Interplay between Ku, Artemis, and the DNA-dependent Protein Kinase Catalytic Subunit at DNA Ends*

Repair of DNA double strand breaks (DSB) by the nonhomologous end-joining pathway in mammals requires at least seven proteins involved in a simplified two-step process: (i) recognition and synapsis of the DNA ends dependent on the DNA-dependent protein kinase (DNA-PK) formed by the Ku70/Ku80 heterodimer and the catalytic subunit DNA-PKcs in association with Artemis; (ii) ligation dependent on the DNA ligase IV·XRCC4·Cernunnos-XLF complex. The Artemis protein exhibits exonuclease and endonuclease activities that are believed to be involved in the processing of a subclass of DSB. Here, we have analyzed the interactions of Artemis and nonhomologous end-joining pathway proteins both in a context of human nuclear cell extracts and in cells. DSB-inducing agents specifically elicit the mobilization of Artemis to damaged chromatin together with DNA-PK and XRCC4/ligase IV proteins. DNA-PKcs is necessary for the loading of Artemis on damaged DNA and is the main kinase that phosphorylates Artemis in cells damaged with highly efficient DSB producers. Under kinase-preventive conditions, both in vitro and in cells, Ku-mediated assembly of DNA-PK on DNA ends is responsible for a dissociation of the DNA-PKcs·Artemis complex. Conversely, DNA-PKcs kinase activity prevents Artemis dissociation from the DNA-PK·DNA complex. Altogether, our data allow us to propose a model in which a DNA-PKcs-mediated phosphorylation is necessary both to activate Artemis endonuclease activity and to maintain its association with the DNA end site. This tight functional coupling between the activation of both DNA-PKcs and Artemis may avoid improper processing of DNA.

DNA double strand breaks (DSB) 2 in cells are produced by exogenous damaging agents like ionizing radiation (IR) or radi-omimetic molecules but also endogenously as by-products of oxidative metabolism or perturbation of the DNA replication fork. In addition, tissue-specific DSB are produced during specialized processes like meiosis in germinal cells or V(D)J recombination in lymphocytes. Improper signaling or repair of DSB in cells can lead to cell death or cancer-prone genomic rearrangements (1,2).
DSB are mainly repaired through two distinct pathways: homologous recombination and nonhomologous end joining (NHEJ), but DSB are mainly processed by the latter pathway in mammalian cells. NHEJ requires several factors that recognize and bind the DSB, catalyze the synapsis of the broken ends, and then process and reseal the break (3)(4)(5). Although alternative subpathways for NHEJ may operate in cells (6,7), the major pathway relies on a set of core proteins, the individual deficiency of which elicits both IR sensitivity and V(D)J recombination defect (2,8). In humans or animals, these defects are responsible for a radiosensitive severe combined immunodeficiency (RS-scid) syndrome (9).
The DSB is recognized and bound by the asymmetric ringshaped heterodimer Ku70/Ku80 that recruits the DNAdependent protein kinase catalytic subunit (DNA-PKcs) (10). The assembled DNA-PK holoenzyme then exhibits serinethreonine protein kinase and DNA end-bridging activities (reviewed in Refs. 11 and 12). One of the kinase functions is to regulate DNA end access to processing enzymes by means of an autophosphorylation operation (13)(14)(15). The XRCC4⅐DNA ligase IV complex is responsible for the ligation step (16,17), and both Ku and DNA-PKcs components are necessary to load this complex to the site of the break (18,19). Recently, a new core NHEJ factor with structural similarity to XRCC4, Cernunnos-XLF, has been identified concomitantly as deficient in a human RS-scid syndrome (20) and as an XRCC4-interacting protein (21). The exact role of this factor is still unknown, but it has been postulated to function in all NHEJ events based on its interaction with XRCC4 (21), the high IR sensitivity of the corresponding deficient cells (20,21), and their complete defect in NHEJ activity in vitro (20,22).
Another NHEJ factor is Artemis, which was originally identified, like Cernunnos, as deficient in a human RS-scid syndrome (23). Cells derived from these patients show an increased sensitivity to IR (24 -26). Gene targeting in mice broadly reproduced these findings (27,28). Artemis exhibits an intrinsic single strand-specific 5Ј to 3Ј exonuclease activity but has also hairpin opening activity in vitro mediated by its DNA-PKcs-dependent phosphorylation (29). This corroborates the findings that a lack of Artemis in vivo leads to V(D)J recombination defects analogous to that produced by a DNA-PKcs deficiency (i.e. impaired coding joining in human cells (26,30) and unresolved coding end hairpins in animals (28)). In addition, DNA-PKcs activates a versatile endonuclease activity of Artemis that can cleave various substrates near the single-to double-stranded transition region (29,31), and the Artemis catalytic core for V(D)J recombination has been mapped (32,33). Although the initial kinetics of DSB repair is normal in Artemis cells (25), Artemis has been proposed to be responsible for the processing of some kinds of IR-generated DNA DSB, since deficient cells have a subtle defect in late repair of DSB (34,35). DNA-PKcs-dependent phosphorylation sites in the C-terminal portion of the Artemis have been suggested to have an important regulatory role in the activity of the protein (31), most of them not being SQ or TQ consensus DNA-PKcs sites (31). In addition, an ATM-and ATR-dependent phosphorylation of Artemis has been reported in cells (34,36,37,38), but its relation to Artemis function in DSB repair is not yet fully understood. Moreover, Artemis could have another role in the regulation of cell cycle progression following DNA damage, including UV irradiation (35,38), although this cell cycle function was not confirmed by others (39). Nevertheless, Artemis clearly has a caretaker function, since fibroblasts from Artemisdeficient mice show genomic instability (27,28), and Artemisdeficient patients as well as Artemis/p53-deficient mice (40) display chromosomal instability and show a predisposition to lymphomas (41).
