Genetic Interactions between BLM and DNA Ligase IV in Human Cells*

BLM has been implicated in DNA double-strand break (DSB) repair, but its precise role remains obscure. To explore this, we generated BLM–/– and BLM–/–LIG4–/– cells from the human pre-B cell line Nalm-6. BLM–/– cells exhibited retarded growth, increased mutation rates, and hypersensitivity to agents that block replication fork progression. Interestingly, these phenotypes were significantly suppressed by deletion of LIG4, suggesting that nonhomologous end-joining (NHEJ) is unfavorable for integrity and survival of cells lacking BLM. We propose that the absence of BLM leads to accumulation of replication-associated, one-ended DSBs, which are deleterious to cells and lead to genomic instability when repaired by NHEJ. In addition, the NHEJ pathway per se was marginally affected by BLM deficiency, as evidenced by x-ray sensitivity and I-SceI-based DSB repair assays. More intriguingly, however, these experiments revealed the presence of an alternative, DNA ligase IV-independent end-joining pathway, which was significantly affected by the loss of BLM. Collectively, our results provide the first evidence for genetic interactions between BLM and NHEJ in human cells.

DNA double-strand breaks (DSBs) 1 are caused by various endogenous and exogenous mechanisms. Efficient and faithful repair of DSBs is crucial for maintenance of genome integrity. If left unrepaired, such lesions may lead to cell death (1). There are two distinct major pathways to repair DSBs, homologous recombination (HR) and nonhomologous end-joining (NHEJ). HR is thought to be active in S to G 2 phases of the cell cycle (2). Recent studies have shown that HR plays a predominant role in the repair of replication-associated DSBs, caused by collision of a replication fork with single-strand breaks (SSBs) or gaps (3). On the other hand, NHEJ is thought to be active throughout the cell cycle, especially G 1 phase (2). We have shown that DNA topoisomerase II (Topo II) inhibitor-induced DSBs are predominately repaired by NHEJ in vertebrate cells (4,5). Mutant cell lines deficient in HR or NHEJ have been shown to be hypersensitive to ionizing radiation, indicating that both pathways are important for the repair of radiation-induced DSBs (2, 6 -9). Together, it seems likely that the nature of DSBs directs which pathway is preferred to repair the damage.
HR, which relies on a number of proteins such as RAD51, RAD51 paralogs, and RAD54 (1), undergoes a precise repair of DSBs by utilizing a homologous sequence as template. In contrast, NHEJ repairs DSBs by simply joining the broken ends, irrespective of sequence homology, with frequent nucleotide loss. Extensive biochemical studies have proposed a model for the mechanism of NHEJ (10,11). The NHEJ reaction is initiated by binding of Ku protein (the heterodimer of Ku70 and Ku86) to broken ends. Then the DNA-dependent protein kinase catalytic subunit-Artemis complex is recruited and the ends are processed to make them ligatable. Finally, the DNA ligase IV (LIG4)-XRCC4 complex is recruited for ligation. Targeted disruption of the LIG4 gene in the human pre-B cell line Nalm-6 has revealed that LIG4 is indispensable for V(D)J recombination that exclusively depends on the NHEJ pathway (6). Knockout mouse models demonstrate that NHEJ plays a critical role in maintaining genome integrity and suppressing carcinogenesis (12,13). Consistent with this, NHEJ-deficient cells show increased chromosome aberrations such as breakage and translocations (14,15).
Bloom syndrome (BS) is a rare genetic disorder characterized by proportional dwarfism, immunodeficiency, sun sensitivity, male infertility, and a high incidence of a wide spectrum of cancer (16). Cells from BS patients (BS cells) show various types of chromosomal instability, including chromosomal breaks, gaps, deletions, and elevated levels of sister chromatid exchanges (SCEs). This instability may be responsible for cancer proneness in BS. The gene mutated in BS, BLM, encodes DNA helicase belonging to the highly conserved RecQ family (17). Escherichia coli RecQ helicase (18,19) and its homologs, Sgs1 in Saccharomyces cerevisiae (20) and Rqh1 in Schizosaccharomyces pombe (21), have been implicated in genetic recombination. There are five RecQ-type helicases in human cells. Mutations in the BLM, WRN, and RTS genes cause genetic disorders, BS (17), Werner syndrome (22), and Rothmund-Thomson syndrome (23), respectively. Biochemical studies have revealed that BLM helicase has a 3Ј-5Ј DNA-unwinding activity in an ATP-dependent manner (24) and that it interacts with a wide variety of proteins involved in DNA replication, recombination, and repair (for review, see Ref. 25). Of note, a physical interaction between BLM and RAD51 suggests a role for BLM in HR. Indeed, BLM is proposed to function in the HR pathway to promote proper intermediate resolution and suppress crossover (26). Inactivation of BLM can lead to enhanced HR activity, as manifested by increased gene targeting efficiencies (27)(28)(29). Aberrant HR in the absence of BLM could account for the genomic instability observed in BS cells. In fact, loss of heterozygosity is prominent in BS cells (27,28). It remains possible, however, that another DSB repair pathway, NHEJ, may contribute more directly to this genomic instability given that NHEJ, unlike HR, is typically an error-prone repair pathway.
