Structural and Functional Interaction between the Human DNA Repair ...

MOLECULAR AND CELLULAR BIOLOGY, June 2009, p. 3163–3172 0270-7306/09/$08.00⫹0 doi:10.1128/MCB.01895-08 Copyright © 2009, American Society for Microbiology. All Rights Reserved.

Vol. 29, No. 11

Structural and Functional Interaction between the Human DNA Repair Proteins DNA Ligase IV and XRCC4䌤† Peï-Yu Wu,1,2 Philippe Frit,1,2* SriLakshmi Meesala,3 Ste´phanie Dauvillier,1,2 Mauro Modesti,4 Sara N. Andres,3 Ying Huang,5 JoAnn Sekiguchi,5 Patrick Calsou,1,2 Bernard Salles,1,2 and Murray S. Junop3,6* CNRS, IPBS (Institut de Pharmacologie et de Biologie Structurale), 205 Route de Narbonne, F-31077 Toulouse, France1; Universite´ de Toulouse, UPS, IPBS, F-31077 Toulouse, France2; Department of Biochemistry and Biomedical Sciences, McMaster University, 1200 Main Street West, Hamilton, Ontario L8N 3Z5, Canada3; CNRS, Unite´ Propre de Recherche 3081, Genome Instability and Carcinogenesis Conventionne´ par l’Universite´ d’Aix-Marseille 2, 31 Chemin Joseph Aiguier, 13402 Marseille Cedex 20, France4; Departments of Internal Medicine and Human Genetics, University of Michigan Ann Arbor, Michigan 481095; and Michael G. DeGroote Institute for Infectious Disease Research, McMaster University, Hamilton, Ontario L8N 3Z5, Canada6 Received 14 December 2008/Returned for modification 27 January 2009/Accepted 24 March 2009

Nonhomologous end-joining represents the major pathway used by human cells to repair DNA double-strand breaks. It relies on the XRCC4/DNA ligase IV complex to reseal DNA strands. Here we report the highresolution crystal structure of human XRCC4 bound to the carboxy-terminal tandem BRCT repeat of DNA ligase IV. The structure differs from the homologous Saccharomyces cerevisiae complex and reveals an extensive DNA ligase IV binding interface formed by a helix-loop-helix structure within the inter-BRCT linker region, as well as significant interactions involving the second BRCT domain, which induces a kink in the tail region of XRCC4. We further demonstrate that interaction with the second BRCT domain of DNA ligase IV is necessary for stable binding to XRCC4 in cells, as well as to achieve efficient dominant-negative effects resulting in radiosensitization after ectopic overexpression of DNA ligase IV fragments in human fibroblasts. Together our findings provide unanticipated insight for understanding the physical and functional architecture of the nonhomologous end-joining ligation complex. cessing enzymes, such as the structure-dependent nuclease Artemis (19); and its requirement for the stable recruitment of the XRCC4/DNA ligase IV (LigIV) complex that catalyzes the final ligation step in NHEJ (7). Deficiency in any of these key factors gives rise to radiosensitive severe combined immunodeficiency syndromes in human patients and animal models as a consequence of the dual function of NHEJ machinery in both V(D)J recombination and DSB repair (34). Recently, an additional NHEJ core factor, termed Cernunnos or XRCC4-like factor (Cer-XLF), was identified as an XRCC4 partner; it is also deficient in a human radiosensitive severe combined immunodeficiency syndrome (1, 6). Ligation is central to DSB repair by the NHEJ pathway and requires the concerted action of LigIV, XRCC4, and Cer-XLF. In vivo, LigIV associates tightly with XRCC4 (9, 21, 22). XRCC4 serves as a multipurpose partner for LigIV, not only stimulating its adenylation (30) and perhaps promoting stable interactions with DNA, but also protecting it from degradation (5, 16). The stoichiometric ratio of the XRCC4/LigIV complex is 2:1 as revealed by both biochemical and crystallographic analyses (31, 37). Within the XRCC4/LigIV complex, interactions have been mapped to the central coiled-coil domain of XRCC4 and to the inter-BRCT (BRCA1 [breast cancer associated 1] C terminal) domain linker at the C terminus of LigIV (23, 24, 37). This XRCC4-interacting region (XIR) of LigIV appears necessary and sufficient for XRCC4/LigIV interaction (24, 37). However, the recent crystal structure of the Saccharomyces cerevisiae ortholog Lif1p/Lig4p complex, at 3.9 Å resolution, suggests that flanking sequences might also participate

DNA double-strand breaks (DSBs) represent the most-toxic form of DNA damage in the genome. If left unrepaired, DSBs can result in large-scale loss of genetic information during cell division and, consequently, cell death. DSBs are formed not only in response to endogenous cellular processes, such as V(D)J recombination and oxidative metabolism, but also to various genotoxic agents, such as ionizing radiation, radiomimetic compounds, and topoisomerase inhibitors (38). To cope with such deleterious DNA lesions, cells have evolved various repair mechanisms, among which nonhomologous end-joining (NHEJ) represents the major pathway in mammals (39). NHEJ is a multistep process initiated by the Ku70/Ku80 heterodimer, which binds DNA ends and recruits the DNAdependent protein kinase catalytic subunit (DNA-PKcs) through a direct interaction (15, 20). The resulting DNA-PK holoenzyme (Ku/DNA-PKcs) has a serine/threonine PK activity that is necessary for efficient repair (27). The pivotal role played by DNA-PK in NHEJ is further emphasized by its DNA end-bridging activity (10); its regulatory function toward pro-

* Corresponding author. Mailing address for Philippe Frit: IPBS, 205 Route de Narbonne, F-31077 Toulouse cedex, France. Phone: (33) 561-17-59-37. Fax: (33) 561-17-59-33. E-mail: [email protected]. Mailing address for Murray S. Junop: Department of Biochemistry and Biomedical Sciences, McMaster University, 1200 Main Street West, Hamilton, Ontario L8N 3Z5, Canada. Phone: (905) 525-9140. Fax: (905) 522-9033. E-mail: [email protected]. † Supplemental material for this article may be found at http://mcb .asm.org/. 䌤 Published ahead of print on 30 March 2009. 3163

3164

MOL. CELL. BIOL.

