The Interaction between Polynucleotide Kinase Phosphatase and the DNA Repair Protein XRCC1 Is Critical for Repair of DNA Alkylation Damage and Stable Association at DNA Damage Sites*

Background: XRCC1 interacts with multiple DNA repair proteins. Results: Identification of mutant versions of XRCC1 that are defective in binding with a different single partner. Conclusion: Interaction between XRCC1 and polynucleotide kinase 3′-phosphatase is critical for the retention of XRCC1 at DNA damage sites and DNA damage repair. Significance: Insights into function of one of three DNA end processing factors that bind to the same region of XRCC1. XRCC1 plays a key role in the repair of DNA base damage and single-strand breaks. Although it has no known enzymatic activity, XRCC1 interacts with multiple DNA repair proteins and is a subunit of distinct DNA repair protein complexes. Here we used the yeast two-hybrid genetic assay to identify mutant versions of XRCC1 that are selectively defective in interacting with a single protein partner. One XRCC1 mutant, A482T, that was defective in binding to polynucleotide kinase phosphatase (PNKP) not only retained the ability to interact with partner proteins that bind to different regions of XRCC1 but also with aprataxin and aprataxin-like factor whose binding sites overlap with that of PNKP. Disruption of the interaction between PNKP and XRCC1 did not impact their initial recruitment to localized DNA damage sites but dramatically reduced their retention there. Furthermore, the interaction between PNKP and the DNA ligase IIIα-XRCC1 complex significantly increased the efficiency of reconstituted repair reactions and was required for complementation of the DNA damage sensitivity to DNA alkylation agents of xrcc1 mutant cells. Together our results reveal novel roles for the interaction between PNKP and XRCC1 in the retention of XRCC1 at DNA damage sites and in DNA alkylation damage repair.

XRCC1 plays a key role in the repair of DNA base damage and single-strand breaks. Although it has no known enzymatic activity, XRCC1 interacts with multiple DNA repair proteins and is a subunit of distinct DNA repair protein complexes. Here we used the yeast two-hybrid genetic assay to identify mutant versions of XRCC1 that are selectively defective in interacting with a single protein partner. One XRCC1 mutant, A482T, that was defective in binding to polynucleotide kinase phosphatase (PNKP) not only retained the ability to interact with partner proteins that bind to different regions of XRCC1 but also with aprataxin and aprataxin-like factor whose binding sites overlap with that of PNKP. Disruption of the interaction between PNKP and XRCC1 did not impact their initial recruitment to localized DNA damage sites but dramatically reduced their retention there. Furthermore, the interaction between PNKP and the DNA ligase III␣-XRCC1 complex significantly increased the efficiency of reconstituted repair reactions and was required for complementation of the DNA damage sensitivity to DNA alkylation agents of xrcc1 mutant cells. Together our results reveal novel roles for the interaction between PNKP and XRCC1 in the retention of XRCC1 at DNA damage sites and in DNA alkylation damage repair.
A characteristic feature of Chinese hamster ovary (CHO) xrcc1 mutant cell lines is the increased frequency of spontane-ous and DNA damage-induced sister chromatid exchanges (1)(2)(3)(4). This and other DNA repair abnormalities in xrcc1 mutant cells suggested that XRCC1 participates in base excision and single-strand break (SSB) 2 repair. The human XRCC1 gene was cloned by complementation of the hypersensitivity of CHO EM9 mutant cells to DNA alkylating agents (1). Although XRCC1 has no known enzymatic activity, it interacts with a number of proteins that are involved in the repair of base damage and single-strand breaks, suggesting that XRCC1 acts as a scaffold that coordinates the assembly and activity of DNA repair enzymes (5).
In accord with the predicted role of XRCC1 as a scaffold that coordinates activities of multiple DNA repair enzymes (5), the binding sites of many but not all interacting proteins have been mapped to discrete regions along the entire length of the XRCC1 polypeptide (Fig. 1). XRCC1 has been identified as a subunit of high molecular weight multiprotein complexes by gel filtration and immunoprecipitation experiments, suggesting that this protein can indeed simultaneously interact with multiple protein partners (28,29). However, the binding of three DNA repair enzymes, PNKP, aprataxin, and APLF, which are involved in processing DNA termini, appear to be mutually exclusive (24,26,27). This notion is supported by the characterization of two distinct XRCC1 protein complexes that contain either PNKP or aprataxin (29).