The fact that BS cells exhibit normal V(D)J recombination which exclusively depends on NHEJ (30) argues against a possible role of BLM in NHEJ. Consistent with this, BLM-null chicken cells show no increased sensitivity to the Topo II inhibitor VP-16 (29). Recently, however, genetic interactions between BLM and NHEJ have been reported in Drosophila (31,32). This may suggest an as-yet undefined functional interplay between BLM and the NHEJ pathway in mammalian cells. Several groups have investigated NHEJ activity in BS cells. Using a plasmid-based in vivo joining assay, Rü nger and Kraemer (33) reported earlier a reduced and error-prone end-joining in BS cells. However, recent in vitro end-joining assays with cell extracts have revealed confusing results; namely, whereas the efficiency and fidelity of DSB repair in BS and control cells are comparable when ligatable ends are present (34), BS cells display a highly increased and error-prone end-joining accompanied by frequent large deletions (35). More recently, Onclercq-Delic et al. (36) described an accurate joining of linear plasmid introduced into BS cells. Taken together, the results and conclusions concerning the interaction between BLM and NHEJ are contradictory and unsettled. Thus, a role of BLM in the NHEJ pathway still remains elusive in mammalian cells.
In this study, we generated BLM Ϫ/Ϫ and BLM Ϫ/Ϫ LIG4 Ϫ/Ϫ cells by gene targeting using the human Nalm-6 cell line and its derivative null for the LIG4 gene, respectively. BLM Ϫ/Ϫ cells exhibited phenotypes characteristic of BS cells reported before (16). Intriguingly, we found that some of the phenotypes such as retarded growth, increased mutation rates, and hypersensitivity to methyl methanesulfonate (MMS), UV, and hydroxyurea (HU) are significantly suppressed by LIG4 deficiency. These results suggest an unfavorable impact of NHEJ on growth and viability of BLM-deficient cells. Using an I-SceIbased DSB repair assay system designed in this study, we found that BLM Ϫ/Ϫ cells have an NHEJ activity similar to that of wild type. In sharp contrast, LIG4 Ϫ/Ϫ cells displayed aberrant end-joining events such as frequent nucleotide loss and rearrangements, which were much more pronounced in a BLM Ϫ/Ϫ background. These results suggest the existence of an alternative end-joining pathway that is independent of LIG4 and highly error prone. Based on these results, we propose that BLM deficiency leads to accumulation of replication-associated, one-ended DSBs, which are deleterious to cells when repaired by NHEJ. Our results suggest the interdependence of the endjoining pathway and the function of BLM helicase.

EXPERIMENTAL PROCEDURES
Vector Construction-A partial BLM cDNA was isolated by RT-PCR with primers 5Ј-TGATGCCAGTCTTCTTGGCTCA-3Ј and 5Ј-CCTT-GATGGGATAGGCAGC-3Ј. Using the cDNA as probe, a genomic fragment containing BLM exons 6 -12 was obtained by screening a human genomic library (Stratagene). Two BLM-targeting constructs, pBLMhyg and pBLM-his, were made by replacing a 1.8-kb region containing exons 11 and 12 with a hygromycin resistance or histidinol resistance gene flanked by loxP sequences, respectively. The diphtheria toxin A expression cassette, excised from pMC1DT-ApA (KURABO, Osaka), was added to the 5Ј-terminal BLM region. To generate an hypoxanthine phosphoribosyltransferase (HPRT) targeting construct, an 8.9-kb human HPRT sequence containing exons 2 and 3 was excised from the BAC clone bWXD187 (Washington University School of Medicine, Department of Molecular Microbiology, Genome Sequencing Center, St Louis, MO). With this sequence, the targeting construct pHPRT-Sce was made by inserting the hygromycin resistance gene, which has an I-SceI site upstream of its promoter, into the XhoI site in exon 3. The diphtheria toxin A expression cassette was also added to the 5Ј terminus of the HPRT sequence.
Cell Culture and DNA Transfection-The human pre-B cell line Nalm-6 and its LIG4 Ϫ/Ϫ derivative were maintained in ES medium (Nissui Seiyaku Co., Tokyo) supplemented with 50 M 2-mercaptoethanol, 10% calf serum (Hyclone) (growth medium) at 37°C. For colony formation, cells were grown for 2-3 weeks in ES medium supplemented with 50 M 2-mercaptoethanol, 20% calf serum, and 0.15% agarose (agarose medium). DNA transfection for gene targeting was performed as described previously (7). Briefly, 4 ϫ 10 6 cells were electroporated with 4 g of linearized DNA constructs, cultured for 22 h, and replated at a density of 5-10 ϫ 10 5 per 90-mm bacterial dish into agarose medium containing either 0.4 mg/ml hygromycin B (Wako, Osaka) or 1.2 mg/ml L-histidinol (Sigma-Aldrich). After a 2-3-week incubation, the resulting drug-resistant colonies were isolated and expanded to store frozen and to prepare genomic DNA for Southern blot analysis.
