Smarcal1 promotes double-strand-break repair by nonhomologous end-joining

Smarcal1 is a SWI/SNF-family protein with an ATPase domain involved in DNA-annealing activities and a binding site for the RPA single-strand-DNA-binding protein. Although the role played by Smarcal1 in the maintenance of replication forks has been established, it remains unknown whether Smarcal1 contributes to genomic DNA maintenance outside of the S phase. We disrupted the SMARCAL1 gene in both the chicken DT40 and the human TK6 B cell lines. The resulting SMARCAL1−/− clones exhibited sensitivity to chemotherapeutic topoisomerase 2 inhibitors, just as nonhomologous end-joining (NHEJ) null-deficient cells do. SMARCAL1−/− cells also exhibited an increase in radiosensitivity in the G1 phase. Moreover, the loss of Smarcal1 in NHEJ null-deficient cells does not further increase their radiosensitivity. These results demonstrate that Smarcal1 is required for efficient NHEJ-mediated DSB repair. Both inactivation of the ATPase domain and deletion of the RPA-binding site cause the same phenotype as does null-mutation of Smarcal1, suggesting that Smarcal1 enhances NHEJ, presumably by interacting with RPA at unwound single-strand sequences and then facilitating annealing at DSB ends. SMARCAL1−/−cells showed a poor accumulation of Ku70/DNA-PKcs and XRCC4 at DNA-damage sites. We propose that Smarcal1 maintains the duplex status of DSBs to ensure proper recruitment of NHEJ factors to DSB sites.


INTRODUCTION
Smarcal1 (a SWI/SNF-related, matrix associated, actindependent regulator of chromatin a-like 1) is a SWI/SNF family protein that carries an ATPase domain and the binding site for replication protein A (RPA), the singlestrand-DNA-binding protein (1,2). Smarcal1 is an ATPdriven annealing helicase that catalyzes the formation of double-strand DNA from complementary single-strand DNA strands associated with RPA. Thus, Smarcal1 could counteract unwinding by DNA helicases at DSB sites. The annealing activity is stimulated by substrate DNA containing both single-strand and double-strand regions, such as a chicken foot structure (3). Mutations in the SMARCAL1 gene cause a rare autosomal recessive disease, Schimke immuno-osseous dysplasia (SIOD), which is characterized by short stature, kidney disease and a severely compromised immune system (4)(5)(6)(7). Phenotypic analysis of Smarcal1depleted cells suggests that Smarcal1 stabilizes replication forks when cells are exposed to aphidicolin, hydroxyurea and camptothecin (a topoisomerase 1 poison) (1,2,(8)(9)(10).
The two major double-strand-break (DSB) repair pathways, homologous recombination (HR) and nonhomologous end-joining (NHEJ) (11)(12)(13) significantly contribute to cellular tolerance to anti-malignant therapies. First, both pathways contribute to cellular tolerance to radiotherapy, HR in the S to G 2 phases and NHEJ throughout the cell cycle. Second, HR plays the dominant role in repairing DSBs generated during DNA replication by chemotherapeutic agents such as camptothecin and poly[ADP ribose]polymerase inhibitor (olaparib). These chemotherapeutic agents cause the accumulation of single-strand breaks, which are converted by DNA replication to DSBs called one-end breaks. These DSBs are repaired by HR but not by NHEJ (14)(15)(16). Third, NHEJ plays the dominant role in repairing DSBs caused by chemotherapeutic topoisomerase 2 inhibitors such as ICRF193 and etoposide (15,17). Measuring the sensitivity of gene-disrupted cells to various anti-malignant therapies allows us to define the role of the gene in HR, NHEJ or both. In addition to the above, the capability of canonical NHEJ is evaluated by examining the V(D)J recombination of Immunoglobulin (Ig) V genes, which requires a collaboration between NHEJ and V(D)J recombinase encoded by the recombinationactivating-genes 1 and 2 (Rag1/Rag2) (18)(19)(20).
We disrupted the SMARCAL1 gene in the chicken DT40 and human B lymphoblastoid TK6 cell lines (38,39). The resulting SMARCAL1 −/− clones exhibited sensitivity to camptothecin, suggesting that Smarcal1 plays a role in DNA replication, as indicated previously (9,10). Remarkably, Smarcal1 is also required for efficient NHEJ in human as well as in chicken cells. This conclusion is in agreement with the fact that SIOD patients exhibit reduced V(D)J recombination products in peripheral lymphocytes as well as increased chromosomal breakage (40,41). We propose that the decreased efficiency of NHEJ in V(D)J recombination as well as the compromised maintenance of replication fork progression result in severe lymphocytopenia in SIOD patients (4,40,41).

