DNA Ligase IV and XRCC4 Form a Stable Mixed Tetramer That Functions Synergistically with Other Repair Factors in a Cell-free End-joining System*

Repair of DNA double-strand breaks in mammalian cells occurs via a direct nonhomologous end-joining pathway. Although this pathway can be studied in vivo and in crude cell-free systems, a deeper understanding of the mechanism requires reconstitution with purified enzymes. We have expressed and purified a complex of two proteins that are critical for double-strand break repair, DNA ligase IV (DNL IV) and XRCC4. The complex is homogeneous, with a molecular mass of about 300,000 Da, suggestive of a mixed tetramer containing two copies of each polypeptide. The presence of multiple copies of DNL IV was confirmed in an experiment where different epitope-tagged forms of DNL IV were recovered simultaneously in the same complex. Cross-linking suggests that an XRCC4·XRCC4 dimer interface forms the core of the tetramer, and that the DNL IV polypeptides are in contact with XRCC4 but not with one another. Purified DNL IV·XRCC4 complex functioned synergistically with Ku protein, the DNA-dependent protein kinase catalytic subunit, and other repair factors in a cell-free end-joining assay. We suggest that a dyad-symmetric DNL IV·XRCC4 tetramer bridges the two ends of the broken DNA and catalyzes the coordinate ligation of the two DNA strands.

DNA double-strand breaks (DSBs) 1 are produced when ionizing radiation creates closely spaced single-strand breaks on opposite strands of a duplex DNA. DSBs can also be produced by the action of radiomimetic drugs and by certain recombination endonucleases. DSBs are among the most lethal forms of DNA damage. Unrepaired DSBs create acentric chromosomal fragments that are lost in subsequent cell divisions, leading to aneuploidy or cell death. All eukaryotic cells have efficient mechanisms for DSB repair. This repair can proceed by a recombination mechanism, mediated by the Rad51 protein and other gene products, or it can proceed via a nonhomologous end-joining pathway (1)(2)(3)(4)(5).
In human cells, NHEJ involves the concerted action of several proteins. The heterodimeric Ku protein carries out the initial recognition of broken DNA ends, forming a platform for the recruitment of other repair proteins (6 -12). The DNA-dependent protein kinase catalytic subunit binds to the initial Ku-DNA complex, inducing an inward translocation of Ku, and leaving DNA-PKcs in direct contact with the DNA termini (8,13). The next steps in the repair pathway are not as well understood. A complex of three proteins, MRE11, RAD50, and NBS1, may be involved in resection of the ends to remove damaged nucleotides and to expose regions of microhomology on opposite sides of the DNA break (14 -18). A complex of DNA ligase IV and the accessory protein, XRCC4, carries out the actual ligation (19 -22). There may also be involvement of other, as yet unidentified, proteins that are required for assembly of the repair complex, anchoring of the free DNA ends, or coupling of repair to cellular signaling pathways. The biochemistry of the NHEJ reaction remains incompletely understood. Several systems have been described that permit re-creation of the reaction in cell free extracts from somatic cells or Xenopus eggs (23)(24)(25)(26)(27). However, crude systems have a limited potential for mechanistic studies, and it is therefore an important goal to obtain all of the components of the NHEJ system in pure and active form.
The current study focuses on the DNA ligase component. DNL IV is one of three isoforms of nuclear DNA ligase found in human cells (28 -30). DNA ligase I is a replicative ligase required for the joining of lagging-strand DNA fragments, and DNA ligase III is required for nucleotide excision repair (31)(32)(33)(34). The only known function of DNL IV is in NHEJ (20,35,36). Although all three ligases are ATP-dependent and share a conserved catalytic core, DNL IV has several unique features (30,34,37,38). Most significantly, DNL IV has an obligate partner protein, XRCC4, is required for biological function (39 -41). Cells lacking either DNL IV or XRCC4 are radiationsensitive and are unable to complete V(D)J recombination, which involves a DSB intermediate and is required for the development of the adaptive immune system (20,22,42); these phenotypes are similar to those arising from mutations in other genes in the NHEJ pathway (43)(44)(45)(46)(47). Mice lacking DNL IV or XRCC4 display an additional phenotype of a high level of apoptosis in the developing nervous system (22,42). The reason for this phenotype is unclear, but it could be due to high sensitivity to unrepaired DSBs in developing neurons.
The native molecular weight of the DNL IV⅐XRCC4 complex is uncertain and has been the subject of conflicting estimates (29,39). Moreover, although XRCC4 stabilizes DNL IV and enhances its enzymatic activity, it is not clear if it plays a fundamental role in the reaction mechanism, and if so, why DNL IV requires this cofactor and other ligases do not. In the present study, we use multiple, independent approaches to demonstrate that the active form of the DNL IV⅐XRCC4 complex is a heterotetramer. Cross-linking suggests that an XRCC4 dimer forms the structural core of the complex and is therefore responsible for maintaining the tetrameric state. We suggest that the presence of two ligase active sites within the complex is an adaptation that permits the two strands of DNA to be ligated in a coordinated manner.