Western Blot and Fluorescence-activated Cell Sorter Analysis-Western blot analysis was performed as described previously (37). The rabbit polyclonal antibody raised against the C-terminal 376 amino acids of human BLM protein (Abcam, Cambridge, UK) was used. Fluorescenceactivated cell sorter analysis was performed with logarithmically growing cells. Cells were harvested, resuspended with 0.1% Triton X-100 (Wako) in Ca 2ϩ -and Mg 2ϩ -free phosphate-buffered saline, treated with RNase, stained with propidium iodide, and analyzed in a Coulter EP-ICS XL cytometer (Coulter). The significance of the differences was analyzed by t test.
Clonogenic Assay-To determine sensitivity to genotoxic agents, 1 ϫ 10 2 -2 ϫ 10 5 cells were plated in 60-mm bacterial dishes containing 5 ml of agarose medium with various concentrations of the drugs. All the drugs were dissolved in dimethyl sulfoxide. To determine sensitivity to x-rays, 1ϫ 10 2 -2 ϫ 10 5 cells were plated in 60-mm dishes containing 5 ml of agarose medium and exposed to various doses of x-rays as described (7). To determine UV sensitivity, cells were exposed to different UV doses, inoculated in agarose medium, and cultured as above. After a 2-3-week incubation, resultant colonies were counted, and surviving fractions were calculated. The significance of differences at the highest dose point was analyzed by t test.
SCE Assay-Cells were plated at a density of 10 5 /ml and cultured for approximately two cell cycle periods in growth medium containing 2 g/ml 5Ј-bromodeoxyuridine (Sigma). During the final 1-2 h, colcemid (0.05 g/ml, Wako) was added to collect metaphase cells. The cells were harvested, washed once with Ca 2ϩ -and Mg 2ϩ -free phosphate-buffered saline, treated with 75 mM KCl, and fixed with methanol-acetic acid (3:1). The samples were then air-dried, stained with 10 g/ml Hoechst 33258 (Sigma) for 15 min, rinsed with 2ϫ SSC (0.3 M NaCl and 30 mM sodium citrate (pH 7.0)), exposed to UV light for 10 min at 65°C, and stained with 2% Giemsa for 5 min.
Fluctuation Test for 6-Thioguanine (6TG) Resistance-The spontaneous mutation rate was determined by the fluctuation test of Luria and Delbrü ck (38). Briefly, 24-well plates, each containing 20 -40 cells/well in 2 ml of growth medium, were incubated for 12-24 days until the cell number reached ϳ1-2 ϫ 10 6 cells/well. Then, aliquots (1 ml) of individual cultures were replated into 90-mm dishes containing 10 ml of agarose medium with 20 M 6TG (Sigma). After a 2-3 week incubation, resultant colonies were counted. The spontaneous mutation rate was calculated by the mean method formulated by Capizzi and Jameson (39).
I-SceI-based DSB Repair Assay-pHPRT-Sce was linearized with NotI and transfected into wild-type and mutant cell lines as described above; the cells were selected doubly with 0.4 mg/ml hygromycin B and 10 M 6TG. Targeted integration was screened by Southern blotting. The resulting targeted lines WTS, LIG4 Ϫ/Ϫ S, BLM Ϫ/Ϫ S, and LIG4 Ϫ/Ϫ BLM Ϫ/Ϫ S were used to assess chromosomal DSB repair. Each cell line was electroporated with 4 g of the I-SceI expression vector pSceI (40), plated into agarose medium, and cultured for 2-3 weeks to allow colony formation. The resulting single clones were isolated, expanded to mass cultures, and used to prepare genomic DNA. Genomic PCR analysis was carried out using primers Sce-F (5Ј-TGAGGGCAA-AGGATGTGTTACG-3Ј) and Sce-R (5Ј-GTGAGTTCAGGCTTTTTCAT-GG-3Ј), located in intron 2 (immediately upstream of exon 3) of the HPRT gene and in the hygromycin resistance gene, respectively. Then, 1-l aliquots of PCR products were digested with 1 unit of I-SceI (New England Biolabs) at 37°C for 2 h and electrophoresed in 2% agarose gels. PCR products resistant to digestion with I-SceI were cloned into pGEM-T Easy (Promega) and sequenced with a Multi-capillary DNA Sequencer CEQ2000 (Beckman Coulter).

