The S-Phase Cyclin Clb5 Promotes rRNA Gene (rDNA) Stability by Maintaining Replication Initiation Efficiency in rDNA

Regulation of replication origins is important for complete duplication of the genome, but the effect of origin activation on the cellular response to replication stress is poorly understood. The budding yeast rRNA gene (rDNA) forms tandem repeats and undergoes replication fork arrest at the replication fork barrier (RFB), inducing DNA double-strand breaks (DSBs) and genome instability accompanied by copy number alterations.

We previously performed a genome-wide screen to identify genes that function to maintain the stability of the rDNA region (13). In that screen, we used a collection of mutants in which each of ;4,800 nonessential genes of budding yeast is deleted, isolated the genomic DNA, and analyzed the size and size heterogeneity of chromosome XII of each mutant by pulsed-field gel electrophoresis (PFGE). The screen identified 708 mutants that showed instability of the rDNA region relative to that of wild-type (WT) cells (13). Many of the genes deleted in these mutants were found to function in nucleic acid transactions such as DNA replication, repair, and recombination. Unexpectedly, however, we also identified genes that had annotated functions in biological processes that seem to be unrelated to maintaining genome stability, such as cell and organelle morphogenesis (13). In this study, we sought to understand how these latter genes contribute to rDNA stability and found that Clb5 is a novel factor that is required for maintaining rDNA stability.
Here, we show that rDNA instability in the clb5D mutant is mostly suppressed when programmed replication fork arrest at the RFB site in rDNA is inhibited by fob1 mutation. Deletion of the CLB5 gene resulted in a reduction in the efficiency of replication origin firing in rDNA by half relative to that of WT cells. Both the rDNA instability and origin firing defects in clb5D cells were suppressed by the additional deletion of CLB6. The level of arrested forks was comparable in the tested strains, regardless of the presence or absence of Clb5. In the absence of Clb5, therefore, fewer replication forks initiate DNA replication in the rDNA and stall at the RFB, but these forks are more stably arrested because the arrival of the converging forks is delayed. Although the absence of Clb5 did not influence the level of DSBs or resected DSBs, it resulted in a higher level of recombination intermediates. rDNA instability in clb5D cells was dependent on Rad52, which is essential for homologous recombination. Thus, persistently arrested forks at the RFB site may lead to DSB-independent, recombination-dependent rDNA instability. Furthermore, absence of Clb5 results in an increase in the distance between active origins due to inefficient replication initiation in rDNA, potentially leading to replication stress that causes damage at non-RFB sites and recombination-mediated rDNA instability.

RESULTS
The rDNA region is destabilized in the absence of Clb5. We previously identified 708 candidate genes that may contribute to maintaining rDNA stability and conducted gene ontology (GO) analysis (13). We have tested the reproducibility of rDNA instability in some mutants lacking genes known to function in nucleic acid metabolism, such as DNA replication, recombination, repair, and transcription (13,(19)(20)(21)(22). Among the original 708 rDNA-unstable mutants, however, 113 mutants lack a gene that has an annotated function seemingly unrelated to genome stability but instead is involved in biological processes such as membrane invagination, endocytosis, cytokinesis, nucleus organization, conjugation, cell morphogenesis, cell wall organization or biogenesis, cytoskeleton organization, sporulation, membrane fusion, pseudohyphal growth, invasive growth in response to glucose limitation, and exocytosis or cell budding (13).
In our previous genome-wide screen, we assessed the degree of rDNA stability of each of ;4,800 mutants only once by PFGE because the number of mutants was too large to analyze in multiplicate. Furthermore, we used a comb with the thinnest teeth to improve throughput, which might have reduced the sharpness of the band and produced false positives, as discussed previously (23). In addition, we now possess an updated version of the original yeast deletion collection. Here, therefore, we first confirmed the reproducibility of rDNA instability in candidate mutants with a GO annotation unrelated to genome stability that were directly obtained from the updated yeast deletion collection. Because 15 of the 113 mutants were analyzed in our previous study (22), and 1 mutant could not be revived from the glycerol stock, the remaining 97 mutants were subjected to confirmation for their rDNA instability.
To assess the rDNA instability, we isolated genomic DNA from the candidate mutants and separated DNA by PFGE using a comb with wider teeth than those of the comb used in our previous screen to improve the sensitivity of our analysis. We then examined the size heterogeneity of chromosome XII, which carries the rDNA array ( Fig. 1B to E). The SIR2 gene encoding a histone deacetylase is known to promote rDNA stability (10,24,25); thus, in each gel, we included DNA from WT and the sir2D mutant, which exhibits an extremely smeared band of chromosome XII compared with that of WT cells (Fig. 1B to E). PFGE analysis showed that the chromosome XII band was prominently smeared in 15 mutants: ypt7D (Fig. 1B), bud6D, svl3D, clb5D, rho4D, pam1D, swm1D, pep7D (Fig. 1C), hcm1D, ldb17D, siw14D (Fig. 1D), bem2D, vrp1D, pho85D, and gon7D (Fig. 1E).
