Nuclear oscillations and nuclear filament formation accompany single-strand annealing repair of a dicentric chromosome in Saccharomyces cerevisiae

The study of DNA repair mechanisms has been aided by the development of experimental systems that generate DNA double-strand breaks (DSBs) at specific chromosomal locations and during defined periods of the cell cycle. One such system is the budding yeast conditional dicentric chromosome(Hill and Bloom, 1989). This inducible dicentric chromosome was constructed by inserting a second copy of centromere 3 (CEN3) into chromosome III, approximately 45 kb upstream of the endogenous centromere 3 (Fig. 1A). The inserted centromere is under control of the GAL1promoter, allowing for the reversible inactivation of this centromere when cells are grown in the presence of galactose. Studies utilizing dicentric chromosomes are unique in that DSBs are generated by mitotic pulling forces when centromeres on the same sister chromatid are aligned to opposite spindle poles during mitosis. Previous work showed that efficient healing of a broken dicentric chromosome in Saccharomyces cerevisiae requires RAD52, indicating that repair occurs primarily through homologous recombination (Brock and Bloom,1994). In cells lacking RAD52, infrequent dicentric chromosome repair events occur through non-homologous end joining(Kramer et al., 1994).

Resolution of a conditional dicentric chromosome is accompanied by the deletion of one centromere and the intervening DNA. These events were initially thought to result from gene conversion (GC) accompanied by crossover(Jager and Philippsen, 1989). However, other studies of direct repeat recombination have shown that deletions commonly occur without generating the circular product predicted to form through a crossover event(Fishman-Lobell et al., 1992). Moreover, many deletion events are dependent on RAD1 and RAD10 (Klein, 1988; Schiestl and Prakash, 1990; Ivanov and Haber, 1995). The Rad1 and Rad10 proteins form a structure-specific endonuclease that generates 5′ single-stranded ends (Siede et al., 1993). These genes function in a recombination pathway known as single-strand annealing (SSA) (Sugawara and Haber, 1992; Sung et al.,1993; Ivanov and Haber,1995). The recognition of SSA as a major pathway for direct repeat recombination suggests that this mechanism plays a role in the repair of a dicentric chromosome.

Previous studies have shown that recombination frequency is primarily governed by the degree of sequence homology rather than the location of the recombining sequences within the genome(Haber and Leung, 1996). This implies the existence of a homology search mechanism as part of the recombination process. The integration of E. coli lacO and tetO sequences into chromosomes of yeast strains encoding lac repressor-GFP and tet repressor-GFP fusion proteins has recently provided visual evidence of long-distance interactions between like sequences(Aragon-Alcaide and Strunnikov,2000). Such interactions may represent a constitutive homology search process. The mechanisms that bring homologous sequences into close association are unknown. Efficient meiotic recombination in the fission yeast Schizosaccharomyces pombe requires microtubule-dependent oscillatory nuclear movements (Chickashige et al., 1994; Ding et al., 1998; Yamamoto et al., 1999). It has been suggested that such movements promote the alignment of homologous chromosomes. Microtubule-driven nuclear movements could similarly promote the association and recombination of homologous sequences during the vegetative cycle.

In this study we show a requirement for RAD1 in the repair of a dicentric chromosome, which is consistent with recombination via SSA. Dicentric chromosome repair also exhibits a requirement for RAD52;however we show that this process can occur via RAD52-independent SSA in cells containing the adaptation-defective cdc5-ad allele of the yeast polo kinase and the DNA damage checkpoint gene RAD9. Dicentric cdc5-ad rad52 mutant cells additionally experience an extended arrest that is characterized by microtubule and ATP-dependent nuclear oscillations. Finally, recombinational repair of a dicentric chromosome containing direct and inverted centromere repeats, as well as spontaneous recombination events between his3 heteroalleles, also show a dependency on microtubule function. These latter findings suggest that microtubule-dependent processes play a general role in recombination. We discuss ways in which nuclear movements might contribute to chromosome repair.

