Main

Chromosome end-to-end fusions are inhibited by shelterin; a multi-subunit complex anchored to telomeric DNA by two Myb-containing proteins—TRF1 and TRF2 (ref. 4). Telomere fusions are executed by two independent end-joining pathways. Classical non-homologous end-joining (C-NHEJ), mediated by LIG4 and the Ku70/80 heterodimer, is primarily blocked by TRF2 (ref. 5). Conversely, alternative NHEJ (alt-NHEJ), which is dependent on LIG3 (ref. 6) and PARP1 (ref. 7), is repressed in a redundant manner2,3. Alt-NHEJ is fully unleashed after the simultaneous deletion of TRF1 and TRF2, and the creation of shelterin-free telomeres in cells deficient for Ku70 and Ku80 (also known as Xrcc6 and Xrcc5, respectively)2. This error-prone end-joining pathway mediates fusion of naturally eroded telomeres1, joining of switch regions during class-switch recombination8, and formation of chromosomal translocations in mouse cells9,10.

To characterize the differences between C-NHEJ and alt-NHEJ at dysfunctional telomeres, we determined whether the sequence of the junction between two fused telomeres differed depending on the type of repair pathway used. Telomere fusions by C-NHEJ were triggered by Cre-mediated depletion of TRF2 using previously described mouse embryonic fibroblasts (MEFs) (Trf2F/FCre-ERT2)11 (Extended Data Fig. 1a). To induce robust fusions by the alt-NHEJ pathway, we depleted the entire shelterin complex by deleting Trf1 and Trf2 from Trf1F/F Trf2F/F Ku80−/− Cre-ERT2 MEFs2 (Extended Data Fig. 1a). DNA was subjected to next-generation sequencing, and reads corresponding to telomeres were identified on the basis of the presence of at least three consecutive TTAGGG repeats. To detect rare reads containing fusion junctions, we exploited the novel sequence arrangement created by the ligation of the 3′ G-rich strand (TTAGGG-3′) to the 5′ C-rich strand (5′-CCCTAA) (Fig. 1a), and filtered reads that started with at least three G-rich repeats and ended with two or more C-rich repeats. We confirmed that this approach could successfully identify telomere fusions by comparing reads derived from Trf2-proficient and Trf2-deficient cells. Starting with a similar number of telomere-repeat containing reads, we identified >90 fusogenic events in Trf2-knockout MEFs, compared to only three events in wild-type cells (Fig. 1b). Sequence analysis of the junctions highlighted different permutations of TTAGGG/AATCCC sequences. Notably, the spectrum of the fusion junctions was different in shelterin-free settings, in which frequent non-telomeric nucleotide insertions (9 out of 46 events) were identified at fusion breakpoints (Fig. 1b–d and Supplementary Information).

Figure 1: Random nucleotide insertions at the junction of telomeres fused by alt-NHEJ.
figure 1

a, Schematic of the junction of a telomere fusion. The 3′ end of the telomeric G-rich strand of a chromosome (blue) is fused to the 5′ end of the C-rich strand of a different chromosome (red). b, Illumina sequencing to analyse telomere fusion junctions. Reads ≥3XTTAGGG consecutively were scored as derived from telomere fragments. Those with ≥3XTTAGGG on the 5′ end and ≥2XCCCTAA at the 3′ end were scored as telomere fusion junctions (see Supplementary Information). c, Examples of telomere fusions generated by C-NHEJ of TRF2-depleted telomeres. Light grey highlights fusion junctions, dark grey marks the flanking telomere repeats. d, Examples of insertions in shelterin-free Ku80-null MEFs. e, Telomere fusions in metaphase spreads from Trf1F/F Trf2F/F Ku80−/− Cre-ERT2 MEFs. Telomeres in red (peptide nucleic acid (PNA) probe) and chromosomes in blue (4′,6-diamidino-2-phenylindole; DAPI). f, Frequency of telomere fusions after the depletion of candidate polymerases. 4-OHT, 4-hydroxytamoxifen; shCtrl, control shRNA. Bars represent mean of n > 1,000 chromosome ends derived from one experiment.

PowerPoint slide

Source data

Full size image

To identify the enzyme that incorporated nucleotides at dysfunctional telomeres, we depleted known low-fidelity DNA polymerases in shelterin-free cells lacking Ku80, and analysed chromosome-end fusions on metaphase spreads. Notably, we observed a reduction in the frequency of telomere fusions in cells with reduced levels of polymerase theta (Polθ, encoded by Polq in mice) (Fig. 1e, f and Extended Data Fig. 1b). The activity of Polθ is specific to alt-NHEJ as its inhibition in Trf2-knockout cells did not affect the frequency of C-NHEJ (Fig. 2a, b and Extended Data Fig. 2a–c).

