Apoptosis in budding yeast caused by defects in initiation of DNA replication

In eukaryotic cells, DNA replication is coupled to the cell-division cycle by the reciprocal regulation of proteins required for initiation at origins of DNA replication and molecular events regulated by cyclin-dependent kinases (CDKs) that drive the cell cycle forward. Origins of replication are `licensed' for activation in S phase by the assembly in G1 of prereplicative complexes (pre-RCs) containing the origin recognition complex (ORC) and other proteins, including the ORC-interacting initiation protein Cdc6. To ensure that replication of chromosomes occurs just once each cell cycle, activation of origins in S phase coincides with the inactivation and disassembly of pre-RCs, and CDKs and other proteins block their reassembly until the end of mitosis, when CDK activity is destroyed (for a review, see Diffley, 2001). In G1 and early S phase cells, CDKs and/or CDK-dependent molecular events required for completion of S phase and mitosis are, in turn, inhibited by checkpoints that depend on pre-RC proteins, either directly (Clay-Farrace et al., 2003; Oehlmann et al., 2004; Toyn et al., 1995; Weinreich et al., 2001) or indirectly through the establishment of DNA replication forks (Tercero et al., 2003). The reciprocal regulation of pre-RCs, CDKs and/or CDK-dependent events couples DNA replication to the cell cycle, providing a stringent, fail-safe mechanism for ensuring that a single but complete copy of the genome is produced each time a cell divides.

Uncoupling of DNA replication from the cell-division cycle might be an important feature of apoptosis. This was first suggested by similarities between apoptosis and mitotic catastrophe in cells with incompletely replicated DNA (for review, see King and Cidlowski, 1995). Apoptosis also coincides with the destruction of the pre-RC proteins Cdc6 (Blanchard et al., 2002; Burhans et al., 2002; Illenye and Heintz, 2004; Pelizon et al., 2002; Schories et al., 2004; Yim et al., 2003; Yin et al., 2004) and Mcm3 (Schwab et al., 1998), the CDK inhibitors Wee1 (Zhou et al., 1998), p21 and p27 (Levkau et al., 1998) (for a review, see Gartel and Tyner, 2002), and the checkpoint proteins ATM (Smith et al., 1999), Rb (Tan et al., 1997) and Rad9 (Lee et al., 2003). The destruction of Cdc6, p21, Rad9 and Rb contributes to apoptosis because cells can be rescued from apoptotic stimuli, downstream of the activation of caspases, by expressing mutant versions of these proteins that cannot be cleaved by caspases (Gartel and Tyner, 2002; Huang et al., 1999; Lee et al., 2003; Pelizon et al., 2002; Tan et al., 1997; Yim et al., 2003). Apoptosis also often coincides with the premature activation of S phase and mitotic CDKs that are restrained by pre-RC-dependent and other checkpoints (for reviews, see Castedo et al., 2002; Guo and Hay, 1999; King and Cidlowski, 1995; Meikrantz and Schlegel, 1995). In addition to the attenuation of apoptosis that occurs when destruction of CDK inhibitors are blocked, apoptosis is also attenuated when CDK activation is directly inhibited by, for example, dominant-negative CDK mutants (for reviews, see Castedo et al., 2002; Gartel and Tyner, 2002; Guo and Hay, 1999). All these findings indicate that uncoupling DNA replication from the cell cycle by events that simultaneously inhibit initiation of DNA replication and abrogate checkpoints can be an important component of a commitment to cell death during apoptosis.

Several features of apoptosis that were detected in metazoans have also been detected in yeasts, including the production of reactive oxygen species (ROS) and the activation of a recently discovered metacaspase, Yca1p (for reviews, see Burhans et al., 2003; Madeo et al., 2002a; Weinberger et al., 2003). Here, we show that, in budding yeast, the temperature-sensitive orc2-1 mutation, which uncouples DNA replication from the cell cycle by simultaneously inhibiting pre-RC assembly (Santocanale and Diffley, 1996) and checkpoints that depend on DNA replication forks to restrain subsequent cell-cycle events (Shimada et al., 2002; Shirahige et al., 1998b; Weinberger et al., 1999) induces ROS and activation of Yca1p coincidently with the activation of a Rad53p-dependent response to DNA damage. In contrast to a recent report that Yca1p activation signals are not associated with lethality and only occur in lysed cells (Wysocki and Kron, 2004), activation of Yca1p by the orc2-1 mutation was detected in cells with intact membranes and contributed to the lethal effects of this mutation. The results of our experiments suggest that uncoupling of DNA replication from the cell cycle contributes to apoptosis by inducing DNA damage. They also provide definitive evidence that, similar to mammals, DNA damage can induce apoptosis in budding yeast and emphasize the important role that the Yca1p metacaspase plays in budding yeast apoptotic programs.

