PAXX, a paralog of XRCC4 and XLF, interacts with Ku to promote DNA double-strand break repair

See allHide authors and affiliations

Science  09 Jan 2015:
Vol. 347, Issue 6218, pp. 185-188
DOI: 10.1126/science.1261971

A factor for repairing broken DNA

Unprogrammed DNA double-strand breaks are extremely dangerous for genomic stability. Nonhomologous end-joining (NHEJ) repair systems are present in all domains of life and help deal with these potentially lethal lesions. Ochi et al. have discovered a new factor involved in NHEJ by searching for proteins with structural similarities to known NHEJ proteins. Specifically, PAXX, a paralog of XRCC1 and XLF, interacts with a key repair pathway protein, Ku, and helps promote ligation of the broken DNA.

Science, this issue p. 185


XRCC4 and XLF are two structurally related proteins that function in DNA double-strand break (DSB) repair. Here, we identify human PAXX (PAralog of XRCC4 and XLF, also called C9orf142) as a new XRCC4 superfamily member and show that its crystal structure resembles that of XRCC4. PAXX interacts directly with the DSB-repair protein Ku and is recruited to DNA-damage sites in cells. Using RNA interference and CRISPR-Cas9 to generate PAXX−/− cells, we demonstrate that PAXX functions with XRCC4 and XLF to mediate DSB repair and cell survival in response to DSB-inducing agents. Finally, we reveal that PAXX promotes Ku-dependent DNA ligation in vitro and assembly of core nonhomologous end-joining (NHEJ) factors on damaged chromatin in cells. These findings identify PAXX as a new component of the NHEJ machinery.

DNA double-strand breaks (DSBs) are toxic lesions that arise in cells exposed to agents such as ionizing radiation (IR) and are also generated as intermediates during V(D)J (variable, diversity, joining) and class-switch recombination at immune-receptor gene loci (1). If unrepaired or repaired incorrectly, DSBs cause cell death or genome instability, and defects in DSB repair components cause hereditary disorders with symptoms including immunodeficiency, neurodegeneration, infertility, and/or increased cancer predisposition (2). A key DNA repair pathway is nonhomologous end-joining (NHEJ), which involves initial recognition of a DSB by the Ku70-Ku80 heterodimer, followed by the assembly of additional factors including the DNA-dependent protein kinase catalytic subunit (DNA-PKcs), x-ray cross-complementing protein 4 (XRCC4), and XRCC4-like factor (XLF; also called Cernunnos), with XRCC4 playing a prime role in recruiting DNA ligase IV (LIG4) to carry out the DSB-joining reaction (3).

XRCC4 and XLF, together with spindle-assembly abnormal protein 6 (SAS6), comprise a homologous superfamily of structurally related proteins (4). To identify other members of this protein superfamily, we used a bioinformatics approach, which suggested that the 22-kD human protein, C9orf142, could be an XRCC4 paralog (tables S1 and S2). Although little overall primary sequence similarity is seen, sequence alignments indicated that the N-terminal portion of C9orf142 contains the present-in-SAS6 (PISA) motif that is conserved throughout the XRCC4 superfamily (5) and shares a conserved tryptophan in this motif with XRCC4 and XLF (Fig. 1A and fig. S1A). We have therefore named this previously uncharacterized protein PAXX (PAralog of XRCC4 and XLF). Although PAXX has a wide evolutionary distribution (fig. S1B), we could not identify PAXX orthologs in insects or fungi.

Fig. 1 Crystal structure of PAXX.

(A) Domain architecture of human PAXX and other XRCC4 superfamily members. Sequence identities between human PAXX and XRCC4, XLF, and SAS6 are 10.9%, 11.2%, and 10.1%, respectively. (B) Structure of a PAXX dimer with the two polypeptide chains shown in cyan and lavender. The N and C termini of the structure are indicated as N-ter and C-ter, respectively. (C) Residues mediating the PAXX dimerization interface, with β-sheet sandwichlike packing between protomers in adjacent asymmetric units. The packing of each strand from different protomers is shown in surface and stick representations. Black dotted lines are hydrogen bonds between side-chain pairs (S99 and S95, T97 and T97, S95 and S99) and between the side chain of D108 of each protomer with the main chain of A104 of the adjacent protomer in the crystals. (D) ES-MS profile showing that PAXX1-204 is a dimer. Three charge states are observed for the dimer. The main charge state 13+ is labeled in the mass spectrum. (E) Comparison of XRCC4 superfamily members. The head domains of PAXX (cyan), XRCC4 (silver), XLF (magenta), and SAS6 (green) are superimposed.