RESULTS
Targeted Disruption of the BLM Locus-To disrupt two alleles of the BLM gene in Nalm-6 cells, we generated two targeting constructs, pBLM-hyg and pBLM-his, containing a hygromycin resistance gene or histidinol resistance gene, respectively, flanked by the 6.7-and 2.9-kb genomic BLM sequences (Fig. 1A). The constructs also contained a diphtheriatoxin A gene to exclude random integrants. Targeted disruption with the constructs resulted in loss of exons 11 and 12 and led to inactivation of the BLM gene. pBLM-hyg was first transfected into wild-type Nalm-6 cells; the resulting 168 hygromycin-resistant clones were isolated and examined by Southern blot analysis. We obtained 14 heterozygous (BLM ϩ/hyg ) clones. pBLM-his was then transfected into one of the clones, and 274 histidinol-resistant clones were isolated and analyzed. We obtained five doubly targeted (BLM his/hyg ) clones. Subsequently, both of the hygromycin and histidinol resistance genes in BLM his/hyg cells were removed by transient expression of Cre recombinase, finally yielding BLM Ϫ/Ϫ cells. Removal of the marker genes was verified by Southern blotting (Fig. 1B). The disruption of the BLM gene was also confirmed by Western (Fig. 1C) and Northern blot analysis (data not shown).
Likewise, to generate cells null for both LIG4 and BLM, targeted disruption of the BLM gene was done using LIG4 Ϫ/Ϫ cells derived from the Nalm-6 cell line (6). The above two BLM targeting constructs and the Cre expression vector were sequentially transfected into LIG4 Ϫ/Ϫ cells. The disruption of the BLM gene and removal of the selective marker genes were also confirmed by Southern blotting (data not shown).
Severe Growth Defect in BLM Ϫ/Ϫ Cells Is Suppressed by LIG4 Deficiency-To compare the growth properties of wild-type and mutant cell lines, we first monitored growth curves ( Fig. 2A). The doubling times of wild-type, LIG4 Ϫ/Ϫ , BLM Ϫ/Ϫ , and LIG4 Ϫ/Ϫ BLM Ϫ/Ϫ cells were 22, 22, 41, and 35 h, respectively. It is obvious that BLM Ϫ/Ϫ cells have a remarkably slower growth rate than wild-type and LIG4 Ϫ/Ϫ cells (p Ͻ 0.001). Intriguingly, LIG4 Ϫ/Ϫ BLM Ϫ/Ϫ double-mutant cells were partially but significantly rescued from the growth defect seen in BLM Ϫ/Ϫ single mutant (p Ͻ 0.01). To confirm this rescue, we corrected one of the defective LIG4 alleles by gene targeting with a plasmid vector containing a 4.0-kb LIG4 genomic fragment and generated LIG4 ϩ/Ϫ BLM Ϫ/Ϫ clones. We found that the two clones assayed grew at a slightly slower rate than the parental LIG4 Ϫ/Ϫ BLM Ϫ/Ϫ cells (date not shown). This finding supports the suppression of the retarded growth of BLM Ϫ/Ϫ cells by LIG4 deficiency. We also analyzed cell cycle phase distributions of logarithmically growing cells by flow cytometry. Fig. 2B shows that the distributions at G 1 , S, and G 2 /M phases were not significantly different among the cell lines. The sub-G 1 fractions (representing apoptotic cells) in wild-type and LIG4 Ϫ/Ϫ cells were essentially the same (1.1-1.3%), whereas BLM Ϫ/Ϫ showed a 4.4-fold higher percentage than wild-type cells (p Ͻ 0.001) (Fig. 2C). Furthermore, LIG4 Ϫ/Ϫ BLM Ϫ/Ϫ cells exhibited a slightly lowered value compared with BLM Ϫ/Ϫ cells.
Increased Mutation Rates, but Not SCE Levels, in BLM Ϫ/Ϫ cells Are Suppressed by LIG4 Deficiency-Similar to BS cells (16), BLM-deficient cells (27,29) have been shown to manifest increased levels of SCE. We examined the SCE level in wildtype and mutant cell lines. Table I shows  It has been reported that BS cells show elevated mutation rates compared with cells derived from normal individuals (41). We determined spontaneous mutation rates at the HPRT locus in wild-type and mutant cell lines. Table II shows that the mutation rate in wild-type cells was 5.4 ϫ 10 Ϫ7 /cell/generation, whereas LIG4 Ϫ/Ϫ cells showed a 5.7-fold increased rate. As expected, BLM Ϫ/Ϫ cells showed a 6.9-fold higher rate than wild-type, indicating a mutator phenotype associated with BLM deficiency. Interestingly, the mutation rate in LIG4 Ϫ/Ϫ BLM Ϫ/Ϫ cells appeared slightly reduced relative to that in BLM Ϫ/Ϫ cells, implying that LIG4 deficiency may improve genomic stability by suppressing the mutator phenotype associated with BLM deficiency.