Cell clones
All the clones used in this study are summarized in Table 1.

Cell culture
DT40 and TK6 cells were cultured in the same manner as described previously (39,42).

Generation of SMARCAL1 −/− DT40 cells
SMARCAL1 gene disruption constructs were generated from genomic polymerase chain reaction (PCR) products combined with histidinol dehydrogenase (hisD) and puromycin-resistance (puro R ) marker genes. Genomic DNA from wild-type cells was amplified using the F1 and R1 primers for the 5 -arm and the F2 and R2 primers for the 3 -arm. The amplified 3 -arm PCR product was subcloned into pCR2.1-TOPO vector (Invitrogen, US). The 5 -arm PCR product harboring the SacI and BamHI sites at the 5 -and 3 -ends, respectively, was cloned into the SacI and BamHI sites of the pCR2.1-TOPO vector carrying the 3 -arm. The BamHI fragment containing either the hisD or puro R gene was cloned into the BamHI site between the 3 -arm and the 5 -arm in the pCR2.1-TOPO vector. To generate SMARCAL1 −/− cells, the SMARCAL1 gene-disruption constructs carrying hisD and puro R were linearized, using the NotI restriction enzyme, and sequentially transfected by electroporation (Bio-Rad, US). A 0.5 kb probe was generated by PCR of genomic DNA using primers F3 and R3 for Southern blot analysis. The genomic DNA of the candidate clones was digested with SmaI and XhoI for Southern blot analysis. The gene disruption was confirmed by RT-PCR, using primers F4 and R4. When generating SMARCAL1 +/− clones from wild-type cells the targeting efficiency was 20% (4/20), while that of generating SMARCAL1 −/− cells from SMARCAL1 +/− cells was 1.4% (1/71). All primers used here are shown in Supplemental  Table S1.
The intact allele of the SMARCAL1 +/− cell was targeted for deleting the first 30 amino acids, which domain is responsible for RPA binding (10). To generate the Δ30 construct, the 5 -arm was amplified from genomic DNA using the NotI-tagged F5 and the BamHI-tagged R5, and the 3 -arm was amplified using the BamHI-tagged F6, which was designed from 90 bases downstream of start codon ATG in exon1, and the SalI-tagged R6. In the Zero Blunt TOPO vector (Invitrogen, US), the left and right arms were cloned at the site of NotI-BamHI and BamHI-SalI, respectively. The selection-marker gene, puromycin resistance (puro R ) flanked by loxP sequences, was then inserted into the BamHI site between the left and right arms. The resulting Δ30-puro R construct was transfected into the SMARCAL1 +/− and KU70 −/− /SMARCAL1 +/− cells. The puro R gene was popped out by transient expression of cre-recombinase, resulting in the generation of SMARCAL1 Δ30/− and KU70 −/− /SMARCAL1 Δ30/− DT40 cells. To confirm the gene disruption by RT-PCR, the primer set F4 and R4 was used.
To generate a pair of TALEN expression plasmids against the SMARCAL1 gene, we used a Golden Gate TALEN kit and a TAL effector kit (Addgene, US) (43,44). The TALEN target sites are shown in Supplementary Figure  S2. The gene-targeting constructs were generated from the genomic DNA of TK6 cells by amplifying with primers XhoI-flanked F7 and NheI-flanked R7 for the 5 -arm and NotI-flanked F8 and HindIII-flanked R8 for the 3 -arm. The 5 -arm PCR products were cloned into the XhoI and NheI sites found upstream of the puro R and hygro R marker genes of the DT-ApA/puro and DT-ApA/hygro vectors, respectively. The 3 -arm PCR products were cloned into the NotI and HindIII sites found downstream of the puro R and hygro R marker genes of the DT-ApA/puro and DT-ApA/hygro vectors, respectively. 6 g TALEN-expression plasmids and 2 g gene-targeting vectors were transfected into 4×10 6 TK6 cells using the Neon Transfection System (Life Technologies, US) with 3X pulse at 1350 V and  To disrupt the DNA-PKcs gene, we designed a guide RNA targeting the 32 nd exon using the Zhang CRISPR tool (45) and gene-targeting constructs. The CRISPR-target site is depicted in Supplementary Figure S3E. The gene-targeting constructs were generated using SLiCE (Seamless Ligation Cloning Extract). The genomic DNA was amplified with primers F19 and R19 from the DNA-PKcs-gene locus and the PCR product was used as template DNA for amplifying the 5 -and 3 -arms. The 5 -arm was amplified using primers F20 and R20 and the 3 -arm was amplified using primers F21 and R21, where each primer shared 20-base pair-end homology with the insertion site of the vector. Both vectors, DT-ApA/neo and DT-ApA/his, were linearized with AflII and ApaI. All the fragments of the vectors and inserts were purified using a qiaquick gel extraction kit (QIAGEN, Netherlands). The gene-targeting constructs were generated in a single reaction mixture containing DT-ApA/neo or DT-ApA/his vectors, 5 -and 3 -arms, and 2×SLiCE buffer (Invitrogen, US) and incubated for 30 min at room temperature. 6 g of CRISPR and 2 g of each gene-targeting vector were transfected into 4×10 6 TK6 cells using the Neon Transfection System (Life Technologies, US). After electroporation, cells were released into 20 ml drug-free medium containing 10% horse serum. Forty-eight hours later, cells were seeded into 96-well plates for selection with both neomycin and histidinol antibiotics for two weeks. The gene disruption was confirmed by RT-PCR using primers