WU ET AL.

in this interaction (12). The respective Cer-XL and XRCC4 homodimers interact through their N-terminal globular head domains (2, 11, 29), but connections between Cer-XLF and LigIV have so far not been detected. To better characterize the physical and functional interactions between XRCC4 and LigIV, we determined the crystal structure of a functional fragment of human XRCC4, consisting of residues 1 to 203 (XRCC41-203), bound to the tandem BRCT domains of LigIV654-911. The high-resolution structure reveals an extensive LigIV binding interface formed by a clamp-shaped helix-loophelix motif within the inter-BRCT linker region, as well as significant interactions involving the second BRCT domain (BRCT2). Functional analyses substantiated the role of BRCT2 in stabilizing XRCC4/LigIV interaction in vivo. MATERIALS AND METHODS XRCC4-LigIV crystallization, data collection, and structure determination. See the supplemental material for a full description of XRCC4-LigIV crystallization, data collection, and structure determination. Expression constructs. Construction of plasmid vectors expressing human LigIV fragments is described in detail in the supplemental material. Cell lines, cell culture, and transfection. Simian virus 40-immortalized MRC5-SV human fibroblasts (a gift from A. Sarasin, Institut Gustave Roussy, Villejuif, France) were maintained in a humidified atmosphere at 37°C with 5% CO2 in Dulbecco’s modified Eagle’s medium (Invitrogen, Carlsbad, CA) supplemented with 10% fetal calf serum, 2 mM glutamine, 125 units/ml penicillin, and 125 ␮g/ml streptomycin. See the supplemental material for the detailed procedure of transfection. Irradiation and survival assays. Cells were seeded at a density of 104 cells/ 12-well plate and irradiated 16 h later with a 137Cs source (Biobeam 8000) at a dose rate of 4.4 Gy/min. After irradiation, cells were diluted and reseeded in 10-cm dishes. After 2 to 3 subsequent weeks, cell colonies were fixed with 100% ice-cold methanol and stained with 0.2% crystal violet. Colonies larger than 50 cells were counted, and values normalized to those for untreated cells. All experiments were performed in duplicate and repeated at least three times. Antibodies and Western blotting. See the supplemental material for origin of antibodies and detailed description of the Western blotting procedure. Recruitment assay. The recruitment assay was performed essentially as previously described (14). See the supplemental material for details. IP. Immunoprecipitation (IP) experiments were conducted as previously described, with minor modifications (41). See the supplemental material for details. RT-PCR. The primer sequences and the detailed protocol for reverse transcription-PCR (RT-PCR) are described in the supplemental material. In vitro NHEJ assay. Cell extract preparation and in vitro NHEJ reactions were carried out as previously described (6), with minor modifications that are detailed in the supplemental material. Protein structure accession number. Coordinates have been deposited in the Protein Data Bank under accession code 2GQO.

RESULTS Structure of human XRCC4/LigIV complex. In order to better understand the nature of interactions responsible for the formation of the highly stable XRCC4/LigIV complex, we determined the crystal structure of a functional N-terminal fragment of XRCC4, XRCC41-203, bound to the C-terminal tandem BRCT region of LigIV654-911. The complex crystallized in space group P1, with four copies of XRCC41-203 and two of LigIV654-911 in each unit cell. Experimental phases were determined by using selenomethionine-substituted protein and single-wavelength anomalous diffraction. The overall structure was refined to R and Rfree values of 24.6 and 28.9%, respectively (PDB accession code 2GQO). Complete data collection and model refinement statistics are listed in Table SI in the supplemental material.

FIG. 1. Structure of human XRCC41-203/LigIV654-911 complex. (A) Ribbon diagram of complex of C-terminal domain of LigIV bound to XRCC4 homodimer. XRCC4 chains A and B are colored in cyan and green, respectively; the tandem BRCT domains of LigIV are shown in red. (B) Closer view of BRCT domains of LigIV in complex with XRCC4. The four parallel ␤-strands in each BRCT domain are colored in deep teal. The inter-BRCT linker is colored purple. (C) Structural alignment of the present structure of human XRCC41-201/ LigIV654-911 (cyan) with a previously reported structure of XRCC4 (red) bound to a peptide of LigIV755-782 (blue) (37). The black twoheaded arrow indicates the 45° rotation that results upon binding of BRCT2. (D) Structural alignment of human XRCC41-201/LigIV654-911 (cyan) with the yeast Lif1p/Lig4p complex (orange) (12).

As determined biochemically, the structure of human XRCC4/LigIV retained a 2:1 stoichiometry (Fig. 1A). The two BRCT domains of LigIV are angled at 45° relative to the plane perpendicular to the coiled-coil tails of XRCC4, wherein the N-terminal BRCT is positioned below the plane while the C-terminal BRCT is seen oriented above. LigIV654-911 was tightly anchored to XRCC41-203. Surprisingly, this interaction occurred not only through sequence-linking BRCT domains, as had been reported earlier (24, 37), but also through substantial contacts with the C-terminal BRCT (BRCT2) domain of LigIV (Fig. 1B). The linker region forms a clamp-shaped helix-loop-helix structure that wraps tightly around the coiledcoil tails of XRCC4 (Fig. 1B). The clamp interacts symmetrically with both subunits of XRCC4, wherein a single amphipathic ␣-helix lies on either side of the tails of XRCC4 and the connecting loop makes a uniform turn around chain B of XRCC4. Both BRCT domains maintain the conserved fold characteristic of the BRCT superfamily, with a central fourstranded parallel ␤-sheet (in the order of ␤2, -1, -3, -4) flanked by three ␣-helices.

VOL. 29, 2009

ARCHITECTURE OF HUMAN NHEJ LIGATION COMPLEX

Structural comparison of human XRCC4/LigIV. Comparison of our structure with the previously determined structure of XRCC4 bound to a short peptide of LigIV, LigIV755-782 (the so-called XIR), revealed significant differences (37). First, the entire region of XRCC4 N terminal to the ligase binding site (residues 1 to 169) in our structure undergoes a large rotation of ⬃45° that results in the head domains of XRCC4 being translated and rotated by ⬃40 Å and ⬃15°, respectively (Fig. 1C). This altered trajectory results entirely from the formation of a kink in the tail region of XRCC4 (residues 170 to 172) created by the binding of BRCT2. In contrast, the relative conformation of the XRCC4 tails C terminal to the kink region of XRCC4 (residues 173 to 200) remained essentially unaltered between both structures, with C␣ atoms superimposed to a root mean square deviation (RMSD) of less than 1.0 Å. Another interesting difference between these structures involves the transformation in the coiled-coil domain of XRCC4 upon interaction with LigIV. In both structures, binding of LigIV triggers a switch in the heptad repeat of the left-handed coiled coil into a right-handed-coil unadecad. This change in coil handedness has a sharply contrasting outcome in the current structure compared to the outcome in the structure with the XIR. In the structure with the XIR, this altered coil handedness causes the helix in chain A to become straight, while the same helix in chain B develops a kink. In the structure reported here, the opposite is observed (Fig. 1C). The helix in chain A kinks at residue 173, causing the N-terminal region to bend, while the helix in chain B straightens. Therefore, binding of BRCT2, in addition to the inter-BRCT linker of LigIV, appears to be a very important determinant for ensuring structural integrity within the LigIV binding region of XRCC4 (residues 154 to 195). More recently, the structure of the homologous complex Lif1p/Lig4p from Saccharomyces cerevisiae has been reported (12). Comparison of the human and yeast structures of XRCC4/LigIV reveals a surprising lack of structural similarity. At the topological level, helix H2 of the LigIV helix-loop-helix clamp was absent in the Lif1p/Lig4p structure and appeared instead as an extended loop (Fig. 1D). Furthermore, only residues 180 to 200 (C␣ atoms) of XRCC4 can be superimposed on Lif1p (RMSD of 4.44 Å). The remaining portion of human XRCC4 also deviates from the yeast structure, leaning closely toward BRCT2 (Fig. 1D). Moreover, in the yeast complex Lig4p, binding interrupts the heptad repeat within the coiledcoil tails of Lif1p, which leads to an increase in the interhelical distance, while the region N-terminal to the Lig4p-interacting region continues to pursue a linear path. This contrasts with the trajectory observed for XRCC4/LigIV, which depends on the enhanced interaction provided by BRCT2, suggesting that BRCT2 of human LigIV has evolved to interact more stably with XRCC4. One rather puzzling structural difference relates to the Lif1p head domain orientation, which is modeled in an inverted orientation compared to that of all structures of human XRCC4. The meaning of such dramatic structural changes is presently unclear; however, since the structure of Lif1p-Lig4p was determined to 3.9 Å and apparently contained significant disorder in this region, differences in the head region may simply reflect limitations in the data used for model refinement.