Cloning of DNL IV and XRCC4 cDNAs and Creation of Recombinant
Baculoviruses-DNL IV (accession no. AA081632) and XRCC4 (accession no. R14027) cDNA clones were obtained from the Image Consortium (clone ID IMAGE 547947 (5Ј) and IMAGE 26811 (5Ј), respectively). The DNL IV clone contained the complete coding sequence, which was amplified by PCR using Pfu DNA polymerase (Stratagene, La Jolla, CA), primer 1, d(GATCGTCGACACCATGGCTGCCTCACAAACT), and primer 2, d(AATTGTCGACTTAATGATGATGATGATGATGAATCAA-ATACTGGTTTTC). The primers introduced a 6-histidine tag at the 3Ј end of the open reading frame and SalI sites at both ends of the amplified fragment. This fragment was digested with SalI and inserted into the SalI site of pCITE-4a(ϩ) (Invitrogen, Carlsbad, CA). The resulting plasmid was digested with BamHI and NotI, and the DNL IV-encoding fragment was inserted at the corresponding sites of the pVL1392 baculovirus transfer vector (PharMingen, San Diego, CA). The DNL IV(His) 6 polypeptide has a predicted molecular mass of 104,747 Da.
In some experiments, where noted, FLAG and Myc epitope-tagged variants of DNL IV were used. To construct these, a PCR reaction was performed using the DNL IV-His 6 clone in the pCITE-4a(ϩ) vector as template, primer 1, and primer 3, d(CGGACGCGTGAATCAAATACT-GGTT). These amplify the DNL IV coding sequence with deletion of the His 6 tag and the addition of an MluI site at the 3Ј end. To construct the FLAG-tagged variant, the PCR product was inserted into pCR-Blunt II-TOPO (Invitrogen, Carlsbad, CA), and the resulting clone was digested with EcoRI and subcloned into pCITE-4a(ϩ) to provide flanking restriction sites. A short DNA fragment downstream of the 3Ј end of the DNL IV gene was excised by digestion with MluI and NotI, and a double-stranded oligonucleotide consisting of d(CGCGTCCGACTA-CAAGGACGACGATGACAAGTAAGC) and d(GGCCGCTTACTTGT-CATCGTCGTCCTTGTAGTCGGA) was inserted. This results in the addition of a FLAG epitope (DYKDDDDK) at the C terminus of DNL IV. The resulting plasmid was digested with BamHI and NotI, and the DNL IV-FLAG encoding fragment was inserted into the corresponding sites of the pVL1393 baculovirus transfer vector (PharMingen). To construct the Myc-tagged variant, PCR was performed using DNL IV-(His) 6 in pCITE-4a(ϩ), primer 1, and primer 3. The amplified fragment was digested with SalI and MluI and was inserted into the corresponding sites of the pMS211 vector (48). This results in the addition of three tandem Myc epitope sequences (EQKLISEEDL) at the C terminus of the protein. The resulting plasmid was digested with SalI and NotI to release the DNL IV-Myc coding fragment, which was inserted into the corresponding sites of the pCITE-4a(ϩ) vector. A BamHI-NotI fragment was subcloned into pVL1393 as described above.
The XRCC4 clone obtained from the Image Consortium was sequenced, and it was determined that this clone contained only the 3Ј region of the gene. In an attempt to obtain the entire gene, total HeLa cell RNA was prepared using TRIzol (Life Technologies, Inc.) and reverse transcription-PCR was performed using SuperScript II Reverse Transcriptase and Taq polymerase (Life Technologies, Inc.). The primers used for this experiment, d(GATCGGATCCACCATGGGCTAC-CCATACGATGTTCCAGATTACGCTCATATGGAGAGAAAAATAAGC) and d(AGTGGGATCCTTAAATCTCATCAAAGAG), amplify the entire XRCC4 open reading frame and introduce a hemagglutinin epitope tag at the 5Ј end (YPYDVPDYA). The amplified fragment was inserted into the pCR2.1 vector (Invitrogen, Carlsbad, CA). DNA sequencing showed that several base changes had been introduced into the 3Ј region of the XRCC4 gene. To correct these, the plasmid containing the amplified XRCC4 gene was digested with EcoRI, and the resulting fragment was subcloned into pCITE-4a(ϩ). A PstI-NotI fragment encompassing the 3Ј region of XRCC4 was excised and replaced with a corresponding fragment from the Image Consortium XRCC4 clone. The resulting plasmid was digested with BamHI and NotI, and the XRCC4-encoding fragment was subcloned into the corresponding sites of the pVL1393 baculovirus transfer vector (PharMingen). The HA-XRCC4 polypeptide has a predicted molecular mass of 39,389 Da.
In some experiments, where noted, a FLAG-tagged variant of XRCC4 was used. The plasmid containing the complete HA-XRCC4 gene in pCITE-4a(ϩ) was digested with NdeI and NotI to excise the XRCC4 coding sequence (removing the HA tag), and the resulting fragment was subcloned into the corresponding sites of the pSK277 baculovirus transfer vector (49), resulting in the addition of a FLAG tag at the N terminus.
To generate viral stocks, transfer vectors encoding DNL IV and XRCC4 were transfected, together with linearized AcNPV baculovirus DNA (PharMingen), into Sf9 insect cells using a liposome-mediated method (Cellfectin, Life Technologies, Inc.). Recombinant baculovirus stocks were amplified separately to a titer of at least 5 ϫ 10 7 plaqueforming units/ml. Protein expression was confirmed by immunoblotting using anti-histidine antibody (mAb Tetra-His; Qiagen, Valencia, CA) and anti-HA antibody (mAb 12CA5; Roche Molecular Biochemicals).