Generation of SMARCAL1 −/− cells reconstituted with
The SMARCAL1 cDNA was bought from the Kazusa DNA research institute (Chiba, Japan). The SMARCAL1 R764Q cDNA was obtained by site-directed mutagenesis of SMARCAL1 WT (wild-type SMARCAL1) cDNA using primers F18 and R18. The SMARCAL1 Δ30 cDNA with the first 30 amino acids deleted was generated from SMARCAL1 WT cDNA by PCR using primers F10 and R10. The SMARCAL1 WT , SMARCAL1 R764Q and SMARCAL1 Δ30 transgenes were cloned into pMSCV-IRES-GFP retroviral expression vector (Clontech, US). The newly engineered retroviral expression vector was co-transfected into human 293T cells with a helper plasmid (pClampho, US) to produce a viral supernatant, which was collected after 24 hours and used to infect the SMARCAL1 −/− cells. The efficiency of infection was assessed by quantifying the number of cells expressing GFP using flow-cytometric analysis (LSRFortessa, BD Biosciences, US). The cells expressing GFP were enriched using a cell sorter (FACSAria III, BD Biosciences, US) and seeded into 96 well plates to isolate single colonies. The expression level of the transgenes in the SMARCAL1 −/− cells was measured by western blot ( Figure 5D).

Colony-survival assay
To measure sensitivity, cells were treated with camptothecin (Topogen, Inc, US) and ICRF193 (Funakoshi, Japan) (17) and irradiated with ionizing radiation ( 137 Cs). Cell sensitivity to these DNA-damaging agents was evaluated by counting colony formation in methylcellulose plates as described previously (46,47). . The x-axis represents the dose of the indicated DNA-damaging agent on a linear scale; the y-axis represents the survival fraction on a logarithmic scale. Error bars show the SD of the mean for three independent assays. The P-value of (C) was calculated by Student's t-test or two-sample t-test for IR sensitivity at 2 Gy.

Cell-cycle synchronization
Cells were synchronized by centrifugal-counter-flow elutriation (Hitachi Koki, Japan). The cell suspension (∼5×10 7 TK6 cells) was loaded at a flow rate of 15 ml/min into an elutriation chamber rotating at 2000 rpm. Cell synchrony was confirmed by FACS analysis (LSRFortessa, BD Biosciences, US).

Immunostaining and microscopic analysis
Cells were fixed with 4% paraformaldehyde (Nacalai Tesque, Japan) for 10 min at room temperature and permeabilized with 0.5% TritonX-100 (Sigma, St. Louis, US) for 20 min. To exclude S-phase cells from the count shown in Figure 3C and D, a Click-iT EdU imaging kit (Alexa594, Invitrogen, US) was used. Images were taken with a confocal microscope (TCS SP8, Leica Microsystems, Germany) and a BX-61 microscope (Olympus, Japan).

