Crucial role for DNA ligase III in mitochondria but not in Xrcc1-dependent repair.

Mammalian cells have three ATP-dependent DNA ligases, which are required for DNA replication and repair. Homologues of ligase I (Lig1) and ligase IV (Lig4) are ubiquitous in Eukarya, whereas ligase III (Lig3), which has nuclear and mitochondrial forms, appears to be restricted to vertebrates. Lig3 is implicated in various DNA repair pathways with its partner protein Xrcc1 (ref. 1). Deletion of Lig3 results in early embryonic lethality in mice, as well as apparent cellular lethality, which has precluded definitive characterization of Lig3 function. Here we used pre-emptive complementation to determine the viability requirement for Lig3 in mammalian cells and its requirement in DNA repair. Various forms of Lig3 were introduced stably into mouse embryonic stem (mES) cells containing a conditional allele of Lig3 that could be deleted with Cre recombinase. With this approach, we find that the mitochondrial, but not nuclear, Lig3 is required for cellular viability. Although the catalytic function of Lig3 is required, the zinc finger (ZnF) and BRCA1 carboxy (C)-terminal-related (BRCT) domains of Lig3 are not. Remarkably, the viability requirement for Lig3 can be circumvented by targeting Lig1 to the mitochondria or expressing Chlorella virus DNA ligase, the minimal eukaryal nick-sealing enzyme, or Escherichia coli LigA, an NAD(+)-dependent ligase. Lig3-null cells are not sensitive to several DNA-damaging agents that sensitize Xrcc1-deficient cells. Our results establish a role for Lig3 in mitochondria, but distinguish it from its interacting protein Xrcc1.

Biochemical and cell biological experiments implicate the nuclear Lig3-XRCC1 complex in single-strand break repair, short patch base excision repair, and nucleotide excision repair[1]. Lig3 and XRCC1 interact via C-terminal BRCT domains found in each protein[7]. This interaction is important for the stability of Lig3[7] and the recruitment of Lig3 to DNA damage foci[8]. Purified Lig3-XRCC1 is proficient at nick sealing in vitro[9], and the complex associates with several other proteins involved in single-strand break repair, including Parp1[10], aprataxin, and TDP1[1].

Lig3 also has a mitochondrial form due to an alternative translation start site, which results in a mitochondrial leader sequence (MLS)[11]. Mammals differ in this respect from budding yeast, where the Lig1 homolog, Cdc9, is the mitochondrial DNA ligase[12]. In mitochondria, Lig3 appears to act independently of XRCC1, as XRCC1 is not present in this organelle[13]. Disruption of the Lig3 gene, like Xrcc1, results in early embryonic lethality in the mouse[2,5], and Lig3 null cell lines could not be established from these animals[2]. The similar timing of lethality of Lig3 and Xrcc1 null embryos suggests that death could result from similar phenotypic consequences related to Lig3 nuclear functions in DNA repair. Alternatively, or in addition, the mitochondrial function of Lig3 may be critical for survival.

To determine whether Lig3 is an essential gene due to its nuclear and/or mitochondrial function, we developed a pre-emptive complementation strategy in mouse ES cells (Fig 1a). A Lig3 cell line was constructed which contains one conditional Lig3 allele with an intronic Neo selection marker and LoxP sites flanking exons 6 and 14 and a second allele in which these exons were already removed by Cre recombinase (Fig. 1a; Supplementary Fig. 1). These exons encode part of the DNA binding domain and the catalytic core of the protein. Cre recombinase was expressed in the Lig3 cells, and 145 clones were replica-plated in media with or without G418. No G418 sensitive clones (i.e., Lig3) were obtained (Fig. 1b), consistent with the requirement for Lig3 for cellular viability. We then stably integrated transgenes expressing wild-type, mitochondrial, or nuclear Lig3; the nuclear (NucLig3) version lacked the MLS, and the mitochondrial (MtLig3) version contained the MLS but was mutated at the nuclear translation initiation site (M88T) (Fig. 1c; Supplementary Fig. 2). GFP fusions of these proteins were also tested (Supplementary Fig. 3a).

Figure 1

Mitochondrial Lig3 activity is critical for cellular viability

(a) Pre-emptive complementation strategy for Lig3 deletion. Lig3 clones were identified by their lack of growth in G418 (neo–). PCR confirmed the genotype, as indicated.

(b) Mitochondrial Lig3 activity is critical for cellular viability. n, number of independently-derived Lig3 transgenic cell lines analyzed; Total, total number of post-Cre colonies analyzed by replica plating with and without G418; G418S, G418 sensitive; %, ratio of G418S to Total.

(c) Lig3 proteins tested for pre-emptive complementation. MLS, mitochondrial leader sequence; ZnF, zinc finger; BRCT, BRCA1 C-terminal related domain; DBD, DNA-binding domain; CC, catalytic core; *M88T, mutation of the nuclear translation initiation site.

