Ablation of PARP-1 does not interfere with the repair of DNA ... - Nature

Oncogene (2004) 23, 3872–3882

& 2004 Nature Publishing Group All rights reserved 0950-9232/04 $25.00 www.nature.com/onc

Ablation of PARP-1 does not interfere with the repair of DNA double-strand breaks, but compromises the reactivation of stalled replication forks Yun-Gui Yang1, Ulrich Cortes1, Srinivas Patnaik1, Maria Jasin2 and Zhao-Qi Wang*,1 1

International Agency for Research on Cancer (IARC), 150 cours Albert-Thomas, Lyon 69008, France; 2Memorial Sloan-Kettering Cancer Center, New York, NY 10021, USA

Poly(ADP-ribose) polymerase-1 (PARP-1) is an abundant DNA end-sensing and binding molecule. Inactivation of PARP-1 by chemicals and genetic mutations slows cell proliferation, increases sister chromatid exchange (SCE), micronuclei formation and chromosome instability, and shortens telomeres. Given its affinity to DNA breaks and temporal occupation on DNA strand break sites, PARP-1 is proposed to prevent inappropriate DNA recombination. We investigated the potential role of PARP-1 in repair of DNA double-strand breaks (DSBs) and stalled replication forks. PARP-1/ embryonic stem cells and embryonic fibroblast cells exhibited normal repair of DNA DSBs by either homologous recombination (HR) or nonhomologous end-joining (NHEJ) pathways. Inactivation of PARP-1 did not interfere with gene-targeting efficiency in ES cells. However, PARP-1/ cells were hypersensitive to the replication damage agent hydroxyurea (HU)-induced cell death and exhibited enhanced SCE formation. Ablation of PARP-1 delayed reactivation of stalled replication forks imposed by HU and re-entry into the G2-M phase after HU release. These data indicate that PARP-1 is dispensable in HR induced by DSBs, but is involved in the repair and reactivation of stalled replication forks. Oncogene (2004) 23, 3872–3882. doi:10.1038/sj.onc.1207491 Published online 15 March 2004 Keywords: poly(ADP-ribose) polymerase-1; Rad51; homologous recombination; DNA double-strand break repair; replication stall and progression

Introduction Homologous recombination (HR) functions as a major pathway in the repair of DNA double-strand breaks (DSBs) and genome replication damage. DSBs are major pathological DNA structures in dividing cells, which are often generated by stalled or collapsed replication forks due to genetic mutations or exogenous damage. Repair of DSBs is mainly conducted by HR or nonhomologous *Correspondence: Z-Q Wang; E-mail: [email protected] Received 1 October 2003; revised 17 December 2003; accepted 19 December 2003; Published online 15 March 2004

end-joining (NHEJ), and is critical to maintain genomic integrity in eukaryotic organisms (Jackson, 2002). HR, as an accurate mode of DSB repair, restores the damaged genetic information by promoting the broken DNA sequence to interact with its homologous donor sequence. Left unrepaired, DSBs can cause chromosome translocations, or breaks, leading to genomic instability, a hallmark of cancer (Hoeijmakers, 2001). HR provides major nonmutagenic repair to stalled replication forks that are induced when the replication fork encounters DNA lesions induced by endogenous or exogenous DNA damage. Recombination proteins bind to ssDNA generated by stalled replication forks, forming a nucleoprotein filament that is necessary to perform HR, which uses the intact sister duplex with subsequent resolution of the Holliday Junctions (HJ) (Cox, 2001; McGlynn and Lloyd, 2002). This process is believed to be conducted mainly by Rad52 epistasis group molecules. Rad51 is a major repair molecule that promotes HJ formation and interactions between sister chromatids at stalled replication forks (Cox, 2001; Masson and West, 2001). In response to replication arrest, Rad51 translocates to the replication forks (Tashiro et al., 2000; Saintigny et al., 2001; Lundin et al., 2002) to facilitate HR repair of replication abnormalities (Cox, 2001; Masson and West, 2001). Moreover, proper disruption of Rad51 nucleoprotein filaments attenuates recombination and may be required for cells to restart replication (Krejci et al., 2003; Veaute et al., 2003). PARP-1 is an abundant nuclear enzyme, which is catalytically triggered when it binds to DNA breaks. Biochemical and genetic studies have indicated that PARP-1 plays a role in DNA repair, recombination, proliferation and genomic stability, as well as in suppressing tumorigenesis (see reviews D’Amours et al., 1999; Burkle, 2001; Tong et al., 2001a). Given its DNA break-binding activity, PARP-1 may temporarily occupy DNA strand break sites, thereby preventing inappropriate DNA recombination (Lindahl et al., 1995). Chemical inhibitors of the enzyme, as well as dominantnegative or null mutations of PARP-1, reduce cell proliferation and cause a high degree of chromosome aberrations and sister chromatid exchange (SCE), as well as telomere shortening (D’Amours et al., 1999; see also Tong et al., 2001a). Moreover, enhanced chromosomal instability was observed in cells lacking PARP-1