Purification of the DNL IV⅐XRCC4 Complex-Recombinant DNL IV and XRCC4-expressing baculoviruses were used to co-infect 3 L of Sf9 cells at a multiplicity of infection of 5. After 60 h, Sf9 cells were collected by centrifugation and suspended in 90 ml of lysis buffer (50 mM NaH 2 PO 4 , pH 8.0, 500 mM NaCl, 5 mM ␤-mercaptoethanol, 10% glycerol). The suspension was sonicated and then centrifuged for 1 h at 150,000 ϫ g. The supernatant was mixed with 10 ml (packed volume) of Ni ϩ -NTA-agarose (Qiagen, Valencia, CA) and incubated for 4 h at 4°C with rotation. The Ni ϩ -NTA-agarose was packed into a column, washed with lysis buffer, and eluted with a linear gradient of imidazole (0 -500 mM) in lysis buffer. Fractions were analyzed by SDS-PAGE, and the presence of DNL IV and XRCC4 was determined by immunoblotting. To purify FLAG-tagged DNL IV⅐XRCC4, which was used in one experiment where noted, recombinant DNL IV-FLAG and FLAG-XRCC4-expressing baculoviruses were used to co-infect Sf9 cells. The pellet from 0.5 liter of culture was suspended in 30 ml of lysis buffer (as described in the preceding paragraph except with 0.1 M NaCl), sonicated, and centrifuged. The supernatant was loaded onto a 5-ml Q-Sepharose FF column (Amersham Pharmacia Biotech), equilibrated with lysis buffer containing 0.1 M KCl. The column was washed and eluted with lysis buffer containing 0.5 M NaCl. Pooled DNL IV-XRCC4containing fractions were loaded onto a Superdex 200 (16/60) column (Amersham Pharmacia Biotech) that had been pre-equilibrated with lysis buffer containing 0.35 M NaCl. Fractions were pooled that contained the DNL IV⅐XRCC4 complex (excluding fractions containing free XRCC4), and these were subjected to Mono Q chromatography.
Determination of Native Molecular Weight-Gel filtration chromatography was performed using a Superdex 200 (26/60) column (Amersham Pharmacia Biotech) using TM buffer containing 0.275 M KCl. To construct a calibration curve, a set of standard proteins was analyzed (high and low molecular weight gel filtration calibration kit, Amersham Pharmacia Biotech). The K av parameter was determined (K av where V e represents elution volume, V o represents void volume, and V t represents total bed volume). The K av values for standard proteins were plotted as a function of the logarithm of molecular weight, and the resulting calibration curve was used to derive a molecular weight for the DNL IV⅐XRCC4 complex.
Native molecular weight was also determined by the method of Ferguson (50). Native gel electrophoresis was performed in a buffer containing 50 mM Tris base, 40 mM boric acid, and 10 mM EDTA. Standard and unknown proteins were analyzed on a series of gels containing different concentrations of polyacrylamide (4%, 5%, 5.5%, 6%, 8%, and 10%). Relative mobility (R f ) was calculated as the ratio of mobility of a given protein to the mobility of bromphenol blue dye. For each protein, a derived value, 100[log (R f ϫ 100)], was plotted as a function of acrylamide concentration (51,52). The slopes of the lines generated for each standard protein were determined by linear regression analysis and were plotted as a function of the logarithm molecular weight. This calibration curve (a "Ferguson plot") was applied to determine the native molecular weight of the DNL IV⅐XRCC4 complex.
Immunoprecipitation of Myc-tagged DNL IV-Antibody-conjugated Sepharose was prepared by incubation of purified mAb 9E10 (gift of M. Sadofsky, Medical College of Georgia) with CNBr activated Sepharose 6MB (Amersham Pharmacia Biotech). Coupling was performed overnight at 4°C in a buffer containing 0.1 M NaHCO 3 , pH 8.3, and 0.5 M NaCl. The reaction was stopped by washing with the same buffer lacking antibody. The beads were blocked by incubating with 1.2 M ethanolamine for 2 h at room temperature. The beads were washed in 0.1 M NaOAc, pH 4.0, containing 0.5 M NaCl, then with 0.1 M Tris-HCl, pH 8.0, containing 0.5 M NaCl. This cycle of washing was repeated twice more.
A triple coinfection was performed using 1 liter of Sf9 cells and baculoviruses encoding the DNL IV-His 6 , DNL IV-Myc, and HA-XRCC4 polypeptides. Each virus was used at a multiplicity of infection of about 5. An extract was prepared and subjected to Ni ϩ -NTA-agarose chromatography as described above. Protein from a peak fraction was incubated with mAb 9E10 anti-Myc Sepharose overnight at 4°C in a buffer containing 0.5 M NaCl, 50 mM Tris-HCl, pH 7.4, and 0.3% Nonidet P-40. The beads were washed three times with the same buffer, and bound proteins were analyzed by SDS-PAGE.
Adenylation and Nick-ligation Assays-Adenylation reactions were performed in a final volume of 10 l. Reactions contained 2 l of a ligase-containing protein fraction, and in addition contained 60 mM Tris-HCl, pH 8.0, 10 mM MgCl 2 , 5 mM dithiothreitol, 50 g/ml bovine serum albumin, and 1 Ci of [␣-32 P]ATP (6000 Ci/mmol, PerkinElmer Life Sciences). Reactions were incubated for 10 min at room temperature and terminated by addition of SDS-PAGE sample buffer. Adenylated proteins were resolved by SDS-PAGE and visualized by Phospho-rImager analysis.