V(D)J-recombination assay
The V(D)J-recombination assay was performed as described previously (48). Briefly, 4×10 6 TK6 cells were trans-fected (Neon, Life Technologies, US) with 600 ng of circular pJH200 or pJH290, 5.4 g of RAG1-and 6.6 g of RAG2-expression vector. The extrachromosomal plasmids were recovered from cells after 48 hours using a modified Hirt extraction method (49). The mixture of 15 l of ElectroMAX TM DH10B TM (Life Technologies, US) competent bacteria and 300 ng (range of 100-500 ng) of recovered plasmids were added to an electroporation cuvette (0.1 cm gap) and incubated on ice for 10 min. The bacteria were then electroporated at 1.8 kV, 200 and 25 uF for 2 seconds using a Gene PulserII (Bio-Rad, US). Pre-warmed SOC media was added to the bacteria and the reaction was incubated at 37 • C for 2 hours. The reaction was plated on LB-agar plates containing 100 g/ml ampicillin and 10 g/ml chloramphenicol and incubated for 16-24 hours at 37 • C. The ampicillin+chloramphenicol-resistant plasmids were isolated and subjected to ApaLI digestion to examine the fidelity of the signal joints. Digestion of original pJH200 plasmid yields a 4.3 kb band, while that of the correct recombination products yields 3.5 kb + 0.8 kb bands due to the newly generated ApaLI site. Signal-joint and codingjoint sequences were analyzed using the sequencing primer R17 (50).

NHEJ assay of I-Sce1-induced DSBs
A TK6-derived line that is heterozygous for point mutation in exon4 of the thymidine-kinase gene (TK +/− ) was used to measure the frequency of NHEJ events as described previously (51,52). To measure the length of deletion in DSBrepair products formed in wild-type and SMARCAL1 −/− cells, primers Fa and Ra were used. To measure the length of the deletion formed in the LIG4 −/−/− cells, primers Fa and Rb were used.

Chromatin fractionation and chromatin immunoprecipitation
A Subcellular Protein Fractionation Kit from Thermo Scientific (78840) was used for chromatin fractionation. Expression plasmids for a TALEN and I-Sce1 were transfected into TK6 cells using the Neon Transfection System. After 20 hours, transfected cells were analyzed by western blotting and ChIP. ChIP was performed as described previously (53), with some modifications. Briefly, samples were sonicated to generate DNA fragments of <500 bp. The antibody was incubated with Dynabeads Protein G for 3 hours at 4 • C. Sheared chromatin was centrifuged at 15000 rpm for 15 min at 4 • C. After centrifugation, supernatants were incubated with antibody-protein G conjugates for 3 hours at 4 • C. The conjugated beads were washed thoroughly with IP buffer-140, IP buffer-500, IP buffer-750, LiCl/detergent and TE. Real-time PCR was carried out as described previously (53). Sequences of primers are given in Supplemental  Table S1.

