.

Cultures were grown to early logarithmic growth phase and spotted directly from galactose medium to YPD/25% gelatin slabs. Cells were filmed using high resolution VE–DIC and DE–DIC microscopy and analyzed as described in Yeh et al. (1995).

The probability of double strand breakage in a dicentric chromosome is influenced by the behavior of each of the paired centromeres relative to the spindle poles. If the two centromeres on one sister chromatid are oriented to opposite poles, a dicentric bridge is formed, followed by breakage, or chromosome loss. Conversely, if both centromeres from one sister are attached to the same pole, the chromosome will separate without engendering a break. If each centromere behaves independently, we expect to observe two populations of cells. Half of the cells will generate a broken chromosome, and half of the cells may propagate intact, dicentric molecules during cell division (Fig. 1; Koshland et al., 1987). The timing of decatenation of sister chromatids will also affect the behavior of the dicentric chromosome. This potential for different chromosome configurations on the spindle and hence, chromosome breakage or maintenance, has limited previous studies on populations of cells.

VE-DIC microscopy of individual cells allowed us to visualize distinct cellular responses upon dicentric chromosome activation (Fig. 2). Cells containing the facultative dicentric chromosome were grown to mid-logarithmic phase in galactose (monocentric conditions) and switched to glucose (dicentric conditions) for VE-DIC analysis. Two distinct mitotic populations of cells were observed (Figs. 3 and 4) and confirm the expectations predicted from the independent behavior of the two centromeres. One population, 7/29 cells (Fig. 3), mimicked wild-type cells in their kinetic progression through the cell cycle. The initial rapid elongation was completed in 7 ± 2 min, the transition from the sausage shape to bi-lobed nuclear morphology took 9 ± 2 min, and cytokinesis followed 35 ± 5 min later (Yeh et al., 1995).

Aberrant nuclear morphology and a considerable delay between the fast and slow phases of anaphase were observed in 22/29 of the mitotic cells (Figs. 2 and 4). During the initial phases of spindle elongation, a finger-like protrusion of the nucleus into the bud was seen, rather than the typical formation of a sausage-shaped nucleus. This protrusion seemingly snaked its way into the bud for ∼5 ± 1 min (Fig. 4,A). As the spindle elongated to ∼4 μm in length, the nucleus remained rounded in shape as it traversed the neck of the dividing cell (Fig. 2,C). The time from nuclear protrusion to the formation of this rounded sausage nucleus took 10 ± 6 min (Fig. 4 A). Interestingly, the nucleus continued to maintain this round to sausageshaped morphology for 47 ± 18 min before further spindle elongation and the transition to bi-lobed structures. This is significantly longer than the 9 ± 2 min transition in wildtype cells, indicating a significant delay in both the kinetics and morphological transitions that accompany anaphase B. Once the bulk of the nucleus formed a bi-lobed morphology and began the slow phase of spindle elongation, the time to cytokinesis was 38 ± 10 min, similar to that observed in wild-type cells.

Bending of the mitotic spindle in and out of the plane of focus often precludes direct visualization of the spindle by VE-DIC microscopy. To confirm that position of the nucleus during mid-anaphase corresponds to actual spindle elongation through the neck, we compared the length of the nucleus in fixed cells, as determined by DAPI staining, to the length of the mitotic spindle, as visualized by indirect immunofluorescence. Cells containing the conditional dicentric chromosome were grown on galactose-containing medium and switched to glucose-containing medium for 2 h to activate the dicentric chromosome. Cells were fixed and stained for nuclear (DAPI) and spindle (anti–αtubulin immunofluorescence) analysis. In all budded cells with DNA spanning the neck (36/36), the mitotic spindle had elongated through the neck (compare Fig. 5, C and D; E and F). Quantitation of the length of tubulin immunofluorescence and of DAPI staining along the mother bud axis in G2 and mid-anaphase arrested cells shows an excellent correlation between the two measurements. In G2 cells (medium size bud, nucleus near the neck), the mean spindle length was 1.25 ± 0.23 μm, and the mean nuclear length was 1.31 ± 0.19 μm (n = 13) (Fig. 5, A and B). Early stages of anaphase in cells containing an active dicentric chromosome were often characterized by finger-like protrusions of the nucleus into the bud (Fig. 2,B, VE-DIC; Fig. 5,D, DAPI staining). Correspondingly, the spindle spanned the distance from the end of the bulk of chromatin in the mother to the tip of the finger-like protrusion in the bud (Fig. 5,C, 2.54 μm in length). In cells exhibiting a mid-anaphase pause, the mean spindle length was 4.17 ± 0.92 μm and the mean nuclear length was 4.09 ± 0.91 μm (Fig. 5, E and F, n = 31). Thus the fixed images corroborate the progression, as well as the length measurements, in early to mid-anaphase, as determined in the real time analysis, and indicate the close correspondence between nuclear and spindle elongation throughout anaphase.