Growth conditions and media used to evaluate the kinetics of dicentric chromosome repair were as previously described(Brock and Bloom, 1994). Genotypes and construction of the monocentric (J178-1D), dicentric inverted(J178#24) and dicentric direct repeat (J178#27) strains used for benomyl sensitivity assay are likewise detailed elsewhere(Brock and Bloom, 1994). The LS20 and cdc5-ad strains were a gift of Lee Hartwell (Fred Hutchison Cancer Center, Seattle, WA). Strain HUWT(Qiu et al., 1999) was derived from progeny of W1021-7c and W1089-6c(Thomas and Rothstein, 1989). Strains developed for this work include KBY3009, a dicentric direct repeat strain created in the LS20 genetic background (MATΔ ade2 ade3::GalHO lys5 trp1 leu2 his3 ura3 can1 cyh2 his4Δ::URA3GALCEN3); KBY3002(KBY3009 with rad1::KAN^r); KBY3010 (KBY3009 with rad52::LEU2); KBY3006 (KBY3002 with rad52::LEU2); KBY3003(KBY3009 with rad9::TRP1); KBY3011, identical to KBY3009 except that it contains the cdc5-ad point mutation (L251T); KBY3012 (KBY3011 with rad1::KAN^r); KBY3034 (KBY3011 with rad52::LEU2);KBY3014 (KBY3012 with rad52::LEU2); KBY3015 (KBY3034 with rad52::LEU2); KBY3031 (KBY3039 with dhc1::LEU2 and rad52::TRP1); KBY3039 (KBY3011 with HTB1-GFP-KAN^r); KBY3037 (KBY3039 with rad52::LEU2); KBY3038 (KBY3009 with HTB1-GFP-KAN^r); KBY3040 (KBY3038 with rad52::LEU2); and KBY5004 (J178-1D MAT a, ade1, met 14, ura3-52,leu2-3, 112, his3-11, 15, H2B-GFP-KAN^r). KBY5004 was transformed with a GAL-MPS1 construct on a 2 μm plasmid to generate a cell population arrested in mitosis by MPS1overexpression. Strains were constructed by one step gene replacement(Rothstein, 1991) using plasmids or PCR fragments. Analysis of CEN3 recombination following activation of the dicentric chromosome was as previously described(Brock and Bloom, 1994) except that digoxigenin-labeled probes labeled with the DIG PCR synthesis kit (Roche,Indianapolis, Indiana) were used in Southern blotting. Probes were hybridized to genomic DNA bound to nylon membranes using the DIG Easy Hyb system (Roche). Southern blots were developed with the AttoPhos Substrate Kit (Roche) and imaged with a Storm gel analyzer (Molecular Diagnosics). A minimum of two clones of each strain were used for physical and microscopic analyses.

Yeast strains were transformed with an integrating PCR fragment generated with oligos HTB2#1:gggctgttaccaaatactcctcctctactcaagccgcagcagcagtcagcagcagcaatgagtaaaggagaagaa,HT2BN#2: tcatatctggacaagtaacagaaccctaatgttactggcggcgttagtatcg and the plasmid pKanMX2-GFP (K.B., unpublished), which contains the green fluorescence protein coding region and the G418 resistance marker. Cells were added to the surface of a gelatin slab formed on the surface of a microscope slide. Slabs contained 25% gelatin, 2% glucose, 0.5% casamino acids, 50 μg/ml tryptophan, 50μg/ml adenine and 16.5 μg/ml uracil (YCAUT). Coverslips were sealed to slides with Valap (1:1:1 vaseline:lanolin:paraffin). A flow chamber was constructed by attaching a coverslip to a microscope slide with petroleum jelly to observe the effects of microtubule and ATP inhibitors on nuclear dynamics. The coverslip was precoated with a 1 mg/ml solution of poly-L-lysine to facilitate cell attachment. Filling and rinsing of the chamber was accomplished by adding liquid to one end and simultaneously withdrawing liquid from the opposite end with a filter paper wick. Microscopic observations were made at room temperature (22°C). The microscopy system and GFP fluorescence imaging have previously described(Shaw et al., 1997a), with the exception that observations in the current study were made using a 100×/1.4 NA Plan Apo objective. Metamorph software (Universal Imaging)was used to control shutters, focusing and image acquisition. Five z-axis fluorescence images separated by 0.75 μm were collected for each time point and converted to a single 3D image using Metamorph software. A single differential interference contrast image was collected at the medial z-step.

