Visualization of DNA double-strand break repair in live bacteria reveals dynamic recruitment of Bacillus subtilis RecF, RecO and RecN proteins to distinct sites on the nucleoids

We have found that SMC-like RecN protein, RecF and RecO proteins that are involved in DNA recombination play an important role in DNA double-strand break (DSB) repair in Bacillus subtilis. Upon induction of DNA DSBs, RecN, RecO and RecF localized as a discrete focus on the nucleoids in a majority of cells, whereas two or three foci were rarely observed. RecN, RecO and RecF co-localized to the induced foci, with RecN localizing first, while RecO localized later, followed by RecF. Thus, three repair proteins were differentially recruited to distinct sites on the nucleoids, potentially constituting active DSB repair centres (RCs). RecF did not form regular foci in the absence of RecN and failed to form any foci in recO cells, demonstrating a central role for RecN and RecO in initializing the formation of RCs. RecN/O/F foci were detected in recA, recG or recU mutant cells, indicating that the proteins act upstream of proteins involved in synapsis or post-synapsis. In the absence of exogenous DNA damage, RCs were rare, but they accumulated in recA and recU cells, suggesting that DSBs occur frequently in the absence of RecA or RecU. The results suggest a model in which RecN that forms multimers in solution and high-molecular-weight complexes in cells containing DSBs initiates the formation of RCs that mediate DSB repair with the homologous sister chromosome, which presents a novel concept for DSB repair in prokaryotes.     

DNA double strand break end-processing and RecA induce RecN expression levels in Bacillus subtilis

Bacillus subtilis cells respond to double strand breaks (DSBs) with an ordered recruitment of repair proteins to the site lesion, being RecN one of the first responders. In B. subtilis, one of the responses to DSBs is to increase RecN expression rather than modifying its turnover rate. End-processing activities and the RecA protein itself contribute to increase RecN levels after DNA DSBs. RecO is required for RecA filament formation and full SOS induction, but its absence did not significantly affect RecN expression. Neither the absence of LexA nor the phosphorylation state of RecA or SsbA significantly affect RecN expression levels. These findings identify two major mechanisms (SOS and DSB response) used to respond to DSBs, with LexA required for one of them (SOS response). The DSB response, which requires end-processing and RecA or short RecO-independent RecA filaments, highlights the importance of guarding genome stability by modulating the DNA damage responses.     

RecA Protein Recruits Structural Maintenance of Chromosomes (SMC)-like RecN Protein to DNA Double-strand Breaks

Escherichia coli RecN is an SMC (structural maintenance of chromosomes) family protein that is required for DNA double-strand break (DSB) repair. Previous studies show that GFP-RecN forms nucleoid-associated foci in response to DNA damage, but the mechanism by which RecN is recruited to the nucleoid is unknown. Here, we show that the assembly of GFP-RecN foci on the nucleoid in response to DNA damage involves a functional interaction between RecN and RecA. A novel RecA allele identified in this work, recA^Q300R, is proficient in SOS induction and repair of UV-induced DNA damage, but is deficient in repair of mitomycin C (MMC)-induced DNA damage. Cells carrying recA^Q300R fail to recruit RecN to DSBs and accumulate fragmented chromosomes after exposure to MMC. The ATPase-deficient RecN^K35A binds and forms foci at MMC-induced DSBs, but is not released from the MMC-induced DNA lesions, resulting in a defect in homologous recombination-dependent DSB repair. These data suggest that RecN plays a crucial role in homologous recombination-dependent DSB repair and that it is required upstream of RecA-mediated strand exchange.     

Degradation of Escherichia coli RecN aggregates by ClpXP protease and its implications for DNA damage tolerance

Protein degradation in bacteria plays a dynamic and critical role in the cellular response to environmental stimuli such as heat shock and DNA damage and in removing damaged proteins or protein aggregates. Escherichia coli recN is a member of the structural maintenance of chromosomes family and is required for DNA double strand break (DSB) repair. This study shows that RecN protein has a short half-life and its degradation is dependent on the cytoplasmic protease ClpXP and a degradation signal at the C terminus of RecN. In cells with DNA DSBs, green fluorescent protein-RecN localized in discrete foci on nucleoids and formed visible aggregates in the cytoplasm, both of which disappeared rapidly in wild-type cells when DSBs were repaired. In contrast, in DeltaclpX cells, RecN aggregates persisted in the cytoplasm after release from DNA damage. Furthermore, analysis of cells experiencing chronic DNA damage revealed that proteolytic removal of RecN aggregates by ClpXP was important for cell viability. These data demonstrate that ClpXP is a critical factor in the cellular clearance of cytoplasmic RecN aggregates from the cell and therefore plays an important role in DNA damage tolerance.     