We solved crystal structures of PAXX residues 1 to 145, 1 to 166, and the full-length protein (PAXX1-145, PAXX1-166, and PAXX1-204) at resolutions of 2.46, 2.35, and 3.45 Å respectively (table S3). PAXX residues 1 to 113 form a head domain that is structurally closely related to those of XRCC4, XLF, and SAS6 (611) (Fig. 1, A and B). This is followed by a 31–amino acid α helix, forming a coiled-coil with the other protomer to make a PAXX homodimer (in the asymmetric unit) in a manner similar to the formation of dimeric interfaces formed by XRCC4, XLF, and SAS6. Contacts within the PAXX crystal indicate that the protein has potential to form higher-order protein filaments with similarities to those of its paralogs (fig. S2A), in which two β sheets, each made up of strands β5 to β7 from different PAXX dimers, form a β sandwich around a dyad axis running between the sheets and orthogonally to the strands (Fig. 1C). However, small-angle x-ray scattering (SAXS) data of PAXX1-145 in solution are well explained by the scattering curve calculated from the structure of the dimer of the construct (fig. S2B), and electrospray mass spectroscopy (ES-MS) confirmed that PAXX predominantly forms dimers in solution under the conditions used (Fig. 1D and table S4). These data thus suggest that the preferred native state for PAXX is a dimer. Although our analyses indicate that the conformation of PAXX is distinct from those of other XRCC4 superfamily members, its overall structural properties most resemble those of XRCC4 (Fig. 1E and fig. S2C).

To gain insights into PAXX function, we expressed green fluorescent protein (GFP), GFP-tagged wild-type (WT) PAXX (GFP-PAXXWT), and GFP-tagged C-terminally truncated PAXX1-145 in human embryonic kidney HEK293FT cells, purified these, and identified potential binding partners by mass spectrometry. This revealed that the only proteins that bound GFP-PAXXWT, but not GFP-PAXX1-145 or GFP, corresponded to Ku70 and Ku80. Immunoprecipitation and Western blot analyses confirmed this interaction and established that endogenous PAXX and Ku interact (Fig. 2, A and B). Also, in reciprocal experiments, Ku70 (GFP-tagged at its endogenous gene locus) interacted with PAXX (fig. S3A). Given that GFP-PAXX1-145 did not interact with Ku and because the extreme C terminus of PAXX has been highly conserved through evolution (Fig. 2C), we speculated that PAXX C-terminal residues might mediate Ku binding. To test this, we synthesized a biotinylated peptide encompassing PAXX residues 177 to 204 and found that it specifically retrieved two major proteins that were identified by mass spectrometry to be Ku70 and Ku80 (Fig. 2D); this was confirmed by Western blotting (fig. S3B). Furthermore, mutating two of the most highly conserved residues in the PAXX C terminus (V199 and F201) to alanine in GFP-PAXXWT or biotinylated PAXX177-204 abolished interaction with Ku (Fig. 2E and fig. S3B). In line with these findings, addition of PAXX177-204 peptide to cell lysates inhibited the interaction of PAXX with Ku (fig. S3C), and surface plasmon resonance (SPR) studies with purified proteins established that DNA-bound Ku and PAXXWT interacted directly, whereas PAXXV199A/F201A did not bind to Ku-DNA detectably (fig. S3, D and E). Thus, PAXX binding to Ku-DNA is direct and is mediated by the PAXX C terminus.

Fig. 2 The PAXX C terminus interacts with Ku.

(A) GFP pull-down assays showing that GFP-PAXXWT, but not GFP-PAXX1-145, transiently overexpressed in HEK293FT cells interacts with Ku. (B) Coimmunoprecipitation (IP) from HeLa nuclear extracts showing that endogenous PAXX and Ku interact. (C) Sequence alignment of the C termini of PAXX orthologs. Conserved residues are indicated with reverse shading, and similar residues are highlighted in gray. (D) Peptide pull-down assays from HeLa nuclear extracts using control (H2AX) or PAXX177-204 peptides analyzed by silver staining. “M” represents protein markers, and numbers represent molecular masses in kilodaltons. (E) GFP pull-down assays showing that PAXX residues V199 and F201 are required for Ku binding in the context of full-length PAXX.

