The DNA-dependent Protein Kinase Catalytic Activity Regulates DNA End Processing by Means of Ku Entry into DNA*

The DNA-dependent protein kinase (DNA-PK) is required for double-strand break repair in mammalian cells. DNA-PK contains the heterodimer Ku and a 460-kDa serine/threonine kinase catalytic subunit (p460). Ku binds in vitro to DNA termini or other discontinuities in the DNA helix and is able to enter the DNA molecule by an ATP-independent process. It is clear fromin vitro experiments that Ku stimulates the recruitment to DNA of p460 and activates the kinase activity toward DNA-binding protein substrates in the vicinity. Here, we have examined in human nuclear cell extracts the influence of the kinase catalytic activity on Ku binding to DNA. We demonstrate that, although Ku can enter DNA from free ends in the absence of p460 subunit, the kinase activity is required for Ku translocation along the DNA helix when the whole Ku/p460 assembles on DNA termini. When the kinase activity is impaired, DNA-PK including Ku and p460 is blocked at DNA ends and prevents their processing by either DNA polymerization, degradation, or ligation. The control of Ku entry into DNA by DNA-PK catalytic activity potentially represents an important regulation of DNA transactions at DNA termini.

DNA double strand breaks (DSBs) 1 are generated by agents such as ionizing radiation (IR) and also occur as intermediates in certain recombination reactions. In response to the deleterious consequences associated with DSBs, highly efficient mechanisms have evolved for their recognition and repair. Mammalian cells appear to rejoin DBSs primarily by a mechanism of non-homologous recombination (reviewed in Refs. [1][2][3][4]. Insights into the proteins involved in DBSs break repair have been gained from analysis of the molecular defects in mutant rodent cells that are hypersensitive to IR and unable to carry out the V(D)J recombination process in the immunoglobulin and T-cell receptor genes. Recent evidence show that a multiprotein complex, the DNA-dependent protein kinase (DNA-PK) plays a central role in DBS repair in mammalian cells (reviewed in Refs. 5 and 6).
The DNA-PK is a heterotrimeric enzyme composed of a large catalytic subunit of ϳ460 kDa (DNA-PK cs , p460) , a serine/ threonine kinase that belongs to the phosphatidylinositol 3-kinase (p110) family (7), and a regulatory component consisting of the Ku80 and Ku70 proteins (8,9) .
The unique property of DNA-PK is to exhibit a protein kinase activity that depends on the interaction of the protein complex with DNA. Both DNA-PK subunits can interact with nucleic acids (reviewed in Ref. 10). The Ku80 and Ku70 proteins form a heterodimer (11,12) that binds with high affinity to DNA ends of dsDNA and more generally to transitions from doubleto single-stranded DNA (12)(13)(14). Ku represents the major dsDNA end-binding protein as detected by band shift experiments (15) . Both Ku subunits are required for stable binding to DNA (16 -19). Ku is able to translocate along the DNA in an ATP-independent manner, allowing several Ku dimers to bind to a single DNA molecule and form a multimeric complex (15,20). A single-stranded DNA-dependent ATPase and helicase activities have also been reported for Ku (19,(21)(22)(23). The binding of Ku molecules to one another leading to DNA loop formation has been reported (24). Ku can transfer between DNA fragments with compatible ends (25). In addition, Ku bridges DNA ends (26,27) and can stimulate DNA end joining by mammalian DNA ligases (27).
Recently, a direct binding of p460 to DNA leading to kinase activation has been reported under in vitro conditions with purified preparations of the catalytic subunit (28 -30). However, this property of p460 is lost in the presence of Ku on very short dsDNA fragments Ͻ26 bp (29,30). For longer DNA, Ku greatly stabilizes the DNA-PK/DNA complex and leads to optimal kinase activity (29,30), in agreement with the common view that Ku is required to target the catalytic subunit to DBSs under physiological conditions (9).
Once bound at DNA termini, the active DNA-PK complex acquires the capacity, at least in vitro, to phosphorylate many DNA-bound proteins in the vicinity (31) . For example, DNA-PK phosphorylates many transcription factors including p53 (32,33) , Sp1 (34), and also the carboxyl-terminal domain of RNA polymerase II (8) .
Ku protein DNA binding properties have been well characterized with purified native or recombinant proteins. However, the question of the influence of the catalytic subunit on Ku interaction with DNA is poorly documented and remains un-answered. DNA-PK complex does not appear to form stably in the absence of DNA (35,36). Once the complex has formed, Ku70, Ku80, and DNA-PK cs can be phosphorylated by DNA-PK. DNA-PK cs phosphorylation inactivates the kinase (33) that dissociates from Ku (36). Phosphorylated Ku70 binds to DNA in Southwestern analysis (36), but whether phosphorylated Ku dimer still binds DNA is not known. It has been shown that the intrinsic Ku DNA-dependent ATPase activity is greatly stimulated when Ku is phosphorylated by DNA-PK (23), but the subsequent possible changes in the association of Ku with DNA have not been determined. On the other hand, extracts from DNA-PK cs -deficient cells exhibit normal DNA end binding activity (37,38) and mutations in the DNA-PK phosphorylation sites in Ku70 do not modify IR sensitivity in vivo (39).