A novel SMC-like protein,SbcE (YhaN), is involved in DNA double-strand break repair and competence in Bacillus subtilis

Bacillus subtilis and most Gram positive bacteria possess four SMC like proteins: SMC, SbcC, RecN and the product of the yhaN gene, termed SbcE. SbcE is most similar to SbcC but contains a unique central domain. We show that SbcE plays a role during transformation in competent cells and in DNA double-strand break (DSB) repair. The phenotypes were strongly exacerbated by the additional deletion of recN or of sbcC, suggesting that all three proteins act upstream of RecA and provide distinct avenues for presynapsis. SbcE accumulated at the cell poles in competent cells, and localized as a discrete focus on the nucleoids in 10% of growing cells. This number moderately increased after treatment with DNA damaging agents and in the absence of RecN or of SbcC. Damage-induced foci of SbcE arose early after induction of DNA damage and rarely colocalized with the replication machinery. Our work shows that SMC-like proteins in B. subtilis play roles at different subcellular sites during DNA repair. SbcC operates at breaks occurring at the replication machinery, whereas RecN and SbcE function mainly, but not exclusively, at DSBs arising elsewhere on the chromosome. In agreement with this idea, we found that RecN-YFP damage-induced assemblies also arise in the absence of ongoing replication.     

Polynucleotide phosphorylase exonuclease and polymerase activities on single-stranded DNA ends are modulated by RecN, SsbA and RecA proteins

Bacillus subtilis pnpA gene product, polynucleotide phosphorylase (PNPase), is involved in double-strand break (DSB) repair via homologous recombination (HR) or non-homologous end-joining (NHEJ). RecN is among the first responders to localize at the DNA DSBs, with PNPase facilitating the formation of a discrete RecN focus per nucleoid. PNPase, which co-purifies with RecA and RecN, was able to degrade single-stranded (ss) DNA with a 3' → 5' polarity in the presence of Mn(2+) and low inorganic phosphate (Pi) concentration, or to extend a 3'-OH end in the presence dNDP · Mn(2+). Both PNPase activities were observed in evolutionarily distant bacteria (B. subtilis and Escherichia coli), suggesting conserved functions. The activity of PNPase was directed toward ssDNA degradation or polymerization by manipulating the Pi/dNDPs concentrations or the availability of RecA or RecN. In its dATP-bound form, RecN stimulates PNPase-mediated polymerization. ssDNA phosphorolysis catalyzed by PNPase is stimulated by RecA, but inhibited by SsbA. Our findings suggest that (i) the PNPase degradative and polymerizing activities might play a critical role in the transition from DSB sensing to end resection via HR and (ii) by blunting a 3'-tailed duplex DNA, in the absence of HR, B. subtilis PNPase might also contribute to repair via NHEJ.     

RecN and RecG are required for Escherichia coli survival of Bleomycin-induced damage

The sensitivity of a panel of DNA repair-defective bacterial strains to BLM was investigated. Escherichia coli recA cells were far more sensitive than were uvrA, dam-3, and mutM mutY strains, underscoring the importance of RecA to survival. Strains recBCD and recN, which lack proteins required for double strand break (DSB) repair, were highly sensitive to BLM, while recF cells were not. The requirement for DSB-specific enzymes supports the hypothesis that DSBs are the primary cause of bleomycin cytotoxicity. The acute sensitivity of recN cells was comparable to that of recA, implying a central role for the RecN protein in BLM lesion repair. The Holliday junction processing enzymes RecG and RuvC were both required for BLM survival. The recG ruvC double mutant was no more sensitive than either mutation alone, suggesting that both enzymes participate in the same pathway. Surprisingly, ruvAB cells were no more sensitive than wildtype, implying that RuvC is able to perform its role without RuvAB. This observation contrasts with current models of recombination in which RuvA, B, and C function as a single complex. The most straightforward explanation of these results is that DSB repair involves a structure that serves as a good substrate for RecG, and not RuvAB.     

