The DNA replication checkpoint aids survival of plants deficient in the novel replisome factor ETG1

Previously, we have identified 70 conserved plant E2F target genes (Vandepoele et al, 2005). This list holds 40 known regulators of DNA replication and chromatin dynamics, but also 21 genes with unknown function. To identify novel S‐phase regulatory genes, we screened the latter phenotypically with T‐DNA insertion lines. One of the insertion lines showed an endoreduplication phenotype and was designated E2F TARGET GENE 1 (ETG1; At2g40550). To address the role of the ETG1 gene in plant growth and development, we analysed the loss‐of‐function effect of ETG1 with two independent T‐DNA insertion lines. The T‐DNA was inserted into the first intron (etg1‐1; SALK_071046) or last exon (etg1‐2; SALK_145460) of the ETG1 gene (Figure 1A). ETG1 transcripts were not detected in the etg1‐1 mutant, whereas the transcript level was 80% reduced in etg1‐2, when compared to control plants (Figure 1B). In etg1 mutant seedlings, plant growth appeared macroscopically normal (Figure 1C). However, by comparing the DNA ploidy level of wild‐type leaves with that of etg1‐1 and etg1‐2 mutants, the distribution of the C values was slightly, but significantly, changed: in etg1 mutants, the population of cells with an 8C and 16C DNA ploidy level had increased, demonstrating that deficiency in ETG1 stimulated endoreduplication (Figure 1D). A similar effect was seen in root tissue (Supplementary Figure S1). The increase in the DNA ploidy level of etg1 mutants probably originated from an advanced onset of the endocycle, as illustrated by the faster increase in the population of cells with an 8C and 16C DNA content during leaf development (Supplementary Figure S2).

When the first leaf pair from wild‐type and etg1 mutant plants at maturity was compared, the leaf blade area was almost identical for both genotypes (Figure 1F). By contrast, the average abaxial pavement cell area had increased significantly in the mutant plants (Figures 1E and G), accompanied with a decrease in cell number per leaf (Figure 1H). At the bolting stage, younger leaves showed a slightly elongated and serrated leaf phenotype (Figures 1I and J), resembling the phenotype observed in plants whose cell division is inhibited by ectopic expression of the CDK inhibitory gene KRP2 (De Veylder et al, 2001). In addition, the root growth rate of the mutant plants was significantly reduced (Supplementary Figure S3).

To study the effect of loss of ETG1 function on cell cycle progression in more detail, kinematics of leaf growth was analysed. From day 5 until day 22 after sowing, the first leaves of etg1‐1 and wild‐type plants grown side by side under the same conditions were harvested and the leaf blade area and the average cell area of the abaxial epidermal cells was measured by image analysis (De Veylder et al, 2001; Boudolf et al, 2004). The total cell number was extrapolated as the ratio of leaf blade and average cell areas. Although the leaf blade area was similar in the wild‐type and etg1‐1 plants during the whole period of leaf development (Figure 2A), the average cell area, which initially was approximately 100 μm² in both plants, increased significantly faster in the etg1‐1 mutant (Figure 2B). The average surface area of etg1‐1 cells was 155% that of wild‐type cells at maturity (3880±320 versus 2500±255 μm²). Simultaneously, the number of epidermal cells of etg1‐1 was only approximately 60% that of the wild type (6650±530 versus 11 170±1017 cells; Figure 2C). Until day 9 after sowing, the average cell division rate for the whole leaf, calculated on the basis of the increase in cell numbers over time, were constantly lower in the etg1‐1 than in wild‐type leaves (Figure 2D). The average cell cycle duration between days 5 and 9, estimated as the inverse of the cell division rate, was significantly longer in the etg1‐1 mutant (25.3 h) than that in the wild‐type plants (21.1 h). In summary, these data illustrate that ETG1‐deficient plants suffer from a cell cycle delay, with a reduced total cell number that is offset by an enlarged cell size.