Hypersensitivity to MMS, UV, and HU of BLM Ϫ/Ϫ Cells Is Suppressed by LIG4 Deficiency-BS cells and BLM Ϫ/Ϫ DT40 cells are known to be hypersensitive to genotoxic agents such as MMS (29), UV (42), or HU (43). MMS-or UV-induced DNA lesions are mainly repaired by base excision repair (44) or nucleotide excision repair (45), respectively. During the repair, SSBs (nick or gap) may arise as repair intermediates in one strand, whereas a single-strand region also arises in the opposite strand. HU, an inhibitor of ribonucleotide reductase, depletes NTP pools and stops replication fork progression, giving rise to a single-strand region ahead of a replication fork. We examined sensitivity of wild-type and mutant cell lines to these agents by clonogenic survival assays. BLM Ϫ/Ϫ cells were more sensitive to MMS than wild-type cells (p Ͻ 0.001) (Fig. 3A), whereas LIG4 Ϫ/Ϫ cells were not. Importantly, LIG4 Ϫ/Ϫ -BLM Ϫ/Ϫ cells exhibited slightly but significantly reduced sensitivity to the agent compared with BLM Ϫ/Ϫ cells (p Ͻ 0.05), indicating that LIG4 deficiency suppresses the hypersensitivity of BLM Ϫ/Ϫ cells. We also examined sensitivity to UV that produces thymine dimers and blocks replication. BLM Ϫ/Ϫ cells were hypersensitive to UV radiation compared with wild-type cells (p Ͻ 0.02) and that the hypersensitivity was also suppressed in the double-mutant cells (p Ͻ 0.2) (Fig. 3B). Similarly, BLM Ϫ/Ϫ cells were more sensitive to HU treatment   (p Ͻ 0.01) than wild-type cells, and this sensitivity was also suppressed by LIG4 deficiency (p Ͻ 0.02) (Fig. 3C). Taken together, these results indicate that LIG4 deficiency alleviates the cytotoxic effects of those three agents on BLM Ϫ/Ϫ cells.
BLM Ϫ/Ϫ Cells Are Not Hypersensitive to VP-16 and X-rays-We also examined the effect of VP-16, a Topo II inhibitor, which is known to generate DSBs by the formation of cleavable complexes (46). Although BLM Ϫ/Ϫ cells showed no significantly increased sensitivity to the drug, LIG4 Ϫ/Ϫ cells were remarkably hypersensitive compared with wild-type cells (Fig. 3D). These results are consistent with our previous data that Topo II-mediated DSBs are predominantly repaired via the NHEJ pathway in vertebrate cells (4,5). LIG4 Ϫ/Ϫ BLM Ϫ/Ϫ cells showed essentially the same sensitivity to the drug as LIG4 Ϫ/Ϫ single mutant (Fig. 3D). We further examined the sensitivity of wildtype and mutant cell lines to x-ray radiation. BLM Ϫ/Ϫ cells were not hypersensitive to X-rays, whereas LIG4 Ϫ/Ϫ cells were more sensitive than wild type (Fig. 3E). These results suggest that BLM is not directly involved in the repair of DSBs via the NHEJ pathway. However, LIG4 Ϫ/Ϫ BLM Ϫ/Ϫ cells were slightly but significantly more radiosensitive than LIG4 Ϫ/Ϫ single-mutant cells (p Ͻ 0.01), indicating that BLM deficiency further increases the radiosensitivity associated with LIG4 deficiency. Interestingly, this finding contrasts with the results obtained with the doublemutant cells that were exposed to agents that block replication fork progression (see Fig. 3, A-C), suggesting a difference in BLM function on distinct chromosomal lesions.
Impact of BLM Deficiency on the Repair of I-SceI-induced DSBs-To study NHEJ-mediated DSB repair on chromosomal DNA, we designed an I-SceI-based DSB repair assay system (Fig. 4). To generate cell lines carrying the restriction endonuclease I-SceI recognition sequence integrated into the specific chromosomal site, we transfected wild-type and mutant cell lines with the targeting vector pHPRT-Sce, which contains an I-SceI site 5Ј-upstream of a hygromycin resistance gene inserted into exon 3 of the HPRT gene (Fig. 5A). After selection with both hygromycin B and 6TG, correctly targeted clones were confirmed by Southern blotting (Fig. 5B). Because Nalm-6 cells had been derived from a male, in targeted clones a 9.5-kb band was detected in place of a 13.3-kb band of the wild-type HPRT allele (which is located on the X chromosome). The cell lines created from wild-type, LIG4 Ϫ/Ϫ , BLM Ϫ/Ϫ , and LIG4 Ϫ/Ϫ BLM Ϫ/Ϫ cells were designated as WTS, LIG4 Ϫ/Ϫ S, BLM Ϫ/Ϫ S, and LIG4 Ϫ/Ϫ BLM Ϫ/Ϫ S, respectively.