Nick-ligation reactions were performed using a (dT) 16 substrate, which was 5Ј end-labeled using T4 polynucleotide kinase and [␥-32 P]ATP (6000 Ci/mmol). The radiolabeled (dT) 16 was hybridized with a stoichiometric amount of poly(dA) (200-nt average length, Amersham Pharmacia Biotech). Ligation reactions were performed in a volume of 5 l, and contained 1 ng of DNA substrate, 60 mM Tris-HCl, pH 8.0, 10 mM MgCl 2 , 5 mM dithiothreitol, 50 g/ml BSA, and 1 mM ATP. In addition, reactions contained 1 l of ligase-containing fraction, which was diluted with TM buffer containing 0.275 M KCl to give the amount of protein indicated in the legend to Fig. 6. Reactions were incubated at 37°C for 10 min and stopped by addition of 95% formamide loading dye. Products were analyzed by 20% urea-PAGE. The gel was fixed with 10% methanol, 10% acetic acid, and the products were visualized by PhosphorImager analysis.
Chemical Cross-linking and Detection of Cross-linked Complexes by Immunoblotting-Chemical cross-linking reactions were performed using a homobifunctional N-hydroxysuccinimide ester cross-linker that had an 11.4-Å arm length (BS 3 , Pierce). Reactions were assembled by mixing 10 l of protein-containing fractions, 10 l of phosphate-buffered saline (27 mM KCl, 1 mM KH 2 PO 4 , 137 mM NaCl, 43 mM Na 2 HPO 4 ), and 2 l of aqueous BS 3 solution. Final BS 3 concentrations were as indicated in the relevant figure. Reactions were incubated for 1 h at room temperature and terminated by adjusting to 200 mM Tris-HCl, pH 8.0. Cross-linked products were resolved by 5% and 7% SDS-PAGE and analyzed by immunoblotting, using anti-histidine tag (mAb Tetra-His; Qiagen) and anti-HA (mAb 12CA5; Roche Molecular Biochemicals) primary antibodies. Immunoblots were incubated with the appropriately conjugated anti-mouse secondary antibodies and were developed as indicated in the legend to Fig 5. Cell-free Nonhomologous End-joining Assay-Extracts were prepared and assayed essentially as described (25). Whole cell extracts were prepared from 4 ϫ 10 9 human lymphoblasts (GM00558 cell line; Coriell Cell Repositories, Camden, NJ). Cells were harvested by centrifugation, resuspended, and lysed by homogenization in 10 ml of hypotonic lysis buffer (10 mM Tris-HCl, pH 8.0, 0.1 mM EDTA, 5 mM dithiothreitol) using 20 strokes of an A pestle. Proteinase inhibitors (0.17 mg/ml phenylmethylsulfonyl fluoride, 1 g/ml pepstatin A, 1 g/ml leupeptin, 1 g/ml soybean trypsin inhibitor) were added. After 20 min incubation on ice, 3.33 ml of high salt buffer (50 mM Tris-HCl, pH 7.5, 1 M KCl, 2 mM EDTA, 2 mM dithiothreitol) was added. Cell lysate was centrifuged for 3 h at 200,000 ϫ g. The supernatant was dialyzed against 20 mM Tris-HCl, pH 8.0, 0.1 M KOAc, 20% glycerol, 0.5 mM EDTA, and 1 mM dithiothreitol. Aliquots were prepared and stored at Ϫ70°C.
DNA substrate for the NHEJ assay was prepared by digestion of pBluescript II KS(ϩ) vector (Stratagene) with BamHI. The DNA substrate was labeled with T4 polynucleotide kinase and [␥-32 P]ATP (6000 Ci/mmol). NHEJ reaction mixtures contained 10 mM Tris-HCl, pH 7.9,  4 OAc in ethanol was added (500 l), the nucleic acid precipitate was collected by centrifuga-tion, and reaction products were analyzed by 0.6% agarose gel electrophoresis. Ligated products were detected by PhosphorImager analysis.

RESULTS
Purification of a DNL IV⅐XRCC4 Complex-The overall cloning scheme for human DNL IV and XRCC4 is described under "Experimental Procedures." The human DNL IV cDNA was obtained by PCR amplification of a single IMAGE consortium clone containing the full-length gene sequence. Two different For each protein, a K av parameter was derived as described under "Experimental Procedures." The K av for the DNL IV⅐XRCC4 complex is indicated by the arrow. B, native gel electrophoresis (Ferguson plot) analysis. Electrophoresis was performed as described under "Experimental Procedures." Proteins used as molecular size standards were ␣-lactalbumin (14,200 Da), carbonic anhydrase (29,000 Da), ovalbumin (45,000 Da), BSA monomer (66,000 Da), BSA dimer (132,000 Da), urease trimer (272,000 Da), and urease hexamer (544,000 Da). Each standard and unknown protein was run on a series of gels with different polyacrylamide concentrations ranging from 4% to 10%. Relative mobility (R f ) was calculated as the ratio of mobility of a given protein to the mobility of bromphenol blue dye. For each protein, a derived value, 100[log(R f ϫ 100)], was plotted as a function of acrylamide concentration (data not shown). The slopes of the lines generated for each standard protein were determined by linear regression analysis, and the resulting numerical values were plotted as a function of molecular weight to generate the calibration curve (Ferguson plot) shown in panel B. The position of the DNL IV⅐XRCC4 complex on this curve is indicated by the arrow .   