Figure 2: Polθ is required for alt-NHEJ-dependent DSB repair in mammalian cells.
figure 2

a, Metaphases from TRF2-depleted (Trf2F/F Cre-ERT2+ 4-OHT) and shelterin-free (Trf1F/F Trf2F/F Ku80−/− Cre-ERT2+ 4-OHT) MEFs infected with the indicated short hairpin RNA (shRNA). b, Quantification of telomere fusions in MEFs with the indicated treatment (mean values ± s.d. derived from six independent experiments. P = 0.003; two-tailed Student’s t-test). c, Design of the translocation assay in which DSBs are induced by Cas9-gRNA(Rosa26;H3f3b). Joining of DNA ends generates der(6) and der(11), detected by nested PCR9. d, Translocation frequency in Polq+/+ and Polq−/− cells 60 h after Cas9-gRNA(Rosa26;H3f3b) expression. Mean values ± s.d. derived from three independent experiments. P = 0.009; two-tailed Student’s t-test.

PowerPoint slide

Source data

Full size image

Polθ is an A-family DNA polymerase that exhibits low fidelity on templated DNA12, and also displays a terminal transferase-like activity that catalyses nucleotide addition in a template-independent manner13. The relevance of these activities in vivo was highlighted in Drosophila melanogaster, in which Polθ was shown to stimulate nucleotide insertions during double-stranded break (DSB) repair by alt-NHEJ14. More recently, Polθ was shown to promote end-joining of replication-associated DSBs in Caenorhabditis elegans15, preventing large deletions around G-rich DNA16. The exact function of Polθ during DSB repair in mammalian cells remains elusive.

The crucial role for Polθ at dysfunctional telomeres prompted us to test whether it is required for DSB repair at non-telomeric loci. To this end, we tested whether the depletion of Polθ affects chromosomal translocations in the context of mouse pluripotent cells, reported to be mediated by alt-NHEJ10 in a LIG3-dependent manner9. To model chromosomal translocations, we induced DSBs in the Rosa26 and H3f3B mouse loci using the CRISPR/Cas9 system (Fig. 2c). When introduced into Polq+/+ and Polq−/− cells17, the Cas9-gRNA(Rosa26;H3f3b) expression plasmid induced simultaneous cleavage of both loci with comparable efficiencies (Extended Data Fig. 2d, e). Consistent with previous reports, 24% of translocation events in Polq+/+ cells were scarred by random insertions. Interestingly, the overall frequency of translocations in cells lacking Polq was significantly reduced (Fig. 2d). Sequence analysis of residual translocations in Polq−/− cells highlighted the absence of insertions, and a concomitant decrease in micro-homology at junctions (Fig. 2d and Extended Data Figs 2, 3, 4, 5). Notably, we observed similar results when assessing translocation frequency in cells expressing a catalytically inactive form of Polθ (Extended Data Fig. 2g–k). Altogether, our data indicate that the promiscuous activity of Polθ during DSB repair contributes to the increased mutagenicity of alt-NHEJ. Importantly, our results indicate that mammalian Polθ plays a critical part by stimulating the end-joining reaction.

We next investigated the upstream signalling event(s) required for the recruitment of Polθ to DNA damage sites, induced after micro-irradiation of HeLa cells expressing Myc-tagged Polθ. Accumulation of Polθ at laser-induced DNA breaks, discerned by its co-localization with the phosphorylated histone H2AX (γ-H2AX), occurred in ∼25% of cells that stained positive for Myc (Fig. 3a, b), and was independent of either ATM or ATR signalling (Extended Data Fig. 6). Instead, co-localization of Polθ with γ-H2AX was reduced after depletion of PARP1 with short interfering RNAs (siRNAs), or after the inhibition of PARP1 activity (using KU58948) (Fig. 3a, b). In a parallel experiment that used a recently developed U2OS-DSB reporter cell line18, we were able to ascertain the localization of Myc–Polθ to bona fide DSBs, induced after FOK1 cleavage of a LacO-tagged genomic locus (Extended Data Fig. 7). In conclusion, our data suggest that PARP1, previously known to be required for alt-NHEJ7,19, facilitates the recruitment of Polθ to DSBs.

