|Home | About | Journals | Submit | Contact Us | Français|
Eukaryotes replicate DNA once and only once per cell cycle due to multiple, partially overlapping mechanisms efficiently preventing reinitiation. The consequences of reinitiation are unknown. Here we show that the induction of rereplication by mutations in components of the prereplicative complex (origin recognition complex [ORC], Cdc6, and minichromosome maintenance proteins) causes a cell cycle arrest with activated Rad53, a large-budded morphology, and an undivided nucleus. Combining a mutation disrupting the Clb5-Orc6 interaction (ORC6-rxl) and a mutation stabilizing Cdc6 (CDC6ΔNT) causes a cell cycle delay with a similar phenotype, although this background is only partially compromised for rereplication control and does not exhibit overreplication detectable by fluorescence-activated cell sorting. We conducted a systematic screen that identified genetic requirements for the viability of these cells. ORC6-rxl CDC6ΔNT cells depend heavily on genes required for the DNA damage response and for double-strand-break repair by homologous recombination. Our results implicate an Mre11-Mec1-dependent pathway in limiting the extent of rereplication.
Eukaryotic DNA replication must occur only once per cell cycle. This control is enforced at the level of prereplicative complex (pre-RC) formation and origin firing. In the budding yeast Saccharomyces cerevisiae, the origin recognition complex (ORC), composed of six subunits (Orc1 to -6), is constitutively bound to defined DNA replication origin sequences (12). The ORC complex provides a platform for pre-RC assembly, which requires Cdc6- and Cdt1-mediated recruitment of the minichromosome maintenance proteins (MCM) Mcm2 to -7 (3, 7, 10, 58, 59). The assembled pre-RC licenses the origin for recruitment of Cdc45, DNA polymerase alpha, Sld2, and other factors required for firing (initiation) and elongation (reviewed in reference 5). Origin firing requires elevated B-type cyclin-Cdk1 activity (normally Clb5-Cdk1 and Clb6-Cdk1) and Cdc7-Dbf4 activity (13, 14, 19, 52, 53). Sld2 is an essential target of the Clb5/6-Cdk1 activity promoting replication initiation (35).
Several mechanisms are known to contribute to the prevention of origin refiring, and all of them act by inhibiting pre-RC formation. In S. cerevisiae, the phosphorylation of Cdc6 by Cdk1 accelerates its SCFCdc4-mediated ubiquitination and degradation by the 26S proteasome (6, 15-17). The phosphorylation of MCMs by Cln-Cdk1 and Clb-Cdk1 kinases promotes their exit from the nucleus (31, 44). Phosphorylation of Orc2 and Orc6 by Clb-Cdk1 is thought to prevent efficient binding of other pre-RC subunits (45). Lastly, the binding of Clb5 to Orc6 via its RXL cyclin-binding motif contributes to the prevention of rereplication (68). Combining disruptions of these mechanisms allows extensive rereplication when cells are arrested with nocodazole (45, 68). In mammalian cells, rereplication can be induced by the overexpression of Cdc6, Cdt1, and cyclin A-Cdk2 (64). In mammalian cells and in Drosophila, geminin inhibits Cdt1 during late stages of the cell cycle, and the inhibition of geminin leads to rereplication (37, 40, 73). In the fission yeast Schizosaccharomyces pombe, the overexpression of Cdc18 (Cdc6) (42) or the disruption of an interaction between the B-type cyclin Cdc13-Cdc2 (69) kinase and Orc2 leads to successive rounds of replication without an intervening mitosis (endoreduplication).
The existence of multiple mechanisms preventing rereplication suggests that any one mechanism is inefficient (leaky) and that the cell is highly intolerant to even very rare events leading to DNA rereplication. Rereplication could lead to DNA damage, to genomic instability, and to gene amplification, which are hallmarks of cancer. In mammalian cells, the activation of a p53-dependent pathway (64, 74) and the G2/M checkpoint (73) have been implicated in protecting cells against rereplication.
For this study, we used the budding yeast system to characterize the response of cells to a compromise of rereplication controls.
Standard methods were used for mating, tetrad analysis, and transformations. The LEU2::ORC6 (-wt, -ps, -rxl, and -ps,rxl), MCM7-NLS, and ORC2-ps alleles were described previously (45, 68). The URA3::GAL-CDC6ΔNT-HAm and URA3::GAL-CDC6ΔNT-HAs (Δ2-48) alleles were obtained by transforming a pRS305-based plasmid as described previously (68) and were screened by Southern blotting for multiple or single integration. One individual single integrant allele was used to construct the strains used for the experiments presented in Fig. Fig.3,3, ,5,5, and and66 and in Fig. S1 in the supplemental material. One individual multiply integrant allele (unknown copy number of GAL-CDC6ΔNT-HA; since all phenotypes were reproducibly observed in multiple meiotic segregants, we assumed that this copy number was effectively constant) was used to construct the strains used for the experiments presented in Fig. Fig.1.1. For historical reasons, a different multiply integrant allele was used to construct the strains used for the experiments presented in Fig. Fig.77 and in Fig. S2 in the supplemental material. The CDC6ΔNT (Δ2-49) allele used for Fig. Fig.55 and for the tetrad analyses presented in Table S2 in the supplemental material was under the control of the endogenous CDC6 promoter, as described previously (68). Other alleles used were RAD9::LEU2 (for tetrad analyses presented in Table S2 in the supplemental material; lab stock), 6FLAG-RAD53 (from J. Petrini), mre11::URA3 (from J. Petrini), mec1::TRP1 (with sml1::HIS3), rad53::LEU2 (with sml1::URA3), mec1-100::LEU2::mec1::HIS3 (all from J. Diffley), mrc1::HIS3, HIS::mrc1AQ-MYC (both from S. Elledge), sgs1::TRP1, mms4::KanMX, and mus81::KanMX (all from S. Brill) in the W303 background. All other gene deletions were marked by KanMX and obtained from the strain collection in the BY4741 background (MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0) (Invitrogen, Carlsbad, CA).
For the induction of rereplication, cells were grown in YPD medium overnight, washed, and transferred to YPR (raffinose) medium for 4 h. The GAL-CDC6ΔNT-HAm allele was then induced by the addition of galactose (3% final concentration).