Role of PARP-1 in the repair of DSB and stalled replication forks Y-G Yang et al

3873

and other DNA repair or chromosomal guardian molecules such as DNA-PK (Morrison et al., 1997), Ku80 (Tong et al., 2002) or p53 (Tong et al., 2001b, 2003). These data indicate that PARP-1 functions in the suppression of chromosome aberrations, most likely via its role in shuttling in and out of DNA ends, and by interacting with other DNA strand break-processing molecules. Furthermore, although V(D)J recombination and Ig class switching are not altered in PARP-1/ mice (Wang et al., 1997), an elevated level of recombination efficiency due to PARP-1 deficiency has been illustrated by the partial rescue of V(D)J recombination and, thereby, T-cell development in SCID mice in a PARP-1 null background (a mutation in DNA-PKcs; see Morrison et al., 1997). All these data support the notion that PARP-1 may function as an anti-recombinogenic factor. Although PARP-1 binds mainly to single-stranded DNA (ssDNA), many studies suggest its involvement in sensing and/or repairing DSBs. PARP-1 is shown to interact with DNA-PKcs (Ariumi et al., 1999), Ku80 (Galande and Kohwi-Shigematsu, 1999) and MRE11 (Boulton et al., 2002). Mice lacking both PARP-1/ATM (Meńissier-de Murcia et al., 2001) or PARP-1/Ku80 (Tong et al., 2002; Henrie et al., 2003) show early embryonic lethality. Finally, PARP-1 null mutation compromises the proliferation of embryonic fibroblasts (EFs) (Wang et al., 1995, 1997; Tong et al., 2000). Moreover, SCID mice in a PARP-1 null background (Morrison et al., 1997) and PARP-1/Ku80 double null embryos (Tong et al., 2002) exhibit growth retardation. All these data imply the involvement of PARP-1 in the repair of DSBs by HR or NHEJ, and perhaps in replication. However, the molecular pathway used by PARP-1 in stabilization of the genome remains elusive. In the present study, we attempt to define the function of PARP-1 in these processes.

Results PARP-1 is not essential for HR repair in vivo We first tested the role of PARP-1 in recombination repair using a fluorescence-based assay system involving chromosomally integrated substrates (Pierce et al., 1999). PARP-1 þ / þ (A16 and A19) and PARP-1/ EF cell lines (A11 and A12) containing a single intact copy of the reporter substrate DR-GFP were isolated after puromycin selection and verified by Southern blotting (data not shown). Three independent DR-GFP clones from each cell line were transfected with the ISceI expression vector to induce genomic DSBs at the integrated DR-GFP substrate. Although a trend of less recombination events was noticed in PARP-1/ cells, the difference of the number of GFP-positive cells between both genotypes was not statistically significant (Figure 1a, b), indicating that PARP-1 has minimum effects on homologous-directed-repair (HDR). The transfection efficiency in both genotypes was virtually identical (about 75%), monitored by transfection with the GFP-expressing vector (data not shown).

Normal gene-targeting efficiency and HDR in PARP-1 null ES cells To further test HR in defined chromosome positions and in other cell types, we targeted the DR-GFP substrate to the X-linked hypoxanthine phosphoribosyl transferase (hprt) locus (Pierce et al., 2001) in male ES cells. The gene-targeting vector (hprt-DRGFP) was electroporated into two ES cell lines of both PARP1 þ / þ and PARP-1/ genotypes, and targeted clones were obtained after puromycin and 6-thioguanine (puro/ 6-TG) double selection, and verified by Southern blotting and PCR (data not shown). Despite some variations between experiments, gene-targeting efficiency (homologous vs total integration) was similar in PARP-1 þ / þ and PARP-1/ ES cell lines in any given experiment (Table 1). Therefore, PARP-1 null mutation does not seem to influence gene-targeting events in ES cells. To test HR, three clones of each cell line of each genotype (Table 1) were electroporated with the I-SceI-expressing vector and a similar proportion of GFP-positive cells was observed in both wild-type and PARP-1 null cells (Figure 2). The equal acceptance of exogenous DNA by both PARP-1 þ / þ and PARP-1/ cells was confirmed by the GFP expression in B80% of ES cells after transfection of the GFP-expressing vector (data not shown). Taken together, our results indicate that PARP-1 is dispensable in HR repair of DSBs in random and targeted integrated DNA substrates in EF and ES cells, although a downward trend was observed in PARP-1/ cells in these HDR experiments. Recombination repair is unaltered at the I-SceI site in PARP-1/ ES cells We further defined the repair pathways in the I-SceIinduced DSBs in the substrate. In the current recombination system, HR may occur via short tract gene conversion (STGC) or homologous deletion (HD), whereas NHEJ may lead to either precise or imprecise repair (Pierce et al., 2001). HR or imprecise NHEJ will result in I-SceI-site loss, which can be verified by I-SceI endonuclease cleavage (Figure 3a). To this end, DNA was isolated from ES cell lines documented in Figure 2b, and the surrounding sequences of the I-SceI site were amplified by PCR. After I-SceI digestion, 31% of PARP-1 þ / þ and 26% of PARP-1/ ES cells showed I-SceI-site loss (P ¼ 0.310; Figure 3b, c). To examine whether I-SceI-site loss after DSB repair was via HR or imprecise NHEJ, we sequenced I-SceI-resistant fragments (I-SceIR in Figure 3b) from 60 clones derived from PARP-1 þ / þ and 44 clones from PARP-1/ ES cells, and found a similar amount (73–75%) of I-SceIsite loss events in both genotypes contributed by HR (Figure 3d). The remaining 15–18% of I-SceI-site loss events in PARP-1 þ / þ and PARP-1/ ES cells were due to imprecise NHEJ, typically accompanied by loss or insertion of a few nucleotides (Figure 3d), suggesting that the absence of PARP-1 does not greatly change the frequency or character of localized NHEJ events. We Oncogene