3165

FIG. 2. Interface interactions of XRCC4/LigIV complex. (A) Close-up view of XRCC4/LigIV complex interface. Three regions of LigIV (light gray) that interact with XRCC4 are indicated: BRCT2, blue; helix-loophelix (HLH) clamp, red-brown; and ␤-hairpin, orange. The ␣L-adaptor region is shown in yellow. XRCC4 (residues 150 to 201) chains A and B are colored cyan and green, respectively. Amino acid side chains participating in complex formation are illustrated in semitransparent stick mode. (B) Schematic of specific interactions. Residues with direct interactions that contribute to complex formation are shown in boldface and underlined. Color of triangle below sequence indicates binding partner. Color scheme is as described for panel A. (C) Arg814 hydrogen-bonding network. Stereoview of hydrogen bonds formed between E800, S811, and R814 of BRCT2 and E173 of XRCC4.

XRCC4/LigIV interface. The XRCC4/LigIV interaction interface is extensive, burying a total surface area of 4,236 Å2. A detailed description of residues from XRCC4 and LigIV mediating this interaction is provided in Fig. 2, as well as in Fig.

3166

WU ET AL.

S1 in the supplemental material. The interaction surface of XRCC4 is asymmetric, spanning residues 154 to 195 on chain A and residues 155 to 190 on chain B. The interaction interface of LigIV is less contiguous and can be broadly divided into three major sections: (i) the 3:5 ␤-hairpin following BRCT1 (residues 759 to 769), (ii) the central helix-loop-helix clamp (residues 770 to 803), and (iii) portions of BRCT2 (loop 1, residues 804 to 814; ␣1⬘, residues 835 to 848; and ␣3⬘, residues 891 to 900). The segment including the ␤-hairpin and the N-terminal helix of the helix-loop-helix clamp (residues 755 to 782) corresponds to the peptide of LigIV that was previously characterized by structure determination (37). However, the largest interaction surface in LigIV is formed by the helix-loophelix clamp, burying 1,990 Å2 of surface area (see Fig. S1B in the supplemental material). Residues 780 to 788 constitute the loop region of the clamp and are responsible for making polar contacts with chain B of XRCC4. Asn783 and Ser784 are engaged in establishing hydrogen bonds with Lys178 of chain B of XRCC4 (Lys178B) and Arg179B. In contrast, the corresponding loop in the yeast Lif1p/Lig4p structure is predominantly comprised of glycines and prolines, which fail to mediate significant contacts with Lif1p. A long amphipathic ␣-helix (H2) follows this loop region and packs against the hydrophobic surface of XRCC4 that is formed by residues Tyr177A, Ile181A, Val183A, Phe180A/B, and Leu184B. Burying nearly as much surface area as the helix-loop-helix clamp, BRCT2 has numerous interactions with XRCC4 and, therefore, is also expected to play an important functional role in mediating interactions between XRCC4 and LigIV (Fig. 2B; see also Fig. S1A in the supplemental material). The final section is formed by the ␤-hairpin (residues 759 to 769) following BRCT1 and buries ⬃500 Å2 of surface area. An additional helix (␣L, residues 749 to 756) located between BRCT1 and the ␤-hairpin does not have direct interactions with XRCC4 (Fig. 2; see also Fig. S1C in the supplemental material). However, together with contributions from the last few residues (740 to 742) of BRCT1, as well as the short loop region immediately following BRCT1, ␣L forms an adaptor between the body of BRCT1 and the ␤-hairpin. This interaction would be expected to stabilize the ␤-hairpin and further contribute to the formation of a stable XRCC4/ LigIV complex. A remarkable feature of the interaction between XRCC4 and BRCT2 is the presence of numerous hydrogen-bonding networks. The first network is formed by hydrogen bonding between Glu173B and Lys169B of XRCC4 and Arg814, along with Leu810 and Phe847 of BRCT2. Arg814 makes double hydrogen bonds with Glu173B, while Lys169B binds to Glu173B, as well as to the carbonyl groups of Leu810 and Phe847. Given that the buried Arg814 makes contacts with XRCC4 chain B and also engages in hydrogen-bonding contacts with Glu800 and Ser811 of BRCT2, Arg814 emerges as a crucial player in holding the XRCC4 chain B close to BRCT2 (Fig. 2C). Two additional hydrogen-bonding networks result from the packing of helices ␣1⬘ and ␣3⬘ close to XRCC4. Thus, key interactions with XRCC4 are mediated by residues from both ␣1⬘ and ␣3⬘ of BRCT2. Of these helices, ␣1⬘ has a more-significant interaction with XRCC4, forming a parallel triple-helix bundle with both chains of XRCC4 over its entire length (Fig. 2A; see also Fig. S1A in the supplemental material). Taken together, the structural information reveals the importance of BRCT2, in

MOL. CELL. BIOL.