Standard protocols for growth of yeast were used (Guthrie and Fink, 1991). Cells were grown in YPD except where plasmid maintenance required auxotrophic selection in defined medium, and YP-galactose was used to induce galactose-regulated gene expression. Cell cultures were maintained at an early to mid-log-phase state of growth. ROS and propidium iodide (PI) were detected by flow cytometry (FACS) using 2′,7′-dichlorodihydrofluorescein diacetate (H₂DCF-DA, Molecular Probes) to detect ROS as described in (Madeo et al., 1997) and a Becton Dickinson FACScan equipped with a 488 nm argon laser. Emission signals were detected at 525-550 nm. Before FACS analysis, cells were resuspended in 50 mM sodium citrate containing 1 μg ml^-1 PI (Sigma). Caspase activity was detected using FITC-VAD-FMK (Promega) according to the manufacturer's instructions. In brief, 5×10⁶ cells were stained with 100 μM FITC-VAD-FMK and 1 μg ml^-1 PI at room temperature in the dark for 20 minutes. They were then washed and resuspended in 1× PBS before analysis by FACS at an excitation wavelength of 488 nm and emission wavelength of 525-550 nm. Effects of a YCA1 deletion (yca1Δ) on viability were determined by colony-forming assays. Growth in rich medium with proper aeration was crucial for optimal detection of ROS and activated caspase signals, and for detecting the effects of deleting the YCA1 gene on viability. In general, growth in minimal medium (which was required for experiments involving ectopic expression of genes from plasmids) reduced the frequency with which ROS and activated caspase signals were detected compared with growth in rich medium. FACS measurements of DNA content were as described previously (Weinberger et al., 1999), except that Sytox Green (Molecular Probes) was used instead of PI. DAPI staining of cell nuclei was as described previously (Guthrie and Fink, 1991). FACS data were processed and quantified using Cell Quest (Becton Dickinson) and WinList (Verity) software. Values presented in graphs are means of three or more independent experiments or are representative of similar results in multiple independent experiments. Photomicrographs of stained cells were obtained at 400× magnification with a Zeiss Axioskop microscope equipped with an Optronics Magnafire CCD camera.

Standard protocols for mating and random spore analysis were used (Guthrie and Fink, 1991). Cell-viability experiments involving temperature shifts, drug treatments and cell-cycle arrests were as previously described (Weinberger et al., 1999). The initial G1-phase arrest in cell-staging experiments was at 23°C with 12 μM α mating factor (orc2-1 strains) or 6 μM α factor (all other strains). After 2.5 hours, an equivalent amount of α factor was added and aliquots of the G1 arrested cultures were incubated for an additional 3 hours at 23°C and 35°C. Each G1-phase-arrested culture was then split into three additional aliquots. The G1-phase arrest was maintained in one culture and the others were released from G1 phase by filtering the culture at the same temperature as G1 arrest and resuspending cells in medium without α factor. All cultures were incubated for an additional 2 hours at the appropriate temperature before measuring cell viability. The proteosome inhibitor MG132 was added to a final concentration of 250 μM and the antioxidant N-acetyl-L-cysteine (NAC) was added to a final concentration of 4 mM.

Crude lysates were prepared by bead beating in lysis buffer [100 mM NaCl, 25 mM Tris-HCl, 10% glycerol, 0.1% Triton X-100 with one protease inhibitor tablet (Roche) per 5 ml]. 30 μg protein were size fractionated on 12% PAGE gels and blotted proteins were visualized by immunodetection using an enhanced chemiluminescence (ECL) kit from Amersham. All antibodies were obtained from Santa Cruz Biotechnology.

All strains were in the W303 background, and lists of strains and plasmids are provided in Table 2. YCA1 was replaced in wild-type and orc2-1 cells by standard techniques using a PCR-generated G418 resistance cassette (Bahler et al., 1998) and the following primers 5′-ACATAATAAATTGCAGATTTACGTCAATAGGGTGTGACGATGATAATTGTGGTTTTTGAGGAATTCGAGCTCGTTTAAAC-3′ and 5′-ATGAAGATGAGCCTCGAAGTTTATCTAAACTACCACCAAAGAAGACCGACTAGATTTACACGGATCCCCGGGTTAATTAA-3′. Confirmation of the disruption was obtained by Southern blots and sequence analysis.

