End-joining of free radical-mediated DNA double-strand breaks in vitro is blocked by the kinase inhibitor wortmannin at a step preceding removal of damaged 3' termini.

Both mammalian cells and Xenopus eggs possess activities for the joining of nonhomologous DNA ends, and such activities may play a major role in double-strand break repair. In order to dissect the biochemical processing of breaks with oxidatively modified ends, vectors containing various site-specific double-strand breaks with 3′-phosphoglycolate termini were constructed and treated with Xenopus egg extracts. These vectors were rejoined by the extracts at rates 30-100 times slower than comparable 3′-hydroxyl vectors. Vectors with blunt or cohesive 3′-phosphoglycolate ends yielded single repair products corresponding to simple phosphoglycolate removal followed by ligation, while a vector with mismatched ends was also rejoined but yielded a mixture of products. Addition of the kinase inhibitors wortmannin and dimethylaminopurine not only blocked rejoining, but also suppressed phosphoglycolate removal, implying an early, essential, kinase-dependent restriction point in the pathway. The results suggest that double-strand breaks with oxidatively modified ends are repaired in Xenopus eggs by a highly conservative and stringently regulated end-joining pathway, in which all biochemical processing of the breaks is contingent on both end alignment and a specific phosphorylation event. Several lines of indirect evidence suggest DNA-dependent protein kinase as a likely candidate for effecting this phosphorylation.


Both mammalian cells and
Xenopus eggs possess activities for the joining of nonhomologous DNA ends, and such activities may play a major role in double-strand break repair. In order to dissect the biochemical processing of breaks with oxidatively modified ends, vectors containing various site-specific double-strand breaks with 3-phosphoglycolate termini were constructed and treated with Xenopus egg extracts. These vectors were rejoined by the extracts at rates 30 -100 times slower than comparable 3-hydroxyl vectors. Vectors with blunt or cohesive 3-phosphoglycolate ends yielded single repair products corresponding to simple phosphoglycolate removal followed by ligation, while a vector with mismatched ends was also rejoined but yielded a mixture of products. Addition of the kinase inhibitors wortmannin and dimethylaminopurine not only blocked rejoining, but also suppressed phosphoglycolate removal, implying an early, essential, kinase-dependent restriction point in the pathway. The results suggest that double-strand breaks with oxidatively modified ends are repaired in Xenopus eggs by a highly conservative and stringently regulated end-joining pathway, in which all biochemical processing of the breaks is contingent on both end alignment and a specific phosphorylation event. Several lines of indirect evidence suggest DNAdependent protein kinase as a likely candidate for effecting this phosphorylation.
Although most if not all organisims are able to rejoin DNA double-strand breaks by homologous recombination with an intact copy of the damaged sequence (1-3), mammalian cells (4)(5)(6), Xenopus eggs (7-9), and to a lesser extent yeast (10) and bacteria (11), also possess "end-joining" activities which are capable of rejoining broken DNA ends by mechanisms which do not involve homologous recombination and do not require a duplicate copy of the sequence. In the case of free radicalmediated double-strand breaks, such end-joining is complicated by the presence of fragmented sugars and damaged or missing bases at the termini. All of these damaged DNA moieties must be removed and replaced with normal nucleotides, and the specific manner in which they are processed will likely have important consequences for the fidelity of repair. Here we describe an in vitro system for examining such processing, and present evidence that a wortmannin-sensitive kinase plays a key role in regulating this end-joining pathway.

EXPERIMENTAL PROCEDURES
Materials-The preparation of 3Ј-phosphoglycolate-terminated plasmids will be described in detail elsewhere (12). Briefly, 5Ј-32 P-endlabeled 3Ј-phosphoglycolate-terminated single-strand fragments 8 -11 bases in length were generated by bleomycin treatment of short partial DNA duplexes (13,14), and were ligated into 10-or 11-base 3Ј-recessed ends which had been generated in a restriction enzyme-linearized plasmid (pSV56, a 5504-bp 1 derivative of pZ189 (15)) by limited exonucleolytic digestion with T4 DNA polymerase (16). Plasmid molecules which failed to incorporate the 3Ј-phosphoglycolate oligomers were eliminated by treatment with T7 DNA polymerase, which acts as a 3Ј 3 5Ј exonuclease only for 3Ј-hydroxyl-terminated DNA. The remaining 3Ј-phosphoglycolate-terminated full-length double-stranded DNA was isolated on an agarose gel.