The kinetics of spindle elongation were determined using a semi-automated tracking system to quantitate spindle morphogenesis and elongation (Skibbens et al., 1993). The direction of spindle movement could be assessed by determining the position of the leading and lagging edge of the nuclear envelope relative to the neck of the budded cell. The neck provided an accurate frame of reference in standardizing data from individual cells and was set at 0 μm. Movements of the spindle pole into the bud (leading edge) were plotted as negative values and movement into the mother (lagging edge) as positive values. The nucleus abutted the neck immediately preceding spindle elongation, and elongation was directed into the bud. In an example of a dicentric cell that displayed wild-type kinetics, both fast and slow phases of spindle elongation, from 2 to 8–10 μm, were complete in ∼22 min (Fig. 3,B). Cells delayed in anaphase B progression (Fig. 4,B) displayed the initial rapid elongation phase (2–4 μm), which was followed by a pause in spindle elongation in which the nucleus remained in the neck for a significant amount of time (up to 60 min in Fig. 4 B). Only after this pause, was the second phase of elongation able to proceed to the maximal spindle length of 10–12 μm.

The partially elongated spindles indicate a delay in anaphase B spindle pole separation. However, the degree of chromosome separation was not apparent in the VE-DIC images. We therefore used FISH to visualize sister chromatids in cells arrested in mid-anaphase. Asynchronous populations of cells containing the conditional dicentric chromosome were grown on galactose and switched to glucose for 1–2 h, when ∼16% of the cells were arrested as large budded cells with DNA spanning the neck (Table I).

Using chromosomal markers proximal to CEN1 on chromosome I (kindly provided by V. Guacci; Guacci et al., 1994), we determined that 34% of cells containing an active dicentric chromosome contained two spots within the nucleus (Fig. 6). In contrast, 8% of the cells were observed to have two fluorescent spots within the nucleus in wildtype, asynchronous populations using the same DNA probes (Fig. 6). The 8% of cells with separated sisters in wild-type populations can be attributed to the time expected for centromeres to separate within one chromatin mass (Yeh et al., 1995). Anaphase spindle elongation (from the fast phase to the bi-lobed structures) persists for ∼9 min (Fig. 3,A). 9/90 min doubling time predicts that 10% of the population will be caught in the mid-anaphase stage. This value compares quite favorably to the 8% determined by FISH (Fig. 6,B; Guacci et al., 1994). By the same reasoning we can predict what fraction of cells harboring a dicentric chromosome will be in mid-anaphase in a logarithmic growing population. The dicentric delay persists for 57 min (Fig. 4). The doubling time of this population is ∼150 min, resulting in ∼38% of the cells at the midanaphase transition at any one time. However, only about a third of the cells in a population pause in response to activation of the dicentric chromosome (Neff and Burke, 1992; Brock and Bloom, 1994). We therefore estimate 2/3(8%) + 1/3(38%) = 17.8% of cells to be in mid-anaphase. This value compares quite favorably to the 16% of cells found to be in mid-anaphase, as determined by DAPI staining (Table I), but less than that determined by FISH (34%). While the high FISH value is certainly indicative of sister chromatid separation in the population of mid-anaphase cells, it may also reflect centromere separation in a fraction of large-budded cells with DNA at the neck (Table I) as well as cells with separated nuclei (Table I, last column), in which the spindles may have collapsed and given rise to apparently undivided nuclei with two spots. In any case, the centromere regions of sister chromatids have separated in nuclei of cells containing active dicentric chromosomes, indicating these cells traverse the metaphase/ anaphase boundary and become blocked following either anaphase A or anaphase A and B, with partially elongated spindles, i.e., mid-anaphase.