To pinpoint the cell cycle arrest point, we measured the ratio of 4C/2C cells of 8‐day‐old leaves by flow cytometry. At this time point, leaf cells of both genotypes divide (Figures 2C and D); consequently, 2C and 4C cells represent G1 and G2 cells, respectively. By comparing the ploidy level of wild type and etg1‐1, a significant increase in the 4C/2C cell ratio was observed (0.29±0.06 and 0.79±0.04 in wild‐type and etg1‐1 plants, respectively; Figures 3A and B). These data indicate that the G2‐to‐M transition was inhibited in the etg1‐1 mutant. Analysis of the expression levels of cell cycle marker genes by real‐time reverse transcriptase (RT)–PCR revealed that the transcript level of the G2/M‐phase‐specific CYCB1;1, CDKB1;1, and KNOLLE genes had increased in the etg1‐1 mutant, whereas those of the S‐phase‐specific histone H4 and CYCA3;1 genes was unaltered (Figure 3C). Despite the increased transcript levels of G2/M‐phase‐specific genes, a decrease in A‐type CDK activity was observed in the etg1‐1 mutant plants (Supplementary Figure S4), in agreement with the observed growth delay and extended cell cycle duration. Counting the number of root cells within the meristem in either metaphase or anaphase/telophase showed that the amount of M‐phase nuclei was approximately two‐fold lower in the etg1 mutant than that of wild‐type plants (Figures 3D–F). Combined with the observed cell cycle delay, these data indicate a prolonged G2 phase in the etg1 mutant. More particularly, induction of the CYCB1;1 and KNOLLE genes implies that the cell cycle arrest occurs around late G2.

The ETG1 gene had originally been identified by microarray analysis as an induced transcript in plants ectopically expressing the heterodimeric E2Fa‐DPa transcription factor (Vandepoele et al, 2005). This induction was confirmed by quantitative real‐time PCR analysis (Figure 4C). To analyse whether the ETG1 gene was directly regulated by E2F transcription factors, we looked for the presence of consensus E2F‐binding sites in its putative promoter region. Two consensus E2F‐binding elements were found, ATTCCCGC and TTTCCCGC (158 and 136 bp upstream from the putative start codon, respectively), both in a reverse orientation (Figure 4A). To address whether ETG1 is an E2F target gene in vivo, we used ChIP with antibodies against E2Fa, E2Fb, and E2Fc. Promoter fragments of the ETG1 gene were specifically amplified from the anti‐E2Fa and anti‐E2Fb immunoprecipitates. These results indicate that E2Fa and E2Fb can bind directly to the ETG1 promoter in vivo, probably participating in the regulation of its expression. By contrast, the ETG1 promoter sequences were not recovered from the immunoprecipitates with anti‐E2Fc antibodies (Figure 4B). Correspondingly, ETG1 expression levels were not altered in plants that overexpress E2Fc (Supplementary Figure S5).

The regulation of the ETG1 promoter activity through its E2F consensus sites was further analysed with transgenic plants expressing the β‐glucuronidase (GUS) reporter gene under control of the ETG1 promoter. To define the contribution of each of the E2F‐binding sites, we deleted exactly either one (ΔI or ΔII) or both (ΔI, II) of the E2F elements. More than five independent transgenic lines were analysed per gene construct, all with identical results. In 6‐day‐old seedlings, high levels of ETG1 expression were observed in the shoot apical and root meristems (Figure 4D). This expression pattern matched that of the E2Fa and DPa genes (De Veylder et al, 2002). Deletion of either one of the E2F‐binding elements (ΔI or ΔII) led to GUS activity patterns identical to those of plants carrying the wild‐type ETG1 promoter (Figure 4D). In contrast, deletion of both E2F‐binding elements resulted in a drastic decrease in promoter activity (Figure 4D). These results suggest that E2Fa and E2Fb bind both E2F consensus elements in the ETG1 promoter and regulate its expression in dividing tissues.

The ETG1 gene encodes a protein of 598 amino‐acid residues (Supplementary Figure S6). ETG1 is a singleton in Arabidopsis. No specific amino‐acid domain could be identified with the exception of a putative nuclear localization signal, PFKKMKV (amino acids 184–190), suggesting that ETG1 resides in the nucleus. To investigate the subcellular localization of ETG1, a fusion protein of ETG1 and an enhanced green fluorescent protein (eGFP) was transiently produced in leaf epidermal cells of tobacco (Nicotiana tabacum). ETG1–eGFP fluorescence was observed in the nucleus only, illustrating that ETG1 is a nuclear protein (Figure 5A). Orthologous proteins of ETG1 were found in rice (Oryza sativa; Os01g0166800), human (Homo sapiens; C10orf119), mouse (Mus musculus; 1110007A13Rik), Xenopus laevis (CAJ81286), Drosophila melanogaster (CG3430), and fission yeast (Schizosaccharomyces pombe; SPAC1687.04) (Supplementary Figure S6), but none in budding yeast (Saccharomyces cerevisiae). Interestingly, the transcript of the human ETG1 is significantly upregulated in brain cancer, breast cancer, and seminoma (http://www.oncomine.org).