Eggs were collected from gonadotropin-induced Xenopus laevis females (NASCO). Following activation with calcium ionophore A23187, extracts were prepared by ultracentrifugation (17) and frozen in small aliquots in liquid N 2 . Kinase inhibitors and okadaic acid were from Sigma.
End-joining Assays-The 3Ј-phosphoglycolate repair substrates (20 ng) were added to a 20-l reaction containing 15 l of egg extract and incubated, usually for 4 h at 13°C. The reaction was terminated by treatment with proteinase K (300 g/ml, 55°C, 3 h). Samples were extracted with phenol and then with chloroform and ethanol-precipitated. The samples were then treated with either DpnI or with XhoI plus BstXI and electrophoresed on denaturing 22% polyacrylamide gels. Wet gels were subjected to quantitative phosphorimage analysis. In some cases, an aliquot of uncut DNA was transfected into WM1100 (recA) cells by electroporation, and the sequences of the repair joints were determined (18).

RESULTS
Specificity and Kinetics of End-joining-Incubation of either blunt-or cohesive-end 3Ј-phosphoglycolate substrates (see Fig.  1) in Xenopus egg extracts resulted in a 5-to 20-fold increase in the number of transfectants obtained upon electroporation into Escherichia coli (data not shown), suggesting that the plasmids were recircularized during incubation in the extract. Sequence analysis revealed that nearly all (34/38) of the clones derived from extract-treated blunt-ended molecules retained all bases from both DNA ends, to give a repair joint with a (CG) 4 sequence, suggesting a mechanism involving simple phosphoglycolate removal followed by blunt-end ligation (Fig. 1). Nearly all clones (37/39) derived from the cohesive-end substrate had a 2-base-shorter (CG) 3 repair joint, consistent with a mechanism involving annealing of the cohesive ends, phosphoglycolate removal, and ligation. Analysis of radiolabeled plasmid by agarose gel electrophoresis confirmed that a small fraction of the blunt-ended 3Ј-phosphoglycolate substrate, generally 1-10% depending on the substrate and the batch of extract, was converted from linear to nicked circular and supercoiled forms (not shown).
In order to examine more directly the processing of a 3Јphosphoglycolate-terminated DNA strand, a substrate was prepared in which one strand was labeled to high specific activity near the 3Ј-phosphoglycolate terminus (see Fig. 1). By cleaving the repair substrates and products with XhoI and BstXI following incubation in the extracts, it was possible to follow 3Јphosphoglycolate removal and subsequent ligation separately. XhoI releases the unprocessed 3Ј-phosphoglycolate end as part of a 10-or 12-base fragment; phosphoglycolate removal can then be detected as a slight decrease in the mobility of this fragment on a sequencing gel, and fill-in of the recessed end as conversion of the 10-base fragment to 11-and 12-base fragments. Recircularization of the plasmid produces a 39-base XhoI/BstXI fragment for blunt-end ligation (Fig. 1D) or a 37base fragment for cohesive-end ligation (Fig. 1E).
These experiments (Fig. 2) showed unequivocally that a fraction of the initially 3Ј-phosphoglycolate-terminated DNA strands in each of the vectors was ligated to the opposite end of the break, but with much lower efficiency than a strand with a 3Ј-hydroxyl terminus. The cohesive-and blunt-ended 3Ј-phosphoglycolate vectors each yielded predominantly a single 37-or 39-base recircularization product, as predicted from sequencing studies. The cohesive-end vector also yielded trace amounts of an 18-base fragment, presumably resulting from intermolecular "head-to-head" joins ( Fig. 1F). For both substrates, phosphoglycolate termini were surprisingly persistent, with Ն85% remaining unaltered after 4 h of incubation in the extracts (Figs. 2 and 3). (Phosphoglycolates on single-strand break ends were quantitatively removed within 4 h; data not shown.) Vectors from which the 3Ј-phosphoglycolates had been removed by treatment with human abasic endonuclease (gift of B. Demple and D. M. Wilson) (19) were much more efficiently rejoined (70% recircularized within 30 min; data not shown), indicating that the low efficiency of end-joining was due solely to the presence of blocked 3Ј ends.