FIG. 3. Immunoblotting of DNL IV⅐XRCC4 complex with anti-FLAG antibodies. Flag-tagged variants of the DNL IV and XRCC4 polypeptides were co-expressed in Sf9 cells and the DNL IV⅐XRCC4 complex was purified as described under "Experimental Procedures." A, immunoblotting of purified DNL IV⅐XRCC4 complex with anti-FLAG antibody (Sigma-Aldrich). Increasing amounts of DNL IV⅐XRCC4 were loaded in each lane, as indicated in nanograms. Immunoblotting was performed using polyvinylidene difluoride membrane with detection by enzyme-catalyzed fluorescence substrate (Vistra ECF substrate, Amersham Pharmacia Biotech). The immunoblot was imaged using a Storm system (Molecular Dynamics). DNL IV and XRCC4 bands were identified by migration relative to prestained molecular size markers (data not shown) and are denoted by arrows. B, fluorescence of each band was quantitated using the Storm system. AUG codons have been proposed as sites of initiation of translation for DNL IV, making this an enzyme of either 844 or 911 residues (28,39,53). We generated constructs expressing both forms, each with a hexahistidine tag introduced at the C terminus (see "Experimental Procedures"). Preliminary characterization revealed that only the 911-residue form was active in an AMP-adenylation assay (data not shown), in agreement with a previous report (53). Based on the difficulty encountered in our attempts to purify the shorter form, we speculate that the N-terminal region contributes to overall folding of the protein. All subsequent experiments were performed with the longer, 911-residue form. The human XRCC4 cDNA clone was assembled from two sources. The 5Ј portion of the gene was obtained by reverse transcriptase-PCR using HeLa cell RNA, and the 3Ј portion was obtained from an IMAGE consortium clone, as described under "Experimental Procedures." Both the DNL IV and the XRCC4 clones were transferred into baculovirus vectors, which were then amplified to create recombinant viral stocks.
Baculoviruses containing the DNL IV and XRCC4 cDNAs were used to co-infect Sf9 cells, and protein purification was carried out according to the scheme shown in Fig. 1A. After lysis, the infected cell extract was subjected to immobilized metal ion chromatography using Ni ϩ -NTA-agarose. The histidine-tagged DNL IV binds directly to this resin, and the XRCC4 binds indirectly because of its interaction with DNL IV. The Ni ϩ -NTA-agarose column was eluted with a gradient of imidazole, and the DNL IV⅐XRCC4-containing fractions were pooled and applied to a Superdex-200 gel filtration column. DNL IV and XRCC4 eluted together from this column in a single well resolved peak, indicating the presence of a homogeneous, stable complex (Fig. 1B). These fractions were pooled and applied to a Mono Q anion exchange column, from which DNL IV and XRCC4 again eluted together in a single sharp peak (Fig. 1C).
Determination of Native Molecular Weight and Stoichiometry of DNL IV⅐XRCC4 Complex-It was of interest to determine the number of copies of each polypeptide that were present within the DNL IV⅐XRCC4 complex. By definition, DSB repair requires a mechanism for coordinated ligation of the two DNA strands, and one straightforward way to achieve this would be if two copies of the DNL IV polypeptide were present in the complex.
We used two complementary approaches to determine the native molecular weight of the DNL IV⅐XRCC4 complex. In the first of these, we calibrated the Superdex 200 gel filtration column used in the DNL IV⅐XRCC4 purification by determining the K av values for a set of standard proteins of known molecular weight. A calibration curve based on these K av values is shown in Fig. 2A. Comparison of the K av for DNL IV⅐XRCC4 versus the standards suggests a native molecular mass for the complex of approximately 310,000 Da.
As an independent method of determining native molecular weight, we used native gel electrophoresis. The mobility of a given protein on a native gel is a function of both size and charge, making it impossible to determine molecular weight from a single measurement. However, the mobility of the same protein on gels of different porosities varies in a systematic way, which can be related to molecular weight. We measured relative mobility of the DNL IV⅐XRCC4 complex and a group of standard proteins using a set of native gels containing from 4% to 10% polyacrylamide. Data were analyzed as described under "Experimental Procedures." The log-linear relationship between mobility and polyacrylamide concentration was determined for each standard and unknown protein, and the slopes of the best-fit lines were plotted as a function of the logarithm of molecular weight to obtain the calibration curve shown in Fig. 2B. The native molecular mass of the DNL IV⅐XRCC4 complex determined by this method is 290,000 Da, in good agreement with the results of gel filtration. Together, the gel filtration and native gel electrophoresis data are consistent with the existence of a tetrameric complex, (DNL IV) 2 (XRCC4) 2 , which has a predicted molecular mass of 288,272 Da, although the precision of the experiments is such that we cannot rule out other alternatives, such as (DNLIV) 2 (XRCC4) 3 .
As another measure of the stoichiometry of DNL IV and XRCC4, we quantitated the relative amount of each polypeptides using two independent methods. In the first, purified DNL IV⅐XRCC4 was analyzed by SDS-PAGE, stained with Coomassie Blue, and the amount of each stained band was quantitated by densitometry. The results suggested a molar ratio of XRCC4 to DNL IV of 1.03-1.19, as measured in two independent experiments (Table I). Within the likely experimental error, this is most consistent with a composition of (DNL IV) 2 (XRCC4) 2 .