Although cellular studies have established a role for Artemis in the repair of minor DSB and biochemical experiments have documented its nuclease activity, this protein is still the NHEJ factor for which the least is known about its interactions with the other components of the reaction. It has been reported to form a complex with and to be phosphorylated by DNA-PKcs, leading to the activation of its endonucleolytic activity (29). However, under these experimental conditions with purified components, DNA-PKcs was activated without Ku, and no role was observed for Ku in the overhang processing or the hairpin opening by the Artemis⅐DNA-PKcs complex (29,42). This implies that important aspects of the interactions between Artemis and the other core NHEJ components may have been overlooked. With this in view, we have analyzed the interactions of Artemis and NHEJ proteins on DNA ends by incubating human nuclear cell extracts with paramagnetic beads bearing double-stranded oligonucleotides and characterizing proteins bound to DNA ends. We next have challenged and validated in cells our in vitro results by using a detergent-based cellular fractionation protocol that allows us to assess in situ the DSB-induced recruitment NHEJ repair proteins (19).
Chemicals-Calicheamicin ␥1 (Cal), a generous gift from P. R. Hamann (Wyeth Research, Pearl River, NY), was dissolved in ethanol and stored at Ϫ70°C. Neocarzinostatin was a kind gift from Dr. V. Favaudon (Institut Curie, Orsay, France). It was stored as 1 mM stock solution in 10 mM sodium citrate buffer, pH 4.0, at Ϫ80°C and diluted in the same buffer before use. Cisplatin (cis-diamminedichloroplatinum-II) was a gift from Roger Bellon Cie. Cisplatin (3 mM stock solution) was dissolved in 150 mM NaCl and stored at Ϫ20°C. Wortmannin (Sigma) and NU7026 (Calbiochem) were dissolved in Me 2 SO (10 mM stock solution) and stored at Ϫ20°C. Small aliquots of stock solution chemicals were used once.
Cell Culture and Extracts-All culture media were from Invitrogen and were supplemented with 10% fetal calf serum unless indicated, 2 mM glutamine, 125 units/ml penicillin, and 125 g/ml streptomycin. Guetel are SV40T-transformed, telomerase-immortalized Artemis-deficient fibroblasts, and Guetel-A were obtained after transduction of Guetel with pMND-Artemis-Myc-ires-GFP retroviral vector expressing a C-terminal Myc-His-tagged Artemis protein (37); both cell lines were grown in RMPI medium. DNA-PKcs-deficient and -complemented cell lines (Fus9, alias M059J, and Fus1, respectively (43) (gifts from Dr. C. Kirchgessner, Stanford University School of Medicine) were maintained in Dulbecco's modified Eagle's medium/F-12 1:1 medium. All cells were grown in a humidified atmosphere, at 37°C with 5% CO 2 . Nuclear protein extracts were prepared as previously described (18), except that the final dialysis was performed for 3 h at 4°C in an excess volume of dialysis buffer as follows: 50 mM Tris-HCl, pH 7.5, 10% glycerol, 100 mM potassium glutamate, 1 mM EDTA, 1 mM dithiothreitol. After preparation, all of the extracts were immediately frozen and stored at Ϫ80°C.
Peptide and Purified Protein-The recombinant N-terminal hexa-His-Ku70/Ku80 heterodimer was expressed in Sf9 cells using baculovirus-based expression vector as previously described (44) and purified on a Hitrap chelating column (Amersham Biosciences) charged with Ni 2ϩ ions followed by HitrapQ HP (Amersham Biosciences) anion exchange chromatography. The C80 peptide (GSGSEGGDVDDLLDMI) was synthesized by H. Mazarguil (Institut de Pharmacologie et de Biologie Structurale) and corresponds to the last 12 C-terminal amino acids of Ku80 with a GSGS linker to a biotin moiety at the N terminus as published (45). The peptide buffer used as control was 25 mM Tris-HCl, pH 8.8.
Ku Immunodepletions and Artemis Immunoprecipitations-For Ku immunodepletions, anti-Ku70/80 (162) antibodies were coupled to magnetic anti-mouse IgG beads (Dynabeads M-450; Dynal) according to the manufacturer's recommendations. Then 250 g of nuclear protein extracts were incubated at 4°C for 60 min under gentle agitation with 20 l of beads in dialysis buffer. The supernatant was removed over a magnet (Dynal MPC; Dynal). A second depletion was performed immediately under the same conditions. For Artemis immunoprecipitation, anti-Myc antibodies were added to Guetel-A nuclear extracts in IP buffer (25 mM Hepes-KOH (pH 7.5), 100 mM NaCl, 20% glycerol, 5 mM EDTA, 1 mM dithiothreitol, 0.05% Nonidet P-40, 10 mM NaF, 0.2 mM sodium orthovanadate, 1 mM cantharidin (Sigma), and protease inhibitor mixture tablets as recommended by the manufacturer (Roche Applied Science) and incubated for 2 h at 4°C under agitation, and then 20 l of protein A-immunobeads (Dynal) were added per reaction and further incubated for 90 min at 4°C under agitation. Then the beads were washed with IP buffer and used as necessary.