Changes in rDNA copy number are often associated with the production of extrachromosomal rDNA circles (ERCs), which are excised from genomic arrays. Many of the previously studied rDNA-unstable mutants, such as the sir2D mutant, accumulate ERCs at a level higher than that of WT cells (24,25). To examine the level of ERCs in the candidate mutant strains, we separated genomic DNA by conventional agarose gel electrophoresis to resolve ERCs from genomic rDNA and then detected ERCs and genomic rDNA by Southern blotting (Fig. 2). As a reference, genomic DNA samples from WT and the sir2D mutant were included in the analyses. ERCs were not clearly detected in one of the genomic DNA samples prepared from WT cells due to poor sample quality, and faint ERC bands were detected in the other WT sample (Fig. 2). Therefore, we could not compare the ERC level in each mutant to that in WT cells; instead, we focused our analysis on the novel mutants that exhibited extreme rDNA instability and compared the ERC level in each mutant to that in the sir2D mutant as a positive control (Fig. 2). The above-mentioned mutants that exhibited a prominently smeared band of chromosome XII in PFGE did not accumulate ERCs to a level comparable to that in the sir2D mutant ( Fig. 1 and 2). Instead, vac8D, chs7D, cdc10D ( Fig. 2C and D), hub1D ( Fig. 2I and J), and rim11D ( Fig. 2K and L) were the highest-ranked mutants and produced ERCs at levels comparable to the level in sir2D, although they did not show severe rDNA instability in PFGE analysis (Fig. 1). Collectively, these analyses enabled us to narrow down the number of genes to test further for confirmation of rDNA instability.
To examine the involvement of these 20 candidate genes in maintaining rDNA stability, we constructed deletion mutants of each gene in the W303 yeast strain background that differs from the BY strain background in which the yeast deletion collection was generated. For unknown reasons, it was not possible to construct mutants lacking CDC10, PAM1, PEP7, PHO85, or GON7. Among the remaining 15 mutants, only clb5D showed a smeared band of chromosome XII ( Fig. 3A and B). In this mutant, however, the ERC level was similar to that in WT cells ( Fig. 3E and F). None of the other 14 mutants produced ERCs at a level comparable to that in the sir2D mutant ( Fig. 3C to J). Taken together, these findings show that absence of Clb5 causes severe rDNA instability in yeast strains of two different genetic backgrounds, demonstrating that Clb5 plays an important role in maintaining rDNA stability.
A previous study showed that the clb5D mutant exhibits small defects in vacuolar fragmentation under various growth conditions (26). GO analysis thus associates the CLB5 gene with the biological processes of organelle assembly, organelle fission, and regulation of organelle organization, in addition to its best-known function as an Sphase cyclin required for the activity of CDK, which regulates the initiation and progression of DNA replication (15)(16)(17)(18).
Clb5 suppresses rDNA instability in response to Fob1-mediated replication fork arrest. To understand how Clb5 promotes rDNA stability, we examined whether rDNA instability in the clb5D mutant occurs in response to Fob1-mediated replication fork arrest at the RFB. To this end, we constructed a diploid strain heterozygous for clb5D and fob1, induced meiosis, isolated haploid clones by tetrad dissection, and examined rDNA stability in WT, clb5D, fob1, and fob1 clb5D cells by PFGE (Fig. 4A). To assess the degree of rDNA instability in clb5D cells, we simultaneously analyzed DNA samples isolated from sir2D cells in parallel. All six independent clb5D clones showed smearing of the chromosome XII band compared with the band of WT clones (Fig. 4A). To demonstrate qualitative differences in rDNA stability, the signal intensities in each lane were determined in the region from ;1.81 to 3.13 Mbp and normalized to the maximum value of signals derived from chromosome IV as a loading control (Fig. 4A, right). The representative clb5D mutant clone displayed a peak chromosome XII signal with a broader width and lower height compared with that of the WT clone (Fig. 4A, right, black and red lines). The peak of the chromosome XII signal in sir2D, however, was   sir2∆  atg5∆  rny1∆  mkc7∆  xrs2∆  vac8∆  cin8∆  gos1∆  ksp1∆  chs7∆  bik1∆  ste50∆  cdc10∆  rsc1∆  sey1∆  WT  sir2∆  flc2∆  lds1∆  bni1∆  bud28∆  ypt7∆  csi1∆  far8∆  cik1∆  fus2∆  scw10∆  vps21∆  scp1∆  gal11∆    substantially broader than that in clb5D (Fig. 4A, right, red and blue lines). Therefore, absence of Clb5 results in rDNA instability but not to the extent seen in the sir2D mutant.