Although Coomassie Blue staining is a widely accepted method, it is known that certain polypeptides may be relatively overstained or understained. Therefore, we also used an alternative method of detection, which involved re-cloning DNL IV and XRCC4, each with an identical FLAG epitope tag. The two polypeptides were co-expressed in baculovirus-infected Sf9 cells, and the complex was purified by sequential chromatography on Q-Sepharose, Superdex 200, and Mono Q columns as described under "Experimental Procedures." The last two of these columns effectively resolve the DNL IV⅐XRCC4 complex from free XRCC4, which was also present in the co-infected Sf9 cells. Fractions corresponding to the complex were pooled and analyzed by immunoblotting with anti-FLAG primary antibody and alkaline phosphatase-conjugated secondary antibody. The immunoblot was developed with a fluorescent substrate and visualized using a Storm imaging system (Fig. 3A). Because both polypeptides carry the same epitope, the fluorescence signal is proportional to relative molar amounts of each. Quantitation of this signal revealed that the stoichiometry of DNL IV and XRCC4 is 1:1, within an error of about 10% (Fig. 3B).
Co-immunoprecipitation of His 6 -and Myc Epitope-tagged DNL IV Polypeptides-Because of the uncertainty inherent in the measurement of native molecular weight, it was desirable to confirm the proposed heterotetrameric structure by independent means. To verify that at least two copies of the DNL IV polypeptide were present in each complex, we performed a triple mixed infection of Sf9 cells with baculoviruses expressing the previously described DNL IV-His 6 and HA-XRCC4 proteins, together with another baculovirus expressing a DNL IV-Myc epitope protein. Cell lysates were prepared and subjected to Ni ϩ -NTA-agarose chromatography, and protein from a peak fraction was immunoprecipitated using anti-Myc antibody-conjugated Sepharose beads as described under "Experimental Procedures." The immunoprecipitate was analyzed by SDS-PAGE with Coomassie Blue staining, which revealed a prominent doublet at the approximate position expected for the DNL IV polypeptide (Fig. 4A, lane 3).  A and B) or with anti-His tag antibody to detect DNL IV-His 6 . Anti-HA antibody was visualized using horseradish peroxidase-conjugated rabbit anti-mouse IgG secondary antibody and a Luminol-based detection system as described in the legend to Fig. 4. Anti-His tag antibody was visualized alkaline phosphatase-conjugated rabbit anti-mouse IgG secondary antibody and bromochloroindolyl phosphate/nitro blue tetrazolium stain. For each of the panels, the positions of molecular size markers are indicated at the left. Arrowheads at the right mark the positions of non-cross-linked DNL IV and XRCC4, as well as cross-linked complexes denoted as X and Y. In some lanes, complex X was resolved as a doublet, which may reflect cross-linking at multiple sites. E, schematic diagram showing interpretation of cross-linking analysis. Complex X apparently corresponds to a (HA-XRCC4) 2 homodimer. Complex Y apparently corresponds to a (DNL IV)(XRCC4) heterodimer, although the presence of higher order species cannot be ruled out.
only one type of epitope tag (i.e. His 6 or Myc, but not both). Because DNL IV⅐XRCC4 complex was subjected to sequential affinity purification, the recovery of protein at the end of the experiment is possible only if a polypeptide bearing each tag are present in the same physical complex. The results (Fig. 4A,  lane 3) provide strong evidence supporting the presence of at least two DNL IV polypeptides within the proposed heterotetrameric structure.
Chemical Cross-linking of DNLIV⅐XRCC4 Complex-The preceding experiment suggests that there are at least two DNL IV polypeptides in the complex, but does not address the number of XRCC4 polypeptides. To more precisely determine the number of XRCC4 polypeptides, as well as to determine which polypeptides are in direct contact with one another, we performed chemical cross-linking on the DNL IV⅐XRCC4 complex, using BS 3 , a water-soluble homobifunctional N-hydroxysuccinimide compound that reacts with primary amines. BS 3 is expected to react with the N terminus and with lysine residues. Cross-linked products were analyzed by both 5% and 7% SDS-PAGE, in order to obtain maximum resolution over a wide size range. Immunoblotting was performed to determine which polypeptides were present in each cross-linked complex.
Immunoblotting with the anti-HA antibody showed that XRCC4 was present in two cross-linked complexes, designated X and Y (Fig. 5, A and B). Complex X migrates at a position corresponding to about 150,000 Da, relative to the marker proteins, and contains only XRCC4, with no detectable DNL IV (compare panel A with C and panel B with D). Because there was no evidence for intermediates running between the XRCC4 monomer and complex X at the lowest cross-linker concentration (0.4 M), we suggest that complex X must correspond to a cross-linked XRCC4 dimer. This species migrates significantly more slowly than predicted from its molecular mass of 79,000 Da, probably because of the branched geometry of the crosslinked structure. A slightly more rapidly migrating form of complex X appears at higher BS 3 concentrations and might correspond to a more compact, multiply cross-linked form (Fig.  5A).
Complex Y has an apparent molecular mass of greater than 200,000 Da and contains both XRCC4 and DNL IV. We suggest that complex Y corresponds principally to a cross-linked XRCC4⅐DNL IV dimer. Complex Y forms quite a broad band in the 5% SDS-PAGE analysis, and could also include a crosslinked (DNL IV)(XRCC4) 2 trimer. We did not detect species migrating significantly more slowly than complex Y, suggesting that cross-linked DNL IV dimer is probably not present.
In addition to the BS 3 cross-linking, we also performed similar experiments using N-hydroxysulfosuccinimidyl 4-azidobenzoate, a heterobifunctional photo-activated cross-linker that reacts with primary amines at one end of the molecule and reacts nonspecifically with various amino acid side chain groups at the other end. Results were substantially the same as with BS 3 (data not shown).