3874

Figure 1 PARP-1/ cells exhibit normal HDR of genomic DSBs at a random locus. (a) A representative flow cytometric analysis for HDR in PARP-1/ EFs. PARP-1 þ / þ and PARP-1/ 3T3 EF cell lines containing the DR-GFP substrate were electroporated with either mock plasmid pCAGGS (mock) or I-SceI-expressing vector pCBASce (I-SceI), and were analyzed by flow cytometry for HDR. The GFP-positive population (after HDR) is separated from the GFP-negative population as shifting to greenward. (b) Percentage of GFP-positive cells is shown and bars represent the mean of three clones of each genotype. P ¼ 0.48 (t-test). The experiments were repeated using two independent cell lines of each genotype

Table 1 Gene-targeting efficiency in PARP-1+/+ and PARP-1/ ES cells Experiment 1 Cell lines

puroR ( 106)a

Experiment 2

Experiment 3

6-TGR/puroR ( 107)b

puroR ( 106)a

6-TGR/puroR ( 107)b

puroR ( 106)a

6-TGR/puroR ( 107)b

PARP-1+/+ ES9-1 ES32-3

100 220

10 (1.00%)c 55 (2.54%)

231 162

21 (0.90%) 29 (1.79%)

248 565

40 (1.63%) 120 (2.12%)

PARP-1/ ES6-1 ES17-2

160 220

14 (0.83%) 37 (1.68%)

505 188

20 (0.34%) 22 (1.14%)

449 512

14 (0.30%) 58 (1.13%)

a Number of total (puroR) integrants divided by the number of cells electroporated with the hprt-DRGFP targeting vector. bNumber of homologous (6-TGR/puroR) integrants divided by the number of cells electroporated with the hprt-DRGFP targeting vector. cGene-targeting efficiency is calculated by dividing the number of homologous (6-TGR/puroR) integrants with the number of total (puroR) clones

Oncogene


3875

Figure 2 ES cell lines lacking PARP-1 exhibit normal HDR of genomic DSBs at a targeted allele. (a) A representative flow cytometric analysis for HDR of DSBs in ES cells. ES cells containing the targeted hprt-DRGFP substrate (see Table 1) were electroporated with the I-SceI-expressing vector or mock plasmid as described in Figure 1A. (b) Percentage of GFP-positive cells is shown and bars represent the mean of three clones of each genotype. P ¼ 0.66 (t-test). The experiments were repeated using two independent cell lines of each genotype

also found that 3–6% of DNA capture events were associated with the insertion of DNA fragments, which originated from the I-SceI vector, into the DSB site (Figure 3d), an observation also reported previously (Liang et al., 1998). Taken together, the spectra of DSB repair in PARP-1 null cells are similar to wild-type cells, indicating that PARP-1 is dispensable in HR repair of DSBs in chromosomally integrated DNA substrates. Sensitivity of PARP-1/ cells to hydroxyurea (HU) treatment In mammalian cells, stalled replication forks constantly generate ssDNA ends, and their repair involves HR

using sister chromatids (Sogo et al., 2002; see also McGlynn and Lloyd, 2002). Since PARP-1 has major binding activities towards ssDNA breaks, we further tested whether PARP-1 is involved in the repair of Sphase-specific DNA damage. As the collapse of replication forks (generating DSBs) may induce cell death (Nyberg et al., 2002), we first examined the survival of primary PARP-1 þ / þ and PARP-1/ EFs in response to HU treatment, which inhibits DNA synthesis and causes replication fork arrest. Cells were exposed to various doses of HU and cell survival was examined by a colony-formation assay. At all doses tested, the number of colonies was reduced in PARP-1/ EFs compared to wild-type controls after 10 days in culture (Figure 4a). Oncogene


3876

Figure 3 PARP-1 þ / þ and PARP-1/ ES cells contain similar spectra of DSB repair. (a) PCR strategy for the detection of DSB repair. PCR primers 1 and 2 flank the I-SceI site. The presence of a B723bp PCR product indicates that the sequence underwent DSB repair events either by HR or imprecise NHEJ, which would eliminate the I-SceI recognition site (black segment) and create a resistance to I-SceI digestion. HR, homologous recombination; STGC, short tract gene conversion; HD, homologous deletion; NHEJ, nonhomologous end joining. (b) Representative PCR analysis of DSB repair. PCR was performed on genomic DNA isolated from cell lines after electroporation with the mock pCAGGS (data not shown) or I-SceI-expressing vector. I-SceIR fragments originating from HR or imprecise NHEJ are resistant to I-SceI; PCR products after precise NHEJ are sensitive to I-SceI (I-SceIS). (c) Quantification of I-SceI-site loss in gels (e.g. in (b)). The density of DNA bands was quantified and normalized for product length as described previously (Pierce et al., 2001). P ¼ 0.310 (t-test). (d) DSB repair events. The I-SceIR fragments (e.g. in (b)) from 60 clones of PARP-1 þ / þ and 44 clones of PARP-1/ origin were sequenced. The precise boundaries of the DSB repair events by HR, NHEJ deletions, insertions or capture of exogenous DNA were determined with Clustal X