The temperature-sensitive orc2-1 mutation causes lethality at nonpermissive temperatures for reasons that are not clear. Diploid orc2-1 cells undergo DNA fragmentation and Rad9p-dependent cell death at higher temperatures (Watanabe et al., 2002), which suggests the induction of an apoptosis-like cell-death program. Consistent with this possibility, shifting GA1410 cells harboring the orc2-1 mutation, but not isogenic wild-type cells, to 37°C induced another common feature of apoptosis, ROS. ROS were present in cells at several stages of the cell cycle, including S phase, as indicated by different bud sizes (Fig. 1A). Even at the semipermissive temperature of 23°C, at which ORC function is partly compromised in orc2-1 cells (Santocanale and Diffley, 1996), a several-hundredfold increase in numbers of ROS-producing cells was detected above the low (<0.1%) background levels of ROS-producing wild-type cells. FACS measurements of the number of ROS-positive orc2-1 cells (Fig. 1B) showed that ROS production in these cells was temperature and time dependent (Fig. 1C). Parallel assessments of cell death detected by membrane changes that allow for staining of DNA in dead or dying cells by propidium iodide, a membrane-impermeable dye, indicated that both ROS production and cell death were rapid, and suggested that ROS production can precede loss of viability (Fig. 1C).

Induction of many ROS-producing cells by high temperatures depended on the temperature-sensitive orc2-1 mutation in GA1410 cells, because production of ROS in these cells was suppressed by ectopic expression of wild-type Orc2p (Fig. 1D), consistent with the recessive nature of the orc2-1 mutation. ROS induced by the orc2-1 mutation was also partly suppressed by ectopic expression of the co-chaperone protein Mge1p (Fig. 1E,F), which suppresses the initiation defect of orc2-1 cells and stimulates initiation of DNA replication in wild-type cells (Trabold et al., 2005). At semipermissive temperatures, steady-state levels of orc2-1p are approximately ten times less than levels of Orc2p in wild-type cells, owing to the more rapid turnover of orc2-1p by the proteasome (Shimada et al., 2002). Shifting GAL–orc2-1 cells to 37°C when grown in galactose to induce the production of high levels of orc2-1p from a galactose-inducible promoter (in order to more easily detect orc2-1p) also caused levels of orc2-1p to decline, but not when these cells were also treated with the proteasome inhibitor MG132 (Fig. 1G). Thus, higher temperatures accelerate the turnover of orc2-1p even further. Inhibition of the proteasome by MG132 also rescued orc2-1 cells from the lethal effects of high temperatures (Fig. 1H) and suppressed the production of ROS (Fig. 1I). Therefore, loss of viability and increased ROS at nonpermissive temperatures are related to acute loss of orc2-1p.

Shifting GA1410 orc2-1 cells to high temperatures also activated the recently discovered Yca1p metacaspase in budding yeast (Madeo et al., 2002b), as detected by a standard assay used in mammalian cells to detect activated caspases in cells undergoing apoptosis (Fig. 2). Substantially more cells with activated metacaspase (as indicated by staining with FITC-VAD-FMK) were detected in cultures of these orc2-1 cells than in wild-type cells after a shift to the nonpermissive temperature of 35°C (Fig. 2A) [because ROS production was greatest at 35°C (Fig. 1C), Yca1p activation was measured after shifting to this temperature rather than 37°C]. Parallel incubation of these cells with PI indicated a much smaller increase in the number of cells that were stained with both PI and FITC-VAD-FMK (Fig. 2A). Therefore, most of the cells that were stained with FITC-VAD-FMK had not lost membrane integrity. This is in stark contrast to a recent report that the caspase activation signals detected by this assay in yeast cells are an artefact related to cell lysis (Wysocki and Kron, 2004). Furthermore, the number of cells with activated caspase signals was substantially reduced in orc2-1 cells with YCA1 deleted (Fig. 2B). In addition, partial restoration of caspase activation signals occurred in orc2-1 yca1Δ cells, but not in wild-type cells expressing Yca1p ectopically from a plasmid (Fig. 2C). These results unambiguously establish that activation of Yca1p is responsible for the caspase activation signals detected in orc2-1 cells with FITC-VAD-FMK. More cells with activated caspases were also detected in orc2-1 yca1Δ, but not wild-type cells transformed with an empty vector control plasmid in this experiment. This suggests the existence of additional, unidentified caspase-like molecules that are also activated by the orc2-1 mutation. Deletion of YCA1 also partly suppressed the temperature sensitivity associated with the orc2-1 mutation (Fig. 2D). Therefore, Yca1p contributes to the lethal effects caused by loss of Orc2p function. This also contrasts with the results of (Wysocki and Kron, 2004), which failed to establish a role for Yca1p in the death of a different temperature-sensitive budding-yeast mutant.

One function during G1 phase of ORC and other initiation proteins, including the ORC-interacting protein Cdc6, is to establish pre-RCs that `license' origins of replication for activation in S phase (Diffley, 2001). As reported previously (Weinberger et al., 1999), loss of viability in orc2-1 cells shifted to nonpermissive temperatures was significantly greater in cells first arrested in G1 phase with mating pheromone compared with cells synchronized in S phase before the temperature shift (Fig. 3A, left). More ROS-producing cells were also detected in G1-phase-arrested than in S-phase-arrested orc2-1 cells (Fig. 3A, right). Thus, lethality and apoptosis in orc2-1 cells are induced by loss of Orc2p function in G1 phase, and S-phase cells are more refractory to these effects.