Recruitment of Bacillus subtilis RecN to DNA double-strand breaks in the absence of DNA end processing

The recognition and processing of double-strand breaks (DSBs) to a 3' single-stranded DNA (ssDNA) overhang structure in Bacillus subtilis is poorly understood. Mutations in addA and addB or null mutations in recJ (DeltarecJ), recQ (DeltarecQ), or recS (DeltarecS) genes, when present in otherwise-Rec+ cells, render cells moderately sensitive to the killing action of different DNA-damaging agents. Inactivation of a RecQ-like helicase (DeltarecQ or DeltarecS) in addAB cells showed an additive effect; however, when DeltarecJ was combined with addAB, a strong synergistic effect was observed with a survival rate similar to that of DeltarecA cells. RecF was nonepistatic with RecJ or AddAB. After induction of DSBs, RecN-yellow fluorescent protein (YFP) foci were formed in addAB DeltarecJ cells. AddAB and RecJ were required for the formation of a single RecN focus, because in their absence multiple RecN-YFP foci accumulated within the cells. Green fluorescent protein-RecA failed to form filamentous structures (termed threads) in addAB DeltarecJ cells. We propose that RecN is one of the first recombination proteins detected as a discrete focus in live cells in response to DSBs and that either AddAB or RecQ(S)-RecJ are required for the generation of a duplex with a 3'-ssDNA tail needed for filament formation of RecA.     

Dynamics and Cell-Type Specificity of the DNA Double-Strand Break Repair Protein RecN in the Developmental Cyanobacterium Anabaena sp. Strain PCC 7120

DNA replication and repair are two fundamental processes required in life proliferation and cellular defense and some common proteins are involved in both processes. The filamentous cyanobacterium Anabaena sp. strain PCC 7120 is capable of forming heterocysts for N₂ fixation in the absence of a combined-nitrogen source. This developmental process is intimately linked to cell cycle control. In this study, we investigated the localization of the DNA double-strand break repair protein RecN during key cellular events, such as chromosome damaging, cell division, and heterocyst differentiation. Treatment by a drug causing DNA double-strand breaks (DSBs) induced reorganization of the RecN focus preferentially towards the mid-cell position. RecN-GFP was absent in most mature heterocysts. Furthermore, our results showed that HetR, a central player in heterocyst development, was involved in the proper positioning and distribution of RecN-GFP. These results showed the dynamics of RecN in DSB repair and suggested a differential regulation of DNA DSB repair in vegetative cell and heterocysts. The absence of RecN in mature heterocysts is compatible with the terminal nature of these cells.     

Repair of radiation-induced heat-labile sites is independent of DNA-PKcs, XRCC1 and PARP

Ionizing radiation induces a variety of different DNA lesions; in addition to the most critical DNA damage, the DSB, numerous base alterations, SSBs and other modifications of the DNA double-helix are formed. When several non-DSB lesions are clustered within a short distance along DNA, or close to a DSB, they may interfere with the repair of DSBs and affect the measurement of DSB induction and repair. We have shown previously that a substantial fraction of DSBs measured by pulsed-field gel electrophoresis (PFGE) are in fact due to heat-labile sites within clustered lesions, thus reflecting an artifact of preparation of genomic DNA at elevated temperature. To further characterize the influence of heat-labile sites on DSB induction and repair, cells of four human cell lines (GM5758, GM7166, M059K, U-1810) with apparently normal DSB rejoining were tested for biphasic rejoining after gamma irradiation. When heat-released DSBs were excluded from the measurements, the fraction of fast rejoining decreased to less than 50% of the total. However, the half-times of the fast (t(1/2) = 7-8 min) and slow (t(1/2) = 2.5 h) DSB rejoining were not changed significantly. At t = 0, the heat-released DSBs accounted for almost 40% of the DSBs, corresponding to 10 extra DSBs per cell per Gy in the initial DSB yield. These heat-released DSBs were repaired within 60-90 min in all cells tested, including M059K cells treated with wortmannin and DNA-PKcs-defective M059J cells. Furthermore, cells lacking XRCC1 or poly(ADP-ribose) polymerase 1 (PARP1) rejoined both total DSBs and heat-released DSBs similarly to normal cells. In summary, the presence of heat-labile sites has a substantial impact on DSB induction and DSB rejoining rates measured by pulsed-field gel electrophoresis, and heat-labile sites repair is independent of DNA-PKcs, XRCC1 and PARP.     