In addition, we found more TUNEL-positive cells in the PARP-1/ population than in wild-type counterparts (Figure 4b). These data suggest that the absence of PARP-1 compromises cell viability in response to replication fork damage. Compounds that inhibit DNA synthesis (e.g., see Rainaldi et al., 1984; Cortes et al., 1993) and the PARP1 activity (e.g., see Ikushima, 1990) increase SCE formation. We thus tested whether PARP-1 deficiency Oncogene

would render cells susceptible to HU-induced SCE formation. PARP-1/ and PARP-1 þ / þ EF cells were treated or untreated with HU and SCE formation in these cells was scored. Without HU treatment, PARP-1/ EFs contained more spontaneous SCEs when compared with PARP-1 þ / þ EFs (Po0.0001, Figure 4c), consistent with previous reports (Meńissierde Murcia et al., 1997; Wang et al., 1997). After HU treatment, both genotypes exhibited elevated SCE


3877

Figure 4 PARP-1 null cells are hypersensitive to HU. (a) Colony-formation assay of wild-type and PARP-1/ cells in response to HU. Colonies were formed after 10 days in culture and scored after Giemsa staining. Survival was calculated as a percentage of untreated respective genotype. The means and error bars represent two samples from duplicate experiments. (b) Apoptosis induced by HU. After exposure to various doses of HU, TUNEL-positive cells in primary PARP-1 þ / þ and PARP-1/ EF populations were scored as apoptotic events. A total of 900 cells of each genotype were counted for each dosage. Apoptosis was expressed as relative values obtained by dividing the values from each dosage with the values of the respective nontreated group. The mean represents duplicate experiments. (c) SCE formation in EFs with or without HU treatment. In all, 50 chromosomal spreads were counted for both treated and untreated groups

formation, but in greater multitudes in PARP-1/ cells than in PARP-1 þ / þ cells (Po0.0001, Figure 4c), suggesting that PARP-1 null cells are hyper-recombinogenic in response to replication fork arrest. Elevated Rad51 foci formation in PARP-1/ cells in response to HU Hypersensitivity of PARP-1 null cells to HU suggests that PARP-1 may be involved in the repair of stalled replication forks. We next investigated the repair of replication forks arrested by HU in cells lacking PARP1 by analyzing translocation of Rad51 to stalled replication forks, where it is thought to promote strand exchange reaction and conduct HR (e.g., see Tashiro et al., 2000; Saintigny et al., 2001; Lundin et al., 2002). Without HU treatment, Rad51 foci were almost undetectable in confluent cultures (data not shown), whereas in subconfluent cultures, a low, but comparable, number of Rad51 foci were found in PARP-1 þ / þ and PARP-1/ primary EF cells

(Figure 5a, b), suggesting that the presence of Rad51 foci is most likely linked to the replication status. After HU treatment, Rad51 foci were induced in both PARP1/ and PARP-1 þ / þ EF cells, and the number of Rad51 foci was markedly increased in PARP-1/ cells (Figure 5c). Quantitative analysis of Rad51 foci-positive cells showed that in comparison with PARP-1 þ / þ cells, the distribution of PARP-1/ EFs was shifted towards cells containing a higher number of Rad51 foci, with B40% of cells containing 35–54 foci and another 25% of cells containing more than 75 foci (Figure 5d). To further define the kinetics of Rad51 foci during replication arrest release, we measured the Rad51 foci change at 6 h and 12 h after removal of HU. As shown in Figure 5E, while the majority (85%) of PARP-1 þ / þ cells contained 5–34 Rad51 foci, PARP-1 null cells maintained higher numbers of Rad51 foci 6 h after HU release. The number of Rad51 foci in both PARP-1 þ / þ and PARP-1/ cells reduced to a similar level only after 12 h (Figure 5f). These data indicate that Rad51 foci were not efficiently removed in Oncogene


3878

Figure 5 Enhanced Rad51 foci formation in response to replication damage in primary PARP-1/ EFs. Representative images of Rad51 foci in primary PARP-1 þ / þ and PARP-1/ EFs untreated (a) or treated with HU (c) after staining with the anti-Rad51 antibody and DAPI. (b) Quantification of the number of Rad51 foci in PARP-1 þ / þ and PARP-1/ cells without HU treatment. (d) HU treatment induced higher numbers of Rad51 foci in PARP-1/ EFs compared with that in PARP-1 þ / þ cells. The number of Rad51 foci in each nucleus was scored in 60–80 Rad51 foci-positive cells at 0 h (d), 6 h (e) and 12 h (f) after HU release. The vertical axis in (b, d, e, f) represents the distribution of cells containing the given numbers of Rad51 foci. The number of Rad51 foci was classified into groups, shown on the horizontal axis

the absence of PARP-1, or that HR activity mediated by Rad51 in the absence of PARP-1 persists for a longer time, attempting to resolve stalled replication forks. Delayed reactivation of replication in the absence of PARP-1 As dislodgment of Rad51 from nucleoprotein filaments is proposed to facilitate replication restart (Krejci et al., 2003; Veaute et al., 2003), we next tested whether the delay of Rad51 foci clearance in PARP-1 null cells had any impact on the reactivation of replication, by analyzing cell cycle progression into the G2-M phase after HU release. To this end, cells were labeled with BrdU before HU treatment and the progression of replication was monitored by the accumulation of BrdU-positive cells in the G2/M phase after HU release. Oncogene