Hydroxyurea (HU)-arrested cells were obtained by growing KBY5004 to log phase and transferring cells to YPD supplemented with 40 μg/ml adenine and 200 mM HU. A MPS1-arrested population was obtained by growing KBY5004 cells containing a GAL-MPS1 expression plasmid overnight in adenine-supplemented sucrose media lacking uracil to maintain the plasmid(Lauze et al., 1995). Cells were grown to log phase and switched to galactose media lacking uracil for 6 hours to promote overexpression of Mps1. Large budded cells were filmed on YCAUT glucose + 25% gelatin slabs for 2 hours. Images were acquired every 2 minutes with an exposure time of 250 milliseconds. Images of dicentric cdc5-ad rad52, dicentric cdc5-ad and dicentric Rad52 cells were obtained with the same exposure and time intervals described above following 6 hours of exposure to YCAUT glucose media. A cell was scored as oscillation + if during the time course it underwent at least one cycle of nuclear movement from mother to bud to mother, such that at each stage >50% of the nuclear material was in either the mother or bud.

Measurements of nuclear filaments were taken with Metamorph software. Filaments were measured from the filament tip to the geometric center of the nuclear mass. Length measurements were made in triplicate using appropriate calibrations from a stage micrometer and corrected for specimen tilt angle with a Microsoft Excel algorithm containing the formula c²=square root (a² + b²), where a is the measured distance taken from the 3D image and b is the distance between the uppermost and lowest focal planes in which the filament was visible. Growing and shortening rates were calculated from a least squares fit of four or more data points. Criteria for dynamic parameters were based on those used to quantify microtubule dynamics(Tirnauer et al., 1999). Growth and shortening events were defined as a change in length sustained over four time points having a correlation coefficient ≥0.85 and yielding a total length change >0.6 μm. The proportion of time spent in growing,shortening or paused phases was determined by measuring the amount of time spent in a given phase for all filaments and dividing by total time. Dynamicity (Toso et al., 1993)was calculated by dividing the sum of all filament length changes measured by total observation time.

The spontaneous his3 heteroallele recombination frequency of strain HUWT was determined by tabulating the formation of HIS+colonies on synthetic media lacking histidine(Qiu et al., 1999). Cells were plated in triplicate on media with and without added benomyl for each experiment. Cells were plated on media lacking uracil both in the presence and absence of benomyl to correct for differences in the cell growth/viability resulting from drug treatments.