The most common assay for Ku DNA end binding activity is electrophoretic mobility shift assay (EMSA), but it is performed without ATP, which is a nonpermissive condition for the kinase activity. Moreover, of the two modes of Ku binding to DNA, i.e. DNA end-binding and translocation into the DNA helix, EMSA does not allow precise characterization of the second owing to length limitation of the DNA probe and complexity of the retarded bands pattern. We have examined the possible involvement of DNA-PK catalytic activity on the regulation of Ku binding to DNA in a competition EMSA. The protein extracts were first preincubated with various potential competitor DNAs for Ku DNA binding activity under kinase preventive or permissive conditions; varying the length of linear competitor DNA allowed to take both modes of Ku binding to DNA into account. Then, the free remaining Ku DNA binding activity was assessed with a 25-bp radiolabeled DNA probe under standard EMSA conditions. Nuclear human cell extracts were used in order to closely mimic the physiological conditions of molar ratios of the DNA-PK components and of protein environment.
Here, we demonstrate that, although Ku can enter DNA from free double-stranded ends in the absence of p460 subunit, the kinase activity is required for Ku translocation along the DNA helix when the whole Ku/DNA-PK cs assembles on DNA ends. When the kinase activity is impaired, DNA-PK including Ku and DNA-PK cs is blocked at DNA ends and prevents end processing by either DNA polymerization, degradation, or ligation.

Cell Extracts and Proteins
The HeLa S3 cell line was obtained from the stock of European Molecular Biology Laboratories (Heidelberg, Germany). Cells were grown in RPMI 1640 medium (Life Technologies, Inc.), supplemented with 10% fetal calf serum, 2 mM glutamine, 125 units/ml penicillin, and 125 g/ml streptomycin at 37°C, in a humidified atmosphere containing 5% CO 2 .
Nuclear protein extracts were prepared from HeLa cells as described previously (40) except that the final dialysis was performed for 3 h at 4°C in 25 mM Hepes-KOH, pH 7.9, 17% glycerol, 100 mM potassium glutamate, 2 mM EDTA, 1 mM EGTA, and 2 mM dithiothreitol. After preparation, extracts were immediately frozen and stored at Ϫ80°C.
The purified Ku complex is baculoviral expressed recombinant proteins purified as already reported (24).

Plasmids and DNA Fragments
Closed circular 2959-bp pBluescript KS ϩ (pBS, Stratagene) was prepared by the standard alkaline lysis method and cesium chloride gradient centrifugation from Escherichia coli JM109 (relevant genotype: recA1, endA1, gyrA96, hsdR17) followed by two neutral sucrose gradient centrifugations as described (41). When necessary, pBS plasmid was linearized by digestion at a unique site with HindIII.
End radiolabeled pBS was produced by incubation of linearized pBS with E. coli Klenow fragment (0.02 unit) (Life Technologies, Inc.) together with 5 M dNTP and 0.5 Ci of [␣-32 P]dCTP (NEN Life Science Products) for 30 min at 30°C, followed by chromatography through Sephadex G-50 (Amersham Pharmacia Biotech) and phenol/chloroform extraction.
The 173-and 710-bp fragments were prepared by digestion with both EcoRI and HindIII of pGf1 and pCMV plasmids, respectively (Ref. 42; kind gifts of Dr M. S. Satoh, Université Laval, Quebec, Canada), and a mean 1400-bp length fragment was obtained by digestion of pBS with BglI producing two fragments (1691 and 1268 bp). The bands corresponding to the fragments were excised from agarose gel, and the DNA was purified by phenol extraction and ethanol precipitation.
Standard EMSA-EMSA was performed as described previously (15). Briefly, radiolabeled DNA (4 ng, 100,000 cpm) was incubated with nuclear extracts (2 g) in 10 l of buffer (20 mM Tris-HCl, pH 7.5, 100 mM KCl, 0.3 mM dithiothreitol, 2 mM MgCl 2 , 0.1 mM EDTA, 5% glycerol) in the presence of closed circular pBS plasmid as a nonspecific competitor (0.5 g) at 30°C for 15 min. The samples were electrophoresed on a 5% polyacrylamide gel at 4°C, for 2 h at 100 V. The gel was dried on Whatman 3MM paper and exposed to a storage phosphor screen (Molecular Dynamics), followed by processing with a PhosphorImager (Molecular Dynamics, Storm System TM ). The regions of the gel containing the free probe and the retardation complexes were quantitated by the ImageQuant software, version 4.2A (Molecular Dynamics). In the supershift experiments, antibodies were added to the mixture following the binding incubation and the mixture was incubated for an additional 5 min prior to gel electrophoresis. The antibodies used were purified on protein A-Sepharose from human autoimmune antiserum AF (a generous gift from Dr. E. M. Tan, Scripps Research Institute, La Jolla, CA), which recognizes Ku70 and Ku80 (11) and, as a control, from the human serum Sa that contains anti-ribonucleoprotein antibodies (provided by Dr. Y. Takeda, Medical College of Georgia, Augusta, GA).