Figure 3: Polθ is recruited by PARP1 to promote alt-NHEJ at the expense of HDR.
figure 3

a, Myc–Polθ localization to DNA damage was monitored after laser micro-irradiation of HeLa cells. Cells were fixed and stained for γ-H2AX and Myc, 1 h after damage induction. b, Quantification of Polθ accumulation at sites of laser damage (mean values ± s.e.m. derived from two independent experiments). c, To test whether Polθ represses recombination at telomeres, we depleted the polymerase in shelterin-free and Lig4-deficient MEFs2, and both repair pathways were monitored using CO-FISH. White arrows indicate alt-NHEJ events, red arrows highlight HDR-mediated T-SCEs. d, Quantification of telomere fusion (alt-NHEJ) and T-SCE (HDR) in cells transduced with shRNAs against Polq, Lig3 or control shRNA. Error bars denote ± s.d. from three independent experiments. e, Immunofluorescence for RAD51 and γ-H2AX in the indicated MEFs 3 h after irradiation. f, Graph representing quantification of ionizing-radiation-induced RAD51 foci. Mean values ± s.d. derived from three independent experiments. P < 0.05, P < 0.01; two-tailed Student’s t-test.

PowerPoint slide

Source data

Full size image

Homology-directed repair (HDR) is prevalent during the S/G2 phase of the cell cycle, which coincides with the peak of alt-NHEJ activity, and these pathways also share the initial resection step mediated by MRE11 and CtIP20. To test whether inhibiting alt-NHEJ could potentially result in increased HDR, we depleted shelterin in Lig4-deficient MEFs, a genetic setting that is conducive to the activity of alt-NHEJ as well as HDR2. To investigate the relative contribution of the two repair pathways, we used a chromosome-orientation fluorescence in situ hybridization (CO-FISH) assay21, and monitored the exchange of telomeres between sister chromatids by HDR (telomere sister chromatid exchange, T-SCE), and, at the same time, measured the frequency of chromosome end–end fusion by end-joining (Fig. 3c). After depletion of shelterin from Trf1F/F Trf2F/F Lig4−/− Cre-ERT2 MEFs, ∼10% of the telomeres were processed by alt-NHEJ, whereas ∼5% of chromosome ends showed T-SCEs2 (Fig. 3c, d and Extended Data Fig. 8a–c). As expected, we observed a substantial reduction in the frequency of alt-NHEJ at shelterin-free telomeres in Lig4−/− cells that lack Polq or Lig3 (Fig. 3d and Extended Data Fig. 8a–e). Remarkably, Polq-depleted cells exhibited a concomitant increase in T-SCE, which was not evident in cells lacking Lig3 (Fig. 3d), thereby highlighting a unique role for Polθ in counteracting HDR. To gain insight into this novel Polθ function, we show that the promiscuous polymerase is not required for end-resection of DSBs (Extended Data Fig. 8f, g). Instead, its activity counteracts the accumulation of RAD51 foci (Fig. 3e, f and Extended Data Fig. 8h). To corroborate these findings, we used the traffic light reporter (TLR) system, designed to generate a flow-cytometric readout for HDR and end-joining at a site-specific DNA break induced by I-Sce1 (ref. 22). We observed that after knocking down Polq in Lig4−/− cells, resolution of the I-Sce1-induced DNA break by HDR is increased, in conjunction with a significant reduction in the frequency of alt-NHEJ (Extended Data Fig. 9).

Alt-NHEJ is often considered as a back-up choice for DSB repair, operating at the expense of genomic stability. Circumstantial evidence suggests that this pathway could be enhanced when HDR is impaired23,24. We therefore postulated that this error-prone mode of repair has an essential role in cells with compromised HDR activity. We tested this hypothesis by inhibiting Polq in cells lacking the breast cancer susceptibility genes—Brca1 and Brca2. Chromosome analysis revealed a fourfold increase in chromosomal aberrancies after Polq depletion in MEFs lacking either Brca1 or Brca2. Such aberrancies included chromatid and chromosome breaks, in addition to radial chromosome structures characteristic of Lig4-mediated processing of chromatid breaks via the C-NHEJ pathway (Fig. 4a, b and Extended Data Fig. 10a, b). Ultimately, the increased genomic instability in cells co-depleted for Polq and Brca genes compromised cellular survival. We observed that BRCA1-mutated human cells (Fig. 4c, d), and mouse cells lacking Brca1 (Extended Data Fig. 10c–f), displayed significantly reduced colony-forming capabilities after Polq impairment. Although we cannot exclude that Polθ performs additional activities required for the survival of Brca-deficient cells25, our data suggest that Polθ-mediated alt-NHEJ promotes the survival of cells with a compromised HDR pathway. In the absence of a safer means to repair breaks, alt-NHEJ may therefore prevent genomic havoc by resolving unrepaired lesions.