Previous studies have shown that the resolution of a conditional dicentric chromosome III in S. cerevisiae occurs primarily through homologous recombination between centromere (CEN) repeats, resulting in the elimination of one copy of CEN3 and 45 kb of intervening DNA(Fig. 1A,B)(Brock and Bloom, 1994). Deletion of RAD52 prevents this recombination event and results in a significant decrease in viability (Fig. 1C, Table 1)(Brock and Bloom, 1994). The requirement for RAD52 is consistent with either GC accompanied by crossover or SSA. We deleted RAD1 in the dicentric chromosome strain to distinguish between these possibilities. The viability of the rad1deletion strain was 26-fold lower than an isogenic strain containing a functional copy of RAD1. This was similar to the loss in viability that accompanied the deletion of RAD52(Table 1). Physical analysis of CEN3 recombination showed that RAD1, in addition to RAD52, is required for the formation of the CEN3-GALCEN3recombination product (Fig. 1D). The viability of the rad1 rad52 double deletion mutant was the same as either of the single mutants(Table 1), suggesting that both genes function in the same pathway. This result is in contrast to several genetic studies of spontaneous recombination events that have revealed a synergistic effect of the deletion of both RAD1 and RAD52,suggesting that these genes operate in different pathways(Klein, 1988; Schiestl and Prakash, 1988; Thomas and Rothstein,1989; Liefshitz et al.,1995). However, other studies utilizing endonucleases to create DSBs have shown that loss of RAD52 reduces the efficiency of SSA when homologous repeats are 5 kb or less in length(Fishman-Lobell et al., 1992; Plessis et al., 1992; Sugawara and Haber, 1992). It is possible that the structure of the DSB determines the requirement for RAD52. DNA ends produced during dicentric chromosome breakage may be more similar to those generated by endonucleases than they are to DNA ends formed during spontaneous recombination events. Physical studies of recombination events have revealed that rad52 mutant strains show an increased rate of DNA strand excision(White and Haber, 1990; Sugawara and Haber, 1992). This suggests that RAD52 may protect DNA ends from excessive resection by nucleases as opposed to playing a direct role in the SSA process. It is also possible that some dicentric chromosome repair events occur through a one-end strand invasion pathway(Sugawara and Haber, 1992; Prado and Aguilera, 1995).

The DNA damage checkpoint gene RAD9 is required for efficient utilization of some repair pathways(Fasullo et al., 1998). A RAD9-dependent midanaphase delay has been observed in yeast cells containing a dicentric chromosome, but loss of this gene does not adversely affect strain viability (Table 1) (Yang et al.,1997). We examined the kinetics of CEN3 recombination in a rad9 deletion strain and found them identical to the corresponding wild-type strain (Fig. 1E,compare with Fig. 1B). Our findings suggest that dicentric chromosome repair events are completed rapidly and do not require the additional time provided by the RAD9-dependent cell cycle delay.

Cells containing the cdc5-ad allele of the yeast polo-like kinase experience an extended mitotic delay in response to an unrepaired DSB(Toczyski et al., 1997). We considered that such a delay might allow us to detect RAD52-independent SSA events by allowing additional time for completion of repair prior to the end of mitosis. CEN3 recombination induced by dicentric chromosome breakage occurred with the same kinetics regardless of whether cells contained the cdc5-ad or CDC5wild-type allele (Fig. 2A,compare with Fig. 1B). However,in contrast to the dicentric CDC5 rad52 strain(Fig. 1C), CEN3recombination was observed in the dicentric cdc5-ad rad52 strain(Fig. 2B). No CEN3recombination product was detected in a cdc5-ad rad1 rad52 deletion strain (Fig. 2C), indicating that this event occurs through SSA. We additionally showed that RAD52-independent repair in a cdc5ad strain requires RAD9 (Fig. 2D). Analysis of populations derived from individual dicentric cdc5-ad rad52 cells revealed that recombination occurred through a RAD1-dependent mechanism in 19 of 20 clones (data not shown).

Over the course of 24 hours following activation of the dicentric chromosome in the cdc5-ad rad52 strain, 89% of the population arrested as abnormally large budded cells, similar to the phenotype previously reported for cdc5-ad cells containing unrepaired DNA damage(Toczyski et al., 1997). This extended mitotic arrest was not observed in dicentric CDC5 cells,dicentric cdc5-ad cells, dicentric rad52 cells, cdc5-ad RAD52 cells or in cdc5-ad rad9 rad52 cells(Table 2). These findings indicate that the requirements for arrest include a source of DSBs (the activated dicentric chromosome), the cdc5-ad allele, the presence of RAD9 and deletion of RAD52, consistent with the original characterization of the cdc5-ad allele(Toczyski et al., 1997). It was previously shown that that the cell cycle arrest was generated by the cdc13-1 allele (Toczyski et al.,1997). By contrast, cells containing a conditional dicentric chromosome experience a midanaphase delay(Yang et al., 1997). To determine if dicentric cdc5-ad rad52 cells also pause in midanaphase,we replaced the endogenous HTB2 (histone 2B) gene with an HTB2-GFP fusion construct. Images of 25 histone-GFP expressing cdc5-ad rad52 cells were collected 2 hours after activation of the dicentric chromosome. Nineteen of the 25 large-budded cells contained a bilobed nucleus that straddled the bud neck (data not shown), which is indicative of a midanaphase arrest (Yang et al., 1997).