Competition EMSA-Nuclear protein extracts (20 g) or purified Ku protein as indicated in 10 l of buffer (45 mM Hepes-KOH, pH 7.8, 50 mM KCl, 7 mM MgCl 2 , 1 mM dithiothreitol, 0.5 mM EDTA, 40 mM phosphocreatine, 2.5 g of creatine phosphokinase (type I, Sigma), 3.4% glycerol, and 18 g of bovine serum albumin) were first preincubated in the presence of closed circular pBS plasmid as a nonspecific competitor (1 g). When necessary, ATP (from 1 M stock solution in water, potassium salt; Sigma) and wortmannin (from 10 mM stock solution in dimethyl sulfoxide; Sigma) were included in the mixture. When indicated, linear DNA was included in the mixture before the preincubation (sample thereafter named ϩ competitor linear DNA) and in some experiments, an equivalent amount of closed circular pBS was added in control reactions (sample named Ϫ competitor linear DNA). After 1 h at 30°C, 1/10 of the reaction mixture (1 l, equivalent to 2 g of protein extracts) was incubated in a 10-l final reaction volume with the radiolabeled 25-mer DNA probe for 10 min at 30°C under standard EMSA conditions as described above. The samples were electrophoresed and the gel processed as described above.
Immunoprecipitation DNA-PK cs Immunodepletion-Anti-DNA-PK cs mouse monoclonal antibody mAb 25-4 (kind gift of Dr. T. Carter, Dept. of Biological Sciences, St. John's University, Jamaica, NY) was reported previously (44). The mAb in hybridoma supernatant was coupled to magnetic anti-mouse IgG beads (Dynabeads M-450, Dynal), according to the manufacturer recommendations. Under a 20-l final volume, 250 g of HeLa nuclear protein extracts were incubated at 4°C for 60 min with 20 l of wet anti-DNAPK beads (equivalent to Ϸ2 g of anti-DNA-PK IgG) in extract dialysis buffer under gentle agitation. The supernatant was removed over a magnet (Dynal MPC, Dynal). A second depletion was performed immediately under the same conditions. An aliquot of the supernatant was assayed for DNA-PK activity and reached Ͻ10% of the initial activity as reported (45). As control, immunodepletions were run in parallel with uncoupled anti-mouse IgG beads (extracts thereafter referred to as mock-treated) or with beads coupled to anti-XPB mouse monoclonal antibody mAb 1B3 (kind gift from Dr. J. M. Egly, IGBMC, Illkirch, France).
DNA Fragment Pull-down Assay-Nuclear protein extracts (20 g) were incubated in a 10-l volume reaction under competition EMSA preincubation conditions (see above) in the presence of 50 ng of endradiolabeled HindIII-linearized pBS, with 5 M ATP and with or without 3 M wortmannin. After 60 min at 30°C, an excess of unlabeled 25-bp dsDNA probe (2.5 pmol) was added to the mixture. This amount was estimated from EMSA experiments to be sufficient to titrate out any Ku that would not have bound linear pBS during the preincubation period (data not shown; see Fig. 5). Then, half of the reaction (5 l) was incubated in a 20-l final reaction volume with SacI enzyme (20 units, Amersham Pharmacia Biotech) in restriction buffer according to the manufacturer. After 60 min at 37°C, 1/4 of the reaction mixture was incubated at 4°C for 60 min with 10 l of wet anti-DNAPK (mAb 25.4) or control anti-IgG magnetic beads under gentle agitation. Then, the supernatant was removed, the beads were washed twice with 500 l of buffer (6 mM Hepes-KOH, pH 7.9, 12 mM KCl, 2, 5 mM MgCl 2 , 0.25 mM dithiothreitol, 0.02% Nonidet P-40, and 5% glycerol) and resuspended in 15 l of the same buffer. DNA fragments in both beads and supernatant fractions were purified by 60 min incubation at 37°C with 0.5% SDS and proteinase K (200 g/ml final concentration) followed by phenol-chloroform extraction and were electrophoresed on a 5% polyacrylamide gel at 4°C, for 2 h at 100 V.

Enzymatic Reactions at DNA Termini
Exonuclease III Assay-Nuclear protein extracts (20 g) were incubated in a 10-l volume reaction under competition EMSA preincubation conditions (see above) in the presence of 50 ng of end-radiolabeled HindIII-linearized pBS, with 5 M ATP and with or without 3 M wortmannin as indicated. After 60 min at 30°C, 1/3 of the reaction mixture was further incubated with E. coli exonuclease III (Life Technologies, Inc.) for 15 min at 37°C. Reaction was terminated by addition of 10 mM EDTA and 0.5% SDS. The DNA was recovered from the reaction mixture by proteinase K digestion (200 g/ml, 15 min at 37°C), followed by phenol extraction and ethanol precipitation, and then separated on TAE-0.8% agarose gels. After visualization under short wavelength UV light, the gel was dried and radioactive DNA was revealed by exposure to a storage phosphor screen (Molecular Dynamics) followed by processing with a PhosphorImager (Molecular Dynamics, Storm System TM ).