DNA End Binding Assay-Five pmol of double-stranded oligonucleotide as indicated were immobilized on 10 l of streptavidin paramagnetic beads (Dynabeads M280 streptavidin; Dynal) in 50 l of 5 mM Tris-HCl, pH 7.5, 0.5 mM EDTA, 1 M NaCl for 30 min at 20°C under agitation. After washes with the same buffer, DNA-or mock-treated beads were incubated in 25 l of reaction mixture containing 30 g of nuclear extracts in standard reaction buffer (40 mM Hepes-KOH (pH 7.8), 5 mM MgCl 2 , 60 mM KCl, 0.5 mM dithiothreitol, 0.5 mM EDTA, 3.4% glycerol, 0.3 mg/ml bovine serum albumin) and 1 mM ATP when necessary. For conditions without ATP, extracts were first incubated for 10 min at 30°C in standard reaction buffer without ATP, supplemented with 2 mM glucose and 0.2 units of hexokinase (Sigma) in order to remove traces of ATP as described (18). For conditions with wortmannin, extracts were preincubated in the presence of 30 M wortmannin in reaction buffer without oligonucleotide and ATP for 10 min at 30°C. Control extracts were similarly preincubated in parallel. Incubation was for 30 min at 30°C under agitation. The extract supernatant was removed, and the beads were washed twice with 250 l of IP buffer; then the beads were heated in SDS sample buffer, and proteins were separated in an 8% acrylamide Tris-glycine-SDS gel.
Coimmunoprecipitation Assay-After DNA end binding assays as above but with 5 pmol of free C6 double-stranded oligonucleotide, the reaction volume was completed to 100 l with IP buffer. For anti-actin, -Ku, or -DNA-PKcs IPs, the mixture was mixed with 10 l of magnetic anti-mouse IgG immunobead suspension coated with the corresponding primary antibodies according to the manufacturer's protocol (Dynal), and the beads were mixed gently on a wheel for 3 h at 4°C. For anti-Artemis IPs, anti-Myc antibodies were first added to the reaction mixture and incubated for 2 h at 4°C under agitation, and then 20 l of protein A-immunobeads (Dynal) were added per reaction and further incubated for 90 min at 4°C under agitation. The beads were pulled down over a magnet, the extract supernatant was removed, the beads were washed twice with 1 ml of IP buffer, and proteins in the immunoprecipitates were heated in SDS sample buffer and separated in a 8% acrylamide Tris-glycine-SDS gel.
DNA-damaging Treatments and Transfection-Before drug exposure, exponentially growing cells were washed with unsupplemented medium, either mock-treated or treated with chemicals, at the specified concentrations in unsupplemented medium at 37°C in culture dishes and then harvested at the indicated time points. For UV irradiation, cells were washed with phosphate-buffered saline (PBS) and then exposed to UVC irradiation (254 nm) with a germicidal lamp (Bioblock). Immediately after irradiation, unsupplemented medium was added, and cells were postincubated as above. Fus9 and Fus1 cells were transiently transfected with the Nucleofector II apparatus (Amaxa) as follows: 2 ϫ 10 6 cells were transfected with 3 g of pcDNA1.1-Artemis-myc vector (37) in 100 l of Nucleofector V buffer (program A23, about 43% transfection efficiency) and then diluted 15-fold and incubated for 48 h in complete medium.
Biochemical Fractionation and Immunoblotting-Treated or mock-treated cells in culture dishes were washed twice with ice-cold PBS, collected by scraping, and centrifuged. Cell fractionation was carried out by two consecutive extractions. The supernatant was collected at each step and labeled as fraction S. Pellets of about 2 ϫ 10 6 cells were first resuspended for 3 min on ice in 200 l of extraction buffer (50 mM Hepes, pH 7.5, 150 mM NaCl, 1 mM EDTA) containing 0.1% Triton X-100, supplemented with protease inhibitor mixture tablets (Complete Mini TM ; Roche Applied Science) and phosphatase inhibitors (10 mM NaF, 10 mM ␤-glycerophosphate, 1 mM sodium orthovanadate, and 1 mM cantharidin, all from Sigma). Following centrifugation at 14,000 ϫ g for 3 min, the supernatant was collected (fraction S1), and the pellet was washed with extraction buffer without Triton. The pellet was further incubated in 100 l of extraction buffer without Triton but supplemented with 200 g/ml RNase A (Sigma) for 30 min at 25°C under agitation. Following centrifugation at 14,000 ϫ g for 3 min, the supernatant (fraction S2) was separated from the pellet, which was then washed with extraction buffer without Triton (fraction P2). When necessary, the P2 pellet was incubated for 1 h at 37°C in the presence of 100 units of calf intestine phosphatase (New England Biolabs) in 20 mM Tris-HCl, pH 8, 2 mM magnesium chloride, and protease inhibitors as above. Insoluble P2 fraction was resuspended in PBS buffer supplemented with 1% SDS, heated 10 min at 100°C, and sonicated for 10 s (Vibracel, Bioblock Scientific). Whole cell extracts of treated or mocktreated cells were obtained by direct lysis in PBS buffer supplemented with 1% SDS and treatment as above. Concentrated loading sample buffer was added for 1ϫ final concentration in all fractions, and the samples were boiled for 5 min. Equal aliquots of each fraction, derived from equivalent cell numbers, were separated on SDS-PAGE (8% for standard separation or 15% for ␥-H2AX and HP1␣ isolation) and blotted onto polyvinylidene difluoride membranes (Immobilon-P; Millipore Corp.). Membranes were blocked for 1 h in 5% dry milk in PBS containing 0.1% Tween 20 (PBS-T) and incubated for 1 h with primary antibody diluted in PBS containing 0.02% Tween 20 and 1% bovine serum albumin (fraction V; Sigma). After three washes with PBS-T, membranes were incubated for 1 h with secondary antibodies in PBS containing 0.02% Tween 20 and 5% dry milk. Immunoblots were visualized by enhanced chemiluminescence (ImmunofaxA; Yelen). After extensive washing and probing first with polyclonal antibodies, successive immunoblots were performed on the same membranes without stripping. For data presentation, films were scanned and processed with Adobe Photoshop 3.0 software.