Coexpression patterns can reveal networks of functionally related genes and provide a deeper understanding of processes requiring multiple gene products (Stuart et al, 2003; Wei et al, 2006). To predict the ETG1 function, we searched for genes coexpressed with ETG1 by using the ATTED‐II coexpression database (Obayashi et al, 2007). This search revealed that ETG1 is highly coexpressed with genes encoding DNA replication proteins, such as the MCM family proteins (MCM2, MCM3, MCM4, MCM5, and MCM7), proliferating cell nuclear antigen proteins (PCNA1 and PCNA2), DNA primase small subunit protein, and DNA polymerase α subunits (Table I). Moreover, when searching for proteins interacting with orthologous ETG1 proteins with the BioGRID protein interactions database (http://www.thebiogrid.org/index.php), we found that the Drosophila orthologue CG3430 interacted with the MCM5 protein. An identical interaction between the Arabidopsis ETG1 and MCM5 (At2g07690) proteins was demonstrated with the yeast two‐hybrid system (Figure 5B). This protein–protein interaction was confirmed in planta by bimolecular fluorescence complementation (BiFC) experiments (Bracha‐Drori et al, 2004; Walter et al, 2004). The sequences coding for ETG1 and MCM5 were fused in frame with either the N‐terminal or C‐terminal fragment of the yellow fluorescent protein (YFP) (ETG1–YFPN or MCM5‐YFPC, respectively). Interaction between the different fusion proteins was tested by introducing them into tobacco leaf epidermal cells. YFP fluorescence was observed in the nuclei of cells transfected with ETG1–YFPN and MCM5–YFPC, demonstrating that the ETG1 protein interacted with MCM5 in the plant nucleus (Figure 5C; ETG1–YFPN and MCM5–YFPC). As expected, no fluorescence was detected when any combination with empty vectors was introduced into tobacco cells (Figure 5C; ETG1–YFPN and YFPC, YFPN and MCM5–YFPC). When the subcellular localization of MCM5 was examined in plants with MCM5–eGFP, the fusion protein MCM5–eGFP resided in both the nucleus and cytoplasm (Figure 5D).

To identify additional ETG1‐associated proteins, tandem affinity purification (TAP) was confirmed with MALDI‐TOF‐TOF‐MS‐based protein identification (Van Leene et al, 2007). Arabidopsis cell suspension cultures were stably transformed by Agrobacterium tumefaciens‐mediated cocultivation with a Pro35S:TAP‐ETG1 cassette. On the basis of the presence of the double tag, two‐step affinity purification was performed. Besides MCM5, we identified five interacting proteins, including other components of the MCM complex, namely MCM2 (At1g44900), MCM3 (At5g46280), MCM4 (At2g16440), MCM6 (At5g44635), and MCM7 (At4g02060) (Table II; Supplementary Figure S7). Combined with the subcellular localization results, these data indicate that ETG1 assembled into the replisome complex. Therefore, ETG1 depletion was expected to affect the efficiency of DNA replication. This hypothesis was substantiated by the bromodeoxyuridine (BrdU) incorporation rate that was lower in etg1‐1 mutant plants than that in the control plants (Figure 5E). Although this decrease in BrdU incorporation might be unexpected in the light of the observed increase in the endoreduplication level, the reason might be that the endocycle is far slower than the mitotic cell cycle (Beemster et al, 2005). Therefore, because of the short labelling period, the BrdU label probably refers to the mitotic S phase only.