Early steps of double-strand break repair in Bacillus subtilis

All organisms rely on integrated networks to repair DNA double-strand breaks (DSBs) in order to preserve the integrity of the genetic information, to re-establish replication, and to ensure proper chromosomal segregation. Genetic, cytological, biochemical and structural approaches have been used to analyze how Bacillus subtilis senses DNA damage and responds to DSBs. RecN, which is among the first responders to DNA DSBs, promotes the ordered recruitment of repair proteins to the site of a lesion. Cells have evolved different mechanisms for efficient end processing to create a 3′-tailed duplex DNA, the substrate for RecA binding, in the repair of one- and two-ended DSBs. Strand continuity is re-established via homologous recombination (HR), utilizing an intact homologous DNA molecule as a template. In the absence of transient diploidy or of HR, however, two-ended DSBs can be directly re-ligated via error-prone non-homologous end-joining. Here we review recent findings that shed light on the early stages of DSB repair in Firmicutes.     

Cellular responses to DNA double-strand breaks after low-dose γ-irradiation

DNA double-strand breaks (DSBs) are a serious threat to genome stability and cell viability. Although biological effects of low levels of radiation are not clear, the risks of low-dose radiation are of societal importance. Here, we directly monitored induction and repair of single DSBs and quantitatively analyzed the dynamics of interaction of DNA repair proteins at individual DSB sites in living cells using 53BP1 fused to yellow fluorescent protein (YFP-53BP1) as a surrogate marker. The number of DSBs formed was linear with dose from 5 mGy to 1 Gy. The DSBs induced by very low radiation doses (5 mGy) were repaired with efficiency similar to repair of DSBs induced at higher doses. The YFP-53BP1 foci are dynamic structures: 53BP1 rapidly and reversibly interacted at these DSB sites. The time frame of recruitment and affinity of 53BP1 for DSB sites were indistinguishable between low and high doses, providing mechanistic evidence for the similar DSB repair after low- and high-dose radiation. These findings have important implications for estimating the risk associated with low-dose radiation exposure on human health.     

Limited concentration of RecA delays DNA double-strand break repair in Deinococcus radiodurans R1

To evaluate the importance of RecA in DNA double-strand break (DSB) repair, we examined the effect of low and high RecA concentrations such as 2500 and 100 000 molecules per cell expressed from the inducible Pspac promoter in Deinococcus radiodurans in absence or in presence of IPTG respectively. We showed that at low concentration, RecA has a negligible effect on cell survival after gamma-irradiation when bacteria were immediately plated on TGY agar whereas it significantly decreased the survival to gamma-irradiation of DeltaddrA cells while overexpression of RecA can partially compensate the loss of DdrA protein. In contrast, when cells expressing limited concentration of RecA were allowed to recover in TGY2X liquid medium, they showed a delay in mending DSB, failed to reinitiate DNA replication and were committed to die during incubation. A deletion of irrE resulted in sensitivity to gamma-irradiation and mitomycin C treatment. Interestingly, constitutive high expression of RecA compensates partially the DeltairrE sensitization to mitomycin C. The cells with low RecA content also failed to cleave LexA after DNA damage. However, neither a deletion of the lexA gene nor the expression of a non-cleavable LexA(Ind-) mutant protein had an effect on survival or kinetics of DNA DSB repair compared with their lexA+ counterparts in recA+ as well as in bacteria expressing limiting concentration of RecA, suggesting an absence of relationship between the absence of LexA cleavage and the loss of viability or the delay in the kinetics of DSB repair. Thus, LexA protein seems to play no major role in the recovery processes after gamma-irradiation in D. radiodurans.     

The RuvAB branch migration translocase and RecU Holliday junction resolvase are required for double-stranded DNA break repair in Bacillus subtilis

In models of Escherichia coli recombination and DNA repair, the RuvABC complex directs the branch migration and resolution of Holliday junction DNA. To probe the validity of the E. coli paradigm, we examined the impact of mutations in DeltaruvAB and DeltarecU (a ruvC functional analog) on DNA repair. Under standard transformation conditions we failed to construct DeltaruvAB DeltarecG, DeltarecU DeltaruvAB, DeltarecU DeltarecG, or DeltarecU DeltarecJ strains. However, DeltaruvAB could be combined with addAB (recBCD), recF, recH, DeltarecS, DeltarecQ, and DeltarecJ mutations. The DeltaruvAB and DeltarecU mutations rendered cells extremely sensitive to DNA-damaging agents, although less sensitive than a DeltarecA strain. When damaged cells were analyzed, we found that RecU was recruited to defined double-stranded DNA breaks (DSBs) and colocalized with RecN. RecU localized to these centers at a later time point during DSB repair, and formation was dependent on RuvAB. In addition, expression of RecU in an E. coli ruvC mutant restored full resistance to UV light only when the ruvAB genes were present. The results demonstrate that, as with E. coli RuvABC, RuvAB targets RecU to recombination intermediates and that all three proteins are required for repair of DSBs arising from lesions in chromosomal DNA.     