For a vector having mixed incompatible 3Ј-phosphoglycolate ends, one blunt and one with a 2-base 5Ј overhang (Fig. 1C), significant amounts of both the 37-and the 39-base recircularization products, as well as of the 18-base intermolecular cohesive-end-joining product, were formed (Fig. 2C). Formation of the 39-base product implies that the recessed 3Ј end had been filled in prior to ligation, while formation of the 37-base product implies that the 5Ј overhang either had been removed or had displaced two bases of the 3Ј terminus on the blunt end of the break. Most strikingly, however, the intermediate processing of the recessed 3Ј-phosphoglycolate terminus was much different for the mixed-end than for the cohesive-end vector (Figs. 2, B   FIG. 1. Terminally blocked end-joining substrates and repair products. The substrates are linearized 5.5-kilobase plasmids which have been modified by addition of phosphoglycolate ends (q) and labeled to high specific activity at the 5Ј-phosphate of the nucleotide shown (*). The blunt-and cohesive-ended substrates each yielded almost exclusively (CG) 4 and (CG) 3 repair products, respectively, while the mixed-end substrate yielded a mixture of the two. The cohesive and mixed-end substrates also yielded significant amounts of an intermolecular cohesive end-joining product (F).

FIG. 2. End-processing of site-specifically labeled linear vectors in
Xenopus egg extracts. The blunt-end (A), cohesive-end (B), or mixed-end (C) phosphoglycolate-terminated vectors (see Fig. 1) or an analogous blunt-end 3Ј-hydroxyl-terminated vector (D) (20 ng each) was incubated for 4 h (phosphoglycolate-terminated vectors) or 90 min (hydroxyl-terminated vector) at 13°C, then treated with BstXI and XhoI to release terminal fragments which were analyzed on sequencing gels. Additions to the extract were: no addition (1), 2.5 mM dimethylaminopurine (2), 30 M H7 (3), 3 M okadaic acid (4), and 2 M wortmannin (5). Lane 6 is incubation in heat-inactivated extract. Lanes 7 and 8 are 2% and 4% dimethyl sulfoxide solvent controls for wortmannin and okadaic acid, respectively. Numbers indicate the size of single-strand fragments, and pg indicates phosphoglycolate. The 10-pg and 12-pg fragments represent unprocessed phosphoglycolate ends. The 37-and 39-base XhoI/BstXI fragments result from recircularization by cohesiveand blunt-end joining, respectively (see Fig. 1), while the 18-base XhoI/ XhoI fragment results from intermolecular head-to-head cohesive-end joining. The 21-and 23-base fragments are unligated phosphoglycolateterminated BstXI cleavage products from the bottom strand, which are only lightly labeled because the oligonucleotides incorporated into the bottom strand had ϳ100-fold lower specific activity than those in the top strand; thus, as expected, ligation products of the bottom strand are not visible. The 23-mer is not visible in C because the radioactivity of the oligonucleotide used in that construction had decayed below the level of detection. The same batch of egg extract was used for all substrates. and C, and 3, B, C, and E). While the fraction of total ligated repair products was similar, the extent of 3Ј-phosphoglycolate removal was significantly greater for the mixed-end vector. Moreover, the mixed-end vector allowed both the accumulation of unligated 3Ј-hydroxyl intermediates, and partial fill-in of the 5Ј overhang, while the cohesive-end vector did not. These marked differences in processing of the same recessed 3Ј-phosphoglycolate end suggest that alignment of the opposite ends of the break must have preceded biochemical modification of the termini.
Kinetic studies (Fig. 4A) showed that the joining of terminally blocked substrates was 30 -100 times slower than joining of a substrate with normal 3Ј-hydroxyl termini. Nevertheless, extended incubation resulted in ligation of a significant fraction (up to 25%) of the terminally blocked DNA strands, implying that end-joining was not restricted to some minor contaminant in the preparation.
As shown in Figs. 2 and 3, wortmannin (2 M) as well as the broad-specificity kinase inhibitor dimethylaminopurine (2.5 mM) not only dramatically reduced end-joining of terminally blocked substrates, but also strongly inhibited phosphoglycolate removal. This was most apparent with the blunt and mixed-end vectors, where free 3Ј-hydroxyl termini (12-mer and 10-mer bands in Fig. 2, A and C) accumulated in control (lanes  1, 7, and 8) but not in inhibitor-containing samples (lanes 2 and  5). While the extent of both end-joining and phosphoglycolate removal differed for different substrates and different batches of extract, the suppression by the kinase inhibitors was quite reproducible (Fig. 3). Dimethylaminopurine and wortmannin also suppressed the joining of both blunt-and cohesive-end breaks bearing 3Ј-hydroxyl termini (Fig. 3, D and F-J), and this suppression was expressed within 10 min of inhibitor addition. Half-maximal suppression of rejoining required 0.1-0.3 M wortmannin (Fig. 4, B and C), comparable to values reported for inhibition of wortmannin-sensitive protein kinases (20,21). H7 had no effect on end-joining of any of the substrates. Thus, the results suggest that the end-joining pathway requires a specific phosphorylation event, catalyzed by a PIKrelated or other wortmannin-sensitive kinase, as a prerequisite to all biochemical processing of double-strand break ends, blocked or unblocked.