Next, to cleave the I-SceI site in vivo, each targeted cell line was transfected with the I-SceI expression vector pSceI and cultured for colony formation (Fig. 4). The resulting colonies were randomly isolated and used to purify genomic DNA. With the DNA as template, a region flanking the I-SceI site was amplified by PCR, and the products were subjected to in vitro I-SceI digestion; the digestible products that indicated accurate repair (or otherwise, uncut I-SceI site) were discarded, whereas I-SceI-resistant products resulting from inaccurate repair were sequenced (Fig. 4). Table III summarizes the results of these experiments. We found that of 695 clones derived from WTS cells, only 11 (1.6%) had lost the original I-SceI site. In BLM Ϫ/Ϫ S cells, 17 of 697 clones (2.4%) were I-SceI-resistant clones. Therefore, it seems that imprecise joining is slightly more frequent in the absence of BLM. In contrast, LIG4 Ϫ/Ϫ S cells showed a 3-fold higher percentage of inaccurate joining: 10 of 191 clones (5.2%) examined had lost the I-SceI site. Strikingly, LIG4 Ϫ/Ϫ BLM Ϫ/Ϫ S cells showed an extremely elevated percentage as high as 14%. Therefore, it appears that LIG4 deficiency results in elevated inaccurate joining, which is further enhanced by BLM deficiency.
Sequence analysis of the I-SceI-resistant PCR products revealed significant differences in the structure of inaccurate joints among the cell lines. All of 11 clones derived from WTS cells contained small insertions ranging from 1 to 5 bp, one of which (W11) lost 2 and 1 bp at the 5Ј and 3Ј termini of the cleavage site, respectively (Fig. 6B). All of the clones derived from BLM Ϫ/Ϫ S cells contained 1-8-bp insertions, five of which (B13-B17) were accompanied by 2-8-bp deletions. Therefore, it seems likely that BLM Ϫ/Ϫ S cells had slightly larger insertions and deletions than WTS cells. No microhomology at the junctions was seen in the clones from WTS or BLM Ϫ/Ϫ S cells. Importantly, in the clones derived from both lines, no large deletions or rearrangements were found. These results suggest that WTS and BLM Ϫ/Ϫ S cells possess almost similar ability to repair chromosomal DSBs. In contrast, 5 (L1-L5) of 10 clones derived from LIG4 Ϫ/Ϫ S cells had significantly larger (1-15 bp) deletions, and two (L3 and L4) were accompanied by insertions; clone L3 contained an insert as large as 1 kb, derived from other chromosomes (Fig. 6B). 4-bp microhomologies were found at the junctions in two clones (L1 and L2) (Fig. 6C). In addition, it should be noted that the five clones (L6 -L10) underwent gross rearrangements. Intriguingly, larger deletions were consistently observed in the clones derived from LIG4 Ϫ/Ϫ BLM Ϫ/Ϫ S cells. Sequence analysis revealed that 12 clones (LB1-LB12) had deletions ranging from 4 to 295 bp, with 2-16-bp insertions in 4 clones (LB6, LB7, LB9, and LB12) and complex rearrangements in 2 clones (LB13 and LB14) (Fig. 6B). Microhomologies of 1-2 bp were found in 8 clones (Fig. 6C).

DISCUSSION
In this study, we have generated BLM Ϫ/Ϫ cells by gene targeting in the human pre-B cell line Nalm-6 using simple positive-negative selection type-targeting constructs (Fig. 1). The targeting efficiency in the first targeting was as high as 8.3%, and that in the second targeting was as high as 1.8%. We observed similar high efficiencies with the LIG4 Ϫ/Ϫ cell line. This mutant line had been isolated with similar high efficiencies from Nalm-6 cells (6). Thus, it is likely that the Nalm-6 cell line possesses high targeting efficiency compared  The number of colonies randomly selected after transfection with I-SceI expression vector. b Sequence analysis was performed with all these clones. The percentage of I-SceI-resistant clones is shown in parentheses. c All events were accompanied by insertions. d All events were accompanied by deletions.
FIG . 4. Scheme of the I-SceI-based DSB repair assay system. Cell lines harboring an I-SceI site inserted into exon 3 (EX3) of the HPRT locus (HPRT-Sce locus) were generated from wild-type and mutant cell lines by gene targeting. Each cell line was then transfected with I-SceI expression vector pSceI and grown for colony formation. The resulting colonies were independently examined for the repair of a cleaved I-Sce-I site by PCR using two DNA primers corresponding to 358 bp upstream and 292 bp downstream of the I-SceI site. After in vitro I-SceI digestion of the PCR products, clones having the I-SceI site produced two bands of 0.35 and 0.29 kb, whereas clones lacking the site produced a single band of different sizes; these clones were subjected to sequence analysis. See "Results" for details.
with other human cell lines. This intriguing feature of Nalm-6 cells will be reported in detail elsewhere.
As compared with wild-type cells, BLM Ϫ/Ϫ cells exhibited severe growth defects ( Fig. 2A), hypersensitivity to MMS, UV, and HU (Fig. 3, A-C), elevated SCE levels (Table I), and high mutation rates (Table II). These phenotypes are similar to those manifested in BS cells (16). Because gene targeting assures the isogenicity between wild-type and targeted cells, we believe that BLM Ϫ/Ϫ Nalm-6 cells are a better model to dissect BLM functions precisely.