Klenow Assay-This assay follows the same procedure as described above except that the HindIII-linearized plasmid pBS substrate was unlabeled, and the exo III enzyme was replaced by E. coli Klenow fragment (0.02 unit) (Life Technologies, Inc.) together with 5 M dNTP and 0.5 Ci of [␣-32 P]dCTP (NEN Life Science Products). DNA was purified as above. The radioactivity incorporation due to the overhang fill-in was visualized by exposure to a storage phosphor screen (Molecular Dynamics) followed by processing with a PhosphorImager (Molecular Dynamics, Storm System TM ).
Ligase Assay-Nuclear protein extract (40 g) was incubated under conditions as described above but with the following modifications: 200 ng of unlabeled HindIII-linearized pBS plasmid without competitor DNA, 1 mM ATP, and with or without 30 M wortmannin as indicated, for 60 min at 37°C. DNA was purified as above. The ligation products were analyzed by agarose gel electrophoresis and ethidium bromide staining.

ATP Stimulates Nuclear Ku DNA Binding Activity on Large
DNA Fragments-Ku DNA binding activity can be easily detected by using dsDNA fragments in an EMSA (15).
When a 25-bp labeled DNA was mixed with nuclear extracts from HeLa cells, a single slow migrating band appeared in the gel (Fig. 1A). This band was supershifted in the presence of antibodies purified from a human serum able to recognize both Ku subunits, while the control antiserum (that contained antiribonucleoprotein autoantibodies) had no effect. Thus, this retarded band can be confidently ascribed to a Ku DNA complex. Under these standard EMSA conditions, Ku binds to the DNA probe independently from the p460 subunit since it has been demonstrated that a stable Ku-p460 complex fails to form on short fragments (Ͻ26 bp) (28 -30) and, moreover, that Ku competes for p460 binding to such short DNA probes (29,30).
We have tentatively analyzed the co-binding of Ku and p460 on probes Ͼ100 bp under standard EMSA conditions (data not shown). However, the appearance of multiple retarded bands including material in the wells of the gel prevents a precise analysis of the protein-DNA complexes involved. Since the EMSA is performed in the presence of excess supercoiled plasmid DNA, which competes for unspecific DNA binding activity in the extracts, we reasoned that it might be possible to characterize a potential p460-dependent Ku DNA binding activity by preincubating the extracts with unlabeled linear plasmid DNA allowing the co-binding of both DNA-PK subunits. For this purpose, nuclear protein extracts were suitable since, as already observed (33,44), we found that p460 was concentrated in the nucleus and hardly detectable in the corresponding cytoplasmic extracts (data not shown). Indeed, only the nuclear fraction exhibited a significant DNA-PK activity as measured by the conventional pull-down assay on DNA-cellulose beads in the presence of a peptide kinase substrate (46) (data not shown). In addition, since the molar ratio of both DNA-PK components may be important for interaction with DNA, nuclear extracts were preferred over whole cell extracts, since in the latter, cytoplasmic Ku may change the physiologic molar ratio of Ku toward p460 in the nucleus.
Thus, we performed a new EMSA (referred to as competition EMSA under "Experimental Procedures") in which nuclear cell extracts were preincubated with or without linear plasmid DNA, in the presence or absence of ATP. Then, the remaining Ku binding activity was assessed on the labeled 25-bp probe that allows to visualize the DNA binding activity specifically devoted to the Ku subunit of DNA-PK independently from p460 (Fig. 1B). Although under these conditions, 10 ng of linear plasmid slightly competed for Ku (Ͼ95% remaining binding to the probe), ATP dramatically increased this effect (Ͻ20% remaining binding to the probe). A preincubation in the presence of ATP but with only the standard supercoiled competitor DNA had no competitor effect (Fig. 1B, lanes Ϫlinear DNA). The stimulating effect of ATP on the Ku titration by linear DNA was detected all over a range of DNA concentration (Fig. 1C) yielding 10-fold for 100 ng of DNA (7% and 65% remaining binding to the probe in the presence and absence of ATP, respectively). A concentration of ATP down to 1 M in the presence of an ATP-regenerating system was sufficient to promote a significant stimulation of Ku titration (data not shown).