DNA-PK Kinase Activity Is Necessary to Maintain the DNA-PKcs/Artemis Association during Binding to DNA Ends in
Vitro-First, we have focused on the protein fraction bound to DNA ends in vitro in order to evaluate the capacity of Artemis to be recruited to DNA ends in the context of nuclear extracts. Nuclear extracts from the Guetel-A fibroblasts expressing a C-terminal Myc-tagged Artemis protein easily detected on Western blot were incubated with paramagnetic streptavidin beads bearing a double-stranded oligonucleotide modified with a biotin moiety at one 5Ј end as target DNA. Proteins bound to DNA ends were analyzed by Western blotting. In addition, to assess the role of DNA-PKcs in the assembly of repair proteins onto DNA ends, the reaction was performed either with ATP or in the presence of the known DNA-PK inhibitor wortmannin.
As shown in Fig. 1, DNA-PKcs, Ku70/80, and Myc-tagged Artemis were present in the protein fraction retained on DNA with no nonspecific binding to the paramagnetic beads in the absence of DNA. Ku80 bound to DNA was shifted to a slightly slower form under ATP conditions corresponding to phosphorylated Ku80 as reported (18). A quantitative modulation was observed for DNA-associated Artemis and DNA-PKcs; DNA-PKcs accumulated on the DNA beads in the presence of wort-mannin ( Fig. 1, lanes 5, 7, 9, and 11) by stalling of DNA-PK at DNA-ends, as reported (13). In contrast, DNA-PKcs binding to the beads decreased with ATP ( Fig. 1, lanes 4, 6, 8, and 10) most probably by autophosphorylation (46). Strikingly, Artemis exhibited a mirror image, since it accumulated on DNA in the presence of ATP and showed a low binding to DNA beads in the presence of wortmannin. In the presence of ATP, the association of Artemis to DNA beads increased with the length of the 5Ј-protruding DNA end up to four nucleotides, but it was also strong with a 3Ј-protruding double-stranded oligonucleotide (data not shown). In reactions with ATP, Artemis was detected in the DNA end-associated fraction and in the supernatant (data not shown) as a form migrating more slowly than in reactions with wortmannin. The slower migrating Artemis was sensitive to protein phosphatase, indicating that it corresponded to a phosphorylated form (data not shown).
Artemis was reported to form a stable complex with DNA-PKcs (29). However, the present data showing an inverse correlation between DNA-PKcs and Artemis binding to DNA beads could indicate that Artemis alone or associated with a partner other than DNA-PKcs exhibited a DNA binding activity. In order to challenge this hypothesis, we tested under similar conditions the DNA binding activity of either native or phosphorylated Artemis in the absence of Ku and DNA-PKcs. In the absence of DNA-PKcs, phosphorylated or native Artemis lacked binding activity to the double-stranded oligonucleotides, which was not restored by adding purified Ku to DNA beads but was restored upon adding back purified DNA-PKcs (data not shown), indicating that DNA-PKcs is necessary for Artemis loading to DNA.
Another explanation for the data of Fig. 1 was that Ku/DNA-PKcs interaction on DNA ends destabilized the DNA-PKcs⅐Artemis complex under kinase-preventive conditions. In order to analyze the composition of the protein complex assembled on DNA ends under kinase permissive or preventive conditions, we performed protein assembly on free doublestranded oligonucleotides followed by immunoprecipitation experiments with either anti-DNA-PKcs or anti-Ku antibodies and checked for coimmunoprecipitation with Artemis (Fig. 2). None of the three proteins were precipitated by the anti-actin control antibody (Fig. 2, lanes 6 and 11). In contrast, the anti-  DNA-PKcs and anti-Ku antibodies precipitated very efficiently their respective target from the nuclear extracts (Fig. 2, compare lanes 2 and 7 with lane 1). In the absence of added DNA, Ku and DNA-PKcs marginally co-precipitated (Fig. 2, lanes 2  and 7) as expected (43); in contrast, Ku and DNA-PKcs coprecipitated in the presence of free double-stranded oligonucleotides under kinase-preventive conditions (Fig. 2, lanes 3 and 8, without ATP, and lanes 5 and 10, with wortmannin), corresponding to the DNA-PK complex stalled at DNA-ends. On the contrary and as expected, ATP dissociated the DNA-PK complex (Fig. 2, lanes 4 and 9). Artemis co-precipitated with DNA-PKcs but not Ku in nuclear extracts in the absence of DNA (Fig. 2, compare lanes 2 and 7), as already reported (29); note that only a fraction of the protein is engaged in a complex with DNA-PKcs in extracts from Guetel-A cells. Under conditions in which a whole DNA-PK⅐DNA complex was formed, Artemis remained associated with the complex only when DNA-PKcs was active, and it co-precipitated as a phosphorylated form (Fig. 2, lanes 4 and 9). These data confirmed our hypothesis that the Artemis⅐DNA-PKcs complex was destabilized when DNA-PK assembled on DNA unless the kinase was active.
Ku Is Responsible for DNA-PKcs/Artemis Dissociation upon DNA-PK Binding to DNA Ends in Vitro-In order to strengthen these conclusions, we performed a similar immunoprecipitation experiment with standard and Ku-immunodepleted nuclear extracts in parallel and analyzed the DNA-PK components co-immunoprecipitating with Artemis (Fig. 3). The anti-Myc antibody precipitated efficiently Artemis from the nuclear extracts (Fig. 3, compare lane 2 with lane 1). Only a fraction of DNA-PKcs associated stably with Artemis (Fig. 3, lane 2). In control extracts, we found again that Artemis remained associated with DNA-PKcs only when the latter was active (ϩATP), and under these conditions, it co-precipitated as a phosphorylated form, together with Ku and DNA-PKcs (Fig. 3, top, compare lane 4 with lanes 3 and 5). In contrast, no significant variation was detected in the amount of DNA-PKcs co-precipitated with Artemis from the Ku-depleted extracts, whatever the incubation conditions (Fig. 3B, bottom). Low salt conditions promote a Ku-independent DNA-PKcs binding to DNA (47) and Artemis phosphorylation with ATP (29). Thus, we have checked the stability of the DNA-PKcs⅐Artemis complex on anti-Myc immunobeads incubated with DNA under such conditions, in the absence of Ku and in the presence or not of ATP. No significant release of DNA-PKcs from the beads was detected, whatever the incubation conditions (data not shown). Taken together, these results indicate that the Ku-mediated assembly of DNA-PK on DNA ends was responsible for a dissociation of the DNA-PKcs⅐Artemis complex under kinasepreventive conditions.