Real-time observations of dicentric cdc5ad rad52 cells expressing HTB2-GFP revealed an extended period of oscillatory nuclear movements(Fig. 3). This behavior initiated with the retraction of the bud-proximal lobe of a bilobed nucleus,resulting in the formation of a single nuclear mass. In 52% of arrested cells,the nucleus oscillated between the mother cell and bud. This behavior was not observed in dicentric cdc5-ad RAD52 cells or in CDC5 rad52cells (Table 3). Time-lapse images of cells collected at 15 minute intervals over a 24 hour period revealed that more than 90% of cells exhibited nuclear oscillations for at least 14 hours after activation of the dicentric chromosome. Nuclear oscillations were also accompanied by the extension and collapse of filamentous nuclear protrusions (Fig. 3). Approximately 1% of cells that remained arrested longer than 6 hours succeeding in completing nuclear division and cytokinesis (data not shown). The divided cells rebudded and proceeded through the next cell cycle with normal kinetics.

To determine if nuclear oscillations are a general feature of cells that have experienced a mitotic arrest, we characterized nuclear movements following exposure to hydroxyurea (HU) or elevated expression of MPS1. Treatment with HU causes a DNA-damage-induced mitotic block. Overexpression of MPS1, a protein involved in the spindle checkpoint and spindle pole body duplication, also results in a mitotic arrest (reviewed in Winey and Huneycutt, 2002). Nuclear oscillations were not observed in HU-arrested cells(Table 3). Cells overexpressing MPS1 did undergo nuclear oscillations, albeit less frequently than dicentric cdc5ad rad52 cells(Table 3). Our results indicate that the sequestration of cells in a mitotic arrest state is not of itself sufficient to promote nuclear oscillations.

Microtubules are required for the oscillatory nuclear movements that occur during meiotic prophase S. pombe(Ding et al., 1998). To determine if nuclear oscillations in dicentric cdc5-ad rad52 cells also depend on microtubules, we treated cells with media containing 100μg/ml of the microtubule-destabilizing drug benomyl. Real-time observation of 10 benomyl-treated cells over a period of 30 minutes to 1 hour revealed no evidence of the nuclear movements seen in the untreated cells(Fig. 4A-D). Some of the benomyl-treated cells contained nuclear filaments, but no growth or shortening of the filaments was observed and no new filaments formed during the period when cells were exposed to the drug. This result demonstrates that nuclear oscillations in dicentric cdc5-ad rad52 cells require microtubules.

In addition to microtubules, meiotic nuclear oscillations in fission yeast require the microtubule motor dynein(Yamamoto et al., 1999). In budding yeast, dynein provides the major pulling force during anaphase and is responsible for the nuclear oscillations observed prior to anaphase onset in wild-type yeast cells (Yeh et al.,2000). Deletion of the gene encoding dynein heavy chain(DHC1) reduces the viability of cells containing a dicentric chromosome (Table 1). We deleted DHC1 to determine if dynein is required for nuclear oscillations in cdc5-ad rad52 cells. The nuclear dynamics of dhc1 mutant cells were indistinguishable from cells containing a functional copy of DHC1 (Fig. 4E-H). Thus, although we cannot rule out a role for dynein in anaphase nuclear oscillations in dicentric cdc5-ad rad52 cells, it is clearly not the only source of motility.