The functional and biological consequences of many of the protein-protein interactions involving XRCC1 have not been clearly delineated. Here we have used a modified version of the yeast two-hybrid genetic screen (30) to identify mutant versions of XRCC1 that are deficient in binding to PARP1, Pol␤, PNKP, or DNA ligase III␣. Notably, the PNKP interaction mutant of XRCC1 is not defective in binding to either aprataxin or APLF, enabling us to delineate functional and biological consequences of the interaction between XRCC1 and PNKP.

EXPERIMENTAL PROCEDURES
Yeast Two-hybrid Assay-A modified version of the yeast two-hybrid assay to screen for human XRCC1 mutants that are defective in interacting with partner proteins was carried out as described by Krejci et al. (30). Briefly, a DNA fragment encoding XRCC1 was subcloned into plasmid pGADT7 to generate pGADT7-XRCC1 that expresses XRCC1 as a GAL4 activation domain fusion protein. pGADT7-XRCC1 (10 g) was incubated in 0.5 ml of 50 mM sodium pyrophosphate (pH 7), 1 M NaCl, 2 mM EDTA, and 0.5 M hydroxylamine at 75°C for 30 min. After amplification through Escherichia coli TG1, the mutagenized pGADT7-XRCC1 plasmids were transformed into the haploid yeast strain PJ69-4a and each clone was replicaplated on solid media lacking leucine (ϪLeu). DNA sequences encoding PARP1, PNKP, DNA ligase III␣, APE1, and Pol␤ were amplified by PCR and then subcloned into the plasmid pGBK to generate plasmids that express the XRCC1-interacting proteins as GAL4 DNA binding domain fusion proteins. Isogenic haploid yeast PJ69-4␣ cells were transformed with pGBK derivatives encoding different XRCC1-interacting proteins followed by selection on solid media lacking tryptophan (ϪTrp). Each of the PJ69-4␣ strains expressing a different XRCC1-interacting protein was mated with the replica-plated PJ69-4a strains harboring the mutagenized XRCC1 plasmids. After the selection of diploids (ϪTrp, ϪLeu), protein-protein interactions were assessed by growth under either low stringency (15 mM 3-amino-1,2,4-triazole, ϪLeu, ϪTrp, ϪHis) or high stringency (ϪLeu, ϪTrp, ϪHis, ϪAde) medium conditions. XRCC1 plasmids that had lost the ability to interact with one of the 5 bait proteins were recovered using the Yeastmaker yeast plasmid isolation kit (Macherey-Nagel) and then sequenced to identify the mutation site(s).
Protein Purification-The DNA sequence change to replace Ala-482 with Thr was made in pFastBac-XRCC1 by site-di-rected mutagenesis and verified by DNA sequencing. E. coli DH10Bac cells containing the Bacmid (Invitrogen) were transformed with pFastBac-XRCC1A482T plasmid to generate  recombinant baculovirus encoding the A482T mutant form of  human XRCC1. DNA ligase III␣-XRCC1 complexes containing  either wild-type XRCC1 or the A482T mutant version were  purified from insect cells after co-infection with DNA ligase III␣ and XRCC1 baculoviruses as described previously (31). Recombinant human NEIL1, PNKP, and Pol␤ were purified as described previously (19,32).
Pulldown Assays-DNA fragments encoding either wildtype XRCC1 or the A482T mutant were subcloned in-frame into the GST expression vector pGSTag (Addgene). Plasmids encoding GST-APLF and GST-aprataxin were gifts from Drs. A. Yasui and S. West, respectively. After expression in E. coli, GST and GST fusion proteins were purified by glutathione affinity chromatography. Glutathione-Sepharose beads (30 l, Amersham Biosciences) liganded by either GST or GST fusion proteins (2 g of each) were incubated with purified PNKP (3 nM) as described (31). Bound proteins were detected by immunoblotting with PNKP antibody (Abcam, ab3817) after separation by SDS-PAGE. In similar experiments, glutathione-Sepharose beads (30 l, Amersham Biosciences) liganded by GST, GST-APLF, or GST-aprataxin were incubated with wild-type and mutant versions of XRCC1 that had been radiolabeled with [ 35 S]methionine by in vitro coupled transcription and translation (TNT7 Kit, Promega). Labeled polypeptides retained by the GST beads were detected in a PhosphorImager after separation by SDS-PAGE.