Figure 4: Polq inhibition in Brca-mutant cells leads to increased chromosomal aberrancies and reduced cellular survival.
figure 4

a, Analysis of genomic instability in metaphase spreads from Brca1F/FCre-ERT2 MEFs treated with shRNA against Polq or vector control. b, Quantification of breaks (chromatid and chromosome) and radials in Brca1F/FCre-ERT2 and Brca2F/FCre-ERT2 MEFs with the indicated treatment. Mean values are presented with error bars denoting ± s.d. from three independent experiments. c, Clonogenic survival after Polq depletion. Crystal violet staining of BJ-hTERT, MCF7 and HCC1937 cells treated with shRNA against Polq or vector control. d, Quantitative analyses of colony formation assay. Colonies in each control shRNA cell line were set to 100%. Colonies in shPolq-expressing cells are normalized to shCtrl. Mean values ± s.d. derived from three independent experiments. P < 0.05, P < 0.01; two-tailed Student’s t-test. e, Schematic depicting our model for the function of Polθ during DSB repair (see Supplementary Information).

PowerPoint slide

Source data

Full size image

Here we provide direct evidence linking Polθ to alt-NHEJ repair in mammalian cells (Fig. 4e). We also show that while Polθ hinders error-free repair by HDR, its activity is essential for the survival of HDR-deficient cells (Fig. 4e). The question remains as to how this promiscuous polymerase orchestrates DSB repair. After DSB formation in the S/G2 phase of the cell cycle, resection of DSBs by MRE11 and CtIP20 potentially exposes micro-homology that allows spontaneous annealing of broken DNA ends (Fig. 4e). The binding of RPA antagonizes this annealing step to promote HDR-mediated repair26,27. An opposing activity is likely to be exerted by Polθ. Placing our finding in the context of recent biochemical experiments and genetic studies in model organisms13,14,15,16, we ultimately propose a model in which Polθ exploits both its template-independent and template-dependent activities to stabilize the annealed intermediate structure and channel repair towards the alt-NHEJ pathway (Fig. 4e). Our findings that Polθ is critical for alt-NHEJ support this model, which provides a potential explanation as to how this polymerase counteracts HDR.

Finally, it is intriguing that although POLQ expression in normal human tissues is generally repressed28, it is upregulated in a wide range of human cancers and associates with poor clinical outcome in breast tumours29,30. Our findings that cells with compromised HDR activity depend on this mutagenic polymerase for survival establish a rationale for the development of Polθ-targeted approaches for cancer treatment.

Methods

Cell culture procedures

Trf2F/F Cre-ERT2, Trf1F/F Trf2F/F Ku80−/− Cre-ERT2 and Trf1F/F TrfF/F Lig4−/− Cre-ERT2 MEF lines were previously described2,11. Polq+/+ and Polq−/− MEFs were a gift from N. Shima17. Brca1F/FCre-ERT2 and Brca2F/FCre-ERT2 MEFs and U2OS-DSB reporter cells18 were a gift from R. Greenberg. Trf1F/F Trf2F/F Lig4−/− Cre-ERT2 MEFs were derived from mice that were deficient in p53. The remaining MEF lines were immortalized with pBabeSV40LargeT. MEFs were cultured in DMEM supplemented with 10–15% FBS (Gibco), 2 mM l-glutamine (Sigma), 100 U ml−1 penicillin (Sigma), 0.1 μg ml−1 streptomycin (Sigma), 0.1 mM non-essential amino acids (Invitrogen) and 1 mM sodium pyruvate (Sigma). Expression of Cre recombinase was induced by treating MEFs carrying the Cre-ERT2 allele with 0.5 μM 4-OHT (Sigma H7904) for 12 h. The t = 0 time point was set at the time of treatment with 4-OHT. BJ-hTERT and MCF7 cells were grown in DMEM supplemented with 10% FBS. HCC1937 cells were grown in RPMI medium (Gibco) containing 15% FBS. U2OS-DSB reporter cells were grown in DMEM supplemented with 10% BCS. Human HeLa cells were grown in DMEM supplemented with 10% FBS, 100 U ml−1 penicillin, 100 μg ml−1 streptomycin and 2 mM l-glutamine. Mouse embryonic stem cells were grown in DMEM supplemented with 15% FBS (ES-qualified FBS) (Gibco), 2 mM l-glutamine (Sigma), 100 U ml−1 penicillin (Sigma), 0.1 μg ml−1 streptomycin (Sigma), 0.1 mM non-essential amino acids (Invitrogen), leukaemia inhibitory factor (LIF) and 2-β-mercaptoethanol (Gibco 21985). For inhibitor experiments, PARPi (KU58948, Axon medchem), ATMi (KU-55933, Tocris) and ATRi (VE-821, Selleckchem) were all used at a final concentration of 10 μM, and were applied to culture medium 2–4 h before irradiation. For ionizing radiation treatment, cells were exposed to 1–10 Gy ionizing radiation by a Faxitron X-ray system (120 kV, 5 mA, dose rate 5 Gy min−1) and recovered for 4 h before immunofluorescence analysis.