Evidence for Different Pathways during Horizontal Gene Transfer in Competent Bacillus subtilis Cells

Cytological and genetic evidence suggests that the Bacillus subtilis DNA uptake machinery localizes at a single cell pole and takes up single-stranded (ss) DNA. The integration of homologous donor DNA into the recipient chromosome requires RecA, while plasmid establishment, which is independent of RecA, requires at least RecO and RecU. RecA and RecN colocalize at the polar DNA uptake machinery, from which RecA forms filamentous structures, termed threads, in the presence of chromosomal DNA. We show that the transformation of chromosomal and of plasmid DNA follows distinct pathways. In the absence of DNA, RecU accumulated at a single cell pole in competent cells, dependent on RecA. Upon addition of any kind of DNA, RecA formed highly dynamic thread structures, which rapidly grew and shrank, and RecU dissipated from the pole. RecO visibly accumulated at the cell pole only upon addition of plasmid DNA, and, to a lesser degree, of phage DNA, but not of chromosomal DNA. RecO accumulation was weakly influenced by RecN, but not by RecA. RecO annealed ssDNA complexed with SsbA in vitro, independent of any nucleotide cofactor. The DNA end-joining Ku protein was also found to play a role in viral and plasmid transformation. On the other hand, transfection with SPP1 phage DNA required functions from both chromosomal and plasmid transformation pathways. The findings show that competent bacterial cells possess a dynamic DNA recombination machinery that responds in a differential manner depending if entering DNA shows homology with recipient DNA or has self-annealing potential. Transformation with chromosomal DNA only requires RecA, which forms dynamic filamentous structures that may mediate homology search and DNA strand invasion. Establishment of circular plasmid DNA requires accumulation of RecO at the competence pole, most likely mediating single-strand annealing, and RecU, which possibly down-regulates RecA. Transfection with SPP1 viral DNA follows an intermediate route that contains functions from both chromosomal and plasmid transformation pathways.     

The COP9 signalosome is vital for timely repair of DNA double-strand breaks

The DNA damage response is vigorously activated by DNA double-strand breaks (DSBs). The chief mobilizer of the DSB response is the ATM protein kinase. We discovered that the COP9 signalosome (CSN) is a crucial player in the DSB response and an ATM target. CSN is a protein complex that regulates the activity of cullin ring ubiquitin ligase (CRL) complexes by removing the ubiquitin-like protein, NEDD8, from their cullin scaffold. We find that the CSN is physically recruited to DSB sites in a neddylation-dependent manner, and is required for timely repair of DSBs, affecting the balance between the two major DSB repair pathways—nonhomologous end-joining and homologous recombination repair (HRR). The CSN is essential for the processivity of deep end-resection—the initial step in HRR. Cullin 4a (CUL4A) is recruited to DSB sites in a CSN- and neddylation-dependent manner, suggesting that CSN partners with CRL4 in this pathway. Furthermore, we found that ATM-mediated phosphorylation of CSN subunit 3 on S410 is critical for proper DSB repair, and that loss of this phosphorylation site alone is sufficient to cause a DDR deficiency phenotype in the mouse. This novel branch of the DSB response thus significantly affects genome stability.     

Increased sister chromatid cohesion and DNA damage response factor localization at an enzyme-induced DNA double-strand break in vertebrate cells

The response to DNA damage in vertebrate cells involves successive recruitment of DNA signalling and repair factors. We used light microscopy to monitor the genetic dependencies of such localization to a single, induced DNA double strand break (DSB) in vertebrate cells. We used an inducible version of the rare-cutting I-SceI endonuclease to cut a chromosomally integrated I-SceI site beside a Tet operator array that was visualized by binding a Tet repressor-GFP fusion. Formation of γ-H2AX foci at a single DSB was independent of ATM or Ku70. ATM-deficient cells showed normal kinetics of 53Bp1 recruitment to DSBs, but Rad51 localization was retarded. 53Bp1 and Rad51 foci formation at a single DSB was greatly reduced in H2AX-null DT40 cells. We also observed decreased inter-sister chromatid distances after DSB induction, suggesting that cohesin loading at DSBs causes elevated sister chromatid cohesion. Loss of ATM reduced DSB-induced cohesion, consistent with cohesin being an ATM target in the DSB response. These data show that the same genetic pathways control how cells respond to single DSBs and to multiple lesions induced by whole-cell DNA damage.     