SMARCAL1-deficient cells are sensitive to DNA-damaging agents
To analyze the role of Smarcal1 in the DSB-repair pathway, we disrupted the SMARCAL1 gene and generated SMARCAL1 −/− DT40 cells ( Supplementary Figures S1A and S1C). Moreover, to selectively analyze the function of the Smarcal1-RPA interaction, we generated SMARCAL1 Δ30/− DT40 mutant cells by deleting the Nterminal region encoding the RPA-binding site of the endogenous SMARCAL1 gene (9) (Supplementary Figures  S1D and S1E). To define the role played by Smarcal1 in various DNA-repair processes, we measured cellular responses to exogenous DNA damages. SMARCAL1 −/− and SMARCAL1 Δ30/− DT40 cells exhibited increased sensitivities to various DNA-damaging agents, including camptothecin and ICRF193 (Figures 1A and B). The sensitivity profile of the SMARCAL1 Δ30/− cells was very similar to that of the SMARCAL1 −/− cells, suggesting that the physical association of Smarcal1 with RPA is essential for its DNA-damage response (9). The elevated sen- sitivity to camptothecin supports the idea that Smarcal1 helps to prevent replication forks from replication collapse at one-end breaks, as indicated previously (1,2,(8)(9)(10)54). NHEJ-deficient KU70 −/− cells (55), but not HRdeficient BRCA2 −/− cells (46), showed a hypersensitivity to ICRF193 ( Figure 1B). We further analyzed the functional relationship between Smarcal1 and Ku70 by generat- The double-mutant cells showed virtually the same ICRF193 sensitivity as did the KU70 −/− single mutant ( Figure 1D), indicating that Smarcal1 is epistatic to Ku70.
To investigate the role of Smarcal1 in human cells, we disrupted the SMARCAL1 gene in the human TK6 B cell line using a TALEN pair combined with genedisruption constructs (Supplementary Figures S2A and  S2B). We also generated both LIG4 −/−/− ( Supplementary  Figures S3A and S3B) and RAD54 −/− TK6 clones (Supplementary Figures S3C and S3D) as controls deficient in NHEJ and HR, respectively. In addition, we generated DNA-PKcs −/− TK6 clones (Supplementary Figures  S3E and S3F). The TK6 cell line has been widely used by the governments of developed countries to detect environmental mutagens due to the very stable phenotype and karyotype of its cells (56,57). The SMARCAL1 −/− cells proliferated with kinetics (15 h per cell cycle) and plating efficiency (58%) similar to that of wild-type TK6 cells. Like the DT40 mutants, human SMARCAL1 −/− cells were moderately but also significantly sensitive to camptothecin ( Figure S4B). We thus conclude that Smarcal1 promotes the canonical NHEJ pathway in both DT40 and TK6 cell lines.  Figure 2E and Supplementary Figure S4A). To further investigate the role of Smarcal1 in NHEJ, we measured sensitivity to ionizing radiation (IR) in the G 1 phase ( Figure 3A and Supplementary Figure  S5A). SMARCAL1 −/− as well as LIG4 −/−/− cells showed a hypersensitivity to IR ( Figure 3B). As expected, the SMARCAL1 −/− /LIG4 −/−/− cells showed the same hypersensitivity to IR as did the LIG4 −/−/− cells in the G 1 phase ( Figure 3B). We then monitored DSB repair kinetics in the G 1 phase by counting the number of ␥ H2AX and 53BP1 foci over time after exposure to IR. The resolution of 53BP1 and ␥ H2AX foci was significantly delayed in SMARCAL1 −/− cells compared to wild-type cells (Figures 3C, D and Supplementary Figure S5B). These results indicate that SMARCAL1 −/− cells are deficient in DSB repair in the G 1 phase. We therefore conclude that Smarcal1 promotes DSB repair by NHEJ in both human and chicken DT40 cells.