Smearing of the chromosome XII band in the clb5D mutant was largely suppressed by the introduction of a fob1 mutation (Fig. 4A); however, the chromosome XII band in the fob1 clb5D mutant was still broader than that in the fob1 single mutant (Fig. 4B, right, black and red lines). This phenotype was not observed for the sir2D mutant, in which extreme smearing of the chromosome XII band was nearly completely suppressed to the level seen in the fob1 mutant (Fig. 4B, right, black and blue lines). The partial suppression of rDNA instability by fob1 mutation is not specific to the clb5D mutant; we previously reported that fob1 mutation does not completely suppress rDNA instability in cells lacking Ctf4, a component of the replisome (19). Interestingly, there seemed to be two classes of fob1 clb5D clones among six independent clones examined, one that showed a broader chromosome XII band compared with that of fob1 clones (first, third, fourth, and sixth lanes of fob1 clb5D samples) and another that showed a much sharper chromosome XII band (second and fifth lanes of fob1 clb5D samples) (Fig. 4B). The fob1 clb5D clones in the latter class had a smaller chromosome XII band than that of the clones in the former class (Fig. 4B). The biological significance of these two classes is discussed below. Collectively, these results suggest that rDNA instability in the clb5D mutant mostly depends on Fob1.
We also assessed rDNA instability by ERC analysis ( Fig. 4C and D). The sir2D mutant accumulated ERCs at a level ;6-fold higher than that in WT cells. Although the clb5D mutant showed rDNA instability in PFGE, the level of ERCs in the clb5D mutant was comparable to that in WT (Fig. 4D), as observed in Fig. 3E and F. Thus, unlike many previously characterized rDNA-unstable mutants such as sir2D, the absence of Clb5 does not cause an accumulation of ERCs. Consistent with our PFGE analysis (Fig. 4A), the level of ERCs was reduced by 2-fold in the fob1 clb5D mutant relative to that in the clb5D mutant ( Fig. 4C and D); however, the level of ERCs in the fob1 clb5D mutant was still ;3to 5-fold higher than that in the fob1 and fob1 sir2D mutants, although these differences were not statistically significant ( Fig. 4C and D). Thus, in the absence of Clb5, ERCs are produced predominantly in a Fob1-dependent manner, but some are generated independently of Fob1. Taken together, these results suggest that Clb5 functions to maintain rDNA stability mainly by promoting the proper response to Fob1-mediated replication fork arrest, but it may also prevent the generation of Fob1-independent, rDNA-destabilizing DNA damage in non-RFB regions.
rDNA instability in the clb5D mutant is suppressed by deletion of its paralog CLB6. CLB5 has a paralog, CLB6, whose product also acts as an S-phase cyclin. Whereas deletion of CLB5 causes lengthening of S phase, deletion of CLB6 has little effect on Sphase duration (16)(17)(18). In the clb5 clb6 double mutant, the onset of S phase is delayed, but the length of S phase is restored to that seen in WT (18). Thus, Clb5 and Clb6 differentially regulate the initiation and progression of S phase.
Next, we sought to understand how Clb5 and Clb6 coordinately regulate rDNA stability. To this end, we constructed a diploid strain heterozygous for clb5D and clb6D, induced meiosis, isolated haploid clones by tetrad dissection, and examined rDNA stability in WT, clb5D, clb6D, and clb5D clb6D cells. In PFGE, smearing of the chromosome XII band was seen in four independent clb5D mutant clones examined but not in the clb6D mutant clones (Fig. 5A). In contrast to the clb5D single mutant, the clb5D clb6D double mutant showed a homogeneous band of chromosome XII, comparable to that in WT and the clb6D single mutant (Fig. 5A). Suppression of smearing of the chromo-       The sum of monomers and dimers was determined relative to genomic rDNA, which was normalized to the average of the WT clones (bars show mean 6 SD). One-way analysis of variance (ANOVA) was used for multiple comparisons. Asterisks indicate a significant difference at P , 0.05; ns indicates no significant difference (P . 0.05). a.u., arbitrary units. some XII band in the clb5D mutant by clb6D was also evident when we compared the shape of the chromosome XII band in the signal profile of the lanes: the clb5D clone showed a broader and shorter peak of chromosome XII signal relative to that of WT and clb6D, but the peak became narrower in the clb5D clb6D clone and comparable to that seen in WT and the clb6D single mutant (Fig. 5A, right). There were no differences in ERC level among WT, clb5D, clb6D, and clb5D clb6D cells ( Fig. 5B and C). Thus, rDNA instability in the clb5D mutant is suppressed by deletion of CLB6. Clb5 regulates replication initiation in the rDNA region. Previous studies have demonstrated that Clb5 and Clb6 differentially regulate the timing of replication origin firing in non-rDNA regions; that is, Clb5 and Clb6 both promote firing of early replication origins, but only Clb5 activates late origins (15,18). Absence of both Clb5 and Clb6 delays entry into S phase but restores the length of S phase and origin firing to WT levels (15)(16)(17)(18). To understand how Clb5 and Clb6 regulate replication in the rDNA region, we assessed the frequency of origin firing in this region in the clb5D and clb6D mutants (Fig. 6A to D).
