Fig 2: PAXX works in both the same and parallel DNA repair pathways with XLF.(a) Sensitivity curves for PAXX−/− DT40 cells complemented by wild-type, F201A or Nmut variants of human PAXX in G1/S phase. Cells were arrested in G1/S phase before IR treatment. Mutant sites in Nmut are indicated in Supplementary Fig. 1a. (Sensitivity assays to analyse genetic interactions of PAXX with XRCC4, Lig4 (b) and XLF (d) in asynchronous DT40 cells. Survival curves were measured using MTS assay. The mean and s.d. from three independent experiments are shown. (c) A cartoon showing potential genetic relationships of PAXX with XRCC4, Lig4 and XLF under different treatment conditions.

Fig 3: DNAPKcs kinase inhibition causes an increase in HDR and structural variants (SVs), and IR sensitivity that is not further enhanced by XLF loss.a Shown is the LMNA-HDR reporter (not to scale) involving a Cas9/sgRNA that induces a DSB in LMNA exon 1, and a plasmid donor such that HDR causes mRuby2 to be expressed from the LMNA locus. The mRuby2 frequencies are normalized to transfection efficiency with parallel GFP transfections. b BRCA2 is required for HDR by the LMNA-HDR assay, but has modest effects on No Indel EJ. Cells were transfected with siRNA that is non-targeting (siCtrl) or targets BRCA2 (siBRCA2). n = 6 biologically independent transfections. Statistics with unpaired two-tailed t test using Holm–Sidak correction. ****P < 0.0001, **P = 0.00305. Immunoblot shows levels of BRCA2. c DNAPKcs, XLF, and XRCC4 suppress HDR to similar degrees. n = 6 biologically independent transfections. Statistics as in (b). ****P < 0.0001, ***P = 0.00069, **P = 0.00117, n.s. not significant. d Suppression of HDR. n = 6 biologically independent transfections. Statistics as in (b). ****P < 0.0001, Parental EV DMSO vs. 1000 nM M3814 ***P = 0.000493, XLF-KO XLF-WT DMSO vs. XLF-K160D DMSO ***P = 0.000404, XLF-KO EV DMSO vs. 1000 nM M3814 **P = 0.0203, XLF-KO XLF-WT 250 nM M3814 vs XLF-K160D 250 nM M3814 *P = 0.0249, †P = 0.0386 but not significant after correction (unadjusted P-value), n.s. not significant. e Suppression of SVs. Shown is the GFPd2-SV reporter (not to scale) that is chromosomally integrated to an FRT site in HEK293 cells. Stable expression of Cas9 and various sgRNAs (DSB-H, DSB-G, and DSB-L) can induce GFP-negative cells (GFP-neg). n = 6. Statistics as in (b). ****P < 0.0001, **P = 0.00136, n.s. not significant. f IR sensitivity. HEK293 stable cell lines were treated with DMSO or M3814, and 0 Gy or 1 Gy IR dose, and plated to form colonies. Fraction clonogenic survival was determined relative to DMSO 0 Gy for each cell line. n = 6 biologically independent wells of seeded cells. Statistics as in (b). ****P < 0.0001, n.s.= not significant. Error bars = SD. Data are represented as mean values ± SD.

Fig 4: V(D)J recombination and end-joining defects in Xlf−/−PaxxΔ/Δ cells.(a) Schematic of pMX-INV V(D)J recombination substrates15 (diagram modified from Zha et al.12). The unrearranged substrate (UR), hybrid joins (HJs), coding/signal end (CE/SE) intermediates and coding/signal joins (CJs/SJs) are diagrammed. The recombination signal sequence (RSS, triangle), GFP probe (solid lines) and hCD4 probe (dash lines) are indicated. Positions of EcoRV (RV) sites and NcoI (N) sites are shown. The red box shows the isolated SE–CE fragments resulting from complete end-ligation defects. The blue box shows the SJ–CE fragments resulting from isolated CJ formation defects (for example, hairpin opening defects in DNA-PKcs−/− cells)3. (b–d) Southern blot analyses of pMX-INV rearrangement products and intermediates. The digestion and probes are marked on the side of the blots. The WT and the two independently derived PaxxΔ/Δ lines in b have the pMX-INV substrates integrated at the same genomic location. And the Xlf−/− and isogenic Xlf−/−PaxxΔ/Δ derivative have their pMX-INV integrated in the same genomic locus. Xrcc4−/− cells were included as a control. (e) Diagram of PpoI-mediated cleavage in one mouse genomic DNA site on chromosome 1. H3=Hind III. The dark line indicates the position of the probe designed to detect DNA ends. Lower: Southern blot analyses for unrepaired ends from representative cells with HindIII-digested DNA.

Fig 5: Model. Substrate DNAs are shown at top, with ends designated as ‘a’ and ‘b’ and arrow showing a to b polarity. Complexes (I) and (II) depict side-by-side and end-to-end substrate pairing mediated by DNA ligase IV/XRCC4/XLF filament. Pairing is protein-mediated and DNA sequence-independent. Joining occurs with random polarity (equal probability of ab-ab, ba-ba, ab-ba, ba-ab joints). Complexes (III) and (IV) depict an alternative side-by side pairing geometry, which has not been observed directly but is consistent with known biochemical characteristics of SFPQ•NONO (see text). Loop structure allows joining by protomeric DNA ligase IV•XRCC4 complex (XLF not required). Joining occurs with random polarity. Complexes (V) and (VI) depict side-by-side pairing stabilized by sequence-dependent DNA strand invasion. Because substrates are constrained to a parallel geometry, inverted (ab-ba, ba-ab) joints are favored. This model is excluded by data in Figure 3. Complexes also contain Ku protein, although the stoichiometry and position relative to other factors is uncertain. For simplicity, Ku is not shown.