Magnetic Ni 2ϩ beads (10 l, Qiagen) were incubated with or without His-tagged PNKP (20 pmol) for 30 min. After washing, the beads were incubated with purified DNA ligase III␣-XRCC1 complex (5 pmol) with either wild-type XRCC1 or the A482T mutant version for 1 h. Bound proteins were detected by immunoblotting with XRCC1 antibody (GeneTex GTX70262) after separation by SDS-PAGE.
PNKP Phosphatase Activity and SSB Repair Assays-PNKP 3Ј-phosphatase activity was measured in the presence or absence of DNA ligase III␣-XRCC1 as described previously (19). To generate a SSB with 3Ј and 5Ј phosphate termini, an end-labeled 51-mer oligonucleotide containing U at position 26 was annealed to a complementary oligonucleotide and then incubated with uracil-DNA glycosylase and Fapy-DNA glycoslyase (New England Biolabs). The labeled DNA substrate with the SSB (2 pmol) was incubated with Pol␤, (50 fmol), PNKP (25 fmol), and DNA ligase III␣-XRCC1 (25 or 50 fmol) in BER buffer. Labeled oligonucleotides were separated and quantitated as before.
For kinetics studies, the 3Ј-32 P-containing SSB substrate (0.1 nM) and an increasing concentration of unlabeled 3ЈP-containing SSB substrate (2.5-160 nM) were incubated with 0.5 nM PNKP or PNKP plus DNA ligase III␣-XRCC1 (0.5 nM) for 5 min at 37°C. k m , V max , and k cat values were obtained from Lineweaver-Burk double-reciprocal plots using Sigma Plot (33).
Similar repair reactions were carried with a circular duplex DNA substrate containing a single 5-OHU residue that was generated as described previously (18,34). Briefly, the plasmid pUC19CPD containing two single-strand nicking sites on the same strand, 32 nucleotides apart, was digested with N.BstNB1 (New England Biolabs) and heated at 65°C for 10 min to dissociate the 32-mer oligonucleotide (5Ј-GCG GAT ATT AAT GTG ACG GTA GCG AGT CGC TC-3Ј) that was removed by annealing with an of excess biotinylated complementary 32-nucleotide oligonucleotide followed by binding to streptavidinagarose Dynabeads (Sigma). The gapped plasmid was purified by phenol/chloroform extraction and ethanol precipitation, and then annealed with a 5-OHU-containing 5Ј-phosphorylated 32-nucleotide oligo (5Ј-pGCG GAT ATT AAT GTG ACG G 5-OHU A GCG AGT CGC TC-3Ј) at 45°C in TE buffer containing 50 mM NaCl. After incubation with T4 DNA ligase, the 5-OHU-containing form I plasmid was purified by cesium chloride equilibrium ultracentrifugation in the presence of ethidium bromide. The presence of 5-OHU was verified by treating the plasmid with NEIL1, which converted the form I plasmid into nicked form II. Repair reactions with the circular DNA substrate were carried out under the same conditions as reactions with the linear substrate.
Generation and Characterization of EM9 Cell Lines Ectopically Expressing Wild-type and Mutant XRCC1-DNA fragments encoding wild-type XRCC1 and the A482T mutant version were subcloned into the mammalian expression vector pcDNAHisMax4 so that the human XRCC1 proteins were expressed with an N-terminal peptide encoding the Xpress TM epitope and a polyhistidine nickel-binding tag. After transfection, puromycin-resistant cells were selected and screened for XRCC1 expression by immunoblotting. Clones that stably expressed recombinant wild-type XRCC1 and the A482T mutant at levels similar to that of endogenous XRCC1 in parental AA8 cells were selected for further analysis. The steady state levels of XRCC1-interacting proteins, aprataxin, PNKP, Pol␤, and DNA ligase III␣, were determined in these cell lines by immunoblotting.
XRCC1 was immunoprecipitated from nuclear extracts (80 g) of EM9 cells (1 to 3 ϫ 10 7 cells) expressing either wild-type XRCC1 or the A482T mutant version with XRCC1 rabbit poly-clonal antibody (1 l, GeneTex GTX70262) as described previously (31). To detect proteins co-immunoprecipitated with XRCC1, membranes were stripped in PBS containing 2% SDS and 0.1 M ␤-mercaptoethanol for 30 min, washed with PBS for 30 min, and re-probed with a different antibody.