In order to focus on DNA-PK assembly onto DNA, we reconstituted the reaction with purified fractions (Fig. 4). The DNA-PKcs⅐Artemis complex was first immunoprecipitated from nuclear extracts without DNA and then incubated with doublestranded oligonucleotides in the presence or absence of various concentrations of purified Ku heterodimer (Fig. 4A). DNA plus Ku conditions promoted a dose-dependent loss of DNA-PKcs from the Artemis immunoprecipitates, whereas neither DNA nor Ku alone had any effect on DNA-PKcs/Artemis association. The extreme C-terminal domain of Ku80 has been shown to be sufficient for association with DNA-PKcs in the absence of DNA (45,48). Thus, we mimicked the Ku/DNA-PKcs interaction by using a C80 peptide derived from Ku80 C terminus (45). When the DNA-PKcs⅐Artemis complex was immunoprecipitated as above, the addition of the C80 peptide was sufficient to promote DNA-PKcs/Artemis dissociation (Fig. 4B, compare  lanes 1 and 2). In contrast, when the DNA-PKcs/Artemis was immunoprecipitated under conditions promoting DNA-PK assembly and Artemis phosphorylation (ATP plus DNA (Fig.  4B, note the shift of Artemis and the presence of Ku in lanes 3 and 4)), the Ku80 C-terminal peptide instead dissociated Ku from the complex, and most of DNA-PKcs still precipitated with phosphorylated Artemis (Fig. 4B, compare lanes 3 and 4).
Taken together, these data from in vitro experiments establish that Ku binding to DNA-PKcs in the presence of doublestranded DNA, most probably via its extreme C terminus, promotes Artemis dissociation from DNA-PKcs unless the kinase is active.
Artemis Is Recruited to Chromatin Containing DSB in Cells-In order to validate our previous conclusions in cells, experiments were then performed in Guetel-A fibroblasts expressing the Myc-tagged Artemis construct after stable retroviral transduction of Guetel Artemis-deficient cells (37). Guetel-A fibroblasts were either treated or not with drugs producing DSB formation, Cal and neocarzinostatin (Ncs). Cal and Ncs are natural enediyne antibiotics that have been shown to produce DSB with selectivity and efficiency higher than IR (49) and to efficiently induce DSB and cytotoxicity when applied to FIGURE 3. Effect of selective Ku immunodepletion on Artemis/DNA-PK co-immunoprecipitation. Reactions with Guetel-A nuclear extracts immunodepleted with anti-Ku or control magnetic beads were performed with free C6 oligonucleotide as in Fig. 2, and Myc-tagged Artemis protein was immunoprecipitated as described under "Materials and Methods." Then the extracts supernatant was removed, and the beads were washed. Proteins in the immunoprecipitates were heated in SDS sample buffer and separated in 8% SDS-PAGE. Western blotting was performed with antibodies as indicated. The same film exposure was used for the input extracts and the immunoprecipitated protein fraction. Artemis was detected with an anti-Myc antibody. The arrow shows the electrophoretic mobility shift. cells (50,51). Fibroblasts established from Artemis-defective patients as well as Artemis Ϫ/Ϫ mouse embryo fibroblasts from Artemis Ϫ/Ϫ mice show increased sensitivity to IR (26 -28) as well as to the radiomimetic drug bleomycin (27). Accordingly, we observed also a marked increased sensitivity of Guetel fibroblasts to the radiomimetic drugs Cal and Ncs, which was fully restored in Guetel-A cells by expression of Myc-tagged Artemis (data not shown).
We have described recently a detergent-based cellular fractionation protocol allowing us to assess in situ the DSB-induced recruitment of the main NHEJ repair proteins, as visualized by immunoblot analysis (19). This protocol was applied to Guetel-A cells in order to check for Artemis mobilization to chromatin after DSB infliction. Since Cal yields a 1:3 ratio of DNA DSB to single-stranded breaks in vivo, compared with a 1:20 ratio for IR (50), we have chosen this radiomimetic drug to treat the cells. After 1 h of drug treatment, cell nuclei were extracted with a Triton-containing buffer, and the clarified cell extract supernatant was collected (S1) after centrifugation. The cell pellet was treated with RNase A in the same buffer but without detergent as described (19), and the soluble and insoluble fractions were collected after centrifugation (S2 and P2, respectively). A parallel extraction procedure was performed on untreated and damaged cells after Cal treatment. Fig. 5A shows the immunoblot analysis following SDS-PAGE of cell-equivalent aliquots of the three fractions compared with whole cell extracts (WCE), under both untreated and Cal-treated conditions. Proteins were detected by antibodies against Artemis, DNA-PKcs, Ku80, DNA ligase IV, XRCC4, and ␥H2AX, the phosphorylated form on serine residue 139 of the histone H2AX variant, which is admitted to be a quantitative nuclear marker of DSB (52). In addition, an anti-Rad51 antibody was used to probe the homologous recombination (HR) route for DSB repair. WCE of Artemis-deficient cells serve as a negative control for Artemis expression (Fig. 5A, lane 1). As opposed to nontreated cells, WCE from Cal-treated cells contain ␥H2AX (Fig. 5A, lane 2), in agreement with the high DNA doublestranded breaking potency of Cal. In untreated cells, the majority of NHEJ proteins was released during the two extraction steps, and only a marginal amount was detected in the insoluble P2 fraction, whereas on the contrary, the P2 fraction from Caltreated cells was highly enriched for these proteins, including Artemis (Fig. 5A, compare lanes 8 and 9). Also, ␥H2AX was exclusively present in the insoluble P2 fraction from Caltreated cells. In contrast, Rad51 protein was detected identically in the P2 fraction of drug-treated and nontreated cells. In addition, Artemis and XRCC4 in all of the fractions of Caltreated cells were detected essentially as slowly migrating forms that were sensitive to calf intestine phosphatase (data not shown), corresponding to phosphorylated forms, as already reported for XRCC4 (19).