To this end, we first grew cells to stationary phase, diluted them into fresh media, and allowed them to grow until they reentered the cell cycle, enabling us to obtain a The sum of monomers and dimers was determined relative to genomic rDNA, which was normalized to the average of WT clones (bars show mean 6 SD). One-way ANOVA was used for multiple comparisons. There were no statistically significant differences among the strains.  sample enriched in replicating cells. We then prepared genomic DNA, digested it with the restriction enzyme NheI, which has two recognition sites within each rDNA unit (Fig. 6A), and performed two-dimensional (2D) agarose gel electrophoresis. DNA molecules were separated by molecular mass in the first dimension, followed by mass and shape in the second dimension. DNA was transferred to a nylon membrane and analyzed by Southern blotting with a probe that hybridizes to the NheI fragment containing an origin of DNA replication and an RFB site (Fig. 6A).
In this analysis, DNA molecules where replication is initiated from an origin are detected as bubble-shaped molecules, while those where replication passively progresses through the NheI fragment generate a Y arc (Fig. 6B) (27). In addition, replication forks that are paused at the RFB site within the restriction fragment produce an intensive signal along the Y arc (Fig. 6B). We assessed the frequency of origin firing by determining the proportion of signal for bubble-shaped molecules relative to the total signal for replication intermediates (Fig. 6B and C). The efficiency of origin firing in the clb5D mutant was lowered by half relative to that of WT (Fig. 6D), demonstrating that Clb5 promotes firing of replication origins in the rDNA. The reduced origin firing observed in the clb5D mutant was restored in the clb5D clb6D double mutant to almost WT levels (Fig. 6D). Together with the finding that rDNA instability in the clb5D mutant was suppressed in the clb5D clb6D mutant (Fig. 5), these results indicate that rDNA stability is influenced by the efficiency of replication initiation.
Replication forks are more stably arrested at the RFB site in the absence of Clb5. To understand how reduced origin firing in the clb5D mutant leads to rDNA instability, we examined whether absence of Clb5 influences the frequency of replication fork arrest. More than 90% of replication forks that initiate from replication origins stall at the RFB site (9). Because fewer origins fired in the absence of Clb5 (Fig. 6D), we expected that fewer forks would be stalled at the RFB site in the clb5D mutant than in WT cells. We assessed the frequency of arrested forks by determining the proportion of signal for arrested fork intermediates relative to the total signal for replication intermediates in our 2D analysis ( Fig. 6B and C). The fork blocking activity was comparable among WT, clb5D, clb6D, and clb5D clb6D cells (Fig. 6E). This finding implies that, although fewer forks are stalled at the RFB in the clb5D mutant than in WT cells, these forks are more stably arrested, resulting in a higher level of persistently arrested forks.
Replication fork arrest at the RFB site leads to the formation of DSBs, which are thought to be a major trigger of genome instability (2). It had long been thought that DSBs formed at arrested forks are repaired by homologous recombination; however, our previous study demonstrated that DSBs are rarely resected and their repair does not require recombination proteins in WT cells as long as the cells carry a normal rDNA copy number (19). Thus, changes in rDNA copy number can be dictated by the frequency of DSB end resection.
To determine the frequency of DSBs and resected DSBs formed at the RFB site, we digested DNA with the restriction enzyme BglII and separated it by single-dimension agarose gel electrophoresis, followed by Southern blotting (Fig. 6A and H). We assessed the DSB frequency by determining the ratio of signal for DSBs to that for arrested fork intermediates ( Fig. 6H and I). DSB levels were comparable among WT, clb5D, clb6D, and clb5D clb6D cells (Fig. 6I). Resected DSBs in the clb5D mutant were below the detection limit of Southern blotting, but this result indicates that absence of Clb5 does not seem to cause a substantial increase in DSB end resection (Fig. 6H). Thus, Clb5 does not influence the formation or repair of DSBs. The clb5D mutant induces homologous recombination-mediated rDNA instability. In 2D gel analysis, we also detected replication forks converging from both sides at the RFB, as well as Holliday junction recombination intermediates that are formed during homologous recombination-mediated repair of DSBs (27). Both intermediates generate a double Y spot signal on top of the spike signal and were seen in a similar position in our 2D gel ( Fig. 6B and C). As shown in Fig. 6C, the double Y spot signal in the clb5D mutant seemed to be stronger than that in WT, clb6D, or clb5D clb6D cells. As a proportion of total replication intermediate signals, however, the double Y spot signal was similar among these cells (Fig. 6F), possibly because more replication intermediates may exist in clb5D, which has a longer S phase. In contrast, the ratio of double Y spot to bubble signals was clearly increased in the clb5D mutant (Fig. 6G). Because the arrested fork signal was not elevated (Fig. 6E) and the bubble signal was reduced (Fig.  6D) in clb5D relative to that in WT, clb6D, and clb5D clb6D cells, the increased signal ratio of double Y spots in the clb5D mutant is most probably due to a higher level of recombination intermediates but not converging replication forks. Collectively, these results indicate that recombination frequency is increased in the clb5D mutant.