To control the experimental setting, flow cytometry analysis revealed that 34.5% of PARP-1 þ / þ and 25.3% of PARP-1/ cells incorporated BrdU, respectively (Figure 6a). At 0 h after removal of HU, most cells were arrested in the S phase in both PARP-1 þ / þ and PARP-1/ EFs (Figure 6b). While wild-type cells were found to resume DNA synthesis and enter the G2M phase as early as 3 h after HU release, PARP-1/ cells did not enter the G2 phase until 9 h after HU release, with the entry becoming more evident after 12 h (Figure 6b). Furthermore, we counted the number of cells over a 12-h period after HU release (2 mM HU) and found that PARP-1/ EFs did not grow until 8 h after HU release and proliferated at a very low rate during this period, whereas PARP-1 þ / þ EFs doubled their cell numbers after 3 h (Figure 6c). These observations may explain the proliferation defects of PARP-1 null


3879

Figure 6 Retardation of PARP-1 null cells in reactivation of replication and re-entry to G2/M after HU release. (a) Flow cytometry analysis of BrdU incorporation into PARP-1 þ / þ and PARP-1/ EFs before (control, upper panels) and after BrdU labeling. (b) Flow cytometry analysis of cell cycle progression after HU release. The progression of replication was measured by the movement of Sphase cells (BrdU-positive population in R3 in (a) into the G2/M phase. The DNA histogram displaying DNA content was gated from green fluorescent BrdU-positive cells (see (a)). BrdU-positive G2/M populations at 0 h represent cells that passed the ‘S’ phase during the incubation period (2 h) with BrdU. The data represent one of two independent experiments. (c) Proliferation of primary PARP-1 null EF cells after HU release. The number of EFs after HU release (incubated with 2 mM HU for 4 h) was counted at the indicated time points. The data represent one of two independent experiments

cells or after 3-AB treatment (e.g., see Wang et al., 1995, 1997; Tong et al., 2000; see also review by D’Amours et al., 1999). Thus, PARP-1 seems to be necessary for the proper reactivation of DNA synthesis after replication fork arrest.

Discussion In the present study, we demonstrate that the ablation of PARP-1 in mouse EF and ES cells does not interfere

with the repair of DSBs by either HR or NHEJ in a substrate stably integrated at a random or defined (targeted) locus in chromosomes. These findings are consistent with a recent report showing that chemical inhibitors of PARP do not affect HR repair in V79 Chinese hamster cells (Schultz et al., 2003). These results may be surprising because genetic and biochemical studies have implicated PARP-1 in the repair and/or sensing of DSBs. As DSB repair is mainly conducted by the MRE11/Rad50/NBS complex, DNA-PK and the Rad52 epistasis group of proteins (Jackson, 2002), it is possible that PARP-1’s function in HR repair may be Oncogene


3880

masked or tossed aside in the presence of these major pathways. On the other hand, a dispensable function for PARP-1 in DSB repair may be expected because PARP1 has been shown to be mainly involved in binding to ssDNA breaks (see reviews D’Amours et al., 1999; Burkle, 2001). However, PARP-1 seems to play a dual role in the response to replication damage. The stalled replication fork triggers binding of Rad51 to the replication forks to form foci, which is believed to facilitate HR repair of replication abnormalities (Cox, 2002). Although we did not observe a significant increase of spontaneous Rad51 foci in primary PARP-1 null EFs, the absence of PARP1 markedly elevated Rad51 foci formation after HU treatment, implying an elevated recombination activity during the repair of replication abnormalities (Masson and West, 2001; Cox, 2002). In this regard, it is interesting to note that the Rad51 foci-containing cell population was readily increased in immortalized PARP-1 null EFs (Schultz et al., 2003). This increased Rad51 foci is apparently independent of expression levels of the cellular Rad51 protein, because Western blot analysis revealed no change of Rad51 protein levels in wild-type and PARP-1 null EFs after HU treatment (data not shown). Therefore, the elevated number of Rad51 foci in PARP-1 null cells after replication stall may be rather due to Rad51 re-distribution or recruitment to the foci. This phenomenon suggests that PARP1 represses Rad51-mediated HR repair at replication forks, which may be achieved through PARP-1’s affinity for ssDNAs that are generated during the repair of replication forks (Sogo et al., 2002) and that thereby occupy the site or compete with other HR repair molecules, such as Rad51. However, whether PARP-1 directly competes with Rad51 during HR repair of stalled replication forks requires further investigation. An alternative explanation could be that there is a proportion of sites after HU damage that are preferentially repaired by a pathway involving PARP-1, but, in the absence of this enzyme, the repair of these sites is picked up by a fallback mechanism involving Rad51. Since HR repair during replication arrest promotes strand exchange using sister chromatids to repair replication forks, it is reasonable to speculate that this event induces SCEs (Tashiro et al., 2000; see also Masson and West, 2001). Consistent with this notion, high degrees of Rad51 foci in PARP-1-deficient cells correlate well with increased SCEs. It is worth mentioning that cells with defective HR proteins Rad51, Rad54 and NBS1 show a low SCE frequency (Sonoda et al., 1999; Dronkert et al., 2000; Tauchi et al., 2002). On the other hand, the persistence of Rad51 foci and the delayed reactivation of replication in PARP-1 null cells suggest that more time is required to dismantle a higher amount of HR complexes (e.g. Rad51) before replication reactivation. In this regard, recent studies have shown that removal of Rad51 by Srs2 (a DNA helicase) from nucleoprotein filaments minimizes DNA recombination, which is thought to be a prerequisite for resolution and reactivation of stalled replication forks (Krejci et al., 2003; Veaute et al., 2003; see also reviews Oncogene