Our genetic screen was performed using a method modified from a published method (62). The query strains, with the genotype MATα ORC6-rxl::LEU2 URA3::GAL-CDC6ΔNT-HAs mfa::MFA1pr-HIS3 trp1 ade2 can1 leu2 his3 lys2 ura3, were spread onto yeast extract-peptone-dextrose (YEPD) in rectangular plates, the deletion mutant arrays (genotype MATa TRP1 ADE2 met15 leu2 ura3 his3 geneX::kanMX) (DMA) were pinned on top of the query cells, and the plates were incubated for 1 day. The resulting MATa/α zygotes were pinned onto selection medium (SD-Min plus His) to allow the growth of diploid cells, which were then pinned to sporulation plates (sporulation medium plus His) and incubated for 5 days at 22°C. The spores were pinned on to haploid selection medium (SD-His/Ura/Leu/Arg plus canavanine) to select haploid MATa mfa::MFA1pr-HIS3 ORC6-rxl::LEU2 URA3::GAL-CDC6ΔNT-HAs spore progeny (the selection for a-mating-type haploids was accomplished with the mfa::MFA1pr-HIS3 cassette, which results in histidine autotrophy in MATa haploids only) (62). The MATa haploids were then pinned onto YPD medium containing G418 and grown for 1 day (one-half of the meiotic progeny should carry the kanMX G418-R-marked deletion mutation from the DMA strain). Finally, double mutants were placed on either SD-His/Ura/Leu/Arg plus canavanine plus G418 or SG-His/Ura/Leu/Arg plus canavanine plus G418 for 2 days, and the proliferation of the haploid ORC6-rxl GAL-CDC6ΔNT-HAs geneXΔ cells was scored visually. After this initial screening, candidate deletion strains were streaked in long, wide patches on regular round plates, and long, wide patches of query strains with the genotype ORC6 (-WT, -ps, -rxl, or -ps,rxl) GAL-CDC6ΔNT-HAs were cross-replica plated onto the deletion strains. The same protocol was used to select MATa haploids containing KanMX and ORC6 (-WT, -ps, -rxl, or -ps,rxl) GAL-CDC6ΔNT-HAs, as described above. This secondary screening served to confirm the initial hits (we found that the larger patch size of diploids in this protocol significantly reduced the high false-positive rate seen in the initial screen) and to ask if the lethality specifically required one of the mutations in ORC6. In this second screen, we also tested every deletion that gave no signal on the d-Leu-Ura-His plus G418 plate in the first screen. This allowed the detection of interactions between gene deletions and ORC6 mutants alone (without the expression of stabilized CDC6). In a separate targeted screen, genes which are related to cell cycle regulation and DNA damage based on Saccharomyces Genome Database annotation (88 deletion strains; a complete list is found in Table S1 in the supplemental material) were also selected and tested using the large-patch-format screen (some deletions were excluded due to their initial slow growth on glucose or galactose). During this screening, we used the ORC6 (-WT, -rxl, or -ps,rxl) GAL-CDC6ΔNT-HAs strain as the query strain. Because of the higher sensitivity of this format, we found additional positive results that were missed in the first round.
In initial screening tests of the DMA strains by PCR, we found that approximately 25% of the strains in the deletion collection did not contain the indicated deletion. Therefore, we confirmed all of our positive hits by PCR, as well as all of the candidates used in the second targeted screen. This significant error rate also means that we probably missed potential interactions with gene deletions that were absent from the collection.
Cells were grown in YPR (raffinose) medium for 4 h, and the GAL-CDC6ΔNT-HAs strain was induced by the addition of 3% galactose (final concentration). In control cultures, 0.1% methyl methanesulfonate (MMS; Sigma, St. Louis, MO) was added. Cell lysates were resolved in 6% polyacrylamide Novex gels (Invitrogen, Carlsbad, CA) and transferred onto nitrocellulose. Six-Flag-Rad53 was probed using the M2 monoclonal antibody directed against the Flag epitope (Sigma, St. Louis, MO). For the experiment shown in Fig. S3 in the supplemental material, Rad53 was probed with a serum raised against Rad53 (a gift of S. Elledge).
Fluorescence and differential interference contrast images were acquired on a Zeiss Axioplan 2 microscope (Carl Zeiss, Inc., Thornwood, NY) with a 100× 1.4-numerical-aperture Planapochromat objective (Vermont Optechs, Charlotte, VT) fitted with a Hamamatsu Orca ER cooled charge-coupled device camera (Sciscope Instrument Co., Iowa City, IA) controlled by Openlab software (Improvision, Coventry, United Kingdom). To visualize DDC2-green fluorescent protein (GFP) foci, cells were grown to log phase, induced or not as indicated, and imaged using a DeltaVision image restoration microscope system (Applied Precision, Inc.) mounted on an Olympus IX-70 microscope and fitted with a Cool Snap-HQ/ICX285 camera. Cells were imaged using a UplanApo60X 1.35-numerical-aperture oil objective and the standard DeltaVision fluorescein isothiocyanate filter set, using identical exposure times for all samples. Image stacks were deconvolved and presented as maximum projections using SoftWoRx software, and the exported TIFF files were processed with Adobe Photoshop. All images shown here were captured and processed at the same time by an identical method. DNA content analysis by fluorescence-activated cell sorting (FACS) was performed as described elsewhere (19).
Multiple mechanisms that prevent rereplication target the components of the pre-RC (ORC, Cdc6, and MCMs) (11, 45, 68). Strains in which several of these mechanisms are disrupted have been generated previously (45, 68). These disruptions included removal of hypothetical Cdk phosphorylation sites in Orc2 (ORC2-ps; S16A, T24A, T70A, T174A, S188A, and S206A) (45) and Orc6 (ORC6-ps; S106A, S116A, S123A, and T146A) (45), introduction of an artificial nuclear localization sequence (NLS) in Mcm7 (MCM7-NLS) (44, 45), stabilization of Cdc6 by N-terminal truncation (CDC6ΔNT) (45), and removal of a Clb5-binding RXL motif in Orc6 (ORC6-rxl) (68). The removal of these controls increased the cell's propensity to rereplicate during a nocodazole arrest (45, 68). With this study, we found that the same is true in the absence of nocodazole, as shown in Fig. 1A to C, and we demonstrate here the additive character of the mutations on the extent of rereplication. In cells expressing stabilized Cdc6 and constitutively nuclear MCMs, the disruption of cyclin-Cdk-dependent regulation of ORC subunits induced phenotypes ranging from a significant accumulation of cells with 2C DNA (Fig. (Fig.1A,1A, panels 2 and 3) to extensive rereplication (Fig. (Fig.1A,1A, panels 5 and 6, and C). Rereplication was not detectable by FACS in the absence of MCM7-NLS or GAL-CDC6ΔNT (data not shown), consistent with results obtained in the presence of nocodazole (45, 62, 68). Therefore, mechanisms preventing rereplication at the level of the pre-RC are additive and not simply redundant, as rereplication appears when any one mechanism is disrupted. This lack of complete redundancy is consistent with the argument that multiple overlapping mechanisms have evolved by natural selection to control rereplication.