To measure clonogenic survival after DNA alkylation damage, cells were incubated with the DNA alkylating agent ethyl methanesulfonate (EMS, Sigma) for 1 h, washed and then grown in fresh media for 5 to 7 days. Colonies were counted after staining with crystal violet.
Plasmid Construction, Transfection, Laser Microirradiation, and Confocal Microscopy-The plasmid encoding PNKP as an enhanced green fluorescent fusion protein has been described previously (28). DNAs encoding wild-type XRCC1 and the A482T mutant version of XRCC1 were subcloned into the mammalian expression vector, pEYFP-EC1, so that the XRCC1 proteins were expressed as YFP fusion proteins (21). HeLa or EM9 cells (200,000) that had been allowed to adhere to a 35-mm glass bottom culture dish with a 10-mm Microwell (MatTek Corporation, Ashland, MA; number P35G-1.5-10-C) for 24 h in DMEM with 10% FBS, 1% penicillin/streptomycin, and 1% glutamate were transfected with plasmid DNA (1 g) using the Dreamfect reagent (OZ Biosciences, Marseille, France) according to the manufacturer's standard protocol. Forty-eight h after transfection, cells were subjected to localized microirradiation using a Nikon Eclipse TE2000-E microscope equipped with a CCD camera (Hamamatsu, Tokyo, Japan), a SRS NL100 nitrogen pumped dye laser with Micropoint ablation system (Photonics Instruments, St. Charles, IL) adjusted via passage through a dye to generate a wavelength of 365 nm, and a CSU10 spinning disk system (Yokogawa, Japan). Laser power was attenuated using Velocity software version 5.5.0, build 0 (Improvision/PerkinElmer Life Sciences). Specifically, 1.7% laser intensity was directed to deliver pulses to a delineated rectangular region of interest (94 ϫ 20 pixels, 0.16 m/pixel) using a Plan Fluor ϫ60/1.25 numerical aperture oil objective. Galvanometer-driven beam displacers oriented the laser beam, which fired randomly throughout the region until complete exposure was obtained; 300 nm was used for the diffraction limited spot size. At a 10-Hz repetition rate, the laser fired 3-ns pulses with a power of 0.7 nanowatt, measured at the back aperture of the ϫ60 objective at 365 nm. These conditions generate free radical-induced DNA single-strand breaks in the targeted region, but did not induce ␥H2AX foci, an indicator of double-strand breaks, in the region of laser damage (35). Throughout the experiment, adherent cells were maintained in a live cell environmental chamber or CO 2 enhancement workhead (Slonet Scientific, Segensworth, United Kingdom). Following site-directed laser damage, fluorescence recovery after photobleaching was recorded at various intervals until fluorescence returned to background levels. Data were analyzed using the Velocity software described above and "region of interest fluorescent intensity" was measured at various time intervals (35,36).

Identification of Protein-Protein Interaction
Mutants of XRCC1-Using a modified version of the yeast two-hybrid assay (30), we screened randomly mutagenized pGADT7-XRCC1 plasmids for human XRCC1 mutants that were defective in interacting with partner proteins, PARP1, Pol␤, PNKP, APE1, and DNA ligase III␣, expressed as GAL4 DNA binding domain fusion proteins. Plasmids encoding versions of XRCC1 that were defective in interacting with one partner but retained interactions with the others were selected for DNA sequence analysis. Using this approach, we identified a series of single amino acid changes that disrupted the interaction of XRCC1 with individual bait proteins (Table 1). Notably, all single amino acid changes occurred within a region of XRCC1 that had been identified previously as being required for the interaction with the specific protein (Table 1 and Fig. 1A). Because these amino acid changes only disrupt the interaction with a single partner, we conclude that they do not have a large effect on the overall folding of XRCC1.
In this study, we chose to focus on the A482T mutant of XRCC1 that is defective in interacting with PNKP but not DNA ligase III␣ (Fig. 1B) or the other XRCC1-interacting proteins tested above (data not shown). Because PNKP preferentially interacts with XRCC1 that has been phosphorylated by casein kinase II (CK2) (37), it is possible that the A482T mutant was detected in the yeast two-hybrid assays because the amino acid substitution prevented XRCC1 phosphorylation by yeast CK2.