by Cox, 2001; McGlynn and Lloyd, 2002). However, the biochemical pathway used by PARP-1 in the negative regulation of Rad51 nucleoprotein filaments is currently unknown. Nevertheless, it is interesting to note that PARP-1 has been shown to associate with the multiplereplication complex (MRC) in DNA replication and, by poly(ADP-ribosyl)ation, and PARP-1 modifies replication molecules in vitro (Simbulan-Rosenthal et al., 1996). It is thus possible that the accumulation (or persistence) of Rad51 foci is one of the consequences of a lack of poly(ADP-ribosyl)ation which may modulate the ability of the repair complex to solve the stalled replication. Finally, we cannot rule out the possibility that PARP-1 deficiency may affect, either directly or indirectly, replisome or primosome activities to resolve the stalled replication forks and thereby induce replication fork collapse. The delayed clearance of Rad51 foci in PARP-1 null cells may also indicate an enforced effort to repair collapsed replication forks. If stalled replication forks are not repaired, chromosome aberrations are induced and apoptotic machinery is activated by the S-phase checkpoint (Sogo et al., 2002; see also Nyberg et al., 2002). Alternatively, in the absence of PARP-1, cells enter the S phase with a higher load of DNA damage, including ssDNA and DSBs, which will result in more HR that promotes strand exchange to repair replication forks. Consistent with these hypotheses, PARP-1 null cells exhibit a high degree of chromosomal aberrations (Tong et al., 2001a), Rad51 foci persistence and delayed reactivation of stalled replication forks, as well as increased apoptosis in response to HU treatment (the present study). Finally, the delayed re-entry of PARP1/ cells into the G2/M phase after HU removal may be explained by the possibility that PARP-1 directly modulates S-phase completion checkpoint molecules, which will affect S-phase exit and G2/M transition. In this regard, it is interesting to note that the absence of PARP-1 affects the accumulation or induction of p53 in response to DNA damage (Wesierska-Gadek et al., 1999). Taken together, our results suggest that the absence of PARP-1 facilitates Rad51 loading to stalled replication forks; however, inefficient clearance of Rad51 foci delays reactivation of replication and reentry into the cell cycle. It has been well documented that inactivation of PARP-1 results in chromosome instability, a high degree of SCE and a reduced proliferation rate. The present study shows that PARP-1 deficiency does not greatly change the frequency or character of the DSB repair carried out by HR or NHEJ pathways, suggesting a nonessential role for PARP-1 in these pathways. However, PARP-1 deficiency increased the accumulation of Rad51 and SCE frequency in response to replication arrest, implying an elevated recombination activity and/or higher levels of ssDNA breaks in the absence of PARP-1. These observations support the hypothesis that PARP-1 is indeed an anti-recombinogenic factor, and also explains the genomic instability phenotype associated with PARP-1 deficiency.