We noted that extensive rereplication as shown in Fig. Fig.11 required the integration of multiple copies of GAL-CDC6ΔNT-HA. In this paper, we refer to the multiple integration allele as GAL-CDC6ΔNT-HAm. We introduced this allele into our strains by crossing, thus maintaining the same copy number within experiments. In other contexts, we used a single integration allele (GAL-CDC6ΔNT-HAs). In the presence of mutations in ORC2, ORC6, and MCM7, GAL-CDC6ΔNT-HAs caused cells to accumulate with an approximate 2C DNA content instead of causing enough rereplication to allow its detection by FACS (data not shown). This suggests that Cdc6ΔNT-HA became limiting with only a single copy of the GAL promoter construct; we do not know how much Cdc6 is produced by this construct in comparison to endogenous levels.
The induction of rereplication in cycling cells caused them to arrest with a large bud and a short spindle, as detected by Tub1-GFP fluorescence. This is shown for a strain with the genotype GAL-CDC6ΔNT-HAm MCM7-NLS ORC6-ps,rxl ORC2-ps TUB1-GFP as an example in Fig. Fig.1D.1D. Consistent results were obtained by staining of the DNA with DAPI (4′,6′-diamidino-2-phenylindole) or propidium iodide, which showed mostly undivided nuclei (see Fig. S2 in the supplemental material; also data not shown). This morphological phenotype indicates that rereplicating cells do not undergo an abortive mitosis. A cell cycle arrest characterized by large-budded cells with undivided nuclei and short spindles could have been induced by a DNA damage response, the intra-S checkpoint, or the spindle attachment checkpoint. We were interested to determine if any checkpoint pathways were activated in rereplicating cells.
In order to determine how cells respond to the loss of rereplication controls and to determine if they activate a specific checkpoint pathway, we conducted a systematic synthetic gene array (SGA) analysis based on a published method (62). Because extensive rereplication as detected by FACS kills the cells, even in an otherwise wild-type background (68), we sought to conduct our screen using a background that combines only a few mutations disrupting mechanisms preventing rereplication. Our previous study showed that the proliferation of ORC6-rxl CDC6ΔNT strains was impaired (68). We hypothesized that this background would increase the frequency of events leading to reinitiation. We suspected that ORC6-rxl CDC6ΔNT strains accumulated inappropriate chromatin structures resulting from an increased tendency of pre-RCs to engage in a step towards firing. As discussed below, these structures could result from the reassembly of initiation complexes, from aberrant unwinding at origins, or from overt reinitiation. For the last case, there must be limited rereplication, since we did not observe DNA content significantly beyond 2C by FACS (see below).
An alternative interpretation of the poor viability of ORC6-rxl CDC6ΔNT strains is that these mutations impair pre-RC formation or function, rather than promoting excessive or inappropriate pre-RC formation or function. We previously showed that ORC6-ps,rxl cells were able to maintain plasmids almost as efficiently as ORC6-wt cells in an assay sensitive to the efficiency of pre-RC formation (68). Using the same assay, we found no significant plasmid maintenance defect in GAL-CDC6ΔNT-HAs cells cultivated in galactose-containing medium (4.2% ± 0.4% [mean ± standard deviation] loss/generation) compared with wild-type cells grown under the same conditions (3.1% ± 1.5% loss/generation). We confirmed that ORC6-ps,rxl cells have no significant defect (4.6% ± 1.1% loss/generation) in this assay. We found a slightly increased plasmid maintenance defect for ORC6-ps,rxl GAL-CDC6ΔNT-HAs cells (6.8% ± 0.8% loss/generation). In preliminary experiments, we detected at least partial rescue of this defect by adding multiple origins to the reporter plasmid (25), consistent with the loss being due to a defect in pre-RC formation or function (data not shown). Since the only known mechanistic effect of these mutations is the loss of Cdk control over reassembly of the pre-RC (45, 68), aberrant pre-RC formation may cause the plasmid loss phenotype by some indirect mechanism. Alternatively, these results may reflect secondary effects of the mutations on other aspects of origin function (for example, the ORC6-ps,rxl mutation may partially inhibit MCM loading or replication initiation). In this report, we will follow the former interpretation and assume that the effects of these mutations are due to a loss of Cdk control over reassembly of a pre-RC after initiation within a single S phase; however, the possibility of subtle defects in other aspects of origin function cannot be ruled out and is a caveat to our conclusions.
If reinitiating cells require some defined pathways for survival, these pathways would be detected by our synthetic lethal analysis. We screened deletions of nonessential genes that would be lethal in the ORC6-rxl GAL-CDC6ΔNT-HAs and ORC6-ps,rxl GAL-CDC6ΔNT-HAs backgrounds on YEP-Gal. In this medium, the starting strains show only slightly slower proliferation than the wild type. In most or all cases in which an additional synthetic lethal deletion was added, the removal of putative Orc6 Cdk phosphorylation sites (“ps” mutations) added to the severity of the genetic interactions caused by the ORC6-rxl mutation (see below; this is consistent with the initial characterization of the ORC6-rxl mutation in reference 68). We therefore used both backgrounds to increase the dynamic range of our screen, and we collectively refer to them as ORC6-(ps)rxl GAL-CDC6ΔNT-HAs. We carried out the SGA analysis by systematically crossing strains from deletion strain collections and selecting for the progeny with the genotype ORC6-rxl GAL-CDC6ΔNT-HAs geneXΔ (see Materials and Methods). ORC6-(ps)rxl GAL-CDC6ΔNT-HAs cells on YEP-Gal proliferate faster than ORC6-(ps)rxl CDC6ΔNTs cells on YEPD (68) (see below). For that reason, we could detect synthetic lethality but less efficiently detect rescue in the ORC6-(ps)rxl GAL-CDC6ΔNT-HAs cells on YEP-Gal. All of the interactions obtained from the screen were then tested against GAL-CDC6ΔNT-HAs (ORC6-WT, -rxl, and -ps,rxl) query strains on glucose- and galactose-containing media; this secondary screen identified mutations that interacted specifically with the ORC6 alleles, with CDC6ΔNT, or with the two in combination. Because genes recovered in the first part of the screen tended to be involved in DNA metabolism and cell cycle regulation, we included a set of candidate genes involved in these functions based on the GO annotations in the Saccharomyces Genome Database (http://www.yeastgenome.org/) in the second round of testing (see Table S1 in the supplemental material for a list of all genes tested in the second screen). After the second round, the identities of all gene deletions were tested by PCR. All hits obtained in the second round of testing were then confirmed by serial dilutions of the relevant strains and their control genotype strains (obtained by tetrad dissection) on YEPD and YEP-Gal (see Fig. Fig.3;3; see also Fig. S1 in the supplemental material). A complete list of synthetic lethal gene deletions is provided in Table Table1.1. Some of the synthetic lethal interactions were additionally confirmed by a tetrad analysis using the ORC6-rxl and CDC6ΔNT (instead of GAL-CDC6ΔNT-HAs) alleles with endogenous levels of Cdc6ΔNT expression (see Table S2 in the supplemental material).