Inhibition of DNA replication in Arabidopsis results in the simultaneous induction of DNA repair genes and the cell cycle inhibitory WEE1 gene that arrests cells in the G2 phase of the cell cycle (De Schutter et al, 2007). The decreased rate of BrdU incorporation and observed interaction of ETG1 with MCM proteins suggested that the G2 arrest noticed in ETG1‐deficient plants might be the consequence of replication checkpoint activation. The activation of a cell cycle checkpoint might be the reason for the observed decrease in CDK activity as well. To test this hypothesis, we compared the expression levels of the RAD51 (DNA repair) and WEE1 (cell cycle checkpoint) marker genes by real‐time RT–PCR in wild‐type and etg1 mutant plants. Ionizing radiation (for example, γ‐irradiation and UV light), DNA replication inhibitory drugs (for example, hydroxyurea and aphidicolin), and radiomimetic agents (such as bleomycin) are known to induce RAD51 and WEE1 expression (Chen et al, 2003; De Schutter et al, 2007). Expression of both RAD51 and WEE1 was significantly upregulated in the etg1‐1 (Figure 6A) and similarly in etg1‐2 seedlings (data not shown). Activation of the DNA stress checkpoint was confirmed by using plants that carried as transgenes promoters of PARP2 and WEE1 fused to GUS, which are markers for DNA stress and activation of the replication checkpoint, respectively (Babiychuk et al, 1998; Doucet‐Chabeaud et al, 2001; De Schutter et al, 2007). No GUS activity was observed in PARP2:GUS plants grown under non‐stress conditions (Figure 6B). By contrast, treatment of the PARP2:GUS reporter line with bleomycin resulted in a strong induction of GUS activity (Figure 6D), demonstrating the DNA stress‐inducible promoter activity. Similarly, PARP2 promoter activity was induced in an etg1‐1 background in the absence of any external DNA stress stimulus (Figure 6C). Especially, GUS activity was strongly induced in shoot apical meristem and vascular cells. Analogous results were obtained with WEE1:GUS reporter plants. In control plants, WEE1 expression was observed in the shoot apex and vascular cells (Figure 6E; De Schutter et al, 2007). This expression pattern was intensified in the etg1‐1 background (Figure 6F), confirming the real‐time RT–PCR experiments. The ETG1 gene itself was not induced under conditions that cause DNA stress, such as γ‐irradiation or UV‐B treatment (Ulm et al, 2004; Culligan et al, 2006). Additionally, etg1‐1 mutant plants were not hypersensitive to hydroxyurea or bleomycin (Supplementary Figure S8).

DNA replication stress caused by blocking of the replication fork is mainly sensed by the ATR kinase (Culligan et al, 2004). Previously, we have demonstrated that the WEE1 gene is transcriptionally activated upon replication stress in an ATR‐dependent manner and transiently arrests cells in the G2 phase, allowing them to finalize DNA replication before proceeding into mitosis (De Schutter et al, 2007). When the increased cell cycle duration observed in the etg1 mutant plants is assumed to result from the activation of the replication checkpoint, ETG1 deficiency is expected to have a dramatic impact on the development of plants that are unable to arrest their cell cycle in response to DNA stress. To test this hypothesis, double mutants were constructed between etg1‐1 and two DNA stress checkpoint mutants, atr‐2 and wee1‐1. WEE1‐deficient plants fail to arrest their cell cycle when DNA replication is perturbed, but DNA repair genes are still induced (De Schutter et al, 2007; T Cools, DI and LDV, unpublished data). In contrast, atr mutant plants fail both to arrest their cell cycle and to induce repair genes (Culligan et al, 2004, 2006). Both wee1‐1 and atr‐2 mutants were hypersensitive to replication‐blocking or DNA‐damaging drugs, but were viable and developed normally in the absence of exogenous DNA stress treatments (Figures 7A–D). By contrast, etg1‐1 wee1‐1 and etg1‐1 atr‐2 double mutant plants had a dwarf phenotype under non‐stress conditions, illustrating a synthetic interaction between ETG1 and WEE1, or ATR (Figures 7E–H). Scanning electron microscopy revealed severe growth suppression (Figures 7I–K). Especially, the size of trichomes was reduced in the double mutants (Figures 7O–Q). No significant difference in leaf epidermal cell shape was observed in etg1‐1 wee1‐1 double mutants, whereas cells lost their jigsaw‐like shape in etg1‐1 atr‐2 plants (Figures 7L–N). The double mutants arrested at an early stage of development. These results unequivocally indicate that the activation of the DNA replication checkpoint in etg1 mutant plants is essential for their survival.