The mid-anaphase delay could reflect a physical constraint on spindle elongation, with the DNA between the two centromeres constraining the poles from further separation. Alternatively, a cellular response to the dicentric insult could prevent further elongation (Fig. 7). To distinguish between these possibilities we have analyzed the kinetics and spindle length of the mid-anaphase transition in cells containing variants of dicentric chromosome III, in which the centromeres are 3 vs. 45 kb pairs apart. 45 kb of DNA is ∼15 μm in its linear form. Estimation of the packing ratio in chromatin (sevenfold around a nucleosome) or the degree of condensation from FISH (80 fold; Guacci et al., 1994) provide an approximate length of this region to be anywhere from 0.2–2 μm in vivo. Similarly, dicentric chromosomes with their centromeres 3 kb apart would be expected to span ∼0.02–0.2 μm in vivo. If the physical constraint model is correct, the length of the spindle at the arrest point would reflect centromere–centromere distances, and activation of a dicentric chromosome with the centromeres 3 kb apart might result in a 10-fold difference in spindle length at the delay point. As shown in Fig. 8, cells containing the 3-kb dicentric minichromosome exhibit similar kinetics and morphologies as those of a cell containing the 45-kb dicentric chromosome. Thus the midanaphase pause is not a function of the length between the two centromeres.

Defects in chromosome alignment, kinetochore attachment, or DNA breakage have all been shown to be subject to cell cycle surveillance. The distinguishing feature of defects monitored by these checkpoint pathways is that they delay anaphase onset. Previous studies performed by DAPI staining of fixed populations and FACS^® analysis have revealed RAD9 to be involved in mediating the dicentric response (Brock and Bloom, 1994). We have reexamined the role of RAD9 in the real time assay to determine whether the delay in mid-anaphase is also dependent upon this checkpoint gene. A single cellular response was observed upon dicentric chromosome activation in the rad9 deletion strain (Fig. 9, 9/9 cells; Table II). The behavior of the spindle and the timing of nuclear transitions were indistinguishable from wild-type cells.

Time lapse, VE- and DE-DIC analysis of cells containing an active dicentric chromosome reveals a delay in their progression through anaphase. By two criteria, spindle elongation and sister chromatid separation, cells containing an active dicentric chromosome transiently pause in mid-anaphase, before maximal spindle elongation and the exit from mitosis. This mid-anaphase delay is dependent on the RAD9 DNA checkpoint gene, indicative of DNA damage surveillance pathways operative during late stages in the cell cycle (Fig. 10).