conjunction with the helix-loop-helix region of LigIV, for interaction with XRCC4. Attempts to further evaluate the contributions of these interactions toward the overall stability of the XRCC4/LigIV complex in vitro were thwarted by the inability to produce soluble LigIV mutants (K876Q R879Q K880Q R887Q, K876S R879S K880S, K883S R884S K885S K887S, and K883S R884S K885S). However, eight XRCC4 mutants (F180D I181D, V183D L184D, L184Q K187D I191S, K187D K188D, I191D L194E L198E, C165F, K169E, and C165A K169V L172E) have been previously shown to disrupt XRCC4/LigIV interactions both in vitro and in vivo (31). As shown in Fig. S1 in the supplemental material, the effects of these mutants correlate well with residues identified in the structure reported here as being important for XRCC4/LigIV stability (31). The BRCT2 domain of LigIV is essential for XRCC4 interaction in human cells. To directly challenge the functional XRCC4/LigIV interface, we performed ectopic expression of competing fragments of LigIV in human cells. We designed several LigIV fragments fused with enhanced green fluorescent protein (GFP) (Fig. 3A). To examine the ability of these constructs to interact with XRCC4 in vivo, we transiently transfected immortalized MRC5 human fibroblasts and performed co-IP experiments on cell extracts 48 h after transfection. As shown in Fig. 3B, only two constructs, GFP-XIR-BRCT2 (GX2) and GFP-BRCT1-XIR-BRCT2 (G-1X2), pulled down XRCC4 when immunoprecipitated with anti-GFP beads. XIR alone (GFP-XIR [G-X]), as well as the single BRCT domains (GFP-BRCT1 [G-1] and GFP-BRCT2 [G-2]) and the tandem BRCT repeat lacking the intervening linker region (GFPBRCT1-BRCT2 [G-12]), failed to interact with XRCC4. The specificity of XRCC4 interaction with G-X2 and G-1X2 was further confirmed by cross-IP experiments with anti-XRCC4 beads (Fig. 3B, bottom panel). Together, these results indicate that while the so-called XIR (i.e., LigIV755-782) is necessary for XRCC4 binding, it is not sufficient to support a stable interaction in cellulo. Additional flanking residues are required, particularly at the C-terminal end. According to the crystal structure described above, both the H2 helix of the helix-loophelix clamp and the BRCT2 domain could participate in this stabilization. To discriminate between these two possibilities, we extended the G-X construct up to the end of the H2 helix (i.e., Ala755 to Ser804) (Fig. 2B and 3A). This extended XIR failed to coimmunoprecipitate XRCC4 (Fig. 3C), strongly suggesting that the XIR and even the helix-loop-helix clamp motif are insufficient on their own to stably interact with XRCC4 and require the additional presence of the BRCT2 domain. Ectopic overexpression of competing LigIV fragments downregulates endogenous LigIV protein. Since XRCC4 is essential for LigIV stability in mammalian cells (5, 16), we reasoned that the introduction of LigIV fragments able to compete with endogenous LigIV for binding to XRCC4 would result in degradation of endogenous LigIV. In line with this hypothesis, stable overexpression of G-X2 or G-1X2 specifically resulted in dramatically decreased levels of endogenous LigIV protein (Fig. 4A), even though these fragments were expressed at lower levels than the GFP, G-X, and G-1X constructs (Fig. 4A, anti-GFP Western blot). This observation further supports the specific binding of these constructs to XRCC4, as determined by co-IP experiments (Fig. 3B). Consistent with LigIV protein

VOL. 29, 2009


3167

FIG. 3. LigIV BRCT2 domain is required for stable interaction with XRCC4 in cellulo. (A) Schematic representation of the functional domains of human LigIV and of the G-1X2 construct. Numbers refer to amino acid position. NLS, nuclear localization sequence; peGFP-C1, peGFP-C1 vector backbone. (B and C) Co-IP experiments on transiently transfected MRC5 cells. MRC5 cells were transfected with various LigIV fragment-expressing constructs. Two days after transfection, cells were harvested and WCE were incubated with anti-GFP or anti-XRCC4 beads. After a wash step, proteins retained on the pelleted beads were denatured and analyzed by Western blotting (WB) with antibodies against XRCC4, LigIV, or GFP. As indicated by double-headed arrow in panel B, XRCC4 signal is seen again in the subsequent anti-GFP Western blot. M, mock-transfected MRC5 cells; ␣, anti; G, GFP; X, XIR; 1, BRCT1; 2 BRCT2; eX or eXIR, extended XIR, i.e., ␤-hairpin-H1-loop-H2; Ig HC, immunoglobulin heavy chain.

degradation in G-X2 and G-1X2 clones, LigIV expression levels were not affected at the transcriptional level, as assessed by semiquantitative RT-PCR (Fig. 4B). Moreover, XRCC4 protein expression levels were similar in all clones, ruling out an indirect effect through XRCC4 downregulation. LigIV competing fragments impair stable XRCC4/LigIV/ Cer-XLF assembly at DSBs in vivo. To investigate the consequences of LigIV downregulation in the G-X2 and G-1X2 clones, we first examined NHEJ repair complex assembly on damaged DNA in vivo. We carried out a previously described recruitment assay that consists of detergent-based cellular fractionation, allowing in situ assessment of the stable mobilization of NHEJ proteins on DSBs (14). MRC5 clones were treated with 10 nM of the enediyne derivative calicheamicin, one of the most-potent DSB inducers. Treated or mock-treated cells were then either directly lysed, to provide control whole-cell extracts (WCE), or fractionated to remove soluble proteins, as well as proteins loosely bound to chromatin. As shown in the Western blot analysis of WCE (Fig. 5A), all the clones responded similarly to calicheamicin exposure with regard to H2AX phosphorylation and to XRCC4 phosphorylation, which induces a slower electrophoretic mobility. In previous studies, we re-

ported that DNA-PK was responsible for this mobility shift upon treatment with radiomimetic compounds (14, 41). In agreement, several pieces of evidence indicate here that DNA-PK is properly assembled and activated at DSBs in the different MRC5 clones, as assessed by Ku70/Ku80 and DNAPKcs recruitment (Fig. 5A, B, and C), as well as XRCC4 phosphorylation and DNA-PKcs autophosphorylation on Ser2056, which are both sensitive to the specific DNA-PK inhibitor NU7026 (Fig. 5C). In contrast, although XRCC4 and Cer-XLF were correctly phosphorylated in all clones (Fig. 5A and B), their stable recruitment was severely impaired in MRC5 clones overexpressing LigIV fragments G-X2 and G-1X2 (Fig. 5A and B). This demonstrates that LigIV downregulation can alter the functionality of the XRCC4/LigIV/ Cer-XLF complex as a whole. LigIV competing fragments impair NHEJ activity and lead to cell radiosensitization. To further evaluate NHEJ activity in MRC5 clones, we performed an in vitro end-joining assay with cell extracts and linearized plasmids (Fig. 6A). Ligation activity was sensitive to wortmannin (Fig. 6, first lane), a phosphatidylinositol 3-kinase inhibitor, and dependent on LigIV (Fig. 6, seventh lane), indicating that our assay reflected bona fide

3168

WU ET AL.

MOL. CELL. BIOL.