In the course of these experiments, we observed that the XRCC4 in the purified complex, although not the DNL IV, had a marked tendency to undergo spontaneous oxidation. This was manifested as an appearance of variable amounts of a cross- linked XRCC4 dimer in the absence of added chemical crosslinker, and could be avoided by inclusion of 35 mM dithiothreitol in the SDS-PAGE sample buffer. We note that dithiothreitol had no effect on mobility of the DNL IV⅐XRCC4 complex in native gels, indicating that disulfide bonds are not required for maintenance of the tetrameric structure. Fig. 5E summarizes and provides an interpretation of the cross-linking experiments. We suggest that an XRCC4 dimer forms the core of the complex and can be readily cross-linked to give the complex X observed in our gels. Each XRCC4 polypeptide is in contact with one DNL IV polypeptide, and these can be cross-linked to give complex Y. The two DNL IV polypeptides do not appear to be in contact with each other, as there is no evidence for the formation of a cross-linked DNL IV dimer.
Activity of Purified DNLIV⅐XRCC4 Complex in Nick-ligation and NHEJ Assays-Although the biological function of DNL IV is to join DNA double-strand breaks, previous reports indicate that the purified enzyme has only a minimal ability to carry out this reaction in vitro, presumably because of a requirement for additional cofactors (25).
Because of the anticipated lack of direct double-strand break joining activity, preliminary characterization of the purified recombinant DNL IV⅐XRCC4 complex was performed using a nick-ligation assay (54). Results are shown in Fig. 6. There was a progressive increase in the formation of ligated products as increasing amounts of DNL IV⅐XRCC4 were added to the reaction (Fig. 6A). Interestingly, under the conditions used, the number of ligation events was approximately equal to the amount of DNL IV⅐XRCC4 tetramer added to the reaction (Fig.  6B), suggesting that the enzyme might be limited to a single turnover under these conditions. Further kinetic studies will be needed to understand the mechanism of interaction with this substrate in more detail.
We next investigated the activity of DNL IV⅐XRCC4 in the presence of other repair factors. An NHEJ assay system has recently been described that measures plasmid end-joining in the presence of a whole cell extract from human lymphoblasts (25). The system is faithful, in that the reaction is dependent on endogenous Ku, DNA-PK, and DNL IV, but end-joining efficiency is low, typically less than 5%, and in our experience is somewhat variable. We carried out NHEJ assays essentially as described (25) using a BamHI-digested plasmid substrate and adding purified DNL IV⅐XRCC4 complex. Results are shown in Fig. 7. As expected, the purified DNL IV⅐XRCC4 complex alone had little activity. Similarly, the lymphoblast cell extract alone had little activity under these conditions. However, the combination of the DNL IV⅐XRCC4 complex and cell extract showed a high level of activity, with as much as 70% of the monomer substrate being converted to various ligated forms. This result indicates that DNL IV⅐XRCC4 is a limiting component in cell extracts, and that there is strong synergy between exogenous, recombinant enzyme and cellular factors.
We note that DNL IV⅐XRCC4 catalyzes many fewer ligation events, per mole of complex, in the end-joining assay than in the nick-ligation assay (compare Figs. 7 and 6). This presumably reflect the complexity of the interactions in reactions containing crude extract. For example, the DNL IV⅐XRCC4 may participate in nonproductive complexes with endogenous repair proteins or it may interact with specific negative regulatory proteins. Further purification and characterization of the components in the extract will be required to address this issue.
In order to determine whether the DNL IV⅐XRCC4 complex was stimulating the specific, Ku-dependent pathway of endjoining, we tested the effect of several human autoimmune sera on the reaction (Fig. 8). The reaction was strongly inhibited by four different sera (lanes 4 -6 and 8). Two of these contained anti-Ku antibodies (HT and OY), one contained anti-DNA-PKcs antibodies (FT), and one contained a mixture of both types of antibodies (TT). By contrast, there was no inhibition in the absence of serum (lane 9) or in the presence of control sera (lanes 3 and 7). One of the control sera was from a normal individual (NS), and the other was an autoimmune serum (TA) that contained antibodies against another nuclear autoantigen, RNA helicase A (55). In addition to the autoimmune sera, we tested an anti-DNA-PKcs monoclonal antibody (mAb 18-2), which also strongly inhibited the reaction (data not shown).
To determine whether the end-joining occurred perfectly, without insertion and deletion of nucleotides, we digested the products of a control reaction with BamHI, which cleaved substantially all of the ligated DNA (Fig. 8, lane 10). Together, these results indicate that the reaction is faithful and remains dependent on physiologically essential repair proteins even when large amounts of DNL IV complex are present to drive the reaction. We have so far been unable to reconstitute the reaction with a mixture of purified DNL IV⅐XRCC4, Ku, and DNA-PKcs, indicating that at least one other repair factor is likely to be required, in agreement with a previous study (25). DISCUSSION We have expressed and purified an active, recombinant DNL IV⅐XRCC4 complex. Physical characterization of the complex by gel filtration, native gel electrophoresis, quantitative analysis of Coomassie Blue binding, quantitative immunoblotting, and chemical cross-linking strongly suggests that is a mixed tetramer containing two copies of DNL IV and two copies of XRCC4. We were also able to isolate mixed multimers containing DNL IV polypeptides with two different epitope tags, which provides additional confirming evidence that more than one copy of DNL IV is present. Although each experimental approach is subject to specific limitations, together they provide strong evidence in favor of the tetramer model. To our knowledge, DNL IV is the first example of a mammalian DNA ligase that contains two active sites within the active enzyme complex. We suggest that this is a specialized adaptation to the function of DNL IV in double-strand break repair.