Because microtubule motors use ATP hydrolysis to generate movement, we tested the ATP requirement for nuclear oscillations by treating dicentric cdc5-ad rad52 cells with a solution containing 0.02% sodium azide and 1% deoxyglucose. To quantify the nuclear dynamics, length changes for individual nuclear filaments were measured over time. A plot of length changes for one nuclear filament in a cell treated with azide and deoxyglucose is shown in Fig. 5. The azide/deoxyglucose treatment resulted in the immediate cessation of both nuclear oscillations and changes in nuclear morphology. Nuclear dynamics resumed almost immediately upon removal of the azide and deoxyglucose. Growing and shortening rates were calculated, and the percentage of total time that filaments spent in growing, shortening or paused phases was determined (see Materials and Methods). The dynamicity parameter, derived from studies of microtubule dynamics, was also calculated as an overall measure of filament growth and shortening (Toso et al.,1993). The results are shown in Table 4. Depletion of ATP resulted in a significant suppression of filament dynamics. The most notable changes were the increase in pause time from 50% to over 90% of the total time, and the decrease in dynamicity, with post-treatment values 13-fold lower than the pre-treatment values. Interestingly, the rates of nuclear filament formation and collapse in untreated cells (0.8 μm/minute) are similar to rates of yeast cytoplasmic microtubule dynamics during anaphase(Tirnauer et al., 1999). Our findings demonstrate that nuclear oscillations in dicentric cdc5-ad rad52 cells require ATP, which is consistent with a possible role of microtubule motors in this process.

The association between nuclear movements and enhanced CEN repeat recombination suggested that microtubules play a role in other recombination events. We tested this by determining if the survival of cells containing an inverted repeat dicentric chromosome was compromised by exposure to benomyl. Inverted repeat dicentric chromosomes undergo RAD52-dependent recombination between repeated sequences that flank the centromere region,resulting in deletion of one of the CEN repeats(Brock and Bloom, 1994). A reduction in the viability of an inverted dicentric strain following treatment with benomyl would thus indicate a microtubule dependence for recombination between non-centromeric sequences. A range of benomyl concentrations was chosen that did not significantly inhibit cell cycle progression nor result in detectible microtubule depolymerization (data not shown). As shown in Fig. 6, a monocentric strain exhibited only minor suppression of colony forming ability when plated on the highest benomyl concentration (15 μg/ml). By contrast, the viability of both inverted and direct repeat dicentric strains was significantly compromised on plates containing 10-15 μg/ml of benomyl. These results indicate that efficient recombinational repair of a dicentric chromosome requires microtubules whether occurring through CEN repeat recombination or by means of recombination between other repeated sequences.

We further examined the potential role of microtubules in recombination by determining the effect of benomyl on the frequency of spontaneous recombination between two his3 heteroalleles flanking a copy of the URA3 gene (Table 5). Cells were plated on media containing concentrations of benomyl in the range of those used in the dicentric chromosome assay, and the his3recombination frequency was compared with cells that had not been exposed to the drug. Recombination was suppressed 2.3-fold by 15 μg/ml benomyl. These findings indicate that microtubule-dependent processes enhance spontaneous recombination events in budding yeast, in addition to promoting recombination events associated with dicentric chromosome resolution and meiotic recombination in fission yeast.

We have utilized a conditional dicentric chromosome in budding yeast to study the repair of DSBs that occur during mitosis. We found that RAD1 is required for dicentric chromosome repair both in the presence and absence of RAD52 (Figs 1 and 2). The requirement for RAD1 in recombinational deletion events is characteristic of the SSA pathway (Ivanov and Haber,1995). The role of cdc5-ad in RAD52-independent repair of the dicentric chromosome is currently unclear. The presence of the cdc5-ad allele alone does not alter the frequency of SSA events resulting from a DSB caused by HO endonuclease in cells that contain Rad52(Galgoczy and Toczyski, 2001). In addition, we did not observe any difference in dicentric chromosome recombination kinetics when CDC5 RAD52 and cdc5-ad RAD52strains were compared (data not shown). However, it is a reasonable assumption that CEN recombination in the absence of RAD52 is a consequence of the cdc5-ad-dependent mitotic block, given that both require RAD9.