Lentiviral delivery of shRNA

shRNA treatments were carried out before 4-OHT treatment. shRNAs (see below for a list of sequences) were introduced by two lentiviral infections at 12 h intervals using supernatant from transfected 293T cells. Parallel infection with pLK0.1 was used as a negative control. Cells were selected with puromycin for 3 days.

TLR assay

Lentiviral constructs coding for TLR (31482) and I-Sce1 with donor e-GFP (31476) were purchased from Addgene22. To avoid the confounding effect of classical-NHEJ on the repair of I-Sce1-induced DNA breaks, we stably integrated the TLR construct into Ku80−/− and Lig4−/− MEFs. The plasmid was transduced by two lentiviral infections at 12 h intervals using supernatant from transfected 293T cells. Cells with integrated TLR were selected with puromycin for 5 days. Cells were then transduced with concentrated Polq shRNA lentiviral particles followed by I-Sce1. Cells were collected 72 h later without further antibiotic selection and analysed on a BD LSRII. eGFP fluorescence, which reflects HDR repair, was measured using a 488-nm laser for excitation and a 530/30 filter for detection. mCherry fluorescence, indicative of alt-NHEJ was measured by using a 561-nm laser for excitation and a 610/20 filter for detection. Data were analysed using FloJo software.

Detection of telomeric fusions

To enrich for telomeric DNA, genomic DNA was digested with two frequent cutters (AluI and MboI) and fragments greater than 10 kilobases (kb) were isolated. The resulting DNA was used to generate a library using the NEBNext Ultra Library Prep Kit and the NEBNext Multiplex Oligos for Illumina (NEB) following the manufacturer instructions. The resulting library was run on an Illumina HiSeq platform generating 100-base-pair (bp) indexed pair-end reads.

Transient transfection of cells and laser micro-irradiation

Full-length human POLQ was cloned into pLPC-Myc vectors. HeLa cells were plated on glass-bottomed dishes (Willco Wells). Myc–Polθ constructs were transfected into HeLa and U2OS-DSB reporter cells cell with HilyMax (Dojindo) according to the manufacturer’s instruction. Then, 24 h after transfection, cells were pre-sensitized with 10 μM 5-bromo-2′-deoxyuridine (BrdU) in normal DMEM medium for 20 h. After indicated treatments, cells were damaged by laser micro-irradiation as previously described31. After laser micro-irradiation, cells were incubated for 1 h, then fixed and analysed by immunofluorescence and microscopic imaging as described below. For PARP1 siRNA experiments, cells were transfected with siCtrl (non-targeting pool, Thermo Scientific) or siPARP1 (GGGCAAGCACAGUGUCAAAUU, Sigma), for 24 h before POLQ transfections and subsequent treatments as described above.

Immunofluorescence and confocal microscopy

After the indicated treatments, cells were processed and analysed for immunofluorescence as previously described32. In brief, cells were fixed with 2% (v/v) paraformaldehyde for 15 min at room temperature. Cells were washed with PBS, permeabilized with 0.5% (v/v) Triton X-100 for 10 min, and blocked with PBS containing 3% BSA. Cells were incubated with the same buffer containing primary antibodies for 1 h at room temperature followed by secondary antibodies incubations for 1 h at room temp. Cells were imaged and analysed with Z-stacked setting using the FV10-ASW3.1 software on a Fluoview 1000 confocal microscope (Olympus). For laser line quantification, >50 cells were counted for all conditions from two independent experiments. The primary antibodies used for immunofluorescence were γ-H2AX (p Ser139) (rabbit polyclonal, Novus, NB100-384) and c-Myc (mouse monoclonal, Santa Cruz, sc-40). The secondary antibodies used for immunofluorescence were Alexa Fluor 594 (rabbit) (Invitrogen, A11037) and Alexa Fluor 488 (mouse) (Invitrogen, A11029). To analyse the recruitment of Polθ to double-stranded breaks, U2OS-DSB reporter cells expressing Myc–Polθ were analysed 4 h after treatment with shield and tamoxifen. Lastly, to analyse RAD51 foci formation and its co-localization with γ-H2AX (p Ser139) after ionizing radiation treatment, cells were treated with 0.2% Triton X-100 (in PBS) for 5 min on ice before fixation with paraformaldehyde. The primary antibodies used for RAD51 immunofluorescence were γ-H2AX (p Ser139) (mouse monoclonal, Novus, NB100-384) and RAD51 (rabbit polyclonal, Santa Cruz, sc-8349).