Dynamic structures of Bacillus subtilis RecN-DNA complexes

Genetic and cytological evidences suggest that Bacillus subtilis RecN acts prior to and after end-processing of DNA double-strand ends via homologous recombination, appears to participate in the assembly of a DNA repair centre and interacts with incoming single-stranded (ss) DNA during natural transformation. We have determined the architecture of RecN–ssDNA complexes by atomic force microscopy (AFM). ATP induces changes in the architecture of the RecN–ssDNA complexes and stimulates inter-complex assembly, thereby increasing the local concentration of DNA ends. The large CII and CIII complexes formed are insensitive to SsbA (counterpart of Escherichia coli SSB or eukaryotic RPA protein) addition, but RecA induces dislodging of RecN from the overhangs of duplex DNA molecules. Reciprocally, in the presence of RecN, RecA does not form large RecA–DNA networks. Based on these results, we hypothesize that in the presence of ATP, RecN tethers the 3′-ssDNA ends, and facilitates the access of RecA to the high local concentration of DNA ends. Then, the resulting RecA nucleoprotein filaments, on different ssDNA segments, might promote the simultaneous genome-wide homology search.     

Mechanism of cell killing after ionizing radiation by a dominant negative DNA polymerase beta

Several types of DNA lesion are induced after ionizing irradiation (IR) of which double strand breaks (DSBs) are expected to be the most lethal, although single strand breaks (SSBs) and DNA base damages are quantitatively in the majority. Proteins of the base excision repair (BER) pathway repair these numerous lesions. DNA polymerase beta has been identified as a crucial enzyme in BER and SSB repair (SSBR). We showed previously that inhibition of BER/SSBR by expressing a dominant negative DNA polymerase beta (polβDN) resulted in radiosensitization. We hypothesized increased kill to result from DSBs arising from unrepaired SSBs and BER intermediates. We find here higher numbers of IR-induced chromosome aberrations in polβDN expressing cells, confirming increased DSB formation. These aberrations did not result from changes in DSB induction or repair of the majority of lesions. SSB conversion to DSBs has been shown to occur during replication. We observed an increased induction of chromatid aberrations in polβDN expressing cells after IR, suggesting such a replication-dependence of secondary DSB formation. We also observed a pronounced increase of chromosomal deletions, the most likely cause of the increased kill. After H₂O₂ treatment, polβDN expression only resulted in increased chromatid (not chromosome) aberrations. Together with the lack of sensitization to H₂O₂, these data further suggest that the additional secondarily induced lethal DSBs resulted from repair attempts at complex clustered damage sites, unique to IR. Surprisingly, the polβDN induced increase in residual γH2AX foci number was unexpectedly low compared with the radiosensitization or induction of aberrations. Our data thus demonstrate the formation of secondary DSBs that are reflected by increased kill but not by residual γH2AX foci, indicating an escape from γH2AX-mediated DSB repair. In addition, we show that in the polβDN expressing cells secondary DSBs arise in a radiation-specific and partly replication-dependent manner.     

Live cell imaging of XLF and XRCC4 reveals a novel view of protein assembly in the non-homologous end-joining pathway

XLF, also known as Cernunnos, is a newly identified core factor of the non-homologous end-joining (NHEJ) pathway for DNA double-strand breaks (DSBs) repair. XLF is known to stimulate DNA ligase IV in vitro through its interaction with XRCC4. Here, we outline the key findings on the dynamic behavior of XLF and XRCC4 at DSBs in living cells. XLF is quickly recruited to DSBs in the absence of XRCC4 or DNA-PKcs. The recruited XLF molecules constantly exchange at DSBs, and XRCC4 modulates the exchange rate of the recruited XLF. XRCC4 can be recruited to DSBs without DNA-PKcs, but DNA-PKcs stabilizes the recruited XRCC4. These observations are inconsistent with the prevailing concept that NHEJ proteins are sequentially recruited to DSBs, which is mainly supported by in vitro evidence. We propose a novel two-phase model for the assembly of NHEJ factors to DSBs in vivo. XLF, XRCC4, and DNA-PKcs are independently recruited to Ku-bound DSBs. The recruited factors are assembled into a large complex, in which the protein interactions observed in vitro define the stability of the recruited factors. This new view has broad implications for the mechanism of DSB sensing and functional protein assembly in the NHEJ pathway.