FIG. 4. Ectopic expression of LigIV competing fragments downregulates LigIV protein expression. (A) Western blot analysis of stably transfected MRC5 cells. Twenty micrograms of WCE from parental MRC5 cells or derived clones were analyzed by Western blotting (WB) for expression of GFP, Ku70, LigIV, and XRCC4. (B) Semiquantitative RT-PCR was performed. Total RNA from the same MRC5 clones for which results are shown in panel A was reverse transcribed, and the resulting cDNAs were amplified by PCR with various cycle numbers as indicated. The primers used were specific to endogenous LigIV (panel 1), GFP (panel 2), or GFP-LigIV fragments (panel 3). G, GFP; ␣, anti.

NHEJ activity. Overexpression of G-X2 or G-1X2, but not G-1X or G-12, fragments abolished end-joining activity in the respective cell extracts (Fig. 6A, fourth to sixth lanes), indicating that the resulting LigIV protein levels in the corresponding clones (Fig. 4A) are low enough to impair NHEJ activity. As the in vitro NHEJ assay was performed with limiting protein amounts relative to DNA substrates, a possibility existed that the remaining endogenous LigIV protein levels in the G-X2 and G-1X2 clones would be sufficient in vivo for maintaining a level of NHEJ activity able to protect the cells from DSBs (40). To address this question, we evaluated the clonogenic survival of the different MRC5 clones following various doses of ␥-ray irradiation. The results shown in Fig. 6B indicate that MRC5 cells overexpressing the G-1X2 fragment were significantly more radiosensitive than the parental cells or cells transfected with vector alone, with a dose-modifying factor (DMF) of ⬃twofold between 10% and 50% survival. The G-X2-expressing clone was also radiosensitized, although to a lesser extent, suggesting that the BRCT1 domain and/or the ␣L adaptor might enhance the stabilizing effect of BRCT2 on XIR/XRCC4 interaction in vivo. Both radiosensitizing effects matched well with the relative endogenous LigIV downregulation in the corresponding clones (Fig. 4A). The G-1X fragment also elicited a moderate radiosensitizing effect (DMF of ⬃1.2- to 1.3-fold) that may rely on its ability to interact with Cer-XLF (see Fig. S3 in the supplemental material). DISCUSSION The essential role of LigIV’s BRCT2 domain in XRCC4 interaction in vivo. Here we report for the first time the highresolution crystal structure of human XRCC4 bound to the C-terminal tandem BRCT repeat of LigIV. Unexpectedly, the structure is different from that of the previously published homologous complex from yeast (12) and reveals an extensive binding interface formed by a helix-loop-helix structure within the inter-BRCT linker region of LigIV, as well as significant interactions involving the second BRCT domain, which induces a kink in the tail region of XRCC4. This conformational

change in XRCC4 was also absent in the crystal structure of human XRCC4 complexed to the so-called XIR of LigIV (37). The extensive contact surface between XRCC4 and LigIV explains the extremely strong interaction between the two proteins, reported to resist 2 M NaCl or 7 M urea (37). In LigIV, contacting residues form clusters in the ␤-hairpin and helixloop-helix clamp motifs, as well as in the ␣1⬘ and ␣3⬘ helices of the BRCT2 domain. By use of complementary biochemical and cellular approaches, we further demonstrated that interaction with the second BRCT domain of LigIV is necessary to stabilize the interaction between the XIR (LigIV755-782) and XRCC4, while BRCT1 is essentially dispensable. Although these results are in line with our structural data, they appear contradictory to the XIR previously described as being necessary and sufficient for XRCC4 interaction (24, 37). However, those crystallographic and biochemical studies relied on purified proteins expressed in bacteria and on co-IP experiments after vaccinia virus-mediated transduction of XRCC4-deficient XR-1 hamster cells, respectively. Consequently, both studies were based on massive ectopic protein overexpression in heterologous systems lacking XRCC4 and LigIV homologs. It is therefore likely that the XIR, which is sufficient in a purified protein context (37) and necessary in cell extracts (24) to interact with XRCC4, requires additional motifs or domains to stabilize this interaction and efficiently compete with the fulllength protein when present. In this regard, we showed that the entire helix-loop-helix clamp (i.e., XIR plus the H2 helix) is insufficient for stably binding XRCC4, whereas further extending the XIR up to the very C terminus of the protein enables stable interaction with XRCC4. This indicates that the BRCT2 domain is a key element for the stable interaction of LigIV with XRCC4 in cellulo. Specific competition using BRCT2-containing LigIV fragments (G-X2 or G-1X2) resulted in endogenous LigIV degradation upon sustained overexpression in stably transfected cells. Given the very tight interaction between XRCC4 and LigIV, which results in LigIV’s half-life being in the range of days when bound to XRCC4 (16), degradation of endogenous

VOL. 29, 2009


FIG. 5. Ectopic expression of LigIV competing fragments impairs NHEJ ligation complex recruitment at DSBs in cellulo. (A and B) NHEJ protein mobilization in vivo in response to DSBs. Parental MRC5 cells or stably transfected clones were incubated with 10 nM calicheamicin (cali.) or mock incubated for 1 h. Cells were harvested and either immediately resuspended and lysed in denaturing buffer (WCE) or fractionated as described in Material and Methods, leading to insoluble material (Pellets). Proteins from the WCE and the pellet fractions were analyzed by Western blotting. (C) DNA-PK assembly and activity. Stable MRC5 clones were treated with 10 nM calicheamicin (cali.) or mock treated for 1 h in the presence of 30 ␮M NU7026 where indicated. Cells were fractionated, and insoluble material was analyzed by Western blotting as described for panel B. ␣-Ph-S2056, antibody raised against phospho-Ser2056 of DNA-PKcs; ⫹, present; ⫺, absent; ␣, anti; G, GFP.

LigIV most likely relies on competition with complex formation between neosynthetized proteins rather than disruption of preexisting complexes. This is further supported by the absence of detectable endogenous LigIV degradation 48 h after transfection (Fig. 3B). Nevertheless, since XRCC4 and LigIV have

3169

FIG. 6. Ectopic expression of LigIV competing fragments impairs NHEJ activity and promotes cell radiosensitization. (A) DNA endjoining activity. WCE from MRC5 stable clones were incubated with linearized plasmids in the presence of wortmannin (wort.) as indicated (⫹, present; ⫺, absent). DNA ligation products were separated by agarose gel electrophoresis, followed by Sybr gold staining. Positions of plasmid monomers (1P), dimers (2P), and trimers (3P) are indicated. (B) Survival after ionizing radiation. Parental or stably transfected MRC5 clones were irradiated at 0, 2, 4, and 8 Gy 24 h after plating. Clonogenic survival was assessed 2 to 3 weeks after irradiation and normalized to the survival of untreated cells. Results were plotted as mean values ⫾ standard deviations from at least three duplicate experiments (two experiments for shL4 [MRC5 clone stably expressing a short hairpin RNA raised against LigIV] [see Fig. S2 in the supplemental material]). G, GFP.