We examined whether this elevation of the double Y spot signal is unique to the clb5D mutant or a similar phenotype is observed in other replication initiation-defective mutants. To this end, we examined the rARSD-3 strain, in which all the endogenous rARS sequences lack the ARS consensus sequence III element (28). The rARSD-3 strain exhibits a 50% reduction in replication initiation activity compared with that of a strain with the intact rARS sequence (28). We compared the patterns of replication intermediates by performing 2D gel analysis of samples prepared from WT, clb5D, and rARSD-3 cells in parallel. The rARSD-3 cells also displayed a strong double Y spot signal compared with that of WT cells (Fig. 6J). The ratio of double Y signal to bubble signal was 2-fold higher in the rARSD-3 strain than in WT cells, and this ratio was comparable to or slightly lower than that in the clb5D mutant, although the differences were not statistically significant (Fig. 6K). These findings raise the possibility that replication initiation defects in the rDNA lead to an accumulation of recombination intermediates.
In PFGE analysis, smearing of the chromosome XII band in the clb5D mutant was suppressed by deletion of RAD52, which is essential for homologous recombination (Fig. 7A). The clb5D mutant showed an increase in ERCs relative to those in WT cells ( Fig. 7B and C); however, the increase was most probably biologically insignificant because this phenotype was not reproducibly observed in other ERC assays (Fig. 3 to  5). Moreover, the fourth sample of the clb5D clones showed a substantially higher level of ERCs, as well as a sharper chromosome XII band in PFGE, compared with that in the other samples (Fig. 7). Thus, we believe that this clone is most probably an outlier, leading to an artificial increase in the average level of ERCs in the clb5D mutant in this assay (Fig. 7C). ERC production in the clb5D mutant was suppressed in the clb5D rad52D mutant to the level seen in the rad52D mutant (Fig. 7C). These results suggest that homologous recombination-dependent rDNA instability is induced in the clb5D mutant.

DISCUSSION
In this study, we confirmed the reproducibility of rDNA instability in 97 previously identified rDNA-unstable mutants that lack genes with a GO annotation that is seemingly unrelated to genome stability, for example, genes functioning in cell and organelle morphogenesis (13). We demonstrated that a mutant lacking the CLB5 gene exhibits rDNA instability. Based on GO analysis, CLB5 is associated with the biological processes of organelle fission and regulation of organelle organization, because its deletion causes small defects in vacuolar fragmentation under various growth conditions (26). However, the mechanism by which CLB5 regulates vacuolar fragmentation remains unknown. The CLB5 gene and its paralog CLB6 encode S-phase cyclins required for the activity of Cdc28 and are involved in the initiation and progression of DNA replication (15)(16)(17)(18). In this study, we have revealed that Clb5 plays an important role in suppressing homologous recombination-dependent rDNA instability, mainly in response to Fob1-mediated replication fork arrest and partially in response to DNA damage at non-RFB sites, by promoting replication initiation in the rDNA region.
The initiation of replication across the genome is under temporal control. In non-rDNA regions, Clb5 and Clb6 function redundantly to promote the timely activation of early firing origins, while late origin firing is regulated only by Clb5 because Clb6 is expressed in the early S phase (15,29). In the rDNA region, each rDNA copy contains a replication origin sequence with the potential to initiate replication, but only a small proportion of these origins fire during rDNA replication (6). We found that deletion of the CLB5 gene results in a 50% reduction in replication initiation efficiency during exponential growth compared with that of WT cells, whereas deletion of CLB6 has little effect on replication initiation (Fig. 6C and D). Thus, Clb5 plays a dominant role in the activation of replication origins in the rDNA region. Although we did not examine the timing of origin activation, we speculate that, in the absence of Clb5, late origin firing is defective in the rDNA region (Fig. 8).
We previously demonstrated that replication initiation activity influences rDNA stability (28). When a substantial number of rDNA repeats is lost, cells enter an "rDNA expansion" mode until they regain the normal copy number; this is referred to as "gene amplification" and is thought to compensate for the lost copies of rDNA that are necessary for rRNA production (30). To examine whether replication initiation activity is required for this gene amplification, we constructed an rDNA two-copy strain in which the ARS consensus sequence III was deleted. The resulting rARSD-3 strain exhibits lower replication initiation activity in the rDNA region and, relative to the two-copy strain carrying an intact ARS sequence, shows a lower rate of rDNA amplification, suggesting that reduced replication initiation activity is less effective for gene amplification (28). On the other hand, the rARSD-3 strain shows ;6-fold higher frequency of marker gene loss from the rDNA array than that of WT cells. Thus, rARSD-3 cells display enhanced rDNA instability in a marker gene loss assay but reduced rDNA instability in gene amplification. The reasons for these discrepancies remain unclear. It is possible that origin firing activity differentially regulates rDNA stability, depending on the number of rDNA repeats carried by the cells; in other words, origin firing activity promotes rDNA instability during gene amplification from an rDNA low-copy-number strain, but efficient origin activity stabilizes rDNA when cells have a normal rDNA copy number. We have not examined the rate of gene amplification for an rDNA low-copy-number strain in the clb5D mutant; however, our finding that the clb5D