It has been reported that Ku dimer could exhibit DNA-dependent ATPase and ATP-dependent helicase activities (19,(21)(22)(23). It was of interest to test the effect of ATP under the same competition EMSA conditions with purified recombinant Ku heterodimer. Fig. 2 shows that, although linear plasmid DNA efficiently competes for Ku in a concentration-dependent manner, ATP has no effect on this competition.
p460 Kinase Activity Is Necessary for ATP-dependent Nuclear Ku DNA Binding Activity-The previous data imply that there exists an ATP-dependent DNA binding mode that might be mediated by proteins in the nuclear extracts. DNA-PK cs was a strong candidate for the protein regulating Ku DNA binding activity in the presence of ATP. First, at the ends of large DNA fragments, Ku interacts with the catalytic p460 kinase subunit of DNA-PK (8,9). Second, the ATP-dependent Ku DNA binding relied on ATP hydrolysis, since under the competition EMSA conditions, we found that ATP could not be replaced by the non-hydrolyzable analog AMP-PNP (data not shown). Third, we observed no effect of ATP when the unlabeled 25-bp DNA probe was used as competitor (data not shown), which correlates well with the impairment of Ku-p460 complex assembly on short DNA fragments (28,29).
One of the first inhibitor of DNA-PK identified is the fungal metabolite wortmannin, which irreversibly inhibits the kinase activity in vitro (7). The nuclear Ku titration by linear DNA was assessed in the presence of ATP and increasing concentrations of wortmannin. As shown in Fig. 3A, the ATP-dependent competitor effect of linear plasmid DNA was reversed by wortmannin in a dose-dependent manner and completely abolished for concentrations between 1 and 3 M wortmannin (Fig. 3B) (the slight competition observed was equivalent to the effect of 50 ng of linear DNA without ATP). Notably, in the absence of linear DNA (Fig. 3, A and B), wortmannin had no effect on Ku binding to the 25-bp probe, in agreement with the ATP-and p460-independent Ku binding to this short DNA fragment.
DNA-PK activity was assessed in parallel using a standard peptide phosphorylation assay (46) and was found to be inhibited in the same range of wortmannin concentrations (data not shown).
Wortmannin is a general inhibitor of enzymes of the phosphatidylinositol 3-kinase (p110) family (47). In order to demonstrate that p460 was involved in the ATP-dependent Ku binding to linear DNA, we used monoclonal antibodies to immunodeplete HeLa nuclear extracts of DNA-PK. We have already reported that two successive immunoprecipitations with mAb 25-4 removed more than 90% of the DNA-PK activity (45). A control immunodepletion with anti-XPB magnetic beads (against one of the subunits of the basal transcription factor TFIIH) was performed in parallel. Mock-treated nuclear extracts and the two immunodepleted extracts were assayed in the competition EMSA with a fixed amount of linear plasmid DNA (50 ng) and under preincubation conditions without ATP, with ATP, or with both ATP and wortmannin (Fig. 4A). The control and XPB-depleted extracts show the expected profile that is an ATP-dependent stimulation of Ku titration by linear DNA, which is reversed by wortmannin. In sharp contrast, no Ku binding to the labeled probe appeared under the three preincubation conditions with the DNA-PK cs -depleted extracts.
A simple explanation could rely on a co-depletion of Ku along p460 in these extracts. However, this possibility was ruled out because in the absence of linear plasmid, p460-depleted extracts showed Ͻ20% reduction of Ku DNA binding when compared with control-depleted extracts (Fig. 4B, lanes 0 competitor linear DNA). Another explanation for the above result could be that Ku bound better to the linear competitor in the absence of DNA-PK cs , irrespectively of whether ATP or wortmannin were present. Indeed, the competitor effect of a range of linear DNA concentrations was compared between the p460-depleted and the control-depleted extracts in the absence of ATP (Fig.  4B). Linear DNA was much more competitor for Ku in p460depleted extracts than in control extracts; this effect was ATPindependent. In addition, this increased competitor effect of linear DNA for Ku in p460-depleted extracts did not change in the presence of ATP or wortmannin ( Fig. 4A and data not  shown).
Taken together, these results indicate that p460 depletion facilitates Ku binding to the competitor DNA in an ATP-independent mode. In other words, when present in nuclear extracts, DNA-PK cs somehow impairs Ku binding to linear plasmid DNA, unless ATP is provided and the kinase is active.
DNA-PK Catalytic Activity Regulates Ku Entry into Linear DNA-The kinase activity of p460 could modulate Ku binding to DNA by changing either Ku affinity for DNA or the mode of Ku binding. Indeed, Ku interacts with DNA by a two-step mechanism where it first recognizes DNA ends and then translocates along the DNA helix, allowing the binding of several Ku dimers to a single DNA molecule (15,20,48). Both modes of Ku interaction with DNA can operate in nuclear extracts for Ku titration by the linear competitor plasmid DNA. We postulated that the competition EMSA would allow to dissociate these two modes by varying independently two parameters regarding the competitor DNA, i.e. the number of free ends and the length of the DNA molecule. If Ku can just bind to the DNA ends (mode 1), the titration would be only dependent on the number of ends whatever the length of the linear competitor. On the opposite, if Ku can invade DNA (mode 2), the titration would be dependent on both parameters. Fig. 5A shows the result of a typical experiment, which compares the nuclear Ku competitor effect of a fixed molarity of DNA ends for three linear DNAs of increasing length (ranging from 171 bp to 1400 bp), without or with ATP. Due to the stimulating effect of ATP on the competition and in order to still visualize a band shift in each case, the amount of DNA ends for the three fragments was adjusted at 200 fmol or 30 fmol per assay, in the absence or presence of ATP, respectively. In addition, the experiment was repeated for various molarities of DNA ends and the results are shown in Fig. 5B. The data in Fig. 5 (A and B) clearly show that, in nuclear extracts in which both Ku and DNA-PK cs are present, Ku binding to the competitor without ATP is independent of the DNA length and only increases with the number of DNA ends while, in the presence of ATP, the extent of Ku titration is correlated to both the number of ends and the DNA length.