We also tested the effect of introducing additional candidate deletions and mutations in checkpoint pathways into ORC6-rxl CDC6ΔNT cells (with CDC6ΔNT under the control of the endogenous CDC6 promoter). ORC6-rxl CDC6ΔNT cells proliferate slowly on glucose-containing medium (68). We were surprised to discover that deletion of the checkpoint genes MRC1, TOF1, and CSM3 actually rescued the sickness and inviability of those cells (see Fig. Fig.55 and Table Table1;1; see below), in contrast to the many deletions of checkpoint genes that caused synthetic lethality (Table (Table1).1). The proliferation defect of the ORC6-rxl GAL-CDC6ΔNT-HAs strain on galactose medium was subtle, preventing us from screening systematically for additional gene deletions that rescue this background.
The genetic interactions identified in our screen suggest that specific DNA damage response and repair genes are crucial in cells with compromised rereplication controls. Our screen also uncovered several strong interactions with cell cycle regulators; these interactions will be characterized elsewhere. As is typical with such screens (e.g., see reference 62), we also found interactions with some genes thought to function in apparently completely unrelated pathways (Table (Table1;1; for example, EDE1, which functions in endocytosis). We have no explanation at present for these positive results. A few genes interacted with ORC6-rxl or CDC6ΔNT alone (Table (Table11).
Most genes/proteins uncovered in our screen were already known to interact genetically or physically in a tight cluster (Fig. (Fig.2).2). Some of the other genes not found in this cluster are nevertheless thought to function in the same pathways or to belong to the same epistasis groups. This clustering suggests the specificity of our genetic screen and of the pathways involved in response to compromised rereplication controls (Fig. (Fig.2,2, inset).
Strikingly, many genes involved in the DNA damage response were required in the ORC6-(ps)rxl GAL-CDC6ΔNT-HAs background (Table (Table1;1; Fig. Fig.2).2). MEC1 encodes a kinase required for the DNA damage response (reviewed in reference 46). MEC1 is essential, and therefore its deletion was not present in the collection of haploid deletion strains that we screened. The deletion of MEC1 is rescued if SML1 is also deleted. SML1 encodes an inhibitor of ribonucleotide reductase, and its deletion increases the nucleotide pool, eliminating the reliance on MEC1 in the absence of overt DNA damage (72). With a tetrad analysis, we observed that ORC6-rxl CDC6ΔNT mec1 sml1 spores were inviable, while ORC6-rxl CDC6ΔNT MEC1 sml1 spores were often viable and slow-proliferating (see Table S2 in the supplemental material; also data not shown), like ORC6-rxl CDC6ΔNT MEC1 SML1 cells (68). ORC6-rxl GAL-CDC6ΔNT-HAs cells also depended on MEC1 when plated on galactose medium (Fig. 3A and B). The deletion of TEL1, the MEC1 homolog thought to act as a sensor of DNA damage in parallel with MEC1 (23, 41), had little or no effect on the viability or proliferation rate of the ORC6-rxl CDC6ΔNT cells (see Table S2 in the supplemental material; also data not shown).
Mec1 is required for all known DNA damage checkpoints and for the response to replication defects (46). Mec1-dependent responses can regulate replication fork progression, inhibit late origin firing, and prevent spindle elongation, leading to a reversible cell cycle arrest (34, 46, 51, 54, 60). DNA damage responses initiated at the G1/S transition, during S phase, or after S phase are thought to be mediated by distinct Mec1-dependent molecular pathways (48). To explore which of these functions was critical in the ORC6-rxl CDC6ΔNT background, we used the mec1-100 allele, which is a separation-of- a function allele (48). The mec1-100 mutant is defective in the G1/S and intra-S DNA damage checkpoints but is competent at delaying nuclear division after UV irradiation in G2 (G2/M checkpoint) and preventing damage-induced replication fork collapse (48, 61). In a tetrad analysis, the mec1-100 allele failed to kill or rescue ORC6-rxl CDC6ΔNT spores (Table S2 in the supplemental material). ORC6-rxl CDC6ΔNT mec1-100 spores formed slow-growing colonies, like ORC6-rxl CDC6ΔNT spores (data not shown). We concluded that ORC6-rxl CDC6ΔNT cells require Mec1 to signal a G2/M checkpoint and/or to prevent replication fork collapse but that Mec1 is not required in this context to signal a G1/S or intra-S checkpoint.
Ddc2 associates with Mec1 in a complex that is recruited to sites of DNA damage (39). Ddc2-GFP allows the visualization of this recruitment by the formation of nuclear foci upon induction of DNA damage (39). We introduced a DDC2-GFP allele into our strains to test whether Ddc2-GFP foci would be detectable following a loss of rereplication control. We found that GAL-CDC6ΔNT-HAm MCM7-NLS ORC6-ps,rxl rereplicating cells accumulated Ddc2-GFP foci when the expression of Cdc6ΔNT-HA was induced (Fig. 4A and B, strain 2). An increase in Ddc2-GFP foci could also be detected in GAL-CDC6ΔNT-HAm MCM7-NLS cells expressing ORC6-rxl (Fig. (Fig.4,4, strain 3) or ORC6-ps (Fig. (Fig.4,4, strain 4), but not in equivalent cells expressing ORC6-wt (Fig. (Fig.4,4, strain 5). Even GAL-CDC6ΔNT-HAs ORC6-rxl cells (Fig. (Fig.4,4, strain 6) accumulated a significant level of Ddc2-GFP foci when the expression of Cdc6ΔNT-HA was induced. GAL-CDC6ΔNT-HAs ORC6-rxl cells do not overtly rereplicate and are able to proliferate in galactose-containing medium (see above). These results are consistent with the dependence of GAL-CDC6ΔNT-HAs ORC6-rxl cells on Mec1 for survival and further suggest that a loss of rereplication control triggers a Mec1-Ddc2-dependent DNA damage response that promotes survival.
Mrc1 is known to function both in mediating the intra-S checkpoint and in DNA replication (2, 47), in addition to a function in sister chromatid cohesion (71). Consistent with the absence of a genetic interaction between ORC6-rxl CDC6ΔNT and mec1-100, as discussed above, the deletion of MRC1 inflicted no synthetic lethality on ORC6-(ps)rxl GAL-CDC6ΔNT-HAs cells (data not shown). This result supports the conclusion that these cells do not rely on the intra-S checkpoint. Instead, as noted above, the deletion of MRC1 restored the normal proliferation of ORC6-rxl CDC6ΔNT cells in tetrad analysis (Fig. (Fig.5;5; see Table S1 in the supplemental material). ORC6-rxl CDC6ΔNT mrc1 spores were fully viable and proliferated at a rate close to that of wild-type cells, unlike the semi-inviable ORC6-rxl CDC6ΔNT MRC1 background (see Table S2 in the supplemental material) (68). Similar results were obtained by deleting TOF1, which encodes a protein that associates and moves with Mrc1 at replication forks (27), and by deleting CSM3, which encodes a protein that interacts with Tof1 (36) (Fig. (Fig.5;5; see Table S2 in the supplemental material). Therefore, Mrc1, Tof1, and Csm3 are required for a function that slows down the proliferation of ORC6-rxl CDC6ΔNT cells.