Mating pheromone arrests cells in G1 after pre-RCs are assembled (Detweiler and Li, 1997). Therefore, although apoptosis associated with loss of Orc2p function in log-phase cultures of cells could be due to inhibition of the assembly of pre-RCs, apoptosis in pheromone-arrested cells might be caused by a failure to maintain pre-RCs at high temperatures during the pheromone-induced arrest, similar to the failure of Cdc6 to maintain pre-RCs in temperature-sensitive cdc6-1 cells shifted to high temperatures during a pheromone arrest (Detweiler and Li, 1997; Piatti et al., 1995). Loss of this maintenance function in G1-arrested cdc6-1 cells shifted to 37°C is probably responsible for the less-efficient DNA replication detected when these cells were subsequently released into S phase at this temperature, in contrast to the more-efficient DNA replication that occurred in cdc6-1 cells synchronized in G1 and then released into S phase at the permissive temperature of 23°C (Fig. 3B). The orc2-1 cells also replicated DNA less efficiently when shifted to 37°C after they were synchronized in G1 phase with pheromone and then released into S phase at 37°C, but not when they were synchronized and released from the pheromone block at 23°C (Fig. 3C). Less-efficient DNA replication in orc2-1 cells was not related to combined effects of high temperature on orc2-1 function and other processes, because GAL–orc2-1 cells depleted of orc2-1p by glucose repression of transcription for 2 hours at 23°C instead of by high temperature also replicated DNA less efficiently when released from a mating-pheromone block in G1 phase (Fig. 3D). This contrasts with the lack of inhibitory effects on DNA replication previously detected in these same cells after glucose inhibition of orc2-1p production for 30 minutes, which was interpreted as evidence that ORC is not required for maintenance of pre-RCs (Shimada et al., 2002). Our experiments used a 2-hour depletion of orc2-1p from GAL–orc2-1 cells before release into S phase and, in our temperature shift experiments using GA1410 orc2-1 cells, the temperature was raised for 3 hours before release occurred. An extended period of time required for disassembly of pre-RCs was previously observed in cdc6-1 cells shifted to 37°C while blocked in G1 phase (Detweiler and Li, 1997) and in G1-phase cells depleted of the pre-RC protein Mcm2p for 1.5 hours by degron-mediated degradation (Labib et al., 2001). Therefore, similar to Cdc6p and Mcm2p, Orc2p is required for maintenance of pre-RCs in budding yeast over longer periods of time. Consequently, apoptosis induced by high temperatures in orc2-1 cells that were first synchronized in G1 phase with mating pheromone is probably caused by loss of this maintenance function.

In some genetic backgrounds, pre-RC and other initiation proteins are required for a G1/M-phase checkpoint that restrains mitosis in G1 cells (Toyn et al., 1995; Weinreich et al., 2001). Divided chromatin, which is a signature of a catastrophic mitosis, was sometimes detected in unsynchronized cultures of orc2-1 cells shifted to higher temperatures (not shown). However, in many experiments, apoptosis was detected in orc2-1 cells that arrested with undivided chromatin (Fig. 3E) and a range of bud sizes indicating a premitotic arrest in S phase (Fig. 3E and Fig. 1A). Therefore, although G1/M-phase checkpoint defects and catastrophic mitosis could cause apoptosis in orc2-1 cells, apoptosis appears to occur in S-phase cells as well.

In fact, significantly more S-phase cells were present in mating-pheromone-synchronized cultures of orc2-1 cells than in cdc6-1 cells, especially at 37°C (Fig. 3B,C, G1). Even this partial synchronization of orc2-1 cells in G1 required much more mating pheromone than for cdc6-1 or wild-type cells. This probably reflects partial derepression by the orc2-1 mutation of ORC-dependent silencing in the HML mating-type locus, which is expected to cause a defect in the response of the orc2-1 cells used in these experiments to α mating factor. Loss of viability in orc2-1 cells shifted to nonpermissive temperatures has been shown to coincide with their entry into S phase (Bell et al., 1993). Therefore, apoptosis in populations of orc2-1 cells synchronized in G1 phase might actually occur in a subset of cells that lost Orc2p function in G1 phase but then escaped the leaky G1 block by mating pheromone in these cells and entered S phase. Consistent with this possibility, shifting cdc6-1 cells to higher temperatures while maintaining them in G1 phase caused less lethality than in orc2-1 cells under similar conditions (Fig. 3G,H, G1). Moreover, both cdc6-1 and orc2-1 cells were significantly less viable when released into S phase from the G1-phase mating-pheromone arrest at 37°C (Fig. 3G,H, G1>S) than cells that were maintained in mating pheromone (Fig. 3G,H, G1). Therefore, loss of viability in both cdc6-1 and orc2-1 cells at higher temperatures requires loss of function in G1, but is exacerbated by subsequent entry into S phase. The increased lethality of orc2-1 cells released into S phase was accompanied by an increase in the number of ROS-producing cells (Fig. 3I). Therefore, although apoptosis is induced in orc2-1 cells by loss of Orc2p function in G1, it probably also requires subsequent entry into S phase.