In the competition EMSA experiments, an alternative explanation for the apparent ATP-stimulated binding of Ku to the linear competitor would be that, under kinase permissive conditions, phosphorylated Ku molecules no longer bind to the 25-bp probe. However, the present data allow to rule out this simple explanation since, in that case, the extent of Ku titration in the presence of ATP would solely depend on the number of kinase-stimulating DNA ends and in any case on the DNA length.
The data above support the interpretation that, in the presence of DNA-PK cs , the kinase activity can regulate the mode of Ku binding to linear plasmid DNA. If ATP is available and the kinase functional, Ku can both bind to the ends and slide along DNA (mode 2), resulting in high Ku titration rate by the linear competitor. Without ATP or with non-hydrolyzable ATP, or when the kinase activity is impaired by wortmannin, Ku can still bind to DNA ends but is unable to enter the DNA molecule (mode 1), resulting in a poor titration effect. In the absence of kinase subunit, however (immunodepleted extracts), Ku can bind to DNA ends and freely enter DNA as shown by the high Ku titration rate of linear DNA with p460-depleted extracts, with or without ATP (Fig. 4).
Wortmannin Blocks DNA-PK at DNA Ends-In nuclear extracts, Ku cannot enter DNA when the kinase activity is impaired, suggesting that Ku and probably the whole DNA-PK complex may stay bound at the DNA ends. In order to test this hypothesis, we have designed an experiment aimed at determining the proteins bound at the DNA ends under kinase permissive or preventive conditions. An end-labeled linear plasmid DNA was incubated with nuclear extracts under the conditions used in the preincubation step of the competition EMSA, with ATP Ϯ wortmannin. Then, after dilution, the DNA was restricted at a unique site close to one labeled end. This second incubation was carried out in the presence of an excess of cold DNA probe (25 bp) calibrated in order to titrate out the remaining Ku that might bind to the labeled oligonucleotide released (data not shown). Then, the mixture was immunoprecipitated in the presence of control or anti-DNA-PK cs magnetic beads. The DNA retained on the beads or remaining in the supernatant was purified, run on polyacrylamide gel electrophoresis, and analyzed by autoradiography.
As shown in Fig. 6, DNA restriction by SacI released a 66-bp oligonucleotide and a 2893-bp fragment that were easily separated in the gel (compare lanes 1 and 2, without and with restriction). The control magnetic beads retained ϳ25% of the large fragment and Ͻ2% of the small fragment (Fig. 6, A and B,  lanes 2 and 3). The anti-DNA-PK cs beads revealed a very distinct profile of DNA retention with the control nuclear extracts without or with wortmannin; while ϳ65% of the large fragment was immunoprecipitated in both cases, up to 90% of the small DNA end fragment was retained with wortmannin (kinase preventive condition), in sharp contrast with ϳ25%, without wortmannin (kinase permissive condition) (Fig. 6, A and B,  compare lanes 4 and 5). A mirror profile was observed in the supernatant. In order to validate these results, the same experiment was performed with the p460-depleted nuclear extracts. Strikingly, although wortmannin was present, only ϳ25% of the small DNA piece was immunoprecipitated by the anti-DNA-PK cs beads (Fig. 6, A and B, lane 6). In addition, these beads retained only 27% of the large fragment under these conditions, close to the nonspecific binding on the control beads (Fig. 6, A and B, lane 6). The same experiment was performed with anti-Ku beads and gave broadly the same result regarding the small fragment (data not shown), although the yield of immunoprecipitation was lower, probably due to steric hindrance by the large catalytic subunit at the DNA end.
These results demonstrate that in the presence of both Ku and p460 subunits, the kinase activity can regulate Ku entry into DNA. Indeed, when the kinase activity is impaired, DNA-PK as a whole complex stays bound at DNA ends and under certain conditions, almost all the ends may bear a DNA-PK complex.
Regulation of DNA Transactions at DNA Ends by DNA-PK Catalytic Activity-Having demonstrated that a defective kinase activity blocked DNA-PK complexes at DNA ends, it was then of interest to determine the potential enzymatic consequences of such protein/DNA structures on DNA transactions at DNA ends. Therefore, nuclear extracts were incubated with linear plasmid DNA and an excess of closed circular competitor under kinase permissive (ϩATP) or preventive (ϩATP, ϩwortmannin) conditions. After dilution, the mixture was assayed for enzymatic reactions involving DNA ends.