Centromere recombination was consistently detected prior to activation of the dicentric chromosome in cdc5-ad rad52 cells(Fig. 2B, lane 1). By contrast,this was not observed with dicentric CDC5 RAD52 or cdc5-ad RAD52 strains (Fig. 1B,lane 1; Fig. 2A, lane 1). In cells grown on galactose, transcriptional readthrough at the GALCENprevents the activation of the dicentric chromosome. However, it has been shown that the conditional dicentric chromosome III is lost more frequently than its monocentric counterpart on galactose(Neff and Burke, 1992). Thus a functional kinetochore can form, albeit infrequently, even in the presence of galactose. We suggest that the nuclear oscillations associated with the cdc5-ad rad52 strain result in an increase in dicentric chromosome breakage, leading to more frequent recombination on galactose. The reduced functional capability of the conditional kinetochore on galactose does not promote efficient dicentric chromosome breakage but may induce sufficient tension on the chromosome to trigger the DNA damage response. The DNA-damage-induced arrest in dicentric cdc5-ad rad52 cells results in a prolonged mitotic arrest characterized by a sustained period of nuclear oscillations. These oscillations could promote more frequent dicentric chromosome breakage events than would normally occur in the presence of galactose and as a consequence, more frequent CEN3 recombination. Consistent with this, a small portion (<10%) of very large budded cells with nuclear morphology identical to that in Fig. 3R were found in populations of dicentric cdc5-ad rad52 cells grown in galactose media(data not shown). It is unlikely that this mechanism accounts for a significant proportion of the recombination events that occur when dicentric cells are grown on glucose, given the high efficiency of chromosome breakage promoted by an activated GALCEN.

Live cell imaging of GFP-labeled nuclei following dicentric chromosome activation in cdc5-ad rad52 cells allowed us to observe an extended period of nuclear oscillations accompanied by the formation and collapse of nuclear filaments (Figs 3,4,5). The association of these nuclear movements with RAD52-independent dicentric chromosome recombination suggests that this behavior might be responsible for the CEN recombination events observed in these cells. Microtubule-dependent oscillatory nuclear movements are also observed during meiosis in fission yeast and are required for efficient meiotic recombination(Chickashige et al., 1994; Ding et al.,1998; Yamamoto et al.,1999). It has been suggested that nuclear oscillations observed during meiosis in S. pombe facilitate recombination by promoting the alignment of homologous chromosomes(Yamamoto et al., 1999). Although this exact mechanism is unlikely to explain enhanced recombination in dicentric cdc5-ad rad52 cells, the contortions that the nucleus undergoes during this process (Fig. 3) could increase the frequency of collisions between CENrepeats located in the two halves of a broken dicentric chromosome. A previous study revealed that recombination events promoted by the HO endonuclease occur with similar frequency regardless of the location of the recombining sequences within the genome, consistent with the idea that a homology search mechanism may facilitate such events (Haber and Leung, 1996). It has also been demonstrated that increasing the copy number of a sequence can increase the likelihood a sequence will undergo recombination, suggesting that recombination is stimulated by increasing the frequency with which sequences interact(Melamed and Kupiec, 1992). Collision frequency may play a greater role in promoting recombinational repair of a broken dicentric chromosome, where each half of the broken chromosome is actively separated from the other, than it does in repair of other types of DSBs.