To examine the effect of the amino acid change in the absence of phosphorylation, wild-type XRCC1 and the A482T mutant were expressed as GST fusion proteins in E. coli and then purified. In accord with the results of the yeast two-hybrid assays (Fig. 1B), the A482T mutant was less effective at binding to in vitro translated PNKP than wild-type XRCC1 (Fig. 1C, upper  panel), indicating that the amino acid change disrupts the bind-

TABLE 1 Amino acid changes within XRCC1 that disrupt interactions with protein partners
The regions of XRCC1 (middle column) that interact with the DNA repair proteins (left column), APE1 (20), PARP1 (23), Pol␤ (13), PNKP (26), and DNA ligase III␣ (9,11). The DNA sequence changes and resultant amino acid changes in XRCC1 that disrupt binding to the indicated partner protein are shown in the right column.  ing of unmodified XRCC1 with PNKP. Because the regions of XRCC1 that are involved in binding to aprataxin and APLF appear to overlap with each other and the region bound by PNKP (24 -27), we examined the interaction of these two proteins with in vitro translated wild-type XRCC1 and the A482T mutant in pulldown assays. The amino acid change in XRCC1 that reduced its binding to PNKP (Fig. 1C, upper panel) had no effect on the interaction with either GST-APLF (Fig. 1C, middle panel) or GST-aprataxin (Fig. 1C, lower panel). Because XRCC1 participates in base excision and SSB repair as a subunit of the DNA ligase III␣-XRCC1 complex, we co-expressed DNA ligase III␣ with either wild-type XRCC1 or the A482T mutant in insect cells and purified the resultant complexes ( Fig. 2A). As expected, the DNA nick ligation activity of the complex containing the XRCC1 A482T mutant was similar to that of the complex containing wild-type XRCC1 (Fig. 2B). In accord with the protein-protein interaction studies with XRCC1 alone, the binding of the DNA ligase III␣-XRCC1 complex containing the XRCC1 A482T mutant to PNKP beads was greatly reduced (less than 10%) compared with the complex containing wild-type XRCC1 (Fig. 2C).

Effect of XRCC1-PNKP Interaction on PNKP Activity and Repair Reactions Reconstituted with Purified
Proteins-In initial studies, we examined the effect of DNA ligase III␣-XRCC1 on PNKP 3Ј-phosphatase activity, which is required for the repair of single-strand breaks generated by oxidative DNA damage (38). The complex containing wild-type XRCC1 enhanced removal of the 3Ј-phosphate group by more than 6-fold at a 2:1 ratio of DNA ligase III␣-XRCC1 complex to PNKP, whereas the complex containing the A482T mutant did not significantly increase the 3Ј-phosphatase activity of PNKP under similar conditions (Fig. 3A). Thus, the interaction with DNA ligase III␣ does not prevent XRCC1 from interacting with PNKP and enhancing its 3Ј-phosphatase activity (26). To determine the mechanism by which XRCC1 stimulates PNKP, we calculated kinetic parameters for the 3Ј-phosphatase activity of PNKP in the presence and absence of the wild-type DNA ligase III␣-XRCC1 complex ( Table 2). The k m and k cat values for PNKP of 17.9 nM and 25.3 min Ϫ1 , respectively, were similar to those reported previously (19). In the presence of DNA ligase III␣-XRCC1, there was almost a 2-fold decrease in k m but only a moderate increase in k cat (Table 2). Thus, the stimulation of PNKP activity by XRCC1 is primarily due to enhanced substrate binding.
In many instances, ionizing radiation generates SSB breaks with 3Ј-phosphoglycolate or 3Ј-phosphate termini. To mimic this situation, we constructed a DNA substrate that contains a 1-nucleotide gap with 3Ј-and 5Ј-phosphate termini and examined the role of the interaction between PNKP and XRCC1 in the repair of this substrate in reconstituted reactions that require the sequential activities of PNKP, Pol␤, and DNA ligase III␣-XRCC1. The repair reactions with the DNA ligase III␣-XRCC1 complex containing wild-type XRCC1 were significantly more efficient than those with the DNA ligase III␣-XRCC1 complex containing the XRCC1 A482T mutant (about 6-fold, Fig. 3B, compare lanes 3 and 5). Thus, the XRCC1 subunit of the DNA ligase III␣-XRCC1 complex stimulates the first step of the repair reaction, removal of the 3Ј-phosphate group generating the 3Ј-OH termini required for DNA synthesis by Pol␤ and subsequent ligation by DNA ligase III␣-XRCC1.