Some Phenotypes of BLM-null Cells Are Suppressed by LIG4 Deficiency-We found that LIG4 deficiency suppresses some phenotypes associated with the BLM deficiency, including severe growth defect (Fig. 1A), hypersensitivity to MMS, HU, and UV (Fig. 3, A-C) and high mutation rates (Table II). Because LIG4 is specifically required for NHEJ, these results suggest that NHEJ has an unfavorable effect on growth and viability of BLM-deficient cells. This was unexpected, given that NHEJ plays an important role in maintaining genomic integrity (12)(13)(14)(15).
How does NHEJ act negatively in the absence of BLM? DSBs with only one end (one-ended DSB), depicted by an open circle in Fig. 7, arise during S phase, when a replication fork collides with lesions such as SSBs and gaps arising spontaneously (3,47). This type of DSB induces HR via an exchange mechanism, and HR plays a predominant role in the repair of the lesions during late S-G 2 phases. In this process, BLM likely acts to prevent the formation of such replication-associated DSBs by resolution of an HR intermediate containing a double Holliday junction (26,48). Hence, BLM deficiency may cause defects in such functions and give rise to considerable amounts of one-ended DSBs (i.e. partially replicated sister chromatids) (Fig. 7). However, the HR pathway alone might be insufficient to repair all the DSBs. In such a case, the NHEJ pathway would deal with some of the DSBs but fail to repair them properly (49). Taken together, we propose that this one-ended DSB repair by the NHEJ pathway generates deleterious recombination products such as large intrachromosomal deletions or interchromosomal translocations, block replication restart, and become highly cytotoxic, resulting in death of BLM-deficient cells. This may be reflected in their slowed growth rate and increase in sub-G 1 cells (Fig. 2, A-C). Therefore, LIG4 deficiency could rescue BLM Ϫ/Ϫ cells by avoiding NHEJ-mediated deleterious recombination products.
DNA base damage or adducts are usually repaired by base excision repair (44) or nucleotide excision repair (45), respectively, where SSBs or gaps arise transiently as repair intermediates. However, abnormally increased levels of such lesions in cells treated with MMS or UV light would become a strong blocker of replication fork progression during S phase and generate abundant one-ended sister chromatids (50,51). HU treatment also arrests fork progression by nucleotide depletion, leading to stalled forks in the same manner as above (52). Treatment with these agents would rapidly increase the amount of one-ended, chromosomal DSBs, especially in the absence of BLM, so that a concomitant dependence on NHEJ activity would abundantly produce deleterious recombination products, block replication restart, and eventually lead to cell death. This explanation may be compatible with our observations that BLM Ϫ/Ϫ cells are hypersensitive to MMS, UV, and HU and that the hypersensitivity is partially but significantly suppressed by LIG4 deficiency (Fig. 3, A-C). We have reported a similar cytotoxic effect of NHEJ on the repair of camptothecin-induced DSBs in DT40 cells (53).
BLM Ϫ/Ϫ cells Possess an NHEJ Activity Similar to That of Wild-type Cells-BLM Ϫ/Ϫ cells displayed little or no increased sensitivity to DSB-inducing agents VP-16 and X-rays (Fig. 3, D and E). In contrast, LIG4 Ϫ/Ϫ cells were hypersensitive to these agents, consistent with previous studies (5)(6)(7)9). These results strongly suggest that BLM Ϫ/Ϫ cells have almost the same NHEJ activity as wild-type cells. This finding is in good agreement with that observed in BLM Ϫ/Ϫ DT40 cells exposed to VP-16 (29).
The NHEJ pathway is considered to be a direct end-joining reaction of two-ended DSBs. These DSBs, depicted by a shaded circle in Fig. 7, arise spontaneously throughout the cell cycle or when cells are exposed to DSB-inducing agents such as ionizing radiation and Topo II inhibitors or upon I-SceI expression. The NHEJ pathway seems to preferentially repair such damage throughout the cell cycle (2), although it may be backed up by the HR pathway during late S-G 2 phases. The I-SceI-based DSB repair assay system allowed us to examine such NHEJ events occurring on a human chromosome (Fig. 4). WTS and BLM Ϫ/Ϫ S cells showed a similar percentage of inaccurate repair of such a DSB and a similar pattern of joint structures (Table III, Fig. 6B); i.e. all of the I-SceI-resistant clones derived from both cell lines displayed junctions having small insertions alone or with small deletions ranging from 1 to 8 bp. It should be noted that neither large deletions nor rearrangements were found in the junctions of WTS and BLM Ϫ/Ϫ S cells. These results suggest that the NHEJ pathway is not strongly affected by BLM deficiency. Therefore, together with their insensitivity to the DSB-inducing agents (Fig. 3, D and E), it is conceivable that BLM Ϫ/Ϫ cells possess essentially a normal capacity of NHEJ to repair two-ended chromosomal DSBs. This notion may be supported by the previous observation that BS cells display a normal V(D)J recombination activity (30). Therefore, we suggest that the repair of two-ended DSBs by NHEJ hardly contributes to the genomic instability in BS cells. This may argue against earlier results obtained with BS cell extracts (33)(34)(35) but agrees with recent work on end-joining (36).