mutant with reduced replication initiation activity shows rDNA instability is consistent with the phenotype of increased marker gene loss seen in the rARSD-3 strain. Replication initiation activity in the rDNA is compromised by mutations in genes encoding components of the origin recognition complex (ORC) and depletion of the  Sld2 and Sld3 proteins (31)(32)(33). These conditions cause rDNA instability that specifically leads to a reduction in rDNA copy number. Replication initiation defects cause each replication fork to replicate over a longer distance, prolonging S phase to complete rDNA replication. Therefore, it has been suggested that these defects confer a selective advantage on cells with fewer rDNA copies over cells with a normal copy number that have a region of ;1.3 Mbp to replicate (33). This suggestion may explain our observation that low rDNA copy number strains in the fob1 clb5 mutant had sharper bands of chromosome XII relative to those of normal rDNA copy number strains (Fig. 4B). The clb5D cells also showed reduced origin firing activity but exhibited both expansion and contraction of rDNA repeats (Fig. 3-7). We envisage that the efficiency of replication initiation is compromised for both early-and late-firing origins in orc mutants and in cells depleted of Sld2 and Sld3. In contrast, deletion of CLB5 alters the timing of origin activation: early firing origins are activated at the normal time, but late-firing origins are inefficiently activated. The differential impact of mutations on the overall efficiency of origins and replication timing may cause differences in the pattern of rDNA copy number changes, a possibility that should be addressed in future studies. Previous studies have demonstrated that the prolonged S phase and late origin firing defects in the non-rDNA regions seen in the clb5D mutant are suppressed in the clb5D clb6D mutant, although the double mutant has delayed entry into S phase (15,18). It has been proposed that mitotic cyclins may promote S phase in the absence of both Clb5 and Clb6 (15). The rDNA instability and replication initiation defects observed in the clb5D mutant were also suppressed in the clb5D clb6D mutant (Fig. 5 and 6), thereby suggesting that Clb6 either represses late origin firing or is not sufficient to activate late origins in the rDNA region in the absence of Clb5. These findings demonstrate that the proper activation of late-firing origins may be critical for maintaining rDNA stability.
Our study demonstrates that rDNA instability in the clb5D mutant occurs in a manner dependent on a homologous recombination protein, Rad52 (Fig. 7). What might be the mechanism underlying the recombination-mediated rDNA instability in the absence of Clb5? Once replication is initiated, the majority of replication forks (.90%) are immediately stalled at the nearest RFB site in a polar fashion (7,9,34). Thus, we expected that the clb5D mutant, which has fewer active origins, would generate fewer forks arrested at the RFB site. Unexpectedly, however, the level of arrested forks in the mutant lacking CLB5 was comparable to that in WT cells (Fig. 6E), indicating that replication forks are more stably stalled at the RFB in the absence of Clb5. Previous studies have shown that active replication origins in the rDNA are clustered and separated by large gaps without active origins (35)(36)(37). A stalled replication fork can be resolved by the arrival of the converging fork proceeding from the nearest origin (3). For the clb5D mutant, which has fewer firing origins, the converging fork must come from an origin that is further away (Fig. 8). In the absence of Clb5, therefore, more forks remain arrested for a longer time, which is the initiating event for RFB-dependent rDNA instability (Fig. 8).
The replication termination sequence RTS1 induces site-specific fork stalling in fission yeast (1). Previous studies have demonstrated that genome rearrangements are induced by homologous recombination-mediated replication fork restart at stalled forks at the RTS1 site when the arrival of the converging fork is inhibited by the presence of another RTS1 site, but this process does not involve DSB formation (38,39). It remains unknown whether replication forks arrested at the RFB site in budding yeast rDNA undergo similar genome rearrangements induced by replication fork restart. Nonetheless, absence of Clb5 delays the arrival of the converging fork by impairing the firing of upstream late origins, and as a result, the forks arrested at the RFB site persist longer (Fig. 8). These forks may undergo DSB-independent faulty template switches to restart DNA replication, leading to rDNA instability.
The other consequence of persistent replication fork arrest is the generation of DSBs, which have the potential to induce genome rearrangements when not repaired properly by equal sister chromatid exchange (40). We found, however, that the level of DSBs at the RFB site was similar between cells with Clb5 and cells lacking Clb5 (Fig. 6I). Several factors can influence the outcome of DSB repair. First, one of the intergenic regions of the rDNA unit contains an RNA polymerase II-dependent, bidirectional promoter, named E-pro, upstream of the RFB site (41), and transcription of noncoding RNA from E-pro may induce the dissociation of cohesin from rDNA repeats, leading to unequal sister chromatid recombination that alters the rDNA copy number during DSB repair (10,41). Thus, Clb5 may suppress rDNA instability by regulating the transcription of noncoding RNA from E-pro and/or cohesin association, two possibilities that need to be examined in future studies. The occurrence of rDNA instability may also depend on the decision of whether cells undergo DSB end resection and repair by homologous recombination (19). In both WT and clb5D, resected DSBs were below the detection limit of our Southern blotting (Fig. 6H). Therefore, it seems unlikely that absence of Clb5 induces DSB end resection at the RFB.