In addition to participating in the repair of DNA singlestrand breaks, both PNKP and DNA ligase III␣-XRCC1 are components of an APE1-independent BER pathway (19). This repair pathway is initiated by the removal of an oxidized base such as 5Ј-OHU by either NEIL1 or NEIL2 (18,19). Cleavage of the resultant abasic site by the DNA glycosylase generates a 1-nucleotide gap with both 3Ј-and 5Ј-phosphate termini. In reconstituted reactions with both linear (Fig. 4A) and circular DNA substrates (Fig. 4B) containing a single 5Ј-OHU, disruption of the interaction between XRCC1 and PNKP significantly reduced the efficiency of the repair reaction (about 3-fold).
The Interaction between PNKP and XRCC1 Is Critical for Cell Survival following DNA Damage-To determine the contribution of the interaction between XRCC1 and PNKP to cell survival after DNA damage, we isolated derivatives of CHO xrcc1 EM9 cells that stably express tagged versions of either wild-type XRCC1 (Fig. 5A, lane 3) or the A482T mutant (Fig. 5A, lane 4) at levels about 2-fold higher than endogenous XRCC1 (Fig. 5A,  lane 1). As expected (7,16), expression of both wild-type XRCC1 (Fig. 5A, lane 3) and the A482T version (Fig. 5A, lane 4) resulted in increased steady levels of Pol␤ and DNA ligase III␣ but had no significant effect on the steady state levels of PNKP. Because XRCC1 and PNKP co-exist in a large multiprotein complex (28,29), we asked whether they were associated in cell extracts. In immunoprecipitation experiments, less PNKP (about 40%) was co-immunoprecipitated by XRCC1 antibody from extracts of cells expressing the XRCC1 A482T compared with cells expressing wild-type XRCC1 (Fig. 5B, compare lanes  3 and 4), whereas similar amounts of DNA ligase III␣, Pol␤, and  aprataxin were co-immunoprecipitated. As expected, expression of wild-type XRCC1 complemented the sensitivity of CHO xrcc1 EM9 cells to ethyl methanesulfonate (Fig. 5C), a DNA alkylating agent. In contrast, expression of the A482T mutant only partially corrected the DNA damage sensitivity of the xrcc1 cells. Thus, the interaction between XRCC1 and PNKP makes a significant contribution to the repair of DNA damage caused by DNA alkylating agents.

Role of the Interaction between PNKP and XRCC1 in the Recruitment and Retention of These Factors at in Vivo Damage
Sites-In agreement with published studies (28), recruitment of EYFP-tagged PNKP to localized sites of DNA damage induced by laser microirradiation was more or less absent in xrcc1 mutant EM9 cells, as compared with EM9 cells expressing wildtype XRCC1 (Fig. 6). Surprisingly, there was no difference in the initial recruitment of PNKP in EM9 cells expressing the XRCC1 A482T mutant (Fig. 6B). However, the average retention time of EYFP-PNKP was greater in the cells expressing wild-type XRCC1 (6.7 Ϯ 1.2 min) than in cells expressing the A482T mutant (3.7 Ϯ 2.8 min, p ϭ 0.0008 as determined by single factor analysis of variance), although there was noticeable variability among individual cells expressing the mutant version of XRCC1 (Fig. 6C). Thus, the interaction with XRCC1 increases the length of time that PNKP is stably associated at DNA damage sites.
To further explore the role of the XRCC1-PNKP interaction at in vivo DNA damaged sites, we performed similar experiments using HeLa cells that had been transfected with EYFP tagged versions of either wild-type XRCC1 or the A482T mutant (Fig. 7). Although there was no significant difference in the kinetics of recruitment of wild-type XRCC1 and the A482T mutant to the site of laser irradiation within the first minute (Fig. 7B), the point of maximal fluorescent intensity at the site of laser irradiation of the XRCC1 A482T mutant was reached much earlier and the retention time at the damaged site (4.1 Ϯ 1.8 and 25 Ϯ 3.5 min, respectively) was dramatically reduced compared with wild-type XRCC1 (18 Ϯ 4.2 min, p ϭ 5.5 ϫ 10 Ϫ8 and 76 Ϯ 7.0 min, p ϭ 3.9 ϫ 10 Ϫ13 , respectively) (Fig. 7C). Thus, under conditions that introduce DNA base damage and SSBs but not double-strand breaks (28), the interaction between PNKP and XRCC1 is not critical for the initial recruitment of these factors to DNA damage, but does influence the stable association of these factors at DNA damaged sites.