Possible Role of BLM in DSB Repair in the Absence of LIG4 -In the I-SceI-based DSB repair assay (Table III, Fig. 6), LIG4 Ϫ/Ϫ S cells showed an increased percentage of inaccurate repair compared with WTS cells. Importantly, large deletions and rearrangements were frequently found in the LIG4 Ϫ/Ϫ S FIG. 7. Model for generation and repair of two types of chromosomal DSBs. One-ended DSBs may arise during S phase spontaneously or when a replication fork collides with lesions such as SSBs and gaps caused by MMS, UV, or HU. These DSBs may be repaired predominantly by HR (accurate repair). However, in the absence of BLM, the DSBs may be repaired by the classical NHEJ pathway, but the repair would be deleterious to cells, leading to cell death (toxic repair). As a consequence, LIG4 deficiency would be favorable for survival of BLM-null cells. In contrast, two-ended DSBs arise spontaneously throughout the cell cycle or by exposure to DSB-inducing agents including x-ray radiation or after transfection with I-SceI vector in the assay system shown in Fig. 4. These DSBs may be repaired predominantly by the classical NHEJ pathway (error-prone repair), partially by HR (accurate repair), and also partially by an aEJ pathway (highly error-prone repair) that is independent of LIG4.
clones. These results clearly show that LIG4 is required for the NHEJ pathway to repair DSBs and to suppress gross chromosomal aberrations, consistent with the finding that NHEJ deficiency increases genomic instability (14,15). However, the fact that surviving clones were indeed recovered in the absence of LIG4 indicates the existence of an alternative end-joining (aEJ) pathway; that is, a LIG4-independent, error-prone pathway (see Fig. 7). Several groups have suggested an alternative, Ku-independent end-joining pathway (49,54). This pathway might be related to the aEJ pathway described here. One may suspect the involvement of single-strand annealing (SSA), which is one of the HR pathways (55), since SSA can join two DNA ends with direct repeats of Ն10 bp (56). In this study, however, microhomologies more than 4 bp were not observed at the junctions (Fig. 6C). Therefore, it seems unlikely that SSA directly contributes to the repair of I-SceI induced DSBs in our assay system.
Despite essentially a normal NHEJ activity in BLM Ϫ/Ϫ cells, the I-SceI-based DSB repair assay revealed that BLM deficiency further increased the percentage of inaccurate repair and the infidelity of end-joining associated with LIG4 deficiency (Table III, Fig. 6). This may be consistent with the observation that LIG4 Ϫ/Ϫ BLM Ϫ/Ϫ double-mutant cells are more sensitive to X-rays than LIG4 Ϫ/Ϫ single-mutant (Fig.  3E). Together, these results suggest that BLM somehow acts on the repair of two-ended DSBs through the LIG4-independent, aEJ pathway (Fig. 7). Ku protein is known to efficiently associate with DSB ends, thereby allowing a NHEJ reaction by recruiting protein complexes including DNA-dependent protein kinase catalytic subunit/Artemis and LIG4/XRCC4 (11). However, in LIG4 Ϫ/Ϫ cells, the NHEJ reaction cannot proceed; moreover, the HR or aEJ pathway might be prevented or retarded by the Ku protein bound to the ends (7). It is possible that BLM protein helps the aEJ reaction proceed efficiently and faithfully. If so, BLM deficiency would reduce the rate of the aEJ reaction, resulting in a further increase in the percentage of inaccurate repair and the infidelity of end-joining. However, an intriguing possibility remains that an interaction exists between BLM and Ku protein, and loss of this interaction by BLM deficiency would cause an abnormal aEJ reaction, as observed in LIG4 Ϫ/Ϫ BLM Ϫ/Ϫ double-mutant cells. In Drosophila, it has been reported that sterility, nondisjunction, and chromosome loss due to BLM mutations is partially rescued by overproduced Ku70 (31), and it has been suggested that BLM is involved in DSB recognition and/or NHEJ (31,32). It will thus be interesting to study the impact of BLM deficiency using Ku70/80-or DNA-dependent protein kinase catalytic subunitnull human cells.
In conclusion, we have provided the first genetic evidence that NHEJ is unfavorable for survival of human cells lacking BLM. Given that replication-associated, one-ended DSBs accumulate in BLM-deficient cells, it is probable that NHEJ-mediated repair of such DSBs leads to genomic instability seen in BS cells. It should be emphasized, however, that the NHEJ pathway per se is marginally affected by BLM deficiency, implying that the genomic instability of BS cells cannot be attributable to aberrant repair of two-ended DSBs via the NHEJ pathway. More intriguingly, our data suggest that in the absence of LIG4, BLM plays an important role in repairing twoended DSBs, possibly by controlling an alternative, LIG4-independent end-joining pathway in human cells. Further studies will be needed to clarify the relevance of this pathway in the repair of one-ended DSBs and its contribution to the phenotypes associated with BS.