FISH

Cells were collected at 96 h after treatment with 4-OHT to analyse the frequency of telomere fusions. In brief, ∼80% confluent MEFs were incubated for 2 h with 0.2 μg ml−1 colcemid (Sigma). The cells were collected by trypsinization, resuspended in 0.075 M KCl at 37 °C for 30 min, and fixed overnight in methanol/acetic acid (3:1) at 4 °C. The cells were dropped onto glass slides and the slides were dried overnight. The next day, the slides were rehydrated with PBS for 15 min then fixed with 4% formaldehyde for 2 min at room temperature. Slides were digested with 1 mg ml−1 pepsin, pH 2.2, at 37 °C for 10 min, washed three times with PBS and fixed again in 4% formaldehyde for 2 min at room temperature. After three PBS washes, the slides were incubated consecutively with 75%, 95% and 100% ethanol and allowed to air dry for 30 min before applying hybridization solutions (70% formamide, 1 mg ml−1 blocking reagent (Roche), 10 mM Tris-HCl, pH 7.2) containing TAMRA-OO-(TTAGGG)3 PNA probes (Applied Biosystems). Slides were denatured by heating for 3 min at 80 °C and hybridized for 2 h at room temperature. The next day, the slides were washed twice for 15 min each in 70% formamide, 10 mM Tris-HCl, followed by three 5-min washes in 0.1 M Tris-HCl, pH 7.0, 0.15 M NaCl and 0.08% Tween-20. Chromosomal DNA was counterstained with DAPI during the second PBS wash. Slides were mounted in antifade reagent (ProLong Gold, Invitrogen) and images were captured with a Nikon Eclipse TI microscope (see http://delangelab.rockefeller.edu/protocols).

CO-FISH

Cells were labelled with BrdU:BrdC (3:1, final concentration 10 μM) for 14–16 h. Two hours before collection by trypsinization, 0.2 μg ml−1 colcemid was added to the media. To fix the cells and drop metaphases on a glass slide, the same procedure that was applied for FISH was followed. Slides were treated with 0.5 mg ml−1 RNase A (in PBS, DNase-free) for 10 min at 37 °C, incubated with 0.5 μg ml−1 Hoechst 33258 (Sigma) in 2×SSC for 15 min at room temperature, and exposed to 365-nm ultraviolet light (Stratalinker 1800 UV irradiator) for 30 min. The slides were then digested twice with 800 U exonuclease III (Promega) at room temperature for 10 min each, washed with PBS and dehydrated through an ethanol series of 70%, 95% and 100%. After air-drying, slides were hybridized with Tamra-OO-(TTAGGG)3 PNA probe in hybridization solution (70% formamide, 1 mg ml−1 blocking reagent (Roche) and 10 mM Tris-HCl, pH 7.2) for 2 h at room temperature. The slides were then washed for a few seconds with 70% formamide and 10 mM Tris-HCl, pH 7.2, and incubated with FITC-OO-(CCCTAA)3 PNA probe in hybridization solution for 2 h. Slides were washed and mounted as described for FISH (see http://delangelab.rockefeller.edu/protocols).