Deletion of Smarcal1 does not compromise the fidelity of NHEJ
To address the accuracy of individual DSB-repair events, we performed two experiments using TK6 cells, (i) the analysis of V(D)J recombination ( Figure 4) and (ii) the repair of I-Sce1 induced DSBs by NHEJ ( Figure 5). V(D)J recombination is initiated by the Rag1 and Rag2 recombinase proteins, which introduce DSBs at the recombination signal (RS) (58,59) and complete recombination by collaborating with canonical NHEJ. We used two episomal V(D)J-recombination substrates, pJH200 and pJH290, where the recombinase generates the RS and coding-joint products, respectively (50) ( Figure 4A). We transiently transfected expression plasmids encoding RAG1 and RAG2 along with either pJH200 or pJH290 into wildtype and SMARCAL1 −/− and LIG4 −/−/− cells ( Figure  4B). We then recovered the transfected substrate plasmids from the TK6 cells, introduced the plasmids into bacterial cells, and plated them on LB agar plates containing either ampicillin or chloramphenicol. Ampicillin-resistant colonies contained recovered pJH200 or pJH290 plasmids, while colonies resistant to chloramphenicol contained only rearranged pJH200 or pJH290 plasmids. Thus, the frequency of rearrangement can be calculated as the gain of chloramphenicol-resistant (cam R ) colonies relative to the total number of ampicillin-resistant (amp R ) colonies. The signal-joint ends are precisely ligated, whereas the coding ends are joined in a process that can involve nucleotide loss or gain in addition to simple ligation (60). We failed to recover any rearranged signal-joint products from LIG4 −/−/− cells, which agrees with the essential role for DNA Ligase4 in signal-joint formation (61,62). The efficiency of recombination in the signal-joint and coding-joint plasmids was decreased 2.3 and 2.4 times, respectively, in SMARCAL1 −/− cells, when compared with wild-type cells ( Figure 4C). The nucleotide sequence analysis of signaljoint products indicated that only a single product among the 29-analyzed sequences contained one nucleotide deletion in SMARCAL1 −/− cells (data not shown). Analysis of coding joints, on the other hand, showed more frequent deletion events in SMARCAL1 −/− as well as in wild-type cells, with the extent and frequency of deletion being comparable between the two genotypes (Supplemental Table   S2). Thus, the loss of Smarcal1 reduces the efficiency of canonical NHEJ without compromising its fidelity.
We next examined individual DSB-repair events in the human TK6 cell line, which is heterozygous (+/-) for the thymidine kinase (TK) gene and contains an I-Sce1 endonuclease recognition site in the fourth intron of the intact TK allele (51) (Figure 5A). If the repair of I-Sce1-induced DSBs causes the deletion of more than 100 nucleotides downstream from the I-Sce1 site, the deletion would inactivate the fifth exon, leading to the formation of TK −/− cells, which cells are able to form colonies in the negativeselection media containing trifluorothymidine (TFT) (Figure 5A). We found that the number of TFT-resistant clones was reduced by 8 and 5 times in the LIG4 −/−/− and SMARCAL1 −/− cells, respectively, when compared with wild-type cells ( Figure 5B). To analyze deletion range and pattern, we isolated at least 50 individual TFT-resistant clones from each genotype. Genomic PCR amplification over the I-Sce1 site indicated that deletion was two times shorter in the SMARCAL1 −/− cells and approximately 0.5 times longer in LIG4 −/−/− cells, compared with wild-type cells ( Figure 5C and Supplementary Figure S6). In summary, we conclude that Smrcal1 promotes NHEJ without affecting its accuracy.

The RPA-binding domain and the ATPase domains are both required for the promotion of NHEJ by Smarcal1
The phenotypic analysis of SMARCAL1 Δ30/− DT40 cells indicated that the RPA-binding site is essential for Smar-cal1 to function in NHEJ. To confirm the relevance of this finding to human cells, we reconstituted the SMARCAL1 −/− TK6 cells with the RPA-binding-sitedeficient (SMARCAL1 Δ30 ) transgene as well as the wildtype (SMARCAL1 WT ) transgene. The expression level of the Smarcal1 protein in the reconstituted cells was similar to that in the wild-type cells ( Figure 5D, lower panel). As expected, reconstitution of SMARCAL1 −/− cells with the SMARCAL1 WT transgene normalized the NHEJmediated repair of I-Sce1-induced DSBs ( Figure 5E). The SMARCAL1 Δ30 transgene failed to normalize NHEJ, indicating that the RPA-binding site plays a role in promoting NHEJ by Smarcal1 in human cells. Moreover, SMARCAL1 −/− cells and SMARCAL1 Δ30 transgene exhibited similar sensitivity to DNA-damaging agent (Supplementary Figure S4C), demonstrating that the RPA-binding site is critical for Smarcal1 DNA repair function. Next, to investigate the role of annealing helicase activity, we introduced a point mutation (Arginine 764 to Glutamine, R764Q) into the ATPase domain of the SMARCAL1 WT transgene, which mutation has no detectable annealing helicase activity (2) and causes a severe form of Schimke immuno-osseous dysplasia (SIOD) (4). Reconstitution with the resulting SMARCAL1 R764Q transgene did not restore the repair of I-Sce1-induced DSBs ( Figure 5D and E). We therefore conclude that the annealing helicase activity as well as the physical association of Smarcal1 with RPA is required for the promotion of NHEJ by Smarcal1. One possible scenario is that Smarcal1 promotes NHEJ by interacting with unwound DSB ends associated with RPA and facilitate their annealing.  Table S2 shows the nucleotide sequences of coding joints associated with deletion events.