First, we tested various concentrations of the bacterial exonuclease III (exo III) on a reaction performed with a labeled linear plasmid, under mild nuclease conditions. Fig. 7 shows the yield of DNA recovery (upper panel) and the extent of DNA end degradation (lower panel). While wortmannin had no effect on exo III activity in the absence of extracts (data not shown), it partially prevented the dose-dependent degradation of DNA ends that was observed with increasing exo III units in the presence of ATP (ϳ3-fold inhibition of degradation for 0.1 unit with wortmannin).
Second, the polymerizing activity of Klenow fragment of E. coli DNA polymerase I at the ends was assessed in a similar assay by incubating unlabeled linear plasmid with control or p460-immunodepleted extracts. As shown in Fig. 8A, DNA end filling by Klenow was drastically impaired in the presence of wortmannin and control extracts, while this inhibition was partially released by DNA-PK cs immunodepletion. The unchanged background incorporation in the open circular species indicates that wortmannin has no effect on Klenow activity per se.
Finally, the intrinsic DNA ligase activity of the extracts on the linear plasmid substrate was tested under similar condi- tions (Fig. 8B). Control and DNA-PK cs -immunodepleted extracts exhibited the same ligation profile of the linear DNA in the absence of wortmannin, leading to high molecular weight DNA species identified as multimers, as reported previously (49). Strikingly, wortmannin almost completely abolished this activity in control extracts while it remained unchanged in the DNA-PK cs -depleted extracts. DISCUSSION We report here for the first time that the DNA-PK catalytic activity regulates DNA end processing by mean of Ku entry into the helix. This conclusion was drawn from several data in experiments with nuclear extracts from human cells; (i) in the presence of DNA-PK cs , Ku entry into DNA depends on ATP hydrolysis and is blocked by inhibition of the kinase activity, (ii) inhibition of DNA-PK catalytic activity results in a stable Ku/DNA-PK cs complex blocked at DNA ends, (iii) blocked DNA-PK at DNA ends prevents end processing by either DNA polymerization, degradation, or ligation.
Data of the literature and the present results support a model of the successive steps for DNA end processing in the presence of DNA-PK (Fig. 9). Ku and p460 assemble at DNA ends (8,9). Under kinase permissive conditions, interactions between DNA and DNA-PK lead to efficient activation of the kinase, which in turn allows Ku entry into DNA, assembling of new DNA-PK subunits, and/or further DNA transactions at the ends. When the kinase is present but not active, both components of DNA-PK are blocked at the ends, thus impeding their processing. This model raises several questions.
What Might Be The Target of DNA-PK Activity for Ku Entry into DNA?-Several results argue against a prominent role for Ku phosphorylation; (i) extracts from DNA-PK cs -deficient cells or DNA-PK cs -depleted extracts exhibit normal DNA end binding activity (Refs. 37 and 38, and our results), (ii) mutations in the DNA-PK phosphorylation sites in Ku70 do not modify IR sensitivity in vivo (39), (iii) purified recombinant Ku dimer freely enters DNA as judged by multimerization of Ku on DNA probe in EMSA (50). 2 However, Ku cannot be ruled out as a key target of p460 for its entry into DNA since (i) the consequence on IR sensitivity of mutations in Ku80 DNA-PK phosphorylation sites has not been evaluated yet, (ii) the intrinsic Ku DNA-dependent ATPase activity has been reported to be greatly stimulated when Ku was phosphorylated by DNA-PK (23), which could in turn modify its putative helicase activity, (iii) Ku molecules bound to a DNA-PK activating DNA fragment are indeed in a phosphorylated form under kinase permissive conditions. 2 Another possible candidate for the target of DNA-PK is p460 itself. Indeed, DNA-PK cs autophosphorylation inactivates the kinase (33), which dissociates from Ku (36). According to our model (Fig. 9), Ku might then enter freely the DNA helix, and 2 P. Calsou, P. Frit, and B. Salles, unpublished results. other binding, phosphorylation, or processing events could then occur at the free ends.
From our results with nuclear extracts, we cannot exclude another protein as the key target of DNA-PK-catalyzed phosphorylation for DNA end cleaning. Therefore, clarifying the protein phosphorylation events at DNA ends and their consequences deserve further experiments with both purified components of the DNA-PK complex (i.e. Ku and p460). However, the present results with nuclear extracts clearly emphasize an endogenous regulation of DNA end processing by the DNA-PKdependent control of end blocking or cleaning.
What Is the Fate of p460 Subunit after Dissociation of the DNA-PK/DNA End Complex?-Some conclusions can be drawn from immunoprecipitations of proteins/DNA end complexes by anti-DNA-PK cs Abs (Fig. 6). While p460 is mostly localized at DNA ends when the kinase is inhibited by wortmannin, it is bound to internal DNA positions under permissive conditions for the kinase activity (since the large DNA fragment is preferentially precipitated over the short end fragment; Fig. 6, lane 4). It has been shown that p460 has low selectivity for DNA ends and can bind to internal sites on DNA, but without activation (29). Whether the internal p460 DNA binding that we observe is Ku-dependent or -independent remains to be established. Nevertheless, this might indicate that either the DNA-PK cs subunit can slide inside the helix together with Ku or that it can reassociate at internally localized Ku dimers. The significance of such internal DNA-PK complexes for DNA-PK functions remains to be determined.