been described as residing in separate subcellular compartments during mitosis (35), we cannot rule out the possibility that dominant-negative LigIV fragments also operate when the endogenous proteins reassemble in late mitosis or early G1 phase. Following the expression of competitor fragments, endogenous LigIV degradation was sufficient to compromise NHEJ at a late stage of the reaction, leading in turn to cell radiosensitization. Moreover, we showed that LigIV competitor overexpression impaired the stable recruitment of XRCC4 to DSBs. This is reminiscent of what we observed previously in LigIVdeficient cells (14), suggesting that in the present study, low levels of endogenous LigIV protein in the cells rather than direct neutralization of XRCC4 by LigIV fragments were responsible for the reduced XRCC4 recruitment. Similar recruit-

3170

WU ET AL.

ment impairment was observed with Cer-XLF, demonstrating that LigIV is critical for the functionality of the XRCC4/LigIV/ Cer-XLF complex as a whole, not only because of its enzymatic activity but also to direct the complex to DSBs, in agreement with our recent report (41). In this regard, DNA-PK has been shown to be required for stable XRCC4/LigIV recruitment to DNA ends in vivo (14). Detailed analyses uncovered morespecific interactions between DNA-PKcs and XRCC4, as well as between Ku and LigIV (8, 25). However, none of the LigIV fragments tested here were specifically recruited to DSBs in vivo. Consequently, the requirement for full-length endogenous LigIV to carry out this task relies both on its C terminus for direct binding to XRCC4 (Fig. 3) and Cer-XLF (see Fig. S3 in the supplemental material) and on additional domains whose delineation and precise function remain to be established. Despite the direct interaction between Cer-XLF and G-1X, Cer-XLF is apparently normally recruited to damaged chromatin in these cells, which might reflect an alternative mechanism of mobilization of the Cer-XLF/G-1X complex to DSBs. Consistent with this, Ku has been shown to interact directly with both Cer-XLF (42) and LigIV through its BRCT1 domain (8). Further experimentation is now required to confirm this hypothesis. Implications for human LIG4 syndrome. Our definitive demonstration of the LigIV BRCT2 domain requirement in vivo to stabilize the interaction with XRCC4 provides insight to understand the molecular defect underlying the phenotype associated with some LIG4 patients. For instance, patients 2303/2304 have been reported to carry heterozygous mutations on both alleles of the LIG4 gene, resulting in the expression of truncated forms of the protein in which stop codons replace arginine residues at positions 580 and 814, respectively (34). Even the least-affected form of LigIV (R814X), which only lacks the BRCT2 domain but retains the XIR, exhibited a dramatically reduced interaction with XRCC4 in cell extracts (17, 34). To be in line with the notion that the XIR was purportedly sufficient for XRCC4 binding, it was proposed initially that part of the linker region (i.e., where the XIR is located) was also absent in the resulting truncated protein (33) and, later, that the region encompassing residues 800 to 814 (i.e., between the XIR and BRCT2) exerted an inhibitory effect on the interaction with XRCC4 (17). Our structural, biochemical, and cellular data now establish that the loss of the BRCT2 domain in LigIVR814X weakens its interaction with XRCC4 and compromises NHEJ activity. Notably, the XRCC4/LigIV1-813 recombinant complex could still be purified from insect cells (17), suggesting again that it is only in the context of mammalian cells and/or normal expression levels that LigIV’s BRCT2 domain is specifically required for XRCC4 interaction. Furthermore, the phenotype exhibited by LigIVR814X/R580X cells, together with our data, strongly suggests that the BRCT2 stabilizing effect mainly stems from interactions through the ␣1⬘ and ␣3⬘ helices, with only a minor participation of the N-terminal loop (residues Ser804 to Arg814). Nevertheless, we cannot strictly rule out a potentially important role for the bordering Arg814 (Fig. 3C). Consequently, our crystallographic and functional findings complementarily provide a comprehensive rationale for explaining the phenotypic characteristics of a subset of NHEJ-deficient LIG4 patients.

MOL. CELL. BIOL.

Pharmacological prospects. Given its crucial role in DSB repair, the NHEJ machinery represents a potentially important target for adjuvant treatments aimed at radiosensitizing tumor cells and improving radiotherapy efficacy. In this respect, most of the strategies developed to date have focused on targeting DNA-PK subunits, particularly by devising specific DNA-PKcs kinase inhibitors. Being much less abundant than DNA-PK and possibly the limiting component of the NHEJ reaction (28), the ligation complex appears to be an appealing alternative target. Moreover, XRCC4 and LigIV function proved to be more specifically restricted to NHEJ and V(D)J processes than Ku70/Ku80 and DNA-PKcs, which are also involved in several additional cellular functions, including telomere maintenance, transcription, and apoptosis (13). Consequently, targeting XRCC4/LigIV could achieve a more-selective toxicity on irradiated tumoral tissues and, hopefully, less adverse effects on unirradiated areas. Here, we demonstrated that ectopic overexpression of XIR-BRCT2 containing LigIV fragments could compete efficiently with endogenous LigIV for XRCC4 binding, resulting in LigIV degradation and subsequent NHEJ impairment and radiosensitization. The radiosensitivity achieved was higher than that obtained with our short hairpin RNA against LigIV, although quite efficient, and similar to that observed in 2303 cells from LIG4 patients when compared with that of the rescued cells (33). A symmetrical approach using a dominant-negative XRCC4 fragment has been reported (26). Although promising, the radiosensitizing effect obtained was modest (DMF of ⬃1.2 to 1.3) and not mechanistically correlated to NHEJ interference. Due to the stoichiometry and the structure of the complex, it is possible that LigIV is a more-appropriate target than XRCC4, at least for a dominant-negative strategy. Moreover, it is remarkable that to avoid concatenation of their genome in infected cells, adenoviruses have evolved a striking DSB repair inhibitory strategy based on proteasomal degradation of LigIV, further emphasizing its critical role in the NHEJ process (3). Based on the high-resolution crystal structure reported here, one can now consider a rational screening assay for compounds that will interfere with XRCC4/LigIV assembly by targeting the helix-loop-helix or BRCT2 interaction surface. Further investigation is now required to optimize the design and use of such competitors. Tandem BRCT domains of LigIV mediate a unique mode of protein-protein interaction. When present in tandem, BRCT domains within known structures are found directly associated with one another in a conserved head-to-tail arrangement which forms an interface that mediates interaction with phosphoserine-containing binding partners (18). The BRCT repeat of human LigIV has been shown to similarly bind phosphoserine-containing peptides in vitro (36, 43). Moreover, the phosphoserine-binding pocket (P1) found in BRCT domains and formed by three conserved residues (SGK motif) is also present within BRCT1 of LigIV (S668 and G669 from the ␤1-␣1 loop and K710 from ␣2) (4). To further assess this motif, we superimposed each LigIV BRCT domain individually onto the BRCT domains of BRCA1 (see Fig. S4 in the supplemental material). Importantly, C␣ atoms of residues specifically involved in phosphoserine binding superimpose with an RMSD of only 1.4 Å. Together, the sequence and structural conservation of the P1 phosphoserine-binding pocket suggest a sim-