I-Sce1
Exon 4 Thymidine kinase (TK) gene alleles 5 6 7 I-Sce1-induced DSB 100 bases Exon 4 5 6 7 Deletion of exon 5 Annealing of double-strand DNA by purified Smarcal1 (2) suggests that Smarcal1 facilitates NHEJ by stabilizing the physical interaction between Ku70/Ku80/DNA-PKcs proteins and DSB ends, thereby activating DNA-PKcs. To test this hypothesis, we analyzed the phosphorylation status of threonine 2609 (63,64) after treatment with the topoisomerase 2 inhibitors ICRF193 (Supplemen-tary Figures S7A and S7B) and etoposide (Supplementary Figure S7C). Strikingly, the phosphorylated threonine 2609 was detectable only in the wild-type and not in the SMARCAL1 −/− cells ( Supplementary Figures S7B  and S7C). The compromised phosphorylation of DNA-PKcs may account for the decreased efficiency of NHEJ, as the substitution of the threonine 2609 site to alanine causes radio-sensitivity and reduced V(D)J recombination efficiency, which sensitivity is less prominent than that of DNA-PKcs null-mutant cells (64). We propose that Smar-cal1 promotes NHEJ, presumably at an initial step, by facilitating the physical association of Ku70/Ku80/DNA-PKcs proteins and DSB ends.

Smarcal1 is required for the recruitment of XRCC4 to DNAdamage sites
To investigate early and late steps of NHEJ, we monitored the dynamics of Ku70 and XRCC4 proteins, respectively, in the chromatin-bound fraction, following exposure of cells to ICRF193 for one hour ( Figure 6). This exposure did not affect the purification of the chromatin-bound or the nuclear-soluble fractions ( Figure 6A). Exposure to ICRF193 caused a marked increase in the amounts of both Ku70 and XRCC4 in the chromatin-bound fraction of wildtype cells. In marked contrast, accumulations of Ku70 and XRCC4 in the SMARCAL1 −/− as well as DNA-PKcs −/− cells were, significantly smaller compared with wild-type cells ( Figure 6B and C).
To validate the significant reduction in the accumulation of XRCC4 near DSB sites, we conducted chromatin immunoprecipitation (ChIP) following transient transfection of empty vector and expression vectors encoding the I-Sce1 restriction enzyme and a TALEN towards the p53 locus ( Figure 6D). XRCC4 is supposed to accumulate at the TK-I or p53-I sites, which are adjacent to the DSB sites, but not at the TK-II or p53-II sites, which are distant from the DSB sites ( Figure 6D). Transient expression of I-Sce1 or TALEN caused the accumulation of XRCC4 near the DSB sites in wild-type cells, whereas the extent of the accumulation was significantly reduced in SMARCAL1 −/− cells ( Figure 6E and Supplementary Figure S7D). We checked the overall expression levels of XRCC4 (Supplementary Figure S7B) to confirm that the deletion of Smarcal1 did not alter the level of protein expression. Thus, Smarcal1 is required for recruitment of XRCC4 to DSB sites.
Like reduced accumulation of Ku70 in the chromatin fraction ( Figures 6B and C), the amounts of Ku70 and DNA-PKcs near the I-Sce1 site were approximately 30% smaller in SMARCAL1 −/− cells in comparison with wildtype cells (Figures 6F and G). In summary, Smarcal1 may promote annealing of DSB ends, which stabilizes complex formation of Ku/DNA-PKcs at DSB sites and fully activates DNA-PKcs. The activation may enhance the recruitment of XRCC4 for the completion of DSB repair by canonical NHEJ.