DISCUSSION
Although XRCC1 plays a pivotal role in DNA repair via multiple interactions with different repair enzymes (5), the functional and biological relevance of many of these interactions have not been described. In this study, we have used a modified version of the yeast two-hybrid genetic screen (30) to identify mutant versions of XRCC1 that are selectively defective in interacting with a single protein partner. Notably, the single amino acid changes that disrupt interactions with PARP1, Pol␤, PNKP, APE1, or DNA ligase III␣ occur within regions of the XRCC1 polypeptide that had been previously implicated in the specific protein-protein interaction (8 -10, 13, 20, 23, 26). Although there is atomic resolution structural information for several different regions of XRCC1 (39 -41), only the amino acid change that disrupts the interaction with DNA ligase III␣ occurs within one of these locations, the BRCT II domain (40).  NOVEMBER 9, 2012 • VOLUME 287 • NUMBER 46

JOURNAL OF BIOLOGICAL CHEMISTRY 39239
The amino acid change that disrupts the interaction with Pol␤ occurs in an unstructured region that is immediately adjacent to that which has been shown to make direct contact with Pol␤, suggesting that the interface between Pol␤ and XRCC1 extends beyond that determined by x-ray crystallography and NMR (41,42).
In this study, we have focused on a mutant version of XRCC1 that is defective in interacting with PNKP. PNKP binds to a region in XRCC1 between the two BRCT domains that is phosphorylated at multiple serine and threonine residues by CK2 (26,37). Notably, PNKP binds specifically to phosphorylated XRCC1 via an FHA domain that interacts with a multiply phosphorylated XRCC1 peptide (residues 515-526) (43). Two other XRCC1 partner proteins, APLF and aprataxin, which also participate in the processing of DNA termini, contain similar FHA domains that presumably bind in a similar manner to the phosphorylated XRCC1 peptide (29,43,44). Recently, it has been shown that unmodified and phosphorylated XRCC1 interact with different regions of PNKP but the region of unmodified XRCC1 mediating the interaction with PNKP was not identified (45). Here, we have identified an amino acid change, alanine 482 to threonine, within a region of XRCC1 that is phosphorylated at multiple sites in vivo (37) that disrupts the interaction between unmodified XRCC1 and PNKP. This confirms that the interaction interface between XRCC1 and PNKP extends beyond that involving the XRCC1-phosphorylated peptide (residues 515-526) and the FHA domain of PNKP (43). Notably, the alanine 482 to threonine change has no detectable effect on interactions with either APLF or aprataxin Thus, whereas the interactions of PNKP, APLF, and aprataxin with XRCC1 are almost certainly mutually exclusive because they bind to overlapping sites in XRCC1, the effect of replacing alanine 482 with threonine is specific for the interaction with PNKP, suggesting a specific requirement for Ala-482 within the PNKP interaction interface.
Almost all of the studies describing interactions of XRCC1 with its multiple partner proteins have utilized XRCC1 that has been expressed and purified in the absence of DNA ligase III␣. Here, we have compared the activity of wild-type and mutant versions of XRCC1 in the context of the DNA ligase III␣-XRCC1 complex and shown that, in accord with published studies for XRCC1 alone (26), the 3Ј-phosphatase activity of PNKP is stimulated by interaction with XRCC1 when it is complexed with DNA ligase III␣. The interaction of PNKP with the DNA ligase III␣-XRCC1 complex, which increases PNKP activity primarily by enhancing DNA substrate binding, also increased the efficiency of reconstituted repair reactions utilizing DNA substrates representing key DNA lesion types generated by reactive oxygen species, including nonligatable singlestrand breaks and oxidized base lesions that are repaired via a NEIL1-initiated, APE1-independent BER reaction (19).