Since Artemis-defective mutants are selectively sensitive to DSB-inducing agents, we then analyzed the specificity of protein recruitment toward the class of DNA lesions (Fig. 5B). As expected for the recruitment of a key NHEJ protein, we found that phosphorylated XRCC4 was retained in the P2 insoluble fraction following treatment of cells with the DSB-inducing agents Ncs and Cal, in correlation with the appearance of ␥-H2AX. Notably, the retention of Artemis paralleled that of XRCC4, and the recruitment of Artemis was accompanied by its phosphorylation. In addition, Artemis was similarly recruited after cell treatment with bleomycin and IR (data not shown). In contrast and when compared with the untreated cells, there was no significant retention of both Artemis and XRCC4 proteins when these cells were heavily irradiated with UV-C rays or treated with the cross-linking agent cisplatin (Fig.  5B). Then the time course of protein retention in the extraction-resistant fraction P2 in Guetel-A cells after exposure to Cal was examined. Fig. 5C shows that ␥H2AX formation was detected at 5 min, the earliest time point examined, and that the kinetics of Artemis, Ku, and XRCC4 protein retention was in close synchrony with the appearance of ␥H2AX, as already shown for the core NHEJ proteins (19). In addition, Artemis and XRCC4 showed a similar retention pattern with the appearance of an intermediate migrating form, followed by progressive accumulation of an even slower migrating form, most likely corresponding to multiple phosphorylated forms. Then the extract supernatant was removed, and the beads were washed. The anti-Myc magnetic beads were then incubated in the presence of double-stranded oligonucleotide and/or purified Ku (300 or 600 ng) as indicated in standard reaction buffer for 15 min at 20°C. Then the extracts supernatant was removed, and the beads were washed. Proteins in the immunoprecipitates were heated in SDS sample buffer and separated in 8% SDS-PAGE. Western blotting was performed with antibodies as indicated. Artemis was detected with an anti-Myc antibody. B, reactions with Guetel-A nuclear extracts were performed, and then Myctagged Artemis protein was immunoprecipitated as in Fig. 3. Then the extract supernatant was removed, and the beads were washed. The anti-Myc magnetic beads were then incubated in the presence of 15 nmol of C80 peptide (ϩ) or the equivalent volume of peptide dilution buffer (Ϫ) in standard reaction buffer for 15 min at 20°C. Then the extract supernatant was removed, and the beads were washed. Proteins in the immunoprecipitates were heated in SDS sample buffer and separated in 8% SDS-PAGE. Western blotting was performed with antibodies as indicated. Artemis was detected with an anti-Myc antibody. The arrow shows the electrophoretic mobility shift.
The kinetics of Cal-induced Artemis phosphorylation is in agreement with the one reported after IR (35,37).
DNA-PKcs Is Necessary for the Recruitment of Artemis to DSB in Chromatin-We then tested whether Artemis relied on DNA-PKcs for its damage-induced recruitment as we have found in vitro. Thus, M059J glioblastoma cells that do not express DNA-PKcs (DNA-PKcs-deficient cells, Fus9) and M059J-complemented cells that contain an extra copy of the human gene coding for DNA-PKcs (DNA-PKcs-complemented cells, Fus1) (43) were transfected with a pcDNA1.1 vector expressing the Myc-tagged Artemis protein. After 48-h expression, both cells were treated with Cal, and the recruitment of Artemis to the insoluble chromatin fraction was assessed. As shown in Fig. 6, Artemis was equally expressed from the transfected vector in both DNA-PKcs-proficient and -defective cells, indicating that Artemis is not likely to be stabilized by its interaction with DNA-PKcs (Fig. 6, lanes 1 and  4). After treatment with Cal, Ku80 was similarly recruited to the damaged chromatin in both cell lines, as reported (19), but in sharp contrast, Artemis mobilization to the P2 fraction was only detected in the DNA-PKcs-proficient cells together with DNA-PKcs recruitment, and in addition, it was detected as a phosphorylated form (Fig.  6, compare lanes 3 and 6).
The simplest interpretation of these results is that DNA-PKcs is necessary for Artemis stable recruitment to DSB in chromatin in agreement with our in vitro data.