Apoptosis in metazoans often depends on proteins required for checkpoints (for a review, see Burhans et al., 2003). Checkpoints were previously implicated in the temperature sensitivity of the orc2-1 mutation by partial suppression of this sensitivity at semipermissive temperatures in diploid orc2-1 cells with RAD9 deleted (Watanabe et al., 2002). The mec1-21 mutation in the MEC1 gene, whose product functions upstream of Rad9p in budding-yeast checkpoints, also partly suppresses the temperature sensitivity of the orc2-1 mutation (Desany et al., 1998). Production of wild-type Mec1p stimulated ROS production at both semipermissive and nonpermissive temperatures in the same orc2-1 mec1-21 double-mutant cells that showed partial suppression of orc2-1 lethality (Desany et al., 1998) (Fig. 4A). Therefore, similar to temperature sensitivity, ROS production in orc2-1 cells can partly depend on Mec1p.

The stimulation of ROS by ectopic production of Mec1p suggested that apoptosis in orc2-1 cells might be caused by DNA damage. In wild-type cells, both DNA damage and stalling of DNA replication forks with the ribonucleotide reductase inhibitor hydroxyurea (HU) induce phosphorylation by Mec1p of the checkpoint kinase Rad53p. Rad53p was phosphorylated in orc2-1 but not wild-type cells that had been shifted to 37°C (Fig. 4B). As expected, Rad53 phosphorylation was also observed in wild-type cells treated with the DNA-damaging agent methyl methanesulfonate (MMS) (Fig. 4B, + MMS 37°C). Therefore, loss of Orc2p function in orc2-1 cells induces a Rad53p-dependent DNA-damage response, similar to the response induced in wild-type cells treated with MMS. Activation of Rad53p by DNA damage or HU in wild-type cells induces the expression of HUG1 (Basrai et al., 1999). Rad53p phosphorylation induced by high temperatures in orc2-1 cells coincided with an almost tenfold increase in expression of HUG1 in orc2-1 but not wild-type cells (Fig. 4C), confirming that Rad53p is activated upon loss of Orc2p function.

As reported earlier (Shimada et al., 2002), unlike wild-type cells, increased phosphorylation of Rad53p was not detected in orc2-1 cells treated with HU at 23°C (Fig. 4B, + HU 23°C), owing to their checkpoint defects. Therefore, phosphorylation of Rad53p and induction of HUG1 expression at high temperatures in orc2-1 cells cannot be a simple consequence of replication fork stalling in the absence of additional events, which probably include damage to DNA. Rad53p phosphorylation was also not a consequence of ROS-induced DNA damage, because it was not altered by the antioxidant N-acetyl cysteine (Fig. 4B, NAC), despite a five-times reduction in levels of ROS in orc2-1 cells treated with this compound (data not shown). Furthermore, similar levels of Rad53p phosphorylation were induced by high temperatures in a second strain of orc2-1 cells (YB0057) that produces ROS at one-tenth the frequency detected in the GA1410 orc2-1 strain used in the experiments described above [Fig. 4B, orc2-1 (1/10 ROS)]. These findings strongly suggest that loss of Orc2p function in orc2-1 cells at high temperatures causes DNA damage independently of ROS (which can also damage DNA). Furthermore, because orc2-1 cells are refractory to activation of Rad53p by low but not high levels of DNA damage (Shimada et al., 2002), loss of Orc2p function appears to induce extensive DNA damage, which could be responsible for the apoptotic phenotype of orc2-1 cells at high temperatures.