We next tested whether the checkpoint function of Mrc1 was responsible for the slow growth of ORC6-rxl CDC6ΔNT cells. We used the separation-of-function allele mrc1-AQ, in which all the consensus sites for phosphorylation by Mec1 (SQ and TQ) have been mutated to AQ motifs, uncoupling Mrc1 from Mec1 and disrupting the intra-S checkpoint (47). Using tetrad analysis, we observed that ORC6-rxl CDC6ΔNT mrc1-AQ cells proliferated slowly, like ORC6-rxl CDC6ΔNT cells (i.e., mrc1AQ did not rescue ORC6-rxl CDC6ΔNT cells) (see Table S2 in the supplemental material). Therefore, the higher proliferation rate of ORC6-rxl CDC6ΔNT mrc1 cells was not due to the removal of the intra-S checkpoint signaling function of Mrc1 but was due to the removal of its replication function or sister chromatid cohesion function. Since Tof1 and Csm3 are thought to function together with Mrc1 (27, 36), we expect that the same conclusion will apply to all three proteins.
The 9-1-1 complex, so called because it is composed of Rad9, Rad1, and Hus1 in S. pombe, is made up of Rad17, Ddc1, and Mec3 in S. cerevisiae. This ring-shaped complex is loaded at sites of damage in a Rad24-dependent manner and acts there as a sensor (30, 38). The deletion of RAD17 or DDC1 rendered ORC6-rxl GAL-CDC6ΔNT-HAs cells inviable in our screen and in plating experiments (Table (Table1;1; see Fig. Fig.3C3C for RAD17 and Fig. S1 in the supplemental material for DDC1). This result was also confirmed with ORC6-rxl CDC6ΔNT spores in a tetrad analysis assay (see Table S2 in the supplemental material). The mec3 deletion strain proliferated very slowly and therefore was not tested. We found by PCR that the rad24 deletion strain was absent from our strain collection, so we did not test the effect of deleting RAD24.
Rad53 is an effector kinase that causes cell cycle arrest in response to DNA damage (46). It also may contribute to the stabilization of replication forks (34, 60, 61). As with MEC1, the deletion of RAD53 is lethal by itself but is rescued by a deletion of SML1. We found that RAD53 was required for the viability of ORC6-rxl CDC6ΔNT cells. In the tetrad analysis, ORC6-rxl CDC6ΔNT rad53 sml1 spores were almost always inviable (see Table S2 in the supplemental material). Deletion of CHK1 (RAD53 homolog) had little or no effect on the viability or proliferation rate of ORC6-rxl CDC6ΔNT cells (see Table S2 in the supplemental material; data not shown).
The mediator protein Rad9 acts as an adaptor to couple Mec1 to Rad53. Rad9 is recruited at sites of damage and becomes phosphorylated in a Mec1- or Tel1-dependent manner (18, 65). The binding of Rad9 by Rad53 promotes Rad53's activating trans autophosphorylation, which in turn promotes its dissociation from Rad9 (56, 57). Surprisingly, the deletion of RAD9 had a more moderate effect than the removal of MEC1, RAD53, RAD17, or DDC1 on the viability or proliferation of ORC6-rxl CDC6ΔNT cells (data not shown). One explanation for this result is that Mrc1, Rad9's functional homolog, may compensate for the loss of Rad9 in this context. This idea is supported by the decreased viability of ORC6-rxl CDC6ΔNT rad9 mrc1-AQ spores relative to the control genotypes (see Table S2 in the supplemental material). This is consistent with a previous report, where Rad9 could compensate for the loss of Mrc1 (2).
In summary, these results strongly suggest that cells with compromised rereplication controls rely on the activation of a DNA damage response requiring the sensors Mec1 and the Rad17-Ddc1-Mec3 complex and the effector kinase Rad53 for their viability. The mediator function thought to be required for coupling Mec1 to Rad53 may be provided redundantly by Mrc1 and Rad9.
The requirement of a DNA damage signaling pathway in ORC6-rxl CDC6ΔNT cells suggests that the cells undergo abnormally frequent DNA damage. We were curious to know what kind of DNA damage was induced in this mutant.
The evolutionarily conserved MRX complex, composed of Mre11, Rad50, and Xrs2 in yeast, is required for the signaling and repair of double-strand breaks (DSBs) (reviewed in reference 9). The deletion of MRE11, RAD50, or XRS2 severely reduced the viability or proliferation of ORC6-rxl GAL-CDC6ΔNT-HAs cells in plating experiments (see Fig. Fig.3D3D for MRE11 and Fig. S1 in the supplemental material for RAD50 and XRS2) and of ORC6-rxl CDC6ΔNT cells in a tetrad analysis (see Table S2 in the supplemental material). These genetic results suggest the hypothesis that double-strand breaks may occur in this genetic background. Although we do not have direct evidence for this at present, the additional genetic results presented below support this inference.
Two biochemical reactions can lead to DSB repair, namely, homologous recombination (HR) and nonhomologous end-joining (NHEJ) (28). MRX has been reported to be required for both HR and NHEJ (reviewed in references 9 and 28). While HR requires Rad52 and is often Rad51 dependent (49), NHEJ requires Dnl4 and the Yku70-Yku80 complex (28). Because we speculated that DSBs might accumulate in ORC6-rxl GAL-CDC6ΔNT-HAs cells, we were interested to determine if one of these pathways was preferentially required in this background. Indeed, we observed that rad52 deletion was strongly lethal in the ORC6-rxl GAL-CDC6ΔNT-HAs background (Fig. (Fig.3E).3E). A weaker but reproducible synthetic lethality was obtained with RAD59 (a homolog of RAD52), but this interaction was also observed with ORC6-rxl alone, for unknown reasons (see Fig. S1 in the supplemental material). However, the deletion of RAD51 had no effect (see Fig. S1 in the supplemental material) (Rad52-dependent, Rad51-independent repair is reviewed in reference 49). In contrast, the deletion of DNL4 had no effect on the slow proliferation phenotype of ORC6-rxl CDC6ΔNT cells (see Table S2 in the supplemental material), and deletions of YKU70 and YKU80 had no detectable effect on the growth of ORC6-rxl GAL-CDC6ΔNT-HAs cells on YEP-Gal (Fig. (Fig.3F;3F; see Fig. S1 in the supplemental material).