Has Ku Entry into the DNA Helix a Function?-Ku entry into DNA may serve different role per se. For example, Ku rigidifies DNA structure (48), is able to form a DNA loop by contact of distant Ku dimers (24), and is able to transfer from one DNA molecule to another (25). Alternatively, multiple Ku proteins bound to a single dsDNA may serve the recruitment of additional proteins. Recently, the yeast Ku homolog has been shown to interact with Sir4 protein (51), a regulatory factor in telomere transcriptional silencing that is believed to rely on a condensed heterochromatin-like DNA organization. In addition, it has been demonstrated that Sir proteins are also necessary for Ku-dependent non-homologous end joining (51,52). The spreading of several Ku dimers possibly integrated in multiprotein complexes on DNA may have various consequences on local DNA metabolism. We have shown that Ku is the major determinant of the inhibition of nucleotide excision repair on linear DNA (49,53). Strikingly, we found that repair inhibition could be released in the presence of wortmannin. 2 This may indicate that repair inhibition in the presence of double strand breaks actually relies at least on Ku binding at internal sites in DNA. An identical inhibitory property of Ku has been described since, in EMSA, purified Ku complex was found to strongly compete with transcription factors for their sequence specific binding (50). In addition, in an in vitro transcription assay, Ku was found to inhibit transcription from linear but not from circular template DNA (50).
What Role May Serve a Stable DNA-PK/DNA End Complex for End Processing?-This reminds of the model established for poly(ADP-ribose) polymerase in the repair of single-strand DNA breaks. In the presence of NAD, the DNA-bound enzyme automodifies and dissociates; if the enzyme is not released, it actually inhibits DNA repair (54,55). The transient binding of the enzyme may avoid aberrant recombination initiations at sites of DNA breaks (56). Similarly, a transient DNA-PK/DNA end complex may, for example, protect DNA ends from unspecific degradation or prevent aberrant recombination during the time interval required for downstream events to take place. In addition, this complex might favor end bridging by Ku as described recently (26, 27) by transiently stabilizing Ku location . Lower panel, incorporated radioactivity by autoradiography of the dried gel. B, preincubation was as in A except that the circular competitor pBS was omitted and linear plasmid DNA, ATP, and wortmannin concentrations were adjusted as detailed under "Experimental Procedures." The presence of ligated forms of the substrate (dimer, trimer, tetramer) were detected by agarose gel electrophoresis and ethidium bromide staining. at DNA ends. Conversely, blocking the p460 subunit at the ends might help its postulated role as structural framework (5) by attracting protein actors of subsequent steps for end processing or signaling pathways. Gu et al. reported recently that wortmannin blocked rejoining of double-strand breaks but also repair of oxidatively modified ends by phosphoglycolate removal in Xenopus egg extracts (57). The authors suggest that a specific DNA-PK-catalyzed phosphorylation event might regulate end-joining pathway in Xenopus egg extracts (57). The Xenopus homologue of DNA-PK cs has been identified (58), and DNA-PK activity has been detected in Xenopus egg extracts (59,60). Then, our results in human nuclear extracts regarding the impairment by DNA-PK of DNA end enzymatic processing in the presence of wortmannin can fully explain the above results. Indeed, the phosphorylation event required for end processing as suggested by Gu et al. may not be specific for end joining and/or repair but may rather allow end cleaning from blocked DNA-PK as a prerequisite for any end-specific biochemical pathway.
Although under our experimental conditions we visualized a stable DNA-PK/DNA end complex only by inhibiting the kinase activity, such complex might exist under physiological conditions. The stability of this complex may then depend on either the DNA structure involved or the presence of regulators of DNA-PK catalytic activity. Indeed, DNA structures that bind Ku without DNA-PK activation have been described (61)(62)(63). For example, efficient processing of hairpin coding ends during V(D)J recombination requires DNA-PK cs (64,65). However, hairpin-ended DNA fails to activate DNA-PK even though DNA-PK binds hairpin ends (63). Stable interaction of DNA-PK with this DNA structure might be an important step for its processing or for downstream signaling events, possibly relying on a non-kinase activity of the p460 subunit. On the other hand, ionizing radiation-sensitive cell lines have been established that exhibit a defect in DNA end joining but have normal Ku and DNA-PK activities in vitro (67,68). However, these criteria may not be sufficient under the standard assay conditions used to conclude that DNA-PK interaction with DNA is normal, since an abnormal regulation of Ku dissociation from DNA ends would not be detected. Our model allows prediction of the existence of other potentially important actors like cellular regulators of DNA-PK that may be necessary for DNA end processing and that remain to be identified.