DISCUSSION
We herein show that Smarcal1 significantly contributes to canonical NHEJ. Although the role of Smarcal1 during DNA replication has been well established (1,3,(8)(9)(10)65), it has remained unclear whether Smarcal1 plays a role outside the S phase. We here reveal that Smarcal1 contributes to DSB repair by NHEJ during the G 1 phase. The role played by Smarcal1 in NHEJ is demonstrated by six points, as follows. First, the loss of Smarcal1 increases cellular sensitivity to ICRF193, which induces DSBs repair only by NHEJ and not by HR (17) (Figures 1B and 2B). Second, nullmutation of KU70 is epistatic to SMARCAL1 Δ30 mutation in DT40 cells ( Figure 1D) and null-mutation of LIG4 is epistatic to SMARCAL1 null-mutation in TK6 cells (Figure 2D), in terms of cellular tolerance to ICRF193. Third, the loss of Smarcal1 significantly reduces the efficiency of DSB repair in the G 1 phase in TK6 cells (Figure 3). Fourth, SMARCAL1 null-mutation in TK6 cells significantly compromises V(D)J recombination (Figure 4) as well as I-Sce1induced DSB repair by NHEJ ( Figure 5). Fifth, SMAR-CAL1 null-mutation impairs the phosphorylation of DNA-PKcs at threonine 2609 ( Supplementary Figures S7B and  S7C). Lastly, SMARCAL1 null-mutation diminishes the accumulation of Ku70, DNA-PKcs and XRCC4 at DNAdamage-induced DSB sites ( Figure 6). We therefore conclude that Smarcal1 plays a role in NHEJ.
The molecular mechanisms underlying the severe lymphocytopenia of SIOD patients remain unclear (4,40,41). The lymphocytopenia might result primarily from a severe defect in V(D)J recombination due to the reduced efficiency of NHEJ, according to the following studies. The analysis of V(D)J recombination in peripheral T lymphocytes indicates that the size of the T-cell-antigenic receptor (TCR) repertoire is extremely small in SIOD patients (40). Moreover, the peripheral T lymphocytes of SIOD patients have severely low copy number of episomal circular DNA generated as a consequence of V(D)J recombination and contain signal joints of the T cell receptor genes (40,41). These observations support the following scenario. A moderate defect in NHEJ can cause a very strong decrease in the efficiency of T-cell development in patients, since it requires productive D-J-and V-D-recombination events in both TCR␣ and TCR␤ chain genes in individual thymocytes. The reduced T-cell production by the thymus may be compensated by enhanced proliferation of newly generated peripheral T lymphocytes (66) in SIOD patients, leading to the quick dilution of episomal circular DNA in individual peripheral T lymphocytes. This enhanced proliferation could cause a strong replication stress in SIOD patients, due to their attenuated stabilization of replication forks. In summary, we propose that the severe lymphocytopenia of SIOD patients (4,40,41) is attributable to the significantly reduced efficiency of V(D)J recombination, together with attenuated stabilization of replication forks.
A prominent question is, how does Smarcal1 facilitate the promotion of NHEJ? We have shown that both the loss of the RPA-binding site and the inactivation of AT-Pase activity in Smarcal1 completely abolished the promotion of NHEJ by Smarcal1 ( Figure 5E). Thus, Smarcal1 plays a role in NHEJ by physically interacting with RPA on unwound DSB ends and then facilitating their annealing. The existence of unwound single-stranded sequences at the DSB sites is supported by the presence of RPA foci in the ␥ -ray irradiated G 1 phase cells (67). The annealing by Smarcal1 may stabilize the interaction of DSB sites with DNA-PKcs/Ku70/Ku80, since Ku70/Ku80 associates with duplex DSB ends more stably than with DSBs carrying single-strand tails (21)(22)(23)(24). The stabilization of DNA-PKcs/Ku70/Ku80 at DSB sites by Smarcal1 is verified by data shown in Figure 6 PKcs. DNA-PKcs/Ku70/Ku80 also interacts with Smar-cal1 in vivo (68,69). We here show that DNA-PKcs −/− and SMARCAL1 −/− have an epistatic relationship in cellular tolerance to IR ( Figure 2E). We also show that the loss of Smarcal1 inhibits the phosphorylation of DNA-PKcs at threonine 2609 following exposure of cells to the topoisomerase 2 inhibitors (Supplementary Figure S7). These observations suggest that Smarcal1 may be required for DNA-PKcs/Ku70/Ku80 to function appropriately. We also show that the loss of Smarcal1 reduces the recruitment of XRCC4 to DSB sites by several times (Figure 6). Previous studies indicate that DNA-PKcs is necessary for the stabilization of recruited XRCC4 (70,71), which is consistent with our data ( Figure 6B). Thus, the effect of Smar-cal1 on the recruitment of XRCC4 might be mediated by DNA-PKcs/Ku70/Ku80. We therefore propose that Smar-cal1 maintains duplex DNA status at DSB ends by interacting with unwound single-strand DNA associated with RPA and facilitating their annealing. This annealing then stabilizes DNA-PKcs/Ku70/Ku80 at duplex DNA termini, which is essential for the proper accumulation and stabilization of XRCC4 at DNA damage sites. Future studies should clarify the molecular mechanism.