To determine whether apoptosis occurs in response to the inhibition of initiation complexes by other means, we asked whether it occurs in strains harboring mutations in other proteins required for the assembly of pre-RCs. Cells harboring the temperature-sensitive cdc6-1 mutation also produced ROS when shifted to higher temperatures. However, the frequency with which ROS were produced in cdc6-1 cells, although several hundred times higher than in wild-type cells, was much lower and occurred with delayed kinetics compared with the GA1410 orc2-1 strain (Fig. 5A). ROS were also detected at levels significantly higher than in wild-type cells in several additional orc2-1 strains shifted to 37°C or depleted of orc2-1p by glucose repression (Table 1), as well as in orc2-1 strains in other genetic backgrounds shifted to 37°C (data not shown). However, like cdc6-1 cells, in all of these strains, the frequency of ROS production was lower than in GA1410 orc2-1 cells. The lower frequency of ROS production in cdc6-1 and other orc2-1 strains was not related to the potential accumulation of mutations in these strains that suppressed production of ROS, because ROS levels did not increase after extensive backcrossing of these strains with wild-type cells to remove suppressors (data not shown). By contrast, although most orc2-1 segregants of a backcross between GA1410 orc2-1 and wild-type cells produced ROS at high temperatures at a frequency similar to GA1410 orc2-1 cells, some segregants of this backcross produced ROS at a frequency more similar to cdc6-1 and several other orc2-1 strains (Fig. 5B, orc2-1 segregants). In addition, at high temperatures, many of the wild-type ORC2 segregants of this backcross produced ROS at a low frequency, although this was still significantly higher than the very low frequency (<0.1%) detected in parental wild-type cells (Fig. 5B, ORC2 segregants). These findings confirmed that high levels of ROS in GA1410 orc2-1 cells depend on the orc2-1 mutation but suggested that GA1410 orc2-1 cells harbor additional loci that stimulate ROS and segregate independently of the orc2-1 mutation. This was confirmed by measuring ROS levels in segregants of a cross between YB0057 orc2-1 cells (which produce ROS at a frequency of 3%) and wild-type ORC2 segregants [ORC2 (ROS+)] from the GA1410 orc2-1 × ORC2 mating (which produce ROS independently of the orc2-1 mutation at a frequency of 7%). ROS were detected in orc2-1-containing segregants of this backcross at a much higher frequency (∼35%) than either parent and at a frequency very similar to that detected in GA1410 orc2-1 cells (Table 1, YMW102). Similarly, although ROS production was not detected above background wild-type levels in a different ORC mutant, orc5-1 (Table 1, YB0059), some orc5-1-containing segregants of a cross between orc5-1 and ORC2 (ROS+) cells produced ROS at high frequencies, similar to the frequency of ROS production in GA1410 orc2-1 cells (Table 1, YMW103). The detection of ROS in other orc2-1 strains and in association with mutations in other initiation proteins indicates that apoptosis is a general consequence of defects in the assembly and maintenance of pre-RCs. However, in GA1410 orc2-1 cells, ROS levels are increased by additional temperature-sensitive loci residing in this strain. At high temperatures, these loci produce ROS independently of the orc2-1 mutation but at much lower levels than detected in GA1410 orc2-1 cells.

The amplification of ROS by undefined loci in GA1410 orc2-1 cells in concert with the induction of a Rad53p-mediated DNA damage response suggested that these loci respond to DNA damage. Consistent with earlier reports that ultraviolet radiation (Del Carratore et al., 2002) or the cdc13-1 mutation – which causes the accumulation of DNA damage (Qi et al., 2003) – induce apoptosis in budding-yeast cells, MMS or adozelesin (an experimental antitumor drug that, like MMS, damages DNA) induced ROS in W303 wild-type cells, albeit at a low frequency compared with GA1410 orc2-1 cells (Fig. 5C, W303). ROS were also induced at a low frequency in a wild-type ORC2 segregant of the GA1410 orc2-1 × ORC2 backcross that produces very little ROS at high temperatures, and therefore lacks the additional ROS-amplifying loci [ORC2 ROS (–)]. By contrast, exposure to adozelesin or MMS at 23°C induced ROS at a much higher frequency in GA1410 orc2-1 cells (Fig. 5C, GA1410 orc2-1) and in a wild-type ORC2 segregant of the GA1410 orc2-1 × W303 backcross, both of which contained these loci {Fig. 5C, ORC2 [ROS (+)]}. In fact, despite the performance of this experiment at a lower temperature that produces much less ROS in GA1410 cells, the frequency of ROS induced in these cells by DNA-damaging agents equaled or exceeded the frequency of ROS production in GA1410 orc2-1 cells at high temperatures. Therefore, the additional loci in GA1410 orc2-1 cells that amplify the frequency of ROS induced at high temperatures also enhance the apoptotic response of budding-yeast cells to DNA damage independently of the orc2-1 mutation.