3881 Materials and methods Construction of cell lines for recombination assays 3T3 EF cell lines included A11, A12 (PARP-1/) and A16, A19 (PARP-1 þ / þ ) (Wang et al., 1995). EF cell lines containing a single intact DR-GFP reporter were constructed essentially as described previously (Pierce et al., 1999). PARP1 þ / þ and PARP-1/ embryonic stem (ES) cell lines were established from blastocysts derived by intercrossing heterozygous PARP-1 þ /mice (129/Sv; Wang et al., 1995) as described (Robertson, 1987). Gene targeting of hprt-DRGFP into two PARP-1 þ / þ (ES9-1 and ES32-3) and two PARP1/ (ES6-1 and ES17-2) male ES cell lines was performed as described previously (Pierce et al., 2001; data not shown). For HDR assays, 100 or 50 mg of the I-SceI expression vector pCBASce (Richardson et al., 1998) was electroporated into 1 107 EF or ES cells, respectively. Mock plasmid pCAGGS (without the I-SceI gene) and GFP-expressing plasmid pCAGGS-EGFP containing the same regulatory elements as pCBASce were constructed for mock and transfection efficiency controls, respectively. Flow cytometry analysis was performed using a FACScalibur apparatus equipped with Cellquest software (Becton Dickinson, San Jose, CA, USA). PCR detection of DSB repair and DNA sequencing DNA surrounding the I-SceI site was amplified by PCR from the genome of I-SceI-transfected cells using primer 1 (50 GGTTCGGCTTCTGGCGTG-30 ) and primer 2 (50 GTAGCCTTCGGGCATGGCGG-30 ). The PCR products were digested with I-SceI (New England Biolabs, England, UK) and resolved on 2.5% agarose gels. Following staining with ethidium bromide, the density of DNA bands was quantified as described previously (Pierce et al., 2001). To determine the spectra of DSB repair products, the I-SceIresistant fragments were recovered from the agarose gels and sequenced. P-values for error bars were calculated by GraphPad Prism biostatistics software. Colony formation assay Logarithmically growing EFs were trypsinized and a total of 1000 cells were plated onto a 10 cm culture dish in duplicate 3 h prior to the treatment of various doses of HU for 12 h. After 10 days, the colonies were fixed with methanol and stained with Giemsa. Colonies consisting of more than 50 cells were scored under a microscope and survival was calculated as a percentage of untreated respective genotype. Apoptosis assay EF cells were plated at 5 104 cells for 24 h before treatment with HU (Sigma, St. Quentin Fallavier, France) for 4 h, and apoptotic cells were stained using an In situ Cell Death Detection Fluorecin Kit (Roche, Meylan, France). The TUNEL-positive cells were scored in a total of 900 cells of

each genotype at each HU dosage and the relative apoptosis was obtained by dividing the TUNEL-positive cells after HU treatment with those from the respective nontreated group. SCE assay EFs were cultured in the presence of 10 mM 5-bromo-20 deoxyuridine (BrdU) (Roche) for 18 h, followed by 4 h in the presence or absence of 2 mM HU. After another 24 h in culture, the cells were collected in order to prepare chromosome spreads (Wang et al., 1997). Chromatid exchanges were scored in 50 spreads of each treatment and genotype. Immunofluorescence staining of Rad51 foci To detect spontaneous Rad51 foci, subconfluent primary EFs were fixed and stained with an anti-Rad51 antibody and counterstained with DAPI. For monitor Rad51 foci formation after HU treatment, primary EFs were synchronized with 2 mM of thymidine for 24 h, followed by a 2-h culture with a drug-free medium. Subsequently, cells were exposed to 2 mM HU for 4 h and harvested at various time points after removal of HU for fixation as described (Tarsounas et al., 2003). Cells were then stained with a polyclonal anti-Rad51 antibody (1 : 300 dilution, a kind gift from Dr S West) and visualized by a Cy3-conjugated secondary antibody (1 : 200; Sigma). Slides were mounted in Vectashield containing DAPI (Vector Laboratories, Burlingame, CA, USA) and the number of Rad51 foci from 60–80 Rad51 foci-positive nuclei of each treatment and genotype was scored. Cell cycle progression assay To analyze cell cycle progression after HU release, primary EFs were cultured in a medium containing 10 mM BrdU (Roche) for 2 h before HU treatment (see ‘Immunofluorescence staining of Rad51 foci’). Cells were collected and DNA was denatured with 2 N HCl/0.5% Triton X-100, followed by neutralization with 0.1 M Na2B4O7 (pH 8.5). BrdU incorporation was analyzed by flow cytometry after co-staining cells with the FITC-conjugated anti-BrdU monoclonal antibody (Becton Dickinson) and propidium iodide (PI, Sigma). The profile of cell proliferation was analyzed with Cellquest software (Becton Dickinson), and the proportion of replicative cells in the cell cycle was quantified with ModFIT LT software (VERITY, Maine, USA). Acknowledgements We thank Dr S West for the Rad51 antibody. We also thank Ms Jocelyne Michelon for technical assistance and Drs Z Herceg, W-M Tong and E Van Dyck for critical reading of the manuscript. Y-G Yang and S Patnaik are supported by a postdoctoral fellowship from the International Agency for Research on Cancer (IARC). This study is supported by the Association pour la Recherche sur le Cancer (ARC), France and the Association for International Cancer Research (AICR), UK.

References Ariumi Y, Masutani M, Copeland TD, Mimori T, Sugimura T, Shimotohno K, Ueda K, Hatanaka M and Noda M. (1999). Oncogene, 18, 4616–4625. Boulton SJ, Gartner A, Reboul J, Vaglio P, Dyson N, Hill DE and Vidal M. (2002). Science, 295, 127–131. Burkle A. (2001). BioEssays, 23, 795–806. Cortes F, Pinero J and Ortiz T. (1993). Mutat. Res., 303, 71–76.