We have identified the ETG1 protein as an important replication factor. We propose that the ETG1 protein is a component of the replication fork that is necessary to maintain genome stability during S phase and that deficiency of ETG1 might generate DNA damage and genome instability (Figure 8). ETG1, originally discovered as a transcript induced in E2Fa‐DPa‐overexpressing plants, was demonstrated to be a direct target gene of the activating E2Fa/E2Fb transcription factors. Two E2F‐binding sites in the ETG1 promoter redundantly control ETG1 expression. Interestingly, not only is the ETG1 gene highly conserved among species but also E2F‐binding sites can be found in the promoter region of ETG1 orthologues of rice, Xenopus, Drosophila, mouse, and human. Moreover, in E2F1‐depleted cells, transcripts of the Drosophila ETG1 orthologue (CG3430) are decreased (Dimova et al, 2003). Thus, besides the gene product, also the transcriptional control mechanisms that drive ETG1 expression are conserved, suggesting that E2F‐dependent regulation of the ETG1 expression and of its orthologues is highly important during the replication process.

Our data show that ETG1 is part of the replisome. First, etg1 mutants display a clear replication phenotype. Second, both ETG1 and its Drosophila orthologue (CG3430) associate with MCM proteins that bind to the replication origin and travel along the DNA with the remainder of the replication machinery (Kearsey and Labib, 1998; Tye, 1999; Labib et al, 2000). Moreover, while this work was in progress, the human orthologous ETG1 has been reported to bind chromatin in a cell cycle phase‐dependent manner with the highest affinity for DNA during G1/S and S (Sakwe et al, 2007). When the subcellular localization of MCM5 in plants was examined with an MCM5–eGFP fusion protein, MCM5–eGFP was found to reside in both the nucleus and cytoplasm. In yeast, the subcellular localization of the MCM5 protein changes with the cell cycle: it is nuclear between the end of M phase and G1‐to‐S transition, but is cytoplasmic in other phases of the cell cycle (Hennessy et al, 1991). Its dual localization in the nucleus and cytoplasm indicates that a similar situation might hold true for the Arabidopsis MCM5 protein. However, by BiFC experiments, we observed that MCM5 and ETG1 interacted only in nuclei and not in the cytoplasm, suggesting an interaction solely when the MCM proteins dock to the replication origins.

ETG1‐deficient plants do not only show a replication defect but also a prolonged G2 phase because of the activation of the DNA replication checkpoint, as illustrated by the upregulation of the DNA repair genes PARP2 and RAD51 and the cell cycle checkpoint gene WEE1. Despite the prolonged cell cycle duration, resulting in a strongly reduced total cell number, ETG1 single mutants had a relatively minor developmental phenotype. In contrast, in an atr or wee1 mutant background, ETG1 deficiency has a huge impact on plant growth with dwarfism, improper cell differentiation, and a developmental arrest as a consequence. These results illustrate that ATR and WEE1 make an essential contribution to the survival of plants without ETG1 function. Previously, lack of ATM or ATR activity has been demonstrated to emphasize the phenotypic defects of telomerase‐deficient (tert) mutants (Vespa et al, 2005). Whereas in a wild‐type background, the tert mutants display growth and developmental phenotypes from the sixth generation onwards, tert atm and tert atr double mutants have misshapen leaves, no apical dominance, delayed flowering, and significantly reduced fertility in earlier generations. These data indicate that the activation of the ATM and ATR pathways is required for survival of plants suffering genomic instability. However, it is impossible to attribute the attenuation of the phenotype by functional ATM and ATR to the induction of DNA repair genes, or the ability to arrest cell cycle progression. As wee1 mutant plants are defective in checkpoint activation but are still able to induce DNA repair genes (De Schutter et al, 2007; T Cools, DI, and LDV, unpublished data), the severe phenotype of the etg1 wee1 double mutants indicates that the activation of the G2 checkpoint contributes mainly to the survival process. Nevertheless, compared to the etg1 wee1 mutant, the observed growth defects were slightly more outspoken for the etg1 atr double mutant, illustrating that the phenotypes resulting from the inability to arrest the cell cycle are partly offset by DNA repair.