VOL. 29, 2009


ilar interaction for BRCT1 of LigIV. Amino acid substitutions within the conserved SGK motif failed, however, to exhibit any obvious defect when tested for their ability to ligate coding sequence in a V(D)J recombination assay (see Tables SII and SIII in the supplemental material). Nonetheless, the V(D)J recombination assay in a murine cellular system might not be fully appropriate to reveal all the potential functions associated with a BRCT1-mediated phosphoserine interaction. Importantly, we show here that the region of LigIV that is amino terminal to the XIR and encompasses BRCT1 and the ␣Ladaptor interacts with Cer-XLF. BRCT1 has also been shown to directly interact with Ku (32). Further investigation is now required to precisely determine the role of BRCT1 in the architecture of the XRCC4/LigIV/Cer-XLF complex and/or in contacting additional partners in either a phosphodependent or independent manner. Together our findings provide important structural and functional clues for understanding the human XRCC4/LigIV/ Cer-XLF complex architecture by uncovering an unexpected dual mode of interaction of the LigIV C-terminal tandem BRCT repeats, where BRCT1-XIR and XIR-BRCT2 interact separately with Cer-XLF and XRCC4, respectively. ACKNOWLEDGMENTS We are indebted to T. Lindahl (Cancer Research UK London Research Institute) and D. Biard (Commissariat à l’Energie Atomique) for the gift of plasmids, P. R. Hamann (Wyeth Research) for the gift of calicheamicin-␥1, and A. Sarasin (Institut Gustave Roussy) for the gift of immortalized MRC5 cells. This work was supported by a grant (MOP-53209) to M.S.J. from the Canadian Institutes of Health Research and, partly, by grants from the Association pour la Recherche sur le Cancer, the Ligue Nationale Contre le Cancer (e´quipe labeliseé), the Commissariat à l’Energie Atomique, and a Radiobiology grant from Electricite´ de France. M.M. was supported by the Association for International Cancer Research (grant 01-215). Patrick Calsou is a scientist from INSERM, France. REFERENCES 1. Ahnesorg, P., P. Smith, and S. P. Jackson. 2006. XLF interacts with the XRCC4-DNA ligase IV complex to promote DNA nonhomologous endjoining. Cell 124:301–313. 2. Andres, S. N., M. Modesti, C. J. Tsai, G. Chu, and M. S. Junop. 2007. Crystal structure of human XLF: a twist in nonhomologous DNA end-joining. Mol. Cell 28:1093–1101. 3. Baker, A., K. J. Rohleder, L. A. Hanakahi, and G. Ketner. 2007. Adenovirus E4 34k and E1b 55k oncoproteins target host DNA ligase IV for proteasomal degradation. J. Virol. 81:7034–7040. 4. Birrane, G., A. K. Varma, A. Soni, and J. A. Ladias. 2007. Crystal structure of the BARD1 BRCT domains. Biochemistry 46:7706–7712. 5. Bryans, M., M. C. Valenzano, and T. D. Stamato. 1999. Absence of DNA ligase IV protein in XR-1 cells: evidence for stabilization by XRCC4. Mutat. Res. 433:53–58. 6. Buck, D., L. Malivert, R. de Chasseval, A. Barraud, M. C. Fondaneche, O. Sanal, A. Plebani, J. L. Stephan, M. Hufnagel, F. le Deist, A. Fischer, A. Durandy, J. P. de Villartay, and P. Revy. 2006. Cernunnos, a novel nonhomologous end-joining factor, is mutated in human immunodeficiency with microcephaly. Cell 124:287–299. 7. Calsou, P., C. Delteil, P. Frit, J. Drouet, and B. Salles. 2003. Coordinated assembly of Ku and p460 subunits of the DNA-dependent protein kinase on DNA ends is necessary for XRCC4-ligase IV recruitment. J. Mol. Biol. 326:93–103. 8. Costantini, S., L. Woodbine, L. Andreoli, P. A. Jeggo, and A. Vindigni. 2007. Interaction of the Ku heterodimer with the DNA ligase IV/Xrcc4 complex and its regulation by DNA-PK. DNA Repair (Amsterdam) 6:712–722. 9. Critchlow, S. E., R. P. Bowater, and S. P. Jackson. 1997. Mammalian DNA double-strand break repair protein XRCC4 interacts with DNA ligase IV. Curr. Biol. 7:588–598. 10. DeFazio, L. G., R. M. Stansel, J. D. Griffith, and G. Chu. 2002. Synapsis of DNA ends by DNA-dependent protein kinase. EMBO J. 21:3192–3200.