Cox MM. (2001). Annu. Rev. Genet., 35, 53–82. Cox MM. (2002). Mutat. Res., 10, 107–120. D’Amours D, Desnoyers S, D’Silva I and Poirier GG. (1999). Biochem. J., 342, 249–268. Dronkert ML, Beverloo HB, Johnson RD, Hoeijmakers JH, Jasin M and Kanaar R. (2000). Mol. Cell. Biol., 20, 3147–3156. Oncogene


3882 Galande S and Kohwi-Shigematsu T. (1999). J. Biol. Chem., 274, 20521–20528. Henrie MS, Kurimasa A, Burma S, Menissier-de Murcia J, de Murcia G, Li GC and Chen DJ. (2003). DNA Repair, 2, 151–158. Hoeijmakers JH. (2001). Nature, 411, 366–374. Ikushima T. (1990). Chromosoma, 99, 360–364. Jackson SP. (2002). Carcinogenesis, 23, 687–696. Krejci L, Van Komen S, Li Y, Villemain J, Reddy MS, Klein H, Ellenberger T and Sung P. (2003). Nature, 423, 305–309. Liang F, Han M, Romanienko PJ and Jasin M. (1998). Proc. Natl. Acad. Sci. USA, 95, 5172–5177. Lindahl T, Satoh MS, Poirier GG and Klungland A. (1995). Trends Biochem. Sci., 20, 405–411. Lundin C, Erixon K, Arnaudeau C, Schultz N, Jenssen D, Meuth M and Helleday T. (2002). Mol. Cell. Biol., 22, 5869–5878. Masson JY and West SC. (2001). Trends Biochem. Sci., 26, 131–136. McGlynn P and Lloyd RG. (2002). Nat. Rev. Mol. Cell. Biol., 3, 859–870. Meńissier-de Murcia J, Mark M, Wendling O, Wynshaw-Boris A and de Murcia G. (2001). Mol. Cell. Biol., 21, 1828–1832. Meńissier-de Murcia JM, Niedergang C, Trucco C, Ricoul M, Dutrillaux B, Mark M, Oliver FJ, Masson M, Dierich A, LeMeur M, Walztinger C, Chambon P and de Murcia G. (1997). Proc. Natl. Acad. Sci. USA, 94, 7303–7307. Morrison C, Smith GC, Stingl L, Jackson SP, Wagner EF and Wang Z-Q. (1997). Nat. Genet., 17, 479–482. Nyberg KA, Michelson RJ, Putnam CW and Weinert TA. (2002). Annu. Rev. Genet., 36, 617–656. Pierce AJ, Hu P, Han M, Ellis N and Jasin M. (2001). Genes Dev., 15, 3237–3242. Pierce AJ, Johnson RD, Thompson LH and Jasin M. (1999). Genes Dev., 13, 2633–2638. Rainaldi G, Sessa MR and Mariani T. (1984). Chromosoma, 90, 46–49. Richardson C, Moynahan ME and Jasin M. (1998). Genes Dev., 12, 3831–3842. Robertson EJ. (1987). Teratocarcinomas and Embryonic Stem Cells: A Practical Approach. Oxford IRL Press: Oxford, pp. 71–112.

Oncogene

Saintigny Y, Delacote F, Vares G, Petitot F, Lambert S, Averbeck D and Lopez BS. (2001). EMBO J., 20, 3861–3870. Schultz N, Lopez E, Saleh-Gohari N and Helleday T. (2003). Nucleic Acids Res., 31, 4959–4964. Simbulan-Rosenthal CM, Rosenthal DS, Hilz H, Hickey R, Malkas L, Applegren N, Wu Y, Bers G and Smulson ME. (1996). Biochemistry, 35, 11622–11633. Sogo JM, Lopes M and Foiani M. (2002). Science, 297, 599–602. Sonoda E, Sasaki MS, Morrison C, Yamaguchi-Iwai Y, Takata M and Takeda S. (1999). Mol. Cell. Biol., 19, 5166–5169. Tarsounas M, Davies D and West SC. (2003). Oncogene, 22, 1115–1123. Tashiro S, Walter J, Shinohara A, Kamada N and Cremer T. (2000). J. Cell Biol., 150, 283–291. Tauchi H, Kobayashi J, Morishima K, Van Gent DC, Shiraishi T, Verkaik NS, Van Heems D, Ito E, Nakamura A, Sonoda E, Takata M, Takeda S, Matsuura S and Komatsu K. (2002). Nature, 420, 93–99. Tong W-M, Cortes U, Hande M-P, Ohgaki H, Cavalli LR, Lansdorp PM, Haddad BR and Wang Z-Q. (2002). Cancer Res., 62, 6990–6996. Tong W-M, Cortes U and Wang Z-Q. (2001a). Biochim. Biophys. Acta-Rev. Cancer, 1552, 27–37. Tong W-M, Galendo D and Wang Z-Q. (2000). Cold Spring Harb. Symp. Quant. Biol., 65, 583–591. Tong W-M, Hande MP, Lansdorp PM and Wang Z-Q. (2001b). Mol. Cell. Biol., 21, 4046–4054. Tong W-M, Ohgaki H, Huang H, Granier C, Kleihues P and Wang Z-Q. (2003). Am. J. Pathol., 162, 343–352. Veaute X, Jeusset J, Soustelle C, Kowalczykowski SC, Le Cam E and Fabre F. (2003). Nature, 423, 309–312. Wang Z-Q, Auer B, Stingl L, Berghammer H, Haidacher D, Schweiger M and Wagner EF. (1995). Genes Dev., 9, 509–520. Wang Z-Q, Stingl L, Morrison C, Jantsch M, Los M, SchulzeOsthoff K and Wagner EF. (1997). Genes Dev., 11, 2347–2358. Wesierska-Gadek J, Wang Z-Q and Schmid G. (1999). Cancer Res., 59, 28–34.