Nuclear oscillations are not unique to dicentric cdc5-ad cells. We have observed cells exhibiting this behavior as a consequence of MPS1overexpression (Table 3), and nuclear transits between the yeast mother cell and bud have previously been reported in cdc13, cdc16, and cdc23 mutant cells following their release from a mitotic arrest(Palmer et al., 1989). The proteins encoded by CDC16 and CDC23 are part of the anaphase-promoting complex (APC) that facilitates entry into and exit from anaphase. The cdc13 arrest occurs through a RAD9-dependent DNA damage response pathway that prevents progression into anaphase by inhibiting the APC-dependent degradation of Pds1/Securin. Recent studies have shown that a reduction in Pds1 levels at anaphase onset is required for anaphase spindle stability (Jensen et al.,2001; Severin et al.,2001). Spindle instability could result in alterations in spindle length and could account for the conversion of elongated/bilobed nuclei to spherical structures observed in dicentric cdc5-ad rad52 cells (see Fig. 3H-I). Interestingly, APC complexes immunoprecipitated from cdc5-ad mutant cells exhibit a reduced degree of ubiquitination activity(Charles et al., 1998). Ubiquitination of Pds1 is a prerequisite for its degradation, and the persistence of elevated levels of Pds1 in cdc5-ad mutant cells could lead to spindle instability and contribute to nuclear oscillations. Future experiments will be required to determine whether nuclear oscillations result from spindle microtubule dynamics, cytoplasmic microtubule dynamics or a combination of the two.

The association of microtubule-dependent nuclear oscillations with CEN recombination raised the question of whether microtubules play a role in recombination events involving other homologous repeats. To test this hypothesis we examined the effect of low concentrations of benomyl on a dicentric chromosome repair process that involves recombination between non-centromere sequences (Fig. 6) and the effect of this drug on spontaneous recombination between of a pair of his3 heteroalleles(Table 5). We have observed that concentrations of benomyl in the range used in our experiments suppress the dynamic instability behavior of cytoplasmic microtubules and yeast kinetochore movements during mitosis (K.B. and Chad Pearson, unpublished). The growth and shortening of microtubules is responsible, at least in part, for nuclear movements that occur throughout the cell cycle in budding yeast(Shaw et al., 1997b) and for chromosome movements during mitosis(Pearson et al., 2001). Thus it is likely that the significant reduction in viability of cells containing a dicentric chromosome in response to exposure to benomyl is due to decreased chromosome breakage resulting from suppressed microtubule dynamics. The suppression of recombination revealed by the his3 assay system indicates that microtubules may play a more general role in facilitating efficient recombination. It is unlikely that this effect was the result of the moderate decrease in growth rate that accompanied the benomyl treatments given that previous studies have shown that cell cycle delays have either a positive effect or no significant impact on DNA repair(Weinert and Hartwell, 1988; Brock and Bloom, 1994). Previous studies have shown that spontaneous intrachromosomal recombination occurs at a higher rate than spontaneous recombination between sequences located on different chromosomes (Lichten and Haber, 1989; Jinks-Robertson et al., 1993). This indicates that a collision-based homology search may not play a significant role in this type of recombination. It is assumed from the similar genetic requirements of recombination induced by agents that produce DSBs and those required for spontaneous recombination that the latter events are likewise promoted by DSBs. The source of these cryptic DSBs has not been definitively identified. On the basis of our finding that low concentrations of benomyl suppress recombination, we suggest that a portion of the DSBs involved in these recombination events result from microtubule-generated nuclear movements. Further, the modest effect of the drug treatments on this process indicates that both dependent and microtubule-independent sources of DSBs give rise to spontaneous recombination events.

We thank David Toczyski and Lee Hartwell for the LS20 and cdc5-adstrains, Junzhuan Qiu and Binghui Shen for strain HUWT. We also thank Chad Pearson and Leana Topper for critical reading of the manuscript. This work was supported by Department of Energy grant DE-FG02-99ER62746 (K.B. and D.T.) and National Institutes of Health grant GM32238 (K.B.).