Western blot analysis

Cells were collected by trypsinization, lysed in 2× Laemmli buffer (100 mM Tris-HCl, pH 6.8, 200 μM dithiothreitol, 3% SDS, 20% glycerol and 0.05% bromophenol blue) at 1 × 104 cells per microlitre. The lysate was denatured for 10 min at 95 °C, and sheared by forcing it through a 28-gauge insulin needle ten times. Lysate from 1 × 105 cells was loaded on an SDS–PAGE and transferred to a nitrocellulose membrane. The membrane was blocked in 5% milk in TBS with 0.1% Tween-20 and incubated with primary antibody in TBS, 5% milk and 0.1% Tween-20 for 2 h at room temperature. The following primary antibodies were used: Polθ (ab80906, Abcam); TRF1 (1449, rabbit polyclonal); RAP1 (1252, rabbit polyclonal); phospho-CHK2 (Thr68) (rabbit polyclonal, Cell Signaling); CHK2 (rabbit polyclonal, Cell Signaling); phospho-CHK1 (Ser 345) (mouse monoclonal, Cell Signaling); CHK1 (mouse monoclonal, Santa Cruz); LIG3 (mouse monoclonal, Santa Cruz); Myc (9E10; Calbiochem); and γ-tubulin (clone GTU-88, Sigma); PARP1 (polyclonal, Cell signaling). (See http://delangelab.rockefeller.edu/protocols.)

Chromosomal aberrancies

Cells were collected and dropped on microscope slides as described for the FISH protocol. After the slides had dried overnight, they were re-hydrated in PBS, stained with 0.25 μg ml−1 DAPI, dehydrated in a 70%, 95% and 100% ethanol series, mounted and imaged using Nikon Eclipse TI microscope. Aberrancies were scored as a percentage of chromatid breaks, chromosome breaks, and chromosome radial structures compared to total number of chromosomes.

Chromosomal translocation assay

Induced pluripotent stem cells were derived from primary Polq+/+ and Polq−/− MEFs according to standard Yamanaka protocol33. To perform the translocation assay, Polq+/+ and Polq−/− induced pluripotent stem cells were transfected with 2 μg of Cas9-gRNA(Rosa26;H3f3b) plasmid per million cell. We constructed Cas9-gRNA(Rosa26;H3f3b) by introducing two guide RNAs (5′-GTTGGCTCGCCGGATACGGG-3′ for H3f3b; 5′-ACTCCAGTCTTTCTAGAAGA-3′ for Rosa26) into pX330 (Addgene, 42230). After transfection, 1 × 104 cells were seeded per well in a 96-well plate, and lysed 3 days later in 40 μl lysis buffer (10 mM Tris, pH 8.0, 0.45% Nonidet P-40 and 0.45% Tween 20). The lysate was incubated with 200 μg ml−1 proteinase K for 2 h at 55 °C. Translocation detection was performed according to previously established protocol9, using nested PCR. The primers used for the first PCR reaction were Tr6-11-Fwd: 5′-GCGGGAGAAATGGATATGAA-3′ and Tr6-11-Rev: 5′- TTGACGCCTTCCTTCTTCTG-3′ for der(6), and Tr11-6-Fwd: 5′-AACCTTTGAAAAAGCCCACA-3′ and Tr11-6-Rev: 5′-GCACGTTTCCGACTTGAGTT-3′, for der(11). For the second round of PCR amplification we used the primers Tr6-11NFwd: 5′-GGCGGATCACAAGCAATAAT-3′ and Tr6-11NRev: 5′-CTGCCATTCCAGAGATTGGT-3′, and Tr11-6NFwd: 5′-AGCCACAGTGCTCACATCAC-3′ and Tr11-6NRev: 5′TCCCAAAGTCGCTCTGAGTT-3′. The number of PCR-positive wells was used to calculate the translocation frequency as previously described9. Amplified products from positive wells were sequenced to verify translocations and determine the junction sequences.

Surveyor assay

Forty-eight hours after transfection, genomic DNA was extracted with GE Healthcare Illustra Genomic Prep Mini Spin Kit (28-9042-76). The genomic region encompassing the guide RNA target sites was amplified using Q5 High-Fidelity DNA polymerase (New England BioLabs) with the primers Rosa26-Fwd: 5′-TAAAACTCGGGTGAGCATGT-3′ and Rosa26-Rev: 5′-GGAGTTCTCTGCTGCCTCCTG-3′, and H3f3b-Fwd: 5′-GCGGCGGCTTGATTGCTCCAG-3′ and H3f3b-Rev: 5′-AGCAACTTGTCACTCCTGAGCCAC-3′. PCR fragments were gel purified and the surveyor assay was performed using a detection kit (Transgenomic), according to manufacturer’s instructions. Agarose gels (2%) were used to visualize the bands after surveyor digestion.

Colony formation assay

After lentiviral transduction with shCtrl or (sequences listed below), cells were selected with puromycin (BJ: 0.5 μg ml−1; MCF7 and HCC1937: 1 μg ml−1; MEFs: 2 μg ml−1) for 72 h and plated in 6-cm dishes (1,000 and 10,000 cells per plate). After 10–14 days, colonies were fixed with 3% paraformaldehyde (5 min), rinsed with PBS, and stained with crystal violet (Sigma-Aldrich).