The mid-anaphase delay was distinguished by both kinetic and morphological criteria. Upon activation of the dicentric chromosome, the bulk of the nucleus remained in the mother cell as the spindle began to elongate (Fig. 2,B). Even as the spindle doubled its length the nucleus remained rounded and only later became noticeably sausage shaped (Fig. 2,C). In contrast, the nucleus in wild-type cells underwent a rounded to sausage-shaped transition as it traversed the neck and elongated to 4 μm (Fig. 3; Yeh et al., 1995). The morphological transitions characteristic of dicentric chromosome activation were accompanied by a 30–120-min temporal delay in the subsequent elongation of the mid-anaphase spindle. This 4 μm stage of spindle morphogenesis is likely to represent a critical juncture in spindle function in wild-type cells as well. In an elegant three-dimensional, serial section reconstruction of a yeast spindle, populations of anaphase spindles sorted into two classes based on their lengths: 3.7 and 9.4 μm (Winey et al., 1995). Both populations of spindles were shown to straddle the neck, and in either population, sister chromatids had already migrated to the poles based upon the length of the presumptive kinetochore microtubules. These data corroborate our findings that sister chromatids have separated at the 4 μm mid-anaphase delay (Fig. 6). The preponderance of 3.7 μm spindles reported by Winey et al. (1995) also reflects the brief pause which often occurs in wild-type cells between the fast and slow phases of anaphase spindle elongation (Yeh et al., 1995). In addition, the oscillatory motion observed in the 4 μm spindle in wild-type cells (Yeh et al., 1995) is indicative of a natural transition in spindle progression. Thus, both serial reconstruction of fixed cells and VE-DIC microscopy of live preparations indicate a clear transition from medial to elongated nuclear structures in wild-type cells. The dicentric chromosome insult simply prolongs this natural transition in the development of the anaphase spindle.

There are several aspects of dicentric chromosome activation that may contribute to the delay. One is that tension generated by the chromosome bridge could restrain spindle elongation. The finger-like protrusions of the nucleus into the bud (Figs. 4,A and 5,D), appear to be futile attempts by the spindle/nucleus to initiate anaphase spindle pole separation. This period (5 min duration) has no correlate in the kinetic progression of a wild type cell. These protrusions, and the rounded morphology of the nucleus as it traversed the neck of the dividing cell (Fig. 2,C), probably reflect transient physical constraints on the spindle. The time lapse studies with centromeres at different distances (3 vs. 45 kb apart) indicate that the arrest in spindle elongation is not exclusively due to a DNA bridge constraining anaphase pole separation. Cells containing either dicentric chromosome display virtually identical spindle morphologies and lengths (4 μm) for similar durations (47–59 min on average; Table II). Thus the mid-anaphase delay likely represents surveillance mechanisms that effectively inactivate the microtubule assembly and motility mechanisms which normally produce further spindle elongation.

The dependency of the mid-anaphase delay on the RAD9 gene reveals that the DNA damage checkpoint continues to monitor chromosome integrity well after the start of anaphase, as well as before anaphase onset (Fig. 10). Thus RAD9-dependent DNA damage arrests the cell at any stage (G1, S, G2, and now mid-anaphase). This notion is similar to recent findings that checkpoint genes (RAD9, Siede et al., 1993; and the SAD1/RAD53 protein kinase, Allen et al., 1994) indeed control multiple checkpoints. Whether the damage leads to the same or different signal in each stage, or different signals in the same stage, is not well understood, but several genes are clearly involved at multiple stages of the cycle. Rad9p is constitutively expressed throughout the cell cycle (Schiestl et al., 1989), consistent with the proposal that this protein is involved in DNA surveillance throughout G1/S, G2/M, and anaphase/ cytokinesis. It is also likely that DNA lesions, in addition to a dicentric-induced aberration, are also subject to RAD9 surveillance at mid-anaphase. However, due to the rapidity of spindle elongation and the relatively extended lengths of G1/S and G2/M, the mid-anaphase delay would be difficult to identify in a population study. In this regard, we have examined the effect of a dominant negative form of topoisomerase I (top1-103), which also results in a RAD9-dependent cell cycle arrest (Levin et al., 1993). As reported previously, a large fraction of cells exhibited the large budded RAD9-dependent arrest morphology following 8 h of top1-103 induction. However, DAPI staining of the cells revealed two classes of arrest morphologies. About two-thirds of the cells displayed a single nucleus at the neck of large budded cells, while one-third exhibited elongated nuclei (sausage-shaped) spanning the neck region between the mother and bud (Beach, D., and K. Bloom, unpublished results). This second class of cells is identical in morphology to cells which arrest in midanaphase and indicates that a variety of DNA lesions can lead to delays in mid-anaphase.