Chemical cross-linking experiments suggest that a XRCC4⅐XRCC4 interface forms the core of the DNL IV⅐XRCC4 tetramer. Consistent with this model, previous studies have shown that XRCC4 self-aggregates when expressed in the absence of DNL IV (56 -58). As has been reported by others, XRCC4 has a propensity for formation of disulfide linked dimers (56). Because the interior of the cell is a reducing environment, it is unlikely that such disulfide linkages exist in vivo, and we suggest that they may form during storage or handling of the purified protein. The propensity for oxidation probably reflects chance juxtaposition of cysteine residues at the XRCC4⅐XRCC4 interface. Two-hybrid studies suggest that this interface involves amino acids 115-204. Three cysteine residues are present within this region (58).
The DNL IV⅐XRCC4 complex has been subjected to gel filtration analysis in several previous studies, with conflicting results. In one case, the DNL IV⅐XRCC4 complex was shown to coelute with a DNL III⅐XRCC1 complex, in another, it was estimated to have a molecular weight of about 200,000, and in a third report, the molecular weight was estimated to be 400,000 (29,39,59). Various authors have suggested the existence of a DNL IV-XRCC4 dimer, a (DNL IV)(XRCC4) 2 trimer and, in a very recent study, a (DNL IV) 2 (XRCC4) 2 tetramer (3,29,39,53,59). One explanation for the discrepancy in the results is that, in the earlier studies, the gel filtration columns may not have been sufficiently well calibrated in the high molecular range to distinguish between different higher order oligomeric forms. In any case, the variability underscores the need for multiple approaches in order to accurately determine oligomeric structure.
Although the isolated DNL IV⅐XRCC4 complex had virtually no ability to join double-stranded DNA fragments in an endjoining assay, there was strong synergy between the isolated enzyme and other repair factors present in cell extracts. This provides strong evidence that the tetrameric DNL IV⅐XRCC4 complex is the active form of the enzyme. Our results also demonstrate that the DNL IV⅐XRCC4 complex is the limiting component when the nonhomologous end-joining reaction is reconstituted with a mammalian cell extract prepared under the conditions described (25).
One of the challenges inherent in the reconstitution of complex biological reactions is that multiple, competing reaction pathways may operate simultaneously. In the case of NHEJ, previous workers have observed the operation of both Ku-dependent and Ku-independent pathways in cell-free systems (25,26,60). The Ku-dependent pathway is believed to reflect the major physiological pathway of mammalian NHEJ, as identified by genetic analysis, whereas the Ku-independent pathways are not well characterized and could reflect nonspecific activity of other DNA ligases in the extract. Because the DNL IV⅐XRCC4 complex is highly dependent on Ku and other repair proteins for its activity, the addition of the purified complex to a cell-free NHEJ system provides the ability to increase the flux through the Ku-dependent pathway without increasing background attributable to nonspecific end-joining. This ability will facilitate future mechanistic studies of NHEJ using the in vitro system.
The DNL IV⅐XRCC4 tetramer presumably has a dyad symmetry, which implies that it may play a role in bridging the two DNA ends in NHEJ. There have already been several suggestions that other repair proteins may work to facilitate the synapsis of DNA ends. Ku itself has some capacity to bridge DNA ends, as evidenced by the facilitated kinetics of transfer of Ku protein between DNA fragments with cohesive ends (61) and by direct observations using atomic force microscopy (8,10). DNA-PKcs may also be involved in synapsis, as evidenced by atomic force microscopy and by enzymatic studies showing the superactivation of DNA-bound kinase by single-stranded DNA presented in trans (8,9,62,63). One way to reconcile these various observations is to postulate that occupancy of the DNA termini by Ku, DNA-PKcs, and DNL IV⅐XRCC4 occurs sequentially, rather than simultaneously.
We have recently reported results of a series of binding and photocross-linking studies using oligonucleotides designed to trap repair complexes in defined positions and orientations. On the basis of these and other studies, we suggest that NHEJ involves sequential occupancy of the ends by three different complexes, as shown in Fig. 9.
1) In the initial complex, a single Ku heterodimer is bound in contact with the free DNA. Previous data indicate that this complex occupies 14 nt, and the Ku is oriented, with the Ku70 subunit proximal and the Ku80 subunit distal to the free DNA end (64).
2) The DNA-PK complex is formed by recruitment of DNA-PKcs to the initial complex. This enzymatically active complex occupies a minimum of 28 nt, and the Ku translocates inward, leaving DNA-PKcs in direct contact with about 1.5 turns of the DNA helix, proximal to the end (11,13).
3) The ligation complex is formed by recruitment of DNL IV⅐XRCC4 to the DNA-PK complex. We have not been able to observe this complex directly, although we infer its existence from the functional synergy between purified DNL IV⅐XRCC4 and factors in the cell extract.
We suggest that the ligation complex involves the obligatory synapsis of the DNA ends, that the XRCC4 dimer interface is responsible for maintaining this synapsis, and that the two DNL IV polypeptides are in direct physical contact with the DNA termini, so that they are in a position to catalyze covalent strand joining. The minimum site size for the complex is unknown, although it may be quite large, since we have so far been unable to observe factor-dependent in vitro ligation using small oligonucleotide substrates. Moreover, the inability to reconstitute the ligation complex with ligase, Ku, and DNA-PKcs alone suggests that formation of the complex occurs only in concert with other repair factors.