Interestingly, at maturity, the population of cells with a high DNA ploidy level was larger in etg1 mutant than in control plants, suggesting that the replisome complex of endoreduplicating cells differs from that found in mitotically dividing cells and that ETG1 is not required for the endoreduplication cycle. Previously, DNA double‐stranded breaks caused by depletion of the chromatin assembly factor 1 have been found to be accompanied with extra endoreduplication cycles (Endo et al, 2006; Exner et al, 2006; Ramirez‐Parra and Gutierrez, 2007). Additionally, wild‐type plants treated with the DNA‐damaging drug zeocin significantly increased the DNA ploidy level (Ramirez‐Parra and Gutierrez, 2007). The mechanisms that link the maintenance of chromosomal stability with the DNA ploidy level under normal and stressed growth conditions remain unclear. Previously, we have found that the level of A‐type CDK activity determines whether a plant cell divides or endoreduplicates (Verkest et al, 2005). Therefore, the extra endoreduplication cycles under replication stress might be an indirect effect from the activation of the WEE1 expression, inactivating the A‐type CDK by tyrosine phosphorylation (De Schutter et al, 2007). The decrease in CDK activity might prevent entry into mitosis, but still allow S phase to be re‐initiated, and, thus, re‐replication of the genome. The size of cells has often been found to correlate with their DNA content (Melaragno et al, 1993; Folkers et al, 1997; Traas et al, 1998; Sugimoto‐Shirasu and Roberts, 2003). Therefore, in endoreduplicating species the coupling of checkpoint activation with initiation of the endocycle might represent a mechanism by which a reduction in cell number by replication stress is compensated by cell enlargement. Such a coupling might correspond with a survival mechanism allowing plants to achieve a critical biomass required for reproduction, possibly explaining the evolutionary success of the endoreduplication programme among angiosperms.

Leaves were chopped with a razor blade in 300 μl of buffer (45 mM MgCl₂, 30 mM sodium citrate, 20 mM 3‐[N‐morpholino]propanesulphonic acid, pH 7, and 1% Triton X‐100). To the supernatants, 1 μl of 4,6‐diamidino‐2phenylindole (DAPI) from a stock of 1 mg/ml was added, which was filtered over a 30‐μm mesh. The nuclei were analysed with the CyFlow flow cytometer and the FloMax software (Partec, Münster, Germany). At least three biological and two technical replicates were used for each sample analysed.

Roots were fixed in a solution of formaldehyde, ethanol, and acetic acid (2:17:1) for 12 h at 4°C, washed twice in water, and mounted under cover slips. The samples were crushed, snap‐frozen with liquid nitrogen to remove the cover slip, and mounted in Vectashield (Vector Laboratories) containing 1 μg/ml DAPI. The roots were analysed for mitotic stages with a Zeiss Axiovert fluorescence microscope.

TAP experiments were carried out according to Van Leene et al (2007). In short, the ETG1‐coding sequence was cloned by recombination into the pKNTAP vector, generating a Pro35S:TAP‐ETG1 cassette (pKNTAPETG1). Arabidopsis cell suspension cultures were stably transformed by Agrobacterium‐mediated cocultivation with pKNTAPETG1. Transformed Arabidopsis cells were selected and transferred to a liquid medium for upscaling. Expression levels of TAP‐tagged proteins were checked by protein blotting with an anti‐calmodulin‐binding protein antibody (data not shown). In the first round of affinity purification, protein extracts of 15 g plant material were incubated with an IgG resin. Bound complexes were released and eluted from the resin by tag cleavage with TEV protease. In the second affinity step on a calmodulin agarose column, co‐eluting non‐interacting proteins and the TEV protease were removed with the flow‐through. Finally, both the ETG1 bait and interacting proteins were eluted from the calmodulin agarose through EGTA‐mediated removal of calcium. Eluted proteins were separated on 4–12% NuPAGE gels, excised, and analysed by MALDI‐TOF‐TOF‐MS as described (Van Leene et al, 2007). To increase the stringency of the data set, contaminating proteins due to experimental background as determined by Van Leene et al (2007) were systematically subtracted from the lists of co‐purified proteins.

Seedlings (3 day old) grown on MS agar plates were incubated in the labelling solution with 10 μM BrdU (Roche) at room temperature for 3 h. After treatment, the genomic DNA was extracted with a DNeasy^® Plant Mini Kit (Qiagen). The amount of incorporated BrdU was determined by ELISA with an anti‐BrdU‐POD antibody (Roche). Three biological and two technical replicates were used at each time point for ELISA. Here, 50 μl of the extracted DNA (0.2 μg/ml) was placed in each well. The ELISA procedure was chiefly that of the 5‐bromo‐2′‐deoxyuridine Labeling and Detection Kit III (Roche).

Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).