In our 2D analysis, the double Y spot signal was stronger in the clb5D mutant than in WT, clb6D, or clb5D clb6D cells (Fig. 6C). Although we did not see a difference in the ratio of double Y signal to total replication intermediates among these strains, the ratio of double Y signal to bubble arc signal was increased in clb5D but not in the other strains (Fig. 6G). As mentioned above, this signal most probably corresponds to recombination intermediates. Therefore, DNA damage might occur more frequently in clb5D than in the other strains, which might explain why only clb5D shows rDNA instability among the mutants tested. Using 2D analysis, previous studies have analyzed replication intermediates in cells with reduced origin firing efficiency, including the rARSD-3 strain, the orc4 Y232C mutant that mimics the mutation carried by patients with Meier-Gorlin syndrome, and cells depleted of the replication initiation factors Sld2 and Sld3 (28,32,33). Those cells all show a substantial reduction in bubble arc signal compared with that of WT cells or a control condition. We analyzed the rARSD-3 strain in this study and demonstrated that this mutant also displays an increase in the ratio of double Y signal to bubble arc signal ( Fig. 6J and K). Depletion of Sld2 and Sld3 does not lead to an alteration of double Y signals relative to linear molecules, compared with the control condition (32); however, because the bubble arc signal is reduced, the ratio of double Y signal to bubble arc signal is increased in these strains. Whether a similar phenotype is seen for the orc4 Y232C mutant should be investigated in a future study. Overall, it seems that an increase in recombination intermediates may reflect features associated with reduced origin activity.
What increases the damage in the clb5D mutant? Because the signals for arrested forks, DSBs, and resected DSBs at the RFB site were not increased in the clb5D mutant (Fig. 6E, H, and I), we speculate that DNA damage occurs at non-RFB sites in the rDNA region in clb5D. In fact, although in PFGE the chromosome XII band in the clb5D fob1 double mutant was sharper than that in the clb5D single mutant, it was still a little broader than that in the fob1 single mutant (Fig. 4A and B). Moreover, in the ERC assay, more ERCs were detected in the clb5D fob1 double mutant than in the fob1 mutant ( Fig. 4C and D). These results suggest that some FOB1 (RFB)-independent recombination occurs in clb5D. One possibility for triggering this recombination is DNA damage caused by the reduced initiation of replication in the clb5D mutant (Fig. 8). It is known that a longer distance between replication origins induces more genome instability, making a site fragile (42). Moreover, Sanchez et al. observed a broken chromosome XII in the orc4 Y232C mutant with reduced replication initiation activity (33), although it remains to be determined whether this break occurs during S phase as a consequence of replication problems. It has been speculated that the long-lasting forks have more time to cause problems, such as accidental fork arrest, during the course of travel between origins. A similar situation may occur in the rDNA of the clb5D mutant. Moreover, a long-lasting fork is expected to make bigger replication bubbles. This might enhance unequal sister chromatid recombination, which would contribute to the rDNA copy number alteration seen in the clb5D mutant. Further analysis is required to reveal the details of how long-lasting forks cause problems and their resulting DNA damage.

MATERIALS AND METHODS
Yeast strains, growth conditions, and genomic DNA preparation. The mutant strains used in Fig.  1 and 2 are derivatives of the BY4741 background (MATa his3D1 leu2D0 met15D0 ura3D0) and obtained from the Yeast Knockout Collection (YSC1053; Open Biosystems [now at Horizon Discovery]) (43)(44)(45). The WT strain used in Fig. 1 and 2 was BY4741. The other strains used in this study were derived from NOY408-1b, which is in the W303 background (MATa ade2-1 ura3-1 his3-11,15 trp1-1 leu2-3,112 can1-100). Mutant strains in which genes of interest were deleted were constructed by a standard one-step gene replacement method, followed by PCR-based genotyping. The mutant strains used in Fig. 3 were constructed by replacing the open reading frame of the gene of interest with the kanMX marker in the NOY408-1b strain. The sir2D and fob1 sir2D strains used in Fig. 4 were constructed by replacing the SIR2 gene with the kanMX marker. The rARSD-3 strain used in Fig. 6J and K was TAK209F, which was constructed in a previous study (28). The other haploid strains used in Fig. 4 to 7 were obtained by constructing diploid strains heterozygous for clb5D::kanMX and fob1::LEU2 (Fig. 4), clb6D::hphMX ( Fig. 5 and  6), or rad52D::hphMX (Fig. 7), followed by tetrad dissection.
For the PFGE and extrachromosomal rDNA circle (ERC) analyses in Fig. 1 and 2, yeast strains were patched from their glycerol stock onto yeast extract-peptone-dextrose (YPD) plates (1% [wt/vol] yeast extract, 2% [wt/vol] peptone, 2% [wt/vol] glucose, and 2% [wt/vol] agar), and the bulk of cells were grown in 5 ml of YPD medium overnight at 30°C. For other PFGE and ERC analyses, yeast strains were streaked onto YPD plates, a single colony was then inoculated into 5 ml of YPD medium, and cells were grown overnight at 30°C. Cells (5 Â 10 7 cells/plug) were collected and washed twice with 50 mM EDTA (pH 7.5).