3171

11. Deshpande, R. A., and T. E. Wilson. 2007. Modes of interaction among yeast Nej1, Lif1 and Dnl4 proteins and comparison to human XLF, XRCC4 and Lig4. DNA Repair (Amsterdam) 6:1507–1516. 12. Dore´, A. S., N. Furnham, O. R. Davies, B. L. Sibanda, D. Y. Chirgadze, S. P. Jackson, L. Pellegrini, and T. L. Blundell. 2006. Structure of an Xrcc4-DNA ligase IV yeast ortholog complex reveals a novel BRCT interaction mode. DNA Repair (Amsterdam) 5:362–368. 13. Downs, J. A., and S. P. Jackson. 2004. A means to a DNA end: the many roles of Ku. Nat. Rev. Mol. Cell Biol. 5:367–378. 14. Drouet, J., C. Delteil, J. Lefrancois, P. Concannon, B. Salles, and P. Calsou. 2005. DNA-dependent protein kinase and XRCC4-DNA ligase IV mobilization in the cell in response to DNA double strand breaks. J. Biol. Chem. 280:7060–7069. 15. Dvir, A., S. R. Peterson, M. W. Knuth, H. Lu, and W. S. Dynan. 1992. Ku autoantigen is the regulatory component of a template-associated protein kinase that phosphorylates RNA polymerase II. Proc. Natl. Acad. Sci. USA 89:11920–11924. 16. Foster, R. E., C. Nnakwe, L. Woo, and K. M. Frank. 2006. Monoubiquitination of the nonhomologous end joining protein XRCC4. Biochem. Biophys. Res. Commun. 341:175–183. 17. Girard, P. M., B. Kysela, C. J. Harer, A. J. Doherty, and P. A. Jeggo. 2004. Analysis of DNA ligase IV mutations found in LIG4 syndrome patients: the impact of two linked polymorphisms. Hum. Mol. Genet. 13:2369–2376. 18. Glover, J. N., R. S. Williams, and M. S. Lee. 2004. Interactions between BRCT repeats and phosphoproteins: tangled up in two. Trends Biochem. Sci. 29:579–585. 19. Goodarzi, A. A., Y. Yu, E. Riballo, P. Douglas, S. A. Walker, R. Ye, C. Harer, C. Marchetti, N. Morrice, P. A. Jeggo, and S. P. Lees-Miller. 2006. DNA-PK autophosphorylation facilitates Artemis endonuclease activity. EMBO J. 25: 3880–3889. 20. Gottlieb, T. M., and S. P. Jackson. 1993. The DNA-dependent protein kinase: requirement for DNA ends and association with Ku antigen. Cell 72:131–142. 21. Grawunder, U., M. Wilm, X. Wu, P. Kulesza, T. E. Wilson, M. Mann, and M. R. Lieber. 1997. Activity of DNA ligase IV stimulated by complex formation with XRCC4 protein in mammalian cells. Nature 388:492–495. 22. Grawunder, U., D. Zimmer, S. Fugmann, K. Schwarz, and M. R. Lieber. 1998. DNA ligase IV is essential for V(D)J recombination and DNA double-strand break repair in human precursor lymphocytes. Mol. Cell 2:477–484. 23. Grawunder, U., D. Zimmer, P. Kulesza, and M. R. Lieber. 1998. Requirement for an interaction of XRCC4 with DNA ligase IV for wild-type V(D)J recombination and DNA double-strand break repair in vivo. J. Biol. Chem. 273:24708–24714. 24. Grawunder, U., D. Zimmer, and M. R. Leiber. 1998. DNA ligase IV binds to XRCC4 via a motif located between rather than within its BRCT domains. Curr. Biol. 8:873–876. 25. Hsu, H. L., S. M. Yannone, and D. J. Chen. 2002. Defining interactions between DNA-PK and ligase IV/XRCC4. DNA Repair (Amsterdam) 1:225– 235. 26. Jones, K. R., D. A. Gewirtz, S. M. Yannone, S. Zhou, D. G. Schatz, K. Valerie, and L. F. Povirk. 2005. Radiosensitization of MDA-MB-231 breast tumor cells by adenovirus-mediated overexpression of a fragment of the XRCC4 protein. Mol. Cancer Ther. 4:1541–1547. 27. Kurimasa, A., S. Kumano, N. V. Boubnov, M. D. Story, C. S. Tung, S. R. Peterson, and D. J. Chen. 1999. Requirement for the kinase activity of human DNA-dependent protein kinase catalytic subunit in DNA strand break rejoining. Mol. Cell. Biol. 19:3877–3884. 28. Lee, J. W., S. M. Yannone, D. J. Chen, and L. F. Povirk. 2003. Requirement for XRCC4 and DNA ligase IV in alignment-based gap filling for nonhomologous DNA end joining in vitro. Cancer Res. 63:22–24. 29. Li, Y., D. Y. Chirgadze, V. M. Bolanos-Garcia, B. L. Sibanda, O. R. Davies, P. Ahnesorg, S. P. Jackson, and T. L. Blundell. 2008. Crystal structure of human XLF/Cernunnos reveals unexpected differences from XRCC4 with implications for NHEJ. EMBO J. 27:290–300. 30. Modesti, M., J. E. Hesse, and M. Gellert. 1999. DNA binding of Xrcc4 protein is associated with V(D)J recombination but not with stimulation of DNA ligase IV activity. EMBO J. 18:2008–2018. 31. Modesti, M., M. S. Junop, R. Ghirlando, M. van de Rakt, M. Gellert, W. Yang, and R. Kanaar. 2003. Tetramerization and DNA ligase IV interaction of the DNA double-strand break repair protein XRCC4 are mutually exclusive. J. Mol. Biol. 334:215–228. 32. Nick McElhinny, S. A., C. M. Snowden, J. McCarville, and D. A. Ramsden. 2000. Ku recruits the XRCC4-ligase IV complex to DNA ends. Mol. Cell. Biol. 20:2996–3003. 33. O’Driscoll, M., K. M. Cerosaletti, P. M. Girard, Y. Dai, M. Stumm, B. Kysela, B. Hirsch, A. Gennery, S. E. Palmer, J. Seidel, R. A. Gatti, R. Varon, M. A. Oettinger, H. Neitzel, P. A. Jeggo, and P. Concannon. 2001. DNA ligase IV mutations identified in patients exhibiting developmental delay and immunodeficiency. Mol. Cell 8:1175–1185. 34. O’Driscoll, M., and P. A. Jeggo. 2006. The role of double-strand break repair: insights from human genetics. Nat. Rev. Genet. 7:45–54.

3172

WU ET AL.

35. Przewloka, M. R., P. E. Pardington, S. M. Yannone, D. J. Chen, and R. B. Cary. 2003. In vitro and in vivo interactions of DNA ligase IV with a subunit of the condensin complex. Mol. Biol. Cell 14:685–697. 36. Rodriguez, M., X. Yu, J. Chen, and Z. Songyang. 2003. Phosphopeptide binding specificities of BRCA1 COOH-terminal (BRCT) domains. J. Biol. Chem. 278:52914–52918. 37. Sibanda, B. L., S. E. Critchlow, J. Begun, X. Y. Pei, S. P. Jackson, T. L. Blundell, and L. Pellegrini. 2001. Crystal structure of an Xrcc4-DNA ligase IV complex. Nat. Struct. Biol. 8:1015–1019. 38. van Gent, D. C., J. H. Hoeijmakers, and R. Kanaar. 2001. Chromosomal stability and the DNA double-stranded break connection. Nat. Rev. Genet. 2:196–206.

MOL. CELL. BIOL. 39. Weterings, E., and D. J. Chen. 2008. The endless tale of non-homologous end-joining. Cell Res. 18:114–124. 40. Windhofer, F., W. Wu, and G. Iliakis. 2007. Low levels of DNA ligases III and IV sufficient for effective NHEJ. J. Cell. Physiol. 213:475–483. 41. Wu, P. Y., P. Frit, L. Malivert, P. Revy, D. Biard, B. Salles, and P. Calsou. 2007. Interplay between Cernunnos-XLF and nonhomologous end-joining proteins at DNA ends in the cell. J. Biol. Chem. 282:31937–31943. 42. Yano, K., K. Morotomi-Yano, S. Y. Wang, N. Uematsu, K. J. Lee, A. Asaithamby, E. Weterings, and D. J. Chen. 2008. Ku recruits XLF to DNA double-strand breaks. EMBO Rep. 9:91–96. 43. Yu, X., C. C. Chini, M. He, G. Mer, and J. Chen. 2003. The BRCT domain is a phospho-protein binding domain. Science 302:639–642.