The Stabilization of Artemis on DSB-containing Chromatin Is Dependent on the Kinase Activity of DNA-PKcs-Since we have set up conditions allowing us to analyze in the cells the DNA-PKcs-dependent mobilization of Artemis to broken chromatin, we then challenged our conclusions from in vitro experiments by assessing in vivo the effect of the DNA-PKcs kinase activity on the association of Artemis with damaged DNA. The selective DNA-PKcs inhibitor NU7026 has been shown to exhibit a strong DNA-PKcs-dependent radiosensitization effect on cells at 10 M when added 1 h before irradiation (53). Therefore, Guetel-A cells were pretreated or not with 10 M NU7026 for 1 h, and then Cal was added for further incubation at 37°C. The extraction protocol was achieved as above, and the WCE and P2 protein fractions were analyzed by Western blot. As shown in Fig. 7A, the DNA-PKcs inhibitor had no obvious effect on the strong mobilization of Ku and DNA-PKcs to the damaged chromatin (Fig. 7A, compare lanes 3 and 4). XRCC4 was also heavily recruited under DNA-PK activity permissive or preventive conditions, but NU7026 abolished its phosphorylation (Fig. 7A,  compare lanes 3 and 4). Accordingly, we have shown elsewhere that DNA-PKcs-dependent XRCC4 phosphorylation was dispensable for its recruitment to damaged chromatin (19). The lack of XRCC4 phosphorylation in the presence of NU7026 is thus a good indicator of the actual DNA-PKcs inhibition under these conditions. Artemis was clearly mobilized to the P2 frac- Cells were collected and lysed in denaturing buffer (WCE) or fractionated by two consecutive extractions as described under "Materials and Methods," leading to S1 and S2 soluble fractions and P2 insoluble material. Protein samples were denatured and separated on SDS-polyacrylamide gels (8% for standard separation or 15% for ␥H2AX, Rad51, and HP1␣ isolation) followed by electrotransfer on membrane. The membranes were blotted with the antibodies as indicated. Artemis was detected with an anti-Myc antibody. Electrophoresis mobility shift is shown by an arrow. A, cell treatment was with 10 nM Cal. B, Guetel-A cells were either not treated (NT ), irradiated with UV-C light (UV; 200 J/m 2 ), or treated with cis-diamminedichloroplatinum-II (CDDP; 200 mM) or neocarzinostatin (Ncs; 1 mM) for 1 h at 37°C. As a control of protein loading, we used the heterochromatin HP1␣ protein. C, Guetel-A cells were treated with 10 nM Cal for the indicated time before lysis and fractionation. As a control of protein loading, we used the nuclear matrix protein lamin A/C.  Fig. 5, leading to P2 insoluble material. Protein samples were denatured and separated on 8% SDS-polyacrylamide gel followed by electrotransfer on membrane and blotting with the antibodies as indicated. As a control of protein loading, we used the nuclear scaffold protein lamin A/C. Electrophoresis mobility shift is shown by an arrow. tion of Cal-treated cells in which it was detected in a phosphorylated form (Fig. 7A, lane 3). In contrast, Artemis was hardly detectable in the P2 fraction of cells treated with both NU7026 and Cal, yielding the same marginal amount as in the control untreated cells (Fig. 7A, compare lanes 2 and 4). When the WCE were analyzed, Artemis was fully phosphorylated in the Cal-treated cells but, after treatment with Cal in the presence of NU7026, migrated as in the control untreated cells (Fig. 7B). This indicates that the phosphorylation observed under these conditions mostly relied on the NU7026-sensitive DNA-PKcs activity. In conclusion, the stabilization of Artemis on DSB is dependent on the kinase activity of DNA-PK.

DISCUSSION
Artemis protein is the factor of the NHEJ apparatus for which the least is known about its interactions with the other components of the reaction. Here, we have analyzed the interactions of Artemis and NHEJ proteins both in a context of human nuclear cell extracts and in cells.
In untreated cells, most of Artemis belongs to the soluble nucleoplasmic compartment, since it is exclusively found in the soluble protein fraction, as shown here by biochemical analysis. In contrast, DSB induce the mobilization of Artemis together with DNA-PK and XRCC4/ligase IV proteins to a detergentresistant nuclear compartment. The Artemis mobilization to damaged chromatin is specifically initiated by DSB-inducing agents, like that of the other NHEJ factors as detected by us (19) and another group using this technique (14,54). The time course of appearance of this recruitment paralleled that of the DSB-specific induction of H2AX phosphorylation, and for the milderdoses,thedisruptionofArtemisrecruitmentanddephosphorylation of ␥-H2AX occurred concomitantly, 3 compatible with the forming and rejoining kinetics of DSB generated in normal cells (50). Thus, it is most likely that the mobilization of Artemis to a less extractable nuclear compartment observed here corresponds to its loading onto sites of DNA DSB. This may reflect the role of Artemis in the repair of a subclass of non-V(D)J DSB, as previously inferred from the inability of Artemis-deficient cells to rejoin a low percentage of radiationinduced breaks (34,35).
We report that DNA-PKcs is necessary for the loading of Artemis on damaged chromatin in cells, since no Artemis was detected in the detergent-resistant nuclear fraction in Caltreated DNA-PKcs-deficient cells, despite normal mobilization of Ku, as already reported (19). Moreover, in cell extracts devoid of DNA-PKcs, Artemis cannot be recruited to DNA ends even in the presence of Ku. Thus, Ku or another partner possibly associated with Artemis fraction cannot substitute for DNA-PKcs to load Artemis onto DSB. Consequently, the Artemis⅐DNA-PKcs complex reported by Ma et al. (29) that we also found here is most likely to represent the functional complex in DSB repair.
We have established here that under kinase-preventive conditions, Ku-mediated assembly of DNA-PK on DNA ends is responsible for a dissociation of the DNA-PKcs⅐Artemis complex. This result was obtained by monitoring the protein fraction bound to double-stranded oligonucleotides on beads in cell extracts and also the association of Artemis with DNA-PKcs in co-IP experiments. In addition, experiments in cells corroborated this result, since Artemis recruitment in damaged chromatin was not detected in cells pretreated with a DNA-PKcs-specific inhibitor.
It could be argued that the short DNA targets used in vitro may not permit sufficient space for colocalization of DNA-PKcs, Ku, and Artemis. However, this hypothesis is unlikely, since an excess of the Ku80 C-terminal peptide promoted efficiently Artemis/DNA-PKcs dissociation even in the absence of DNA, implying another mechanism.