These results suggest that the disruption of mechanisms preventing rereplication causes DNA lesions, perhaps DSBs, that require RAD52-dependent homologous recombination but not NHEJ.
Several additional genes known to function in maintaining genomic integrity, such as SRS2, MMS1, MMS22, CTF4, and CCR4, were synthetically lethal when deleted from the ORC6-rxl GAL-CDC6ΔNT-HAs background. In contrast, the deletion of other genes involved in DNA repair had no effect on the viability or proliferation rate of ORC6-rxl CDC6ΔNT cells in the tetrad analysis (see Table S2 in the supplemental material). These included deletions of EXO1 (exonuclease, functions in DSB repair by recombination  and the activation of Mec1 ), SGS1 (DNA helicase, required for faithful chromosome segregation) (67), TOP3 (topoisomerase) (29), and MMS4 and MUS81 (endonucleases involved in DNA repair) (70). Thus, only specific DNA damage signaling and repair pathways are vital in cells that have lost controls that prevent rereplication. In particular, we found a strong role for pathways responsible for DSB signaling and repair by HR.
The dependence of ORC6-rxl CDC6ΔNT cells on RAD53 suggested that Rad53 was activated in these cells. The activation of Rad53 by phosphorylation was observed using ORC6-ps,rxl GAL-CDC6ΔNT-HAs FLAG-RAD53 cells induced by the addition of galactose (Fig. (Fig.6A).6A). A slower-migrating signal in an anti-Flag Western blot, presumably corresponding to phosphorylated Flag-Rad53, became visible 2 h following induction and reached a maximum between 2 and 4 h after induction. This shift in the electrophoretic mobility of Flag-Rad53 was similar to that observed when MMS (a DNA alkylating agent) was added (Fig. (Fig.6A).6A). The induction of GAL-CDC6ΔNT-HAs in these cells did not cause them to rereplicate significantly, although it caused an accumulation of cells with a 2C DNA content (Fig. (Fig.6A).6A). The induction of GAL-CDC6ΔNT-HAs in the presence of ORC6-ps or ORC6-rxl, but not ORC6-wt, caused a slight Rad53 activation and accumulation of 2C DNA-containing cells (Fig. (Fig.6A).6A). These results suggest the speculation that the Rad53-mediated response allows ORC6 (-ps, -rxl, or -ps,rxl) GAL-CDC6ΔNT-HAs cells in YEP-Gal to pause the cell cycle during or after DNA replication, probably in order to repair some form of DNA damage (perhaps DSBs; see above) before resuming the cell cycle.
To begin to dissect the pathways involved in signaling and repairing potential DNA damage in rereplication-sensitized cells, we monitored Rad53 activation and cell cycle delays in ORC6-ps,rxl GAL-CDC6ΔNT-HAs FLAG-RAD53 cells in which we introduced selected gene deletions. Deleting MEC1 from ORC6-ps,rxl GAL-CDC6ΔNT-HAs FLAG-RAD53 cells prevented Rad53 activation when galactose or MMS was added (Fig. (Fig.6B).6B). This is consistent with the role of Mec1 in activating Rad53 (56). The deletion of RAD17 or DDC1 largely, but not completely, prevented Rad53 activation, suggesting that only residual DNA damage checkpoint signaling could occur (Fig. (Fig.6B).6B). Interestingly, the deletion of MRE11 from ORC6-ps,rxl GAL-CDC6ΔNT-HAs FLAG-RAD53 cells largely and reproducibly eliminated Rad53 activation induced by galactose but not by MMS (Fig. (Fig.6B).6B). These results suggest that a loss of rereplication controls induces a narrower or different spectrum of DNA damage than that induced by DNA alkylation. The question of the involvement of MRX in checkpoint signaling versus a direct role in DSB repair has been controversial (9). Our observations suggest that in the context of ORC6-ps,rxl GAL-CDC6ΔNT-HAs cells, Mre11 is required for signaling, but this does not exclude an additional function of MRX in repairing DNA damage. A requirement for Mre11 in Mec1 signaling in response to DSBs or UV has been demonstrated recently (43), and this is consistent with our results, especially if DSBs are indeed generated in the ORC6-ps,rxl GAL-CDC6ΔNT-HAs context.
In the absence of RAD52 (Fig. (Fig.6B),6B), ORC6-ps,rxl GAL-CDC6ΔNT-HAs FLAG-RAD53 cells exhibited Rad53 activation. This places Rad52 downstream of or parallel to Rad53 and is consistent with a role for Rad52 in repairing DNA damage in cells with compromised rereplication controls and not in checkpoint signaling. Using an anti-Rad53 antiserum (a generous gift of S. Elledge), we confirmed that Rad53's gel mobility is reduced in ORC6-rxl GAL-CDC6ΔNT-HAs cells incubated in galactose (see Fig. S3 in the supplemental material). Using this assay, we found that Rad53 could still shift in the presence of deletions of CLA4, CTF4, IKI3, RTS1, SRS2, and SWI6, while a deletion of CCR4 abolished (or reduced) the Rad53 shift (see Fig. S3 in the supplemental material). Overall, our analysis places Mec1, Rad17, Ddc1, Ccr4, and Mre11 upstream of Rad53 and Rad52, Cla4, Ctf4, Iki3, Rts1, Lsm1, Srs2, and Swi6 downstream of or parallel to Rad53 in the response to a loss of mechanisms preventing rereplication.
We examined the effect of disrupting the DNA damage response on overt rereplication. As noted above, the detection of overt overreplication by FACS (rather than a delay with an apparent 2C DNA content) required multiple copies of GAL-CDC6ΔNT-HA and the MCM7-NLS allele. ORC6-ps,rxl MCM7-NLS GAL-CDC6ΔNT-HAm mec1 sml1 cells reproducibly showed greater rereplication than MEC1 SML1 or MEC1 sml1 controls (Fig. (Fig.7;7; see Fig. S2 in the supplemental material). The increase in FACS signal could have been due to cells passing through mitosis without cytokinesis; if this were associated with a normal cycle of cyclin degradation, it could be expected that origins would reload normally, followed by normal reinitiation. This idea would be consistent with the checkpoint role of Mec1. However, this explanation seems unlikely, since we did not detect multinucleate cells (indicative of nuclear division without cytokinesis) or rebudded cells (indicative of mitotic exit without cytokinesis) (see Fig. S2 in the supplemental material). These results suggest that the increase in DNA overreplication in the absence of Mec1 occurred during the same S phase and was unlikely to be a gross effect due to a loss of cell cycle checkpoint function. Together, these results suggest that Mec1 restrains rereplication. We do not know why the mec1 mutants seemed to still exhibit a cell cycle arrest, but previous results indicated that the Mad2-dependent spindle checkpoint can be activated in response to DNA damage by an unknown mechanism (21).