The description of a mid to late-anaphase checkpoint leads to the notion that the major cyclin-dependent kinase–cyclin complexes, other regulatory kinases, e.g., DBF2 and 20 (Toyn and Johnston, 1994), cyclin destruction, or other anaphase-progression factors, are operative after anaphase onset (Irniger et al., 1995). In fact, H1 kinase levels remain elevated for prolonged periods in yeast cell populations following activation of the dicentric chromosome (Neff and Burke, 1992; Brock and Bloom, 1994). Cyclins and H1 kinase activity persist throughout anaphase in budding yeast, Drosophila, and nontransformed mammalian cells (Surana et al., 1993; Girard et al., 1995; Sigrist et al., 1995). In budding yeast, fully activated cyclins or overexpression of nondegradable forms of cyclin prevent spindle disassembly at telophase, rather than anaphase onset (Stueland et al., 1993; Surana et al., 1993), and high levels of cdc28/clb are found in cells arrested at telophase (cdc15; Surana et al., 1993; Irniger et al., 1995). The persistence of cyclin/clb complexes throughout anaphase raises the possibility that they may be involved in regulating the progression of anaphase directly, or by delaying anaphase progression in response to perturbation. The coexistence of subpopulations of degraded and protected cyclins (Irniger et al., 1995) is also suggestive of a role for the cdc28/cyclin B complex in mediating the progression of anaphase, spindle disassembly, and cytokinesis. Thus the sequential degradation of different cyclins may drive the progression from metaphase to telophase, or may be required to delay anaphase progression in the event of damage.

The description of a mid-anaphase transition raises two major questions. Upon sister chromatid separation, the cells lose a major pathway for repair of DNA damage. Therefore, why have a mid-anaphase checkpoint? Secondly, what perturbations are monitored by this checkpoint in the normal course of cell division? Several attributes of mitosis in budding yeast shed light on these issues. Yeast chromosomes have not been observed to align on a metaphase plate before anaphase onset. This has been thought to reflect the absence of chromosome condensation. However, Winey et al. (1995) demonstrated that kinetochore microtubules shorten asynchronously, reflecting the asynchronous movement of chromosomes to the poles. Mid-anaphase may therefore represent the last time in the cell cycle where errors in attachment may be rectified. Alternatively, the mid-anaphase transition could represent the last stage for DNA repair between sister chromatids. While FISH experiments (Guacci et al., 1994) and results from analysis of sister chromatids at the midanaphase delay (Fig. 6) indicate that sister centromeres have separated at this stage, the chromosome arms may not be completely disjoined. The uniformity of DAPI staining in cells with 4 μm spindles spanning the neck indicates that the bulk of the chromosomal DNA is evenly distributed within the nucleus. In contrast, the morphological transition to a bi-lobed nucleus results in substantial separation of chromosomal DNA. This most likely represents the final decatenation between sister chromatids and the irreversible commitment to chromosome separation (Fig. 10). In this view, the mid-anaphase transition represents the stage prior to the final commitment to complete DNA separation. The temporal coordination between completion of sister chromatid decatenation and force generation for spindle elongation necessary for preventing chromosome breakage may be the primary reason for the midanaphase pause, and one that may be important for all eukaryotic organisms. Loss of coordination of these events in transformed cells lacking mid-anaphase checkpoints may result in increased chromosome breakage, giving rise to chromosomal rearrangements, loss of genomic integrity, and hence, the neoplastic state.

We thank Dr. Vinny Guacci (Carnegie Institute, Baltimore, MD) for his generous gifts of probes, protocol, and invaluable expertise for the FISH experiments. We thank Drs. Danny Lew (Duke University, Durham, NC) and Jim Allen (University of North Carolina, Chapel Hill, NC) for critical comments. Thank you also to Paul Maddox for assistance with microscopy and image processing.

This work was supported by research grants from the National Institutes of Health General Medical Sciences (K. Bloom and E.D. Salmon).

DE and VE

digital and video enhanced

DIC

differential interference contrast

FISH

fluorescent in situ hybridization