Effects of the broad-specificity phosphatase inhibitor okadaic acid were complex and less definitive. It had no effect on joining of a 3Ј-hydroxyl-terminated vector, but it did partially inhibit joining of phosphoglycolate-terminated vectors. DISCUSSION It has long been postulated that double-strand break repair in higher eukaryotes may be carried out at least in part by simple end-joining rather than by homologous recombination (24,25). Recently, support for this hypothesis has come from the discovery that in mammalian cells DNA-dependent protein kinase (DNA-PK, a PIK-related kinase) and its essential cofactor, the DNA end-binding heterodimer Ku, are required both for efficient repair of radiation-and bleomycin-induced doublestrand breaks and for ostensibly similar end-joining events which occur during V(D)J recombination (26 -32).
Although the extent of overlap between this putative mammalian end-joining repair system and the end-joining activity detected in Xenopus egg extracts is unknown, the present re- with vectors having 3Ј-hydroxyl ends and either 2-base (CG) (I) or 4-base (CGCG) (J) cohesive 5Ј overhangs. These were generated by cleavage of pSV56 with BsaHI or MluI, respectively, followed by 5Ј-end-labeling; analysis was as in Fig. 2. The dimethylaminopurine concentration was 2.5 mM in all cases. The wortmannin concentration was 2 M or 10 M (*). In all cases, the fraction rejoined represents the sum of all end-joined products divided by the total radiolabel in each lane. Since removal of 3Ј-phosphoglycolates must necessarily precede rejoining, the percent phosphoglycolate removal is calculated as the sum of all free 3Ј-hydroxyl ends plus the sum of all end-joined products, divided by the total radiolabel in the lane. DMAP ϭ dimethylaminopurine; ND ϭ not done. sults show that the Xenopus system, like its mammalian counterpart, involves a wortmannin-sensitive kinase, and is capable of processing DNA breaks with oxidatively modified ends. The apparent presence of DNA-PK activity in Xenopus egg extracts (33), its known association with DNA ends, and its sensitivity to dimethylaminopurine (34), are consistent with its being a participant in end-joining. Half-maximal inhibition of protein phosphorylation by isolated human DNA-PK requires 0.1-0.25 M wortmannin (21), similar to concentrations required to inhibit end-joining in vitro (Fig. 4, B and C). These data are clearly more compatible with involvement of a wortmanninsensitive protein kinase than of PIK itself, which is inhibited at ϳ100-fold lower concentrations of wortmannin (20). Nevertheless, several PIK-related kinases have been linked to DNA damage/repair responses (35), and it is certainly possible that the wortmannin sensitivity of end-joining in Xenopus egg extracts may reflect involvement of a kinase(s) other than DNA-PK.
Analysis of end-joining events in both Xenopus egg extracts and mammalian cells has suggested a model wherein DNA ends are first aligned (often on the basis of partial complementarities in 5Ј or 3Ј overhangs), then trimmed and patched as necessary to produce a ligatable substrate (4,9). The present results further suggest that at least in the Xenopus system, not only patching and ligation, but also removal of 3Ј-terminal blocking groups, are all stringently regulated by phosphorylation. Since terminal processing seems to be also dependent on prior completion of the end alignment step (Fig. 2), we speculate that perhaps the wortmannin-sensitive phosphorylation event is contingent on end alignment, and that one of its functions may be to delay biochemical processing of the DNA ends until they have been optimally aligned. Since free radicalmediated double-strand breaks will often have staggered ends with missing bases and damaged termini, the capacity to use partial complementarities for such alignment, along with the capacity to remove and replace fragmented as well as mismatched nucleotides (apparently without disturbing end alignment), makes the Xenopus system seem ideally suited for repair of these complex lesions. Whether end-joining of doublestrand breaks in mammalian chromosomal DNA is carried out by a similar mammalian homologue of this system, remains to be determined; however, the finding that the Xenopus pathway can be arrested in vitro, by kinase inhibitors and by modified DNA ends, should greatly improve prospects for isolation of an intact, active end-joining complex, and for definitive identification of its constituents.