Important but poorly understood connections exist between cell-cycle-regulatory pathways and apoptosis in mammalian cells. In addition to the inappropriate activation of S phase and mitotic CDKs (Castedo et al., 2002; Guo and Hay, 1999; King and Cidlowski, 1995; Lukovic et al., 2003; Meikrantz and Schlegel, 1995), which can be required for apoptosis (for reviews, see Castedo et al., 2002; Gartel and Tyner, 2002; Guo and Hay, 1999), apoptosis is often accompanied by the caspase- and/or proteasome-dependent destruction of multiple DNA replication and/or checkpoint proteins that directly or indirectly inhibit CDK activity (Blanchard et al., 2002; Flygare et al., 2000; Huang et al., 1999; Itoh and Horio, 2001; Lee et al., 2003; Levkau et al., 1998; Pelizon et al., 2002; Smith et al., 1999; Tan et al., 1997; Yim et al., 2003; Zhou et al., 1998). Apoptosis also coincides with the caspase destruction of proteins required for DNA repair (Flygare et al., 2000; Huang et al., 1999; Itoh and Horio, 2001). Caspase-dependent destruction of many of these DNA replication, checkpoint and DNA repair proteins is required for full induction of the apoptotic machinery, because non-caspase-cleavable mutants of some of these proteins can block apoptosis (Gartel and Tyner, 2002; Huang et al., 1999; Lee et al., 2003; Pelizon et al., 2002; Tan et al., 1997; Yim et al., 2003). Therefore, despite the frequent assumption that the commitment to cell death during apoptosis must occur upon activation of caspases, in at least some cases this commitment clearly occurs downstream of the activation of caspases and requires the simultaneous inhibition of DNA replication and checkpoints in concert with the unscheduled activation of CDKs and the induction of DNA damage. The nature of this commitment is not clear.

The findings reported here establish that, in budding yeast, mutations that partly uncouple DNA replication from the cell cycle by simultaneously reducing the number of pre-RCs and attenuating checkpoints induce DNA-damage responses and an apoptosis-like phenotype. This phenotype includes the production of ROS (Fig. 1) and activation of Yca1p, a recently described budding-yeast metacaspase (Fig. 2). A role for Yca1p in yeast apoptosis and the validity of the concept of apoptosis in yeasts was recently challenged by experiments that used extraordinarily lethal conditions allowing for a maximum of 0.001% survival (Wysocki and Kron, 2004). Unlike the Wysocki and Kron study, but similar to other studies that (like our study) used far less lethal conditions more appropriate for studying cell-death pathways (Bettiga et al., 2005; Fahrenkrog et al., 2004; Madeo et al., 2002b; Wadskog et al., 2004) (C. Mazzoni et al., personal communication), our experiments in orc2-1 cells clearly show that disruption of YCA1 in budding-yeast cells undergoing apoptosis-like cell death substantially reduces caspase activation signals in parallel with a significant increase in viability (Fig. 2). In conjunction with the production of ROS (Fig. 1) and an earlier report of DNA degradation and checkpoint-dependent cell death induced by the orc2-1 mutation (Watanabe et al., 2002), these results establish that many of the orc2-1 cells used in our experiments were undergoing an apoptosis-related cell death.

Apoptosis in orc2-1 cells requires loss of Orc2p function in G1 and subsequent entry into S phase or mitosis (Fig. 3). Although loss of Orc2p function is expected to reduce the number of pre-RCs in G1 cells by inhibiting pre-RC formation, our results indicate that, in budding yeast, Orc2p is also required for maintaining pre-RCs once they are formed. Although this conclusion is contrary to that of Shimada et al. based on a similar experiment (Shimada et al., 2002), depletion of orc2-1p in that earlier study might not have occurred for long enough to allow the disassembly of pre-RCs.

Apoptosis in orc2-1 cells is accompanied by activation of the checkpoint protein Rad53p (Fig. 4B). Owing to the attenuation of checkpoints by a reduction in the number of replication forks below a threshold required for full checkpoint activation, orc2-1 cells are refractory to activation of Rad53p by stalled replication forks or low levels of DNA damage (Fig. 4) (Shimada et al., 2002). Therefore, Rad53p activation at high temperatures in orc2-1 cells is probably induced by extensive DNA damage associated with acute loss of ORC function. In metazoans, apoptosis is induced by DNA damage that exceeds the capacity for repair. We consider it likely that extensive DNA damage exceeding capacity for repair is also responsible for the apoptotic phenotype of initiation mutants in budding yeast.

DNA damage caused by initiation defects in our experiments is consistent with recent evidence that, in yeast and mammals, defects in initiation activate DNA-damage responses (Jacobson et al., 2001; Kim et al., 2002; Liang et al., 1999; Watanabe et al., 2002) and cause gross chromosomal rearrangements (GCRs) (Huang and Koshland, 2003; Lengronne and Schwob, 2002; Tercero et al., 2003; Watanabe et al., 2002). Why defects in initiation cause these effects has not been clear. DNA damage could be caused by inappropriate entry of cells into anaphase with partly replicated chromosomes, as has been detected in budding-yeast cells with fewer pre-RCs owing to unscheduled activation of S-phase CDKs (Lengronne and Schwob, 2002). It was proposed that the absence of checkpoint responses observed in these experiments reflected the failure of budding-yeast cells to evolve mechanisms for sensing ongoing DNA replication when an extended period of time is required to finish replicating chromosomes in cells with reduced numbers of origins (Lengronne and Schwob, 2002). Genome instability associated with defects in initiation has also been suggested to arise during S phase in the absence of DNA damage, owing to elevated recombination at intrinsically stalled replication forks that require more time for rescue by forks emanating from more distal origins (Tanaka and Diffley, 2002).