In vivo, PAXX was predominantly nuclear (fig. S4, A and B), and GFP-PAXX localized to DNA damage generated by laser microirradiation of live cells (Fig. 3A). To test if PAXX might be involved in DSB repair by NHEJ, we depleted human U2OS cells (a human osteosarcoma cell line) of PAXX, using multiple small-interfering RNAs (siRNAs) (fig. S4C), then performed clonogenic survival assays after exposing the cells to IR. Cells in which PAXX was depleted were significantly more radiosensitive than control cells and displayed sensitivities similar to those of cells depleted of XRCC4 (Fig. 3B and fig. S4D). Furthermore, expression of PAXXWT, but not PAXXV199A/F201A, restored IR resistance in PAXX-depleted cells (Fig. 3B). As PAXXV199A/F201A is impaired for Ku binding, these data support a model in which the PAXX-Ku interaction is crucial for PAXX function in DNA repair.

Fig. 3 PAXX is required for DSB repair in human cells.

(A) GFP-tagged PAXX accumulates at sites of laser microirradiation in U2OS cells. White arrowheads indicate the path of the laser used to induce DSBs. (B) Clonogenic survival assay showing that PAXX depletion in U2OS cells causes radiosensitivity and that this is rescued by exogenous expression of PAXXWT but not PAXXV199A/F201A. In this experiment and those below, error bars represent the standard error of the mean (SEM) from three independent experiments. (C) Clonogenic survival assay showing that PAXX−/− cells are radiosensitive and that PAXX loss is epistatic with XRCC4 depletion. (D) PAXX−/− cells display persistent γ-H2AX foci after IR. Cells with >5 foci were scored as positive and were costained with cyclin A to eliminate S and G2 cells from analysis. At least 100 cells were scored per condition. “cl.” indicates clone number. (E) PAXX is required for cellular DSB repair as measured by neutral comet assay. R/D ratios represent mean tail length of cells treated with 40 μg/ml phleomycin for 2 hours and allowed to recover (R) for 2 hours over mean tail length of cells damaged (D) for 2 hours without recovery.

To verify and extend the above conclusions, we used CRISPR-Cas9 gene editing (12) in nontransformed human retinal pigmented epithelial (RPE-1) cells to generate PAXX−/− clones (fig. S5). Like siRNA-treated U2OS cells, PAXX−/− cells were hypersensitive to IR and the radiomimetic drug phleomycin (Fig. 3C and fig. S6B). Furthermore, by depleting XRCC4 or XLF in PAXX+/+ or PAXX−/− cells (fig. S6C), we established that combined loss of PAXX and XRCC4, or PAXX and XLF, did not cause IR sensitivity greater than that of PAXX−/− cells or cells depleted of XRCC4 or XLF alone (Fig. 3C and fig. S6D), which implies that PAXX functions epistatically with XRCC4 and XLF to promote IR resistance via classical NHEJ. Given that NHEJ-deficient cells display defective resolution of IR-induced γH2AX foci, we compared the appearance and disappearance of these foci by immunofluorescence microscopy in PAXX+/+ and PAXX−/− cells. Although PAXX loss did not impair γH2AX focus formation, we observed a substantial defect in γH2AX focus resolution in multiple PAXX−/− clones (Fig. 3D). Furthermore, both PAXX−/− RPE-1 and PAXX-depleted U2OS cells were impaired in repairing DSBs, as measured by neutral comet assays (Fig. 3E and fig. S6E), and PAXX-depleted U2OS cells were also defective in random-plasmid integration, which occurs through NHEJ events (fig. S6F) (13). As with other NHEJ factors, such as XRCC4, PAXX loss did not impair checkpoint signaling (fig. S7, A and B).