CRISPR targeting to mutate Polq gene in mouse embryonic stem cells

To generate cells carrying a catalytic dead Polθ, two mutations at residues Asp2494Gly and Glu2495Ser (ref. 34) were introduced in the endogenous Polq locus in mouse embryonic stem cells using CRISPR/Cas9 gene targeting. Two guides RNAs were co-transfected with a Cas9-nickase (pX335-U6-Chimeric_BB-CBh-hSpCas9n(D10A)), and a donor cassette that introduces a SacII restriction site while replacing the two amino acid residues. Clonal cells lines were derived and genotyped to determine successful targeting. Two independent clonally derived lines were used for the analysis of translocation.

shRNA target sequence (pLK0.1 vector)

shPolm: 5′-CTCACCTCTCACACACCATAA-3′; shPolk: 5′-GCCCTTAGAAATGTCTCATAA-3′; shPolh: 5′-GCTCGATTCTCCAGCTTACAA-3′; shPoli: 5′-AGTGAAGAAGATACGTTTAAA-3′; shPolb: 5′-CCAAAGTTGTTACATCGTGTT-3′; shPoln: 5′-CCTACTCACATGAAGGACATT-3′; shLig3 (mouse): 5′-CCAGACTTCAAACGTCTCAAA-3′; shLIG3 (human): 5′-CCGGATCATGTTCTCAGAAAT-3′; shPolq-1 (mouse): 5′-CGGCGGAGTATGAGAACTATT-3′; shPolq-2(mouse): 5′-CCAGGAATCAAAGACGACAAT-3′; shPolq-3(mouse): 5′-CCTGGCTGAATGCTGAACTTT-3′; shPOLQ (human): 5′-CGGGCCTCTTTAGATATAAAT-3′; shBrca1(mouse): 5′-GCTCAGTGTATGACTCAGTTT-3′.

Primers for quantitative PCR

BRCA1Fwd: 5′-CTGCCGTCCAAATTCAAGAAGT-3′ and BRCA1Rev: 5′-CTTGTGCTTCCCTGTAGGCT-3′; BRCA2Fwd: 5′-TGTGGTAGATGTTGCTAGTCCGCC-3′ and BRCA2Rev: 5′-GCTTTTCTCGTTGTAGTACTGCC-3′; POLbFwd: 5′-TGAACCATCATCAACGAATTGGG-3′ and POLbRev: 5′-CCATGTCTCCACTCGACTCTG-3′; POLmFwd: 5′-AGGCTTCCGCGTCCTAGAT-3′ and POLmRev: 5′-GTGGGGAGAGCATCCATGTT-3′; POLkFwd: 5′-AGCTCAAATTACCAGCCAGCA-3′ and POLkRev: 5′-GGTTGTCCCTCATTTCCACAG-3′; POLhFwd: 5′-ATCGAGTGGTTGCTCTTGTAGA-3′ and POLhRev: 5′-CCAAATGCTCGGGCTTCATAG-3′; POLiFwd: 5′-GCAGTCAAGGGCCACCTAC-3′ and POLiRev: 5′-AGGTCTGTCCTTTAATTCTGGGT-3′; POLnFwd: 5′-AGCTGATGGATGCTCTCAAGCAGG-3′ and POLnRev: 5′-GAGTCAGAGTGCTGTTGCCTACATGG-3′; LIG3(mouse)Fwd: 5′-GAAGAAAGCTGCTGTCCAGG-3′ and LIG3(mouse)Rev: 5′-CAGAGTTGTTGGGTTTTGCTG-3′; LIG3(human)Fwd: 5′-GAAGAGCTGGAAGATAATGAGAAGG-3′ and LIG3(human)Rev: 5′-AGTGGTTGTCAACTTAGCCTGG-3′; POLQ(mouse)Fwd: 5′-CAAGGTTTCATTCGGGTCTTGG-3′ and POLQ (mouse)Rev: 5′-CGAGCAGGAAGATTCACTCCAG-3′; POLQ(human) Fwd: 5′-CAGCCCTTATAGTGGAAGAAGC-3′ and POLQ(human) Rev: 5′-GCACATGGATTCCATTGCACTC-3′.

Statistical analysis

Results are presented as mean ± s.d. of two or three independent experiments unless otherwise stated. P < 0.05 was considered statistically significant, and P values were calculated using the two-tailed Student’s t-test. No statistical methods were used to predetermine sample size.