Both these models fail to consider the dual role of initiation complexes in DNA replication and in checkpoints that directly or indirectly depend on initiation complexes (Shirahige et al., 1998a; Tercero et al., 2003; Weinberger et al., 1999). In this context, DNA damage associated with inappropriate entry of initiation mutants into mitosis with incompletely replicated chromosomes is more likely to be caused by checkpoint failure than by a failure to evolve checkpoints in cells with an extended S phase. Similarly, DNA damage caused by checkpoint failure, rather than recombination in the absence of DNA damage, might underlie genome instability that arises during S phase in strains with partial defects in initiation. Checkpoints stabilize stalled DNA replication forks, which collapse in checkpoint-defective cells (Sogo et al., 2002). Induction of DNA damage during S phase, as well as mitosis, is consistent with the Rad9-dependent G2 arrest detected in these strains, including those harboring the orc2-1 mutation (Jacobson et al., 2001; Watanabe et al., 2002).

Our findings suggest a model for how the targeted destruction of highly conserved DNA replication initiation, checkpoint and repair proteins, and activation of CDKs might contribute to apoptosis in mammalian cells. Similar to budding yeast, this probably leads to DNA damage when cells with a reduced number of origins and defective checkpoint and DNA repair mechanisms enter mitosis with incompletely replicated DNA. This is consistent with an extensive literature describing the induction by a variety of stimuli of an apoptosis-like cell death that includes mitotic catastrophe driven by activation of the mitotic CDK, CDK1 (Castedo et al., 2002). The relationship between apoptosis and mitosis is not entirely clear, however, partly because some mitotic events are not always detected in apoptotic cells (Zhou et al., 1998). The detection of apoptosis in orc2-1 cells in S phase (Fig. 1A) in the absence of a mitotic catastrophe (Fig. 3E) suggests an additional possibility – that DNA damage leading to apoptosis can be induced by uncoupling DNA replication from S-phase, as well as mitotic, events. This is consistent with the activation during apoptosis of CDK2/cyclin-A, which functions in S phase and is required for apoptosis in some cases (Meikrantz et al., 1994). Subsequent full engagement of the apoptotic machinery in S phase might lead to abortive mitotic entry or even block mitosis altogether, depending on other factors, such as the status of parallel p53-dependent mitotic checkpoints.

As in budding yeast, simultaneous inhibition of DNA replication and checkpoints during apoptosis is expected to cause DNA damage specifically at replication forks that collapse while stalled in S phase or during mis-segregation of partly replicated chromosomes during mitosis. Interestingly, several studies indicate that, during early stages of apoptosis induced by a range of apoptotic stimuli in mammalian cells, large fragments of DNA (50 kb to 1 Mb) corresponding to looped domains of chromatin are detached from chromosomes before detection of the nucleosome-sized fragments of DNA at later stages of apoptosis (Li et al., 1999; Ploski and Aplan, 2001; Solovyan et al., 2002) (for reviews, see Higuchi, 2003; Nagata et al., 2003). The production of these large replication-sized DNA fragments can amplify an initial low caspase-dependent apoptotic response, thus establishing a causal role for these fragments in the apoptotic program downstream of the activation of caspases (Boulares et al., 2001). Although the mechanism by which these DNA fragments are produced remains unclear, topoisomerases I and II (both of which are essential for DNA synthesis at replication forks) have been implicated in their formation during the execution phase of apoptosis (for a review, see Sordet et al., 2003). Other experiments suggest that stalled replication forks can serve as `suicide' substrates for topoisomerases (Snapka et al., 1991). In the context of our findings and the generally accepted notion that DNA replication takes places at the base of looped domains of chromatin, there is the interesting possibility that some of the large, replicon-sized fragments detected at early stages of apoptosis are produced by topoisomerase-mediated replication-fork collapse when DNA replication becomes uncoupled from subsequent cell-cycle events. This model clearly does not extend to all forms of apoptosis. We think it is possible, however, that it describes a conserved core apoptotic program that might have been augmented or superceded by the evolution of more specialized mechanisms in differentiated cells of multicellular organisms.

This work was supported by a grant from the US National Institutes of Health to W.C.B. (CA084086) and by an NCI Cancer Center Support Grant (P30 CA016056) to Roswell Park Cancer Institute. This Cancer Center Support Grant partially supports the flow cytometry and biopolymers facilities used in this study.