The deletion of RAD17 or MRE11 increased the extent of rereplication relative to control strains, even more than that observed with the mec1 sml1 strain (Fig. (Fig.7A),7A), while the deletion of MRC1 had no or little effect (data not shown). Therefore, upstream signaling through Mec1, Rad17, and Mre11 seems to result in a restraint of rereplication.
In this work, we have examined how cells respond to rereplication and to the partial loss of rereplication controls. We have identified genetic requirements related to the DNA damage response and cell cycle control. We have begun dissecting how DNA damage signaling and repair genes function in promoting survival in response to a loss of pre-RC regulation (Fig. (Fig.8).8). While this paper was being revised, Green and Li reported that rereplication causes extensive DNA damage (22), consistent with our findings. Among other results, Green and Li used pulsed-field gel electrophoresis to demonstrate the occurrence of subchromosomal fragments following rereplication.
Cells with a disruption of rereplication controls relied on genes required for signaling DNA damage. We identified Mec1, Rad17, Ddc1, and the MRX complex (Mre11, Rad50, and Xrs2) as being required for the pathway signaling damage in ORC6-rxl GAL-CDC6ΔNT-HAs cells. Visualization of Ddc2-GFP foci, which are recruited to sites of damage in complex with Mec1 (39), documented an active DNA damage response in these cells, despite their continued proliferation. All of these proteins function upstream of Rad53, which is also required in this pathway, based on an examination of the genetic requirements for Rad53 phosphorylation in response to a loss of rereplication controls. MRX activity was known to be required for an efficient response to DSBs (reviewed in reference 9). Mre11-Rad50-Nbs1 (MRN; the vertebrate ortholog of MRX) was recently found to function in incorporating linear DNA fragments into damage-signaling complexes in Xenopus extracts (8). In these experiments, MRN was required for ATM (Mec1 ortholog) activation. Similarly, Mre11 has been reported to be required for Mec1 signaling in response to DSBs or UV (43). Exo1 has been shown to cooperate with Mre11 upstream of Mec1 (43), but it had no effect on the viability of our strains with compromised rereplication controls, suggesting that our strains experience damage that can activate Mec1 without the need for processing by Exo1. We therefore suggest that an MRX-Mec1 (Ddc2)-Rad53 pathway functions in directly signaling damage induced in cells with compromised rereplication controls. Our results suggest that Rad17 and Ddc1 are likely to be involved in the same pathway, upstream of Rad53. Because the Rad17-Ddc1-Mec3 complex functions as a sensor of DNA damage recruited independently of Mec1-Ddc2 (30, 38), we propose that it functions around the level of MRX and Mec1 in a pathway responding to rereplication. Preliminary experiments suggested that the accumulation of 2C DNA (Fig. (Fig.6A)6A) does not occur in ORC6-ps,rxl GAL-CDC6ΔNT-HAs cells when MEC1, RAD17, DDC1, or MRE11 is deleted but still occurs when RAD52 is deleted (data not shown). Therefore, the [MRX-Mec1 (Ddc2)-(Rad17-Ddc1)]-Rad53 pathway that we identified may be required for a cell cycle delay allowing time to repair DNA damage before mitosis when rereplication controls are compromised.
In addition, we have shown that at least Mre11, Mec1, and Rad17 function to restrain the extent of rereplication (Fig. (Fig.7).7). We speculate that this occurs through the pausing of replication forks in cells that undergo rereplication. Mec1 and Rad53 can mediate fork pausing and stabilization (60). Perhaps Mre11 and Rad17 also participate in this pathway. The deletion of MRE11 caused the most extreme increase in rereplication, more than the deletion of MEC1 or RAD17 (Fig. (Fig.7A).7A). Therefore, we suspect that Mre11 participates in restraining rereplication by an additional process. Mre11 is the most upstream signaling molecule in the pathway that we identified and has been shown to nucleate signaling complexes, including damaged DNA, in Xenopus extracts (8). Such complexes could constitute a physical barrier to the progression of DNA replication in rereplicating cells that accumulate damage.
The dependence of ORC6-rxl GAL-CDC6ΔNT-HAs cells on MRE11, RAD50, XRS2, and RAD52 suggests that DNA damage, probably DSBs, accumulates in these cells. The requirement of RAD52 (and RAD59, to a lesser extent) suggests that the same cells rely on homologous recombination. In contrast, the lack of dependence on NHEJ genes such as DNL4, YKU70, and YKU80 (28) indicates that NHEJ is not required. Our screen also identified MMS1 and MMS22 as being required in ORC6-rxl GAL-CDC6ΔNT-HAs cells. Both of these genes have been found to function in a pathway protecting against replication-dependent DNA damage and in concert with RAD52 (4, 28). The source of homology for repair may be the sister chromatid. This could explain why CTF4, which is required for efficient sister cohesion (24), was recovered in our screen (Table (Table1).1). In fact, it has recently been reported that postreplicative recruitment of cohesin to DSBs is required for DNA repair (55). Cohesins themselves could not be identified in our screen since they are essential for viability. The SRS2 gene encodes a DNA helicase that may function in postreplication repair (1, 32, 50). More recently, specific roles for Srs2 have been described. These include signaling a DNA damage checkpoint in a Mec1-Rad53 pathway (33), the suppression of crossovers during DSB repair (26), and recovery from cell cycle arrest following repair of DSBs (63). We found that deleting SRS2 is synthetically lethal to ORC6-rxl GAL-CDC6ΔNT-HAs cells. Since the deletion of SRS2 did not impair Rad53 activation in those cells (see Fig. S3 in the supplemental material), it is unlikely that an Srs2 checkpoint function upstream of Rad53 is required. We have not explored further which function of Srs2 is required in cells that have lost controls over rereplication. The deletion of SRS2 is synthetically lethal with several genes recovered in our screen (Fig. (Fig.2).2). Among them are MRC1, TOF1, and CSM3, which has been proposed to be required for sister chromatid cohesion to aid the repair of spontaneous damage (71); surprisingly, we found that the deletion of these genes rescued the viability of ORC6-rxl CDC6ΔNT cells (Fig. (Fig.5).5). The deletion of MRC1, TOF1, or CSM3 may channel DNA damage repair to a more efficient and possibly SRS2- and RAD52-dependent pathway in ORC6-rxl GAL-CDC6ΔNT-HAs cells. Indeed, the synthetic lethality between genes that we uncovered in our screen suggests that several different molecular mechanisms might cooperate in response to DNA damage incurred when mechanisms preventing rereplication are lost.