Interestingly, Ku dissociates Artemis from DNA-PKcs in the presence of DNA, whereas the C-terminal Ku80 peptide is efficient without DNA. It has been demonstrated that the conserved C-terminal motif of Ku80 is required for the efficient recruitment of DNA-PKcs to DNA ends in vitro (55). In addition, cells expressing a form of Ku80 lacking the C terminus exhibit a DNA-PKcs minus phenotype despite the fact that the kinase is present and the Ku DNA binding property is conserved, indicating that the extreme C terminus of Ku80 is also required for DNA-PKcs recruitment and activation at DNA DSB in cells (48,55). Moreover, a C-terminal fragment of Ku80 confers a dominant negative effect on DSB repair and radiosensitivity (56), and the Ku80 C-terminal peptide sensitizes cells to DSB (57). Since it has been shown that Ku and DNA-PKcs do not associate in the absence of a DNA terminus (58), whereas a Ku80 C-terminal fragment is sufficient to interact with DNA-PKcs (45), it is most likely that Ku binding to a DNA end exposes the Ku80 C terminus as a docking module for DNA-PKcs, probably necessary for other subsequent stabilizing interactions (55). This region absent in the x-ray structure of the Ku heterodimer with DNA (59) was shown to be flexible in solution by NMR determination and exhibited a high helical propensity, allowing folding upon Ku binding to DNA or to its protein partner (60,61). Thus, Ku binding to DNA or the Ku80 C-terminal motif may displace Artemis from DNA-PKcs by competition on the same domain of the kinase. Previous studies have 3 P. Calsou, P. Frit, and B. Salles, unpublished results. Guetel-A cells were pretreated or not with 10 M NU7026 in unsupplemented medium for 45 min, and then Cal (10 nM) was added or not for 15 min before lysis and fractionation. Protein samples were denatured and separated on 8% SDS-polyacrylamide gels followed by electrotransfer on membrane. The membranes were blotted with the antibodies as indicated. Electrophoresis mobility shift due to phosphorylation is shown by an arrow. A, analysis of the P2 protein fraction. B, analysis of the S1 protein fraction.
established that residues in the FAT domain of DNA-PKcs adjacent to the catalytic domain are involved in its interaction with Ku (62), contrary to the leucine-rich region (54). One possibility is that the DNA-PKcs FAT domain is involved similarly in interactions with Ku80 and Artemis, implying that some domain of Artemis may share a structural homology with the Ku80 C terminus. Another possibility is that Artemis dissociation is elicited by a change of DNA-PKcs conformation. Indeed, several studies have revealed extensive conformational changes of the DNA-bound kinase consisting of domain rearrangements, including the FAT portion to form channels that could accommodate duplex and single-stranded DNA (63)(64)(65)(66). Thus, deciphering whether Ku dissociates the Artemis⅐DNA-PKcs complex via an induced change of DNA-PKs conformation or a competition on the same binding site on DNA-PKcs awaits further experiments. In addition, the role of another DNA-PKcs partner for this dissociation in cells cannot be excluded.
Our present data from experiments in vitro and in cells showed that the activity of the DNA-PKcs kinase prevented Artemis dissociation from the DNA-PK⅐DNA complex. This is in agreement with the results of Ma et al., who showed that DNA-PKcs is physically required for Artemis activity after its phosphorylation (42) and that DNA-PKcs remained associated with Artemis immunobeads after phosphorylation and stringent washes (31). Under our conditions, Artemis accumulation on double-stranded oligonucleotides in vitro was correlated with the length of the 5Ј or 3Ј single-stranded tail. A plausible explanation is that DNA-PKcs activation may be proportional to the length of the single-stranded tail as reported (67) and that the activity level may in turn regulate the extent of the Artemis⅐DNA-PK complex stabilization. The shift from an unstable Artemis⅐DNA-PK⅐DNA complex to a stable association probably relies on a change in the geometry of the protein⅐DNA complex involving Artemis and/or DNA-PKcs phosphorylation. Indeed, differential autophosphorylation of DNA-PKcs on two major clusters has been shown to greatly influence DNA end access to processing enzymes (reviewed in Ref. 11). A current model is that an intermediate phosphorylated state of DNA-PKcs directs a rearrangement of the DNA-PK complex that ensures access to broken ends, whereas complete autophosphorylation dissociates the complex from DNA ends (14,15). Thus, the DNA-PKcs associated with Artemis under kinase permissive conditions may be either unphosphorylated or subphosphorylated and may be necessary to load Artemis to its substrate and/or to maintain a proper configuration of the DNA termini. Alternatively, Artemis phosphorylation per se could change its interactions with DNA-PK and/or DNA. Artemis phosphorylation takes place on the C-terminal domain (31,37), possibly allowing extrusion of the inhibitory C-terminal domain (31), which could also be implicated in maintaining Artemis association with the DNA-PK⅐DNA complex.
In Cal-treated cells, Artemis was detected as a phosphorylatedprotein,andaspecificDNA-PKcsinhibitorabrogatedthisphosphorylation, implying that DNA-PKcs was mainly responsible for Artemis phosphorylation under these conditions. This is in contrast with other reports that also implicated ATM in Artemis IRinduced phosphorylation (35,37). Our results more easily agree with the biochemical evidence of the DNA-PK-mediated activa-tion of the Artemis endonucleolytic function (29,42). This discrepancy may rely on the DSB inflicted, which in the case of Cal may be enriched in the subclass processed by Artemis.
Altogether, our data allow us to propose a model in which a DNA-PKcs-mediated phosphorylation is necessary both to activate Artemis and to maintain its association with the DNA-PK/DNA end site (Fig. 8). DNA-PKcs may constitute a flexible sensor of various unusual structures (68) on which an improperly controlled nuclease activity may cause genetic instability. Thus, a tight functional coupling between the activation of both DNA-PKcs and Artemis is indispensable. Together with the incapacity of the sole Artemis protein to load onto DNA termini, this coupling may avoid undesirable nucleolytic activity on binding sites of DNA-PKcs other than DSB that may not necessarily activate the kinase.