Because both wild-type XRCC1 and the A482T mutant interact with Pol␤ and DNA ligase III␣, expression of these proteins in xrcc1 mutant cells at a level similar to endogenous XRCC1 restored the steady state levels of Pol␤ and DNA ligase III␣ to that observed in the parental wild-type cell line as expected (7,16). Furthermore, similar amounts of aprataxin were co-immunoprecipitated with wild-type XRCC1 and the A482T mutant, whereas less PNKP was co-immunoprecipitated with the A482T mutant. Thus, replacement of alanine 482 with threonine reduces the amount of XRCC1 complexed with PNKP but has no effect on the amount complexed with apraxatin (29). Taken together with our in vitro binding results, we conclude that the replacement of alanine 482 with threonine reduces but does not totally abolish the interaction between XRCC1 and PNKP, presumably because the interaction interface between the phosphorylated peptide segment in XRCC1 and the PNKP FHA domain remains intact. The reduced binding between XRCC1 and PNKP correlates with the inability of the A482T mutant to fully complement the sensitivity of xrcc1 mutant cells to DNA alkylation damage, providing a molecular explanation for previous studies showing that knockdown of PNKP levels resulted in increased sensitivity to methyl methanesulfonate (46). The increased DNA damage sensitivity of the EM9 cells expressing the XRCC1 A482T mutant was only observed at the highest doses of ethyl methanesulfonate, suggesting that PNKP The amount of coimmunoprecipitated PNKP was determined by determining the ratios of coimmunoprecipitated PNKP to immunoprecipitated wild-type and mutant XRCC1 and expressing these ratios as percentage of the ratio obtained with wild-type XRCC1. C, effect of ethyl methanesulfonate (EMS) on the survival of AA8 cells (triangle), EM9 cells (circle), EM9 cells stably expressing Xpresstagged wild-type XRCC1 (square) and EM9 cells stably expressing Xpresstagged A482S XRCC1 (ϫ). The graphs represent the compilation of 4 independent experiments with error bars representing S.D. may only participate in the repair of a subset of lesions generated by high levels of DNA alkylating agents rather than the major base adducts generated by DNA alkylation. Unlike BER initiated by the NEIL DNA glycosylases (18,19), the repair of alkylated bases, which is initiated by the methyl purine DNA glycosylase, involves APE1 not PNKP (47). At high concentrations of DNA alkylating agents, it is possible that the number of abasic sites generated either by the action of the methyl purine  NOVEMBER 9, 2012 • VOLUME 287 • NUMBER 46 DNA glycosylase or increased spontaneous loss of alkylated bases exceeds the capacity of APE1. We suggest that the cleavage of these abasic sites by NEIL DNA glycosylases or other mechanisms generates single-strand breaks with 3Ј-phosphate and/or 5Ј-hydroxyl termini that require PNKP activity prior to rejoining.

Interaction of XRCC1 with PNKP in DNA Damage Repair
There are contradictory reports as to the roles of XRCC1 and PNKP in their recruitment to DNA damaged sites. PNKP did not form H 2 O 2 -induced nuclear foci in xrcc1 mutant EM9 cells (37) and less PNKP was recruited to localized sites of damage induced by laser microirradiation in these same mutant cells (28). In contrast, PNKP was required for the efficient recruitment of XRCC1 to DNA single-strand breaks with modified 3Ј-termini (14). It is likely that differences in the experimental approaches, including the types of DNA damage, account for these apparently contradictory observations. Under the conditions used in this study, the recruitment of PNKP to localized DNA damaged sites was dependent upon XRCC1. Surprisingly, no defect in initial PNKP recruitment was observed in cells expressing the XRCC1 A482T mutant. Because the pro-tein-protein interaction is weakened rather than abolished, it is possible that any modest defect in recruitment was masked because of high levels of expression of the tagged PNKP protein following transient transfection. Interestingly, disruption of the interaction between XRCC1 and PNKP decreased the retention time of PNKP at damaged sites, suggesting that this interaction plays an important role in the stability of the multiprotein repair complex assembled at damage sites.
In our studies, XRCC1 (ϳ76 min) was retained at the damaged site much longer than PNKP (ϳ7 min). This is consistent with studies showing that, whereas Pol␤, DNA ligase III␣, PCNA, and CAF1 dissociate at a similar time, XRCC1 remains at the damaged site (48,49). Remarkably, disruption of the interaction with PNKP dramatically reduced the retention of XRCC1 at the damaged sites, suggesting that this interaction during the initial repair phase is required for the subsequent stable association of XRCC1 at the lesion. It is assumed that the fraction of XRCC1 retained at the damaged site has some as yet undefined function after completion of DNA repair. Because the association of XRCC1 with chromatin prevents its degradation (16), it is possible that the retention of XRCC1 at repaired sites ensures that the level of this key DNA repair protein corresponds to the levels of DNA damage.