Our work was based on the assumption that the ORC6-rxl CDC6ΔNT mutant has reduced proliferation because of an increased propensity towards rereplication. This was inferred from the dominance of both alleles, from the fact that they do not confer lower firing frequencies (68), and from the marked increase in DNA rereplication observed by FACS when these alleles were combined with other alleles (ORC2-ps, MCM7-NLS, and ORC6-ps) (Fig. (Fig.1).1). Our working model is that in the ORC6-rxl CDC6ΔNT context, pre-RCs engage more readily in a step that is part of the molecular rearrangements leading to origin firing. This could lead to occasional refiring events. In this case, the induction of reinitiation could generate double-strand breaks if a second replication fork encounters Okazaki fragments generated at the first fork. Alternatively, if a fork “catches up” with a fork ahead, double-strand breaks could also be generated. This could be especially true for those strains that rereplicate extensively (Fig. (Fig.11).
It is important to note that we lack direct evidence that the ORC6-rxl CDC6ΔNT mutant rereplicates, even at a low level, although this seems like a reasonable speculation based on past (68) and present data. As noted above, we cannot exclude the possibility that the defects in this context are due to the failure of aspects of origin function other than the prevention of pre-RC reassembly, although we consider this unlikely.
However, partially disrupting rereplication controls may not necessarily be sufficient for actual DNA polymerization but may nonetheless activate a cellular response. Helicases that unwind DNA can become uncoupled from replication forks under certain conditions (27, 66). In Xenopus egg extracts, the inhibition of DNA polymerase alpha leads to a marked increase in DNA unwinding due to loading of Cdc45 and RPA (66). Therefore, it is possible that areas of extensive DNA unwinding at origins in the ORC6-rxl CDC6ΔNT strains could arise and be more vulnerable to DSBs, thereby accounting for much of the synthetic lethality involving deletions of DNA damage response and repair genes that we obtained in our screen. ORC6-rxl CDC6ΔNT mutants could accumulate pre-RC-derived structures that impose a block on incoming forks. In this model, the fork stabilization activity of Mec1 could be required to prevent the collapse of forks that struggle to pass through the barriers while they are being resolved. The rescue of ORC6-rxl CDC6ΔNT cells by the deletion of MRC1 or TOF1 could indicate that the Mrc1-Tof1 complex is required to establish the initiation intermediates causing the problem. This scenario would be consistent with the interaction of the Mrc1-Tof1 complex with Cdc45 and with its function in coupling the helicase (MCMs) with the replisome (27).
While reinitiation from an origin of replication within a single cell division cycle is frequent in prokaryotes, it is tightly repressed in eukaryotes. While prokaryotic genomes are mostly contained in one chromosome with a single origin, the complexity and size of eukaryotic genomes, existing on multiple chromosomes, require the establishment of a spindle apparatus that relies on the synthesis of two paired copies of each chromosome for the faithful segregation of the replicated genome. Furthermore, the large size of the genome increases the probability of generating an inviable cell due to uneven replication. These are likely reasons that may have pushed early eukaryotes to evolve mechanisms to prevent rereplication by allowing each origin to fire only once per cell cycle. The fact that multiple mechanisms exist to prevent rereplication suggests that any individual mechanism is leaky and that a leak is detrimental to the cell. The appearance of any new mechanism adding to the prevention of rereplication presumably had to provide an advantage even in the presence of the preexisting mechanisms in order to be selected. Therefore, removing only one control should inflict a selective disadvantage, at least in a wild environment. In this study, we could readily observe active responses in cells combining two mutations (ORC6-rxl CDC6ΔNT). We have not examined if a weaker selective disadvantage or response to rereplication occurred in single mutants with mutations in ORC, MCMs, or Cdc6, but we suspect that one could be observed using more sensitive assays. It is possible that even in a wild-type cell, rereplication events occur at a rate that is sufficient for the selection of machinery capable of responding to DNA damage specific to rereplication. Therefore, studying how cells respond to rereplication, even cells with multiple mutations, is likely to be relevant to regulation in wild-type cells. In any case, cancer cells, which typically require multiple transforming mutations, have been shown to amplify sections of their genomes, making rereplication an important subject of study from a medical perspective.
In mammalian cells, rereplication induced by the overexpression of Cdc6, Cdt1, and cyclin A-Cdk2 triggers a DNA damage response involving ATM/ATR, Chk2, p53, and p21. This response leads to inhibition of cyclin A-Cdk2, which represses rereplication (64). In another case, the inactivation of Cdk2 in S phase led to chromatin loading of MCMs and to the activation of the ATM/ATR-p53 pathway that was required to prevent rereplication (74). Rereplication can also be induced by the depletion of geminin, the Cdt1 inhibitor (73). In this context, a G2/M checkpoint is activated, leading to the inhibition of Cdc2 through Chk1 and Cdc25C signaling, and surprisingly, the p53-dependent pathway is not activated. Therefore, at least two pathways can lead to a repression of rereplication in mammalian cells.
The signaling of DNA damage by an [MRX-Mec1 (Ddc2)-(Rad17-Ddc1)]-Rad53 pathway, the likely Rad52-dependent repair of DSBs (possibly with Srs2, Ctf4, Mms1, and Mms22), and the limitation of the extent of rereplication by at least Mre11, Mec1, and Rad17 are all pathways that differ from the known responses to rereplication in mammalian cells. Because S. cerevisiae does not have a known p53 ortholog and because its Cdc25 (Mih1)-dependent G2/M checkpoint is not as central to cell cycle control as that in mammalian cells, it should be no surprise if other pathways are used to respond to rereplication by inducing a cell cycle arrest and promoting genomic stability and survival. The reliance on DNA damage response pathways that we uncovered could be specific to cells that have an incomplete removal of mechanisms preventing rereplication. We speculate that allowing rereplication without any restraint might not induce DNA damage, but instead might create problems in mitosis or problems due to imbalances in gene dosages. It would be interesting to determine if fork stabilization by ATM/ATR (orthologs of Mec1) and DSB signaling through the MRN (Mre11-Rad50-Nbs1; ortholog of MRX) complex, ATM/ATR, the 9-1-1 complex, and Chk2 (Rad53 ortholog), as well as homologous recombination repair, are important responses that contribute to prevent genome instability in mammalian cells that rereplicate. These conserved pathways could play a central role in this response.
We thank Alison North for help with DeltaVison microscopy. V.A. thanks Michael P. Rout for providing laboratory space during the initial part of this work. We thank Lea Schroeder and Veronica Campbell for technical assistance and John Petrini (Memorial Sloan-Kettering Cancer Research Institute), Xiaolan Zhao (The Rockefeller University), John Diffley (Clare Hall Laboratories, Cancer Research United Kingdom), Stephen Elledge (Harvard University, Cambridge, MA), and Steve Brill (Rutgers University) for strains. We also thank Stephen Elledge for the gift of anti-Rad53 serum. We thank the reviewers for their careful readings of the manuscript and their useful comments.
Funding was provided by PHS grant GM047238 to F. C. A.E.I. was funded by the Charles H. Revson Foundation.
†Supplemental material for this article may be found at http://mcb.asm.org/.