Subsequent biochemical investigations established that, although PAXX does not bind DNA detectably on its own, PAXX retarded the electrophoretic mobilities of DNA complexes containing two Ku molecules (Fig. 4A and fig. S8, A to C). Furthermore, such binding was abrogated in the presence of a large excess of PAXX177-204 peptide or when the extreme PAXX C-terminal region was absent or contained alanine substitutions for V199 and F201 (Fig. 4A and fig. S8D). We next tested whether PAXX affected DNA ligation by LIG4 in vitro in a manner dependent on its ability to bind Ku. Indeed, PAXXWT, but not PAXXV199A/F201A, markedly stimulated double-stranded DNA ligation in reactions containing the XRCC4/LIG4 complex—but only in the presence of Ku (Fig. 4B). We speculated that PAXX might act as a scaffold to stabilize two Ku heterodimers at DNA ends and thus promote assembly and/or stability of the NHEJ machinery at DSB sites. To test this, we treated PAXX+/+ and PAXX−/− cells with phleomycin and examined the association of NHEJ proteins with chromatin by Western blotting. This revealed that PAXX deficiency produced substantial defects in the ability of Ku, DNA-PKcs, XRCC4, and XLF to assemble on chromatin in response to DNA damage, without affecting the overall levels of these proteins (Fig. 4C and fig. S8E).

Fig. 4 PAXX promotes NHEJ in vitro and stabilizes NHEJ proteins on damaged chromatin.

(A) Electrophoretic mobility shift analysis of Ku and PAXX derivatives with 50 base pairs of 6-FAM(6-carboxyfluorescein)–labeled DNA. PAXX (200 nM) and Ku (20 nM) were added where indicated. (B) Stimulation of DNA end ligation by PAXX. pcDNA3.1(–) (50 ng) digested by Eco RV was incubated with PAXXV199A/F201A (250 nM), PAXX (250 nM), Ku (25 nM), and XRCC4/LIG4 (25 nM) as indicated (left). Ligation efficiency (right) was calculated as a percentage of ligated plasmid from four independent assays. (C) Chromatin fractionation of PAXX+/+ and PAXX−/− cells treated with phleomycin as indicated. Note that DNA damage–dependent chromatin recruitment of proteins, such as RPA, that function in DNA repair pathways other than NHEJ, were unaffected by PAXX loss. H3, histone 3. (D) Model of PAXX in NHEJ. Two Ku-bound DNA ends are bound by a PAXX dimer at its C termini, which stabilizes the NHEJ machinery to promote DNA end ligation.

In conclusion, we have identified and characterized an XRCC4 superfamily member, PAXX. We have shown that PAXX binds Ku and promotes DSB repair at the biochemical and cellular levels and stabilizes NHEJ-protein assembly at DSB sites (Fig. 4D), which establishes PAXX as a hitherto uncharacterized NHEJ factor.

Supplementary Materials

Materials and Methods

Figs. S1 to S9

Tables S1 to S4

References (1439)

References and Notes

  1. Acknowledgments: T.O., Q.W. and T.L.B. are supported by the Wellcome Trust (WT 093167). The Jackson laboratory is funded by Cancer Research UK (CRUK) program grant C6/A11224, the European Research Council, and the European Community Seventh Framework Programme grant agreement HEALTH-F2-2010-259893 (DDResponse). Core infrastructure funding to the Jackson lab is provided by CRUK (C6946/A14492) and the Wellcome Trust (WT092096). S.P.J. receives his salary from the University of Cambridge, supplemented by CRUK. V.M.D. is a CRUK Career Development Fellow. The Draviam lab is funded by a CRUK Career Development Fellowship (C28598/A9787). We thank M. Hyvönen, V. Bolanos-Garcia, and L. Hanakahi for reagents; K. Scott for assistance with SPR; beamline scientists at I03 and I22 for their help at the Diamond Light Source; and J. Brown, K. Inoue, Y. Kimata, M. Lamers, B. Luisi, N. Lukashchuk, R. Nishi, C. le Sage, C. Schmidt, and P. Wijnhoven for useful discussions and technical assistance. Crystallization and initial x-ray diffraction experiments were performed in the x-ray crystallographic facility at the Department of Biochemistry, University of Cambridge, with help from the Facility Manager, D. Chirgadze. Protein Data Bank accession numbers of PAXX1-166 and PAXX1-204 are 3WTD and 3WTF, respectively.
View Abstract

Stay Connected to Science

Navigate This Article