For cells subjected to two-dimensional (2D) and DSB analyses, a single colony was inoculated into 5 ml of YPD medium and grown overnight at 30°C until the culture reached the saturation phase of growth. The cells were then inoculated into 100 ml of YPD medium at an optical density at 600 nm (OD 600 ) of 0.1 and grown at 30°C until they reached an OD 600 of 0.4. The cells were immediately treated with 0.1% sodium azide and then collected (5 Â 10 7 cells/plug) and washed twice with 50 mM EDTA (pH 7.5). For PFGE, ERC, 2D, and DSB analyses, genomic DNA was prepared in low-melting-temperature agarose plugs as described previously (19).
PFGE analysis. One-third of an agarose plug, along with Hansenula wingei chromosomal DNA markers (Bio-Rad), was separated by electrophoresis on a 1.0% agarose gel (pulsed-field certified agarose, Bio-Rad) in 0.5Â Tris-borate-EDTA (TBE) buffer (44.5 mM Tris base, 44.5 mM boric acid, and 1 mM EDTA [pH 8.0]) in a Bio-Rad contour-clamped homogeneous electric field DR-III system using the following conditions: 68 h at 3.0 V/cm, 120°included angle, linear ramp from 300 s of initial switch time to 900 s of final switch time. The gel was stained with 0.5 mg/ml of ethidium bromide and photographed.
ERC assay. One-half of an agarose plug, along with 500 ng of lambda HindIII DNA markers, was separated by electrophoresis on a 0.4% agarose gel (15 by 25 cm gel) in 1Â Tris-acetate-EDTA (40 mM Tris base, 20 mM acetic acid, and 1 mM EDTA [pH 8.0]) at 1.0 V/cm for ;48 h at 4°C with buffer circulation in a Sub-cell GT electrophoresis system (Bio-Rad). The buffer was changed every ;24 h.
DNA was transferred to Hybond-XL (GE Healthcare). Southern blotting was then performed with a probe prepared by PCR amplification of genomic DNA using primers 59-CATTTCCTATAGTTAACAGGACATGCC and 59-AATTCGCACTATCCAGCTGCACTC, as described previously (19). The membrane was exposed to phosphor screens for an appropriate amount of time before any signals were saturated, and the radioactive signal was detected using Typhoon FLA 7000 (GE Healthcare). The membrane was reexposed to the phosphor screen for several days and scanned. The scanned images taken after short and long exposures were used to quantify genomic rDNA and ERC bands, respectively, using FUJIFILM Multi Gauge version 2.0 software (Fujifilm). The ratio of ERCs relative to genomic rDNA was determined.
2D gel electrophoresis. 2D gel electrophoresis was performed as described previously with slight modifications (46). In brief, one-half of an agarose plug was placed in a 2-ml flat-bottom tube. The plug was equilibrated twice in 1 ml of 1Â M buffer (TaKaRa) by rotating the tube for 30 min at room temperature. After discarding the buffer completely, the plug was incubated in 160 ml of 1Â M buffer containing 160 units of NheI (TaKaRa) for 7 h at 37°C. The plug and 600 ng of lambda HindIII DNA markers were separated by electrophoresis on a 0.4% agarose gel (SeaKem Agarose LE; Lonza) in 1Â TBE buffer at 1.32 V/ cm for 14 h at room temperature with buffer circulation in a Sub-cell GT electrophoresis system (15 by 20 cm gel; Bio-Rad). The gel was stained with 1Â TBE buffer containing 0.3 mg/ml of ethidium bromide and photographed. Gel slices containing DNA ranging from 4.7 to 9.4 kb were excised, rotated 90°, and cast in a 1.2% agarose gel (SeaKem Agarose LE; Lonza) containing 0.3 mg/ml of ethidium bromide in 1Â TBE. The second-dimension gel electrophoresis was performed in 1Â TBE buffer containing 0.3 mg/ ml of ethidium bromide at 6.0 V/cm for 5 h at 4°C with buffer circulation in a Sub-cell GT electrophoresis system (Bio-Rad). DNA was transferred to Hybond-XL (GE Healthcare).
Southern blotting was performed with a probe prepared by PCR amplification of genomic DNA using primers 59-CATTTCCTATAGTTAACAGGACATGCC and 59-AATTCGCACTATCCAGCTGCACTC, as described previously (19). The membrane was exposed to phosphor screens for several days and the radioactive signal was detected using Typhoon FLA 7000 (GE Healthcare). ImageJ (NIH) was used to quantify bubbles, Y arcs containing RFB spots, RFB spots, and double Y spots.
DSB assay. The DSB assay was performed as described previously (19). In brief, one-third of an agarose plug was placed in a 2-ml flat-bottom tube. The plug was equilibrated four times in 1 ml of 1Â Tris-EDTA (TE; 10 mM Tris base [pH 7.5] and 1 mM EDTA [pH 8.0]) by rotating the tube for 15 min at room