(d) Western blot analysis showing the loss of endogenous Lig3 in Lig3 clones stably expressing GFP-tagged Lig3 or MtLig3. Lig3 is 105 kDa, whereas the GFP fusions are ~135 kDa. neo+, Lig3; KO, Lig3; α-tub, α-tubulin.

To determine which Lig3 transgenes permit the survival of cells deleted for endogenous Lig3, Cre recombinase was used to transform Lig3 cells to Lig3 cells. A large fraction of the post-Cre clones expressing wild-type Lig3 or MtLig3 were G418 sensitive (34 to 50%), whereas no G418 sensitive clones were obtained with NucLig3 (Fig. 1b; Supplementary Fig. 4). We confirmed that G418 sensitive cells were Lig3 (Fig 1a) and that endogenous Lig3 was no longer present, with the only Lig3 present in the cells expressed from the transgene (Fig. 1d). Thus, cellular viability requires mitochondrial Lig3. To determine whether DNA ligase activity was essential for cell survival, we introduced a K508V mutation that abolishes ligase adenylylation and nick-sealing[1] into MtLig3. No G418 sensitive clones were derived from 4 independent transgenic cell lines (Fig. 1b), demonstrating that the requirement for mitochondrial Lig3 depends on its ligase activity.

BRCT domains are frequently involved in protein-protein interactions, and the BRCT domain of Lig3 is known to interact with XRCC1[7]. However, as XRCC1 is not found in mitochondria[13], the role of the BRCT domain for mitochondrial function of Lig3 is uncertain. Loss of the BRCT domain had no effect on the presence of Lig3 in mitochondria (Lig3-ΔBRCT and MtLig3-ΔBRCT; Supplementary Fig. 3a; data not shown). Lig3 clones expressing Lig3-ΔBRCT or MtLig3-ΔBRCT (Fig. 2a) were recovered as a substantial fraction of clones (39 to 49%; Fig. 2b), indicating that the BRCT domain is not required for viability. Thus, MtLig3 does not have a partner protein bound to its BRCT domain that is essential for its function.

Figure 2

Mitochondrial DNA ligase activity can be provided by a variety of DNA ligases

(a) DNA ligases tested for pre-emptive complementation of Lig3 cells. ΔZnF, deletion of Lig3 amino acids 89 to 258; ΔBRCT, deletion of Lig3 amino acids 934 to 1009; ΔNLS, deletion of Lig1 amino acids 135 to147; NES, nuclear export signal from MAPKK[21]; Mt, presence of Lig3 MLS; NLS, nuclear localization signal; HhH, helix-hairpin-helix.

(b) Mitochondrial DNA ligase activity can be provided by a variety of DNA ligases. ChVLig and EcoLigA presumably enter mitochondria without the requirement for an MLS.

(c) Cells expressing exogenous DNA ligases are competent to replicate and maintain mtDNA. mtDNA copy number was quantified by qPCR by amplifying a 117 bp fragment from mtDNA. Values are presented relative to levels of mtDNA in Lig3; Lig3 GFP Tg cells. Data represent the mean of two biological repeats each determined twice by qPCR +/- SEM.

(d) Cells expressing exogenous DNA ligases showed similar capacities for mtDNA repair after oxidative damage. Cells were treated with 175 μM H2O2 for 15 min and allowed to recover for 1.5 h. To measure repair, a 10 kb mtDNA fragment was amplified following damage and quantified by qPCR. Values were normalized to the amplification of a 117 bp mtDNA fragment. Percent repair is the amount of damage remaining after 1.5 h recovery divided by the initial damage. There was no significant difference between cell lines expressing wild-type Lig3 (parental cells and transgene rescued cells) and the other ligase forms. For cells expressing MitLig1-ΔNLS, and MtChVLig two transgenic cell lines were analyzed. Data represent the mean of 2-4 determinations on multiple clones with each qPCR performed twice +/- SEM.

A unique feature of Lig3 compared with the other mammalian DNA ligases is a ZnF at its N-terminus. The Lig3 ZnF interacts with Parp1[10], and this interaction is reported to be important for the association of Lig3 with mitochondrial DNA (mtDNA)[14]. Biochemical studies have shown that the ZnF promotes DNA nick recognition[15] and intermolecular ligation[16]. Nonetheless, Lig3 clones expressing Lig3-ΔZnF or MtLig3-ΔZnF (Fig. 2a) were efficiently recovered after Cre expression (37 to 38%; Fig. 2b).

Our results reveal that the catalytic activity of Lig3 is critical for cell survival, but that the ZnF and BRCT domains, which interact with various proteins, are dispensable, raising the question whether Lig3 itself is critical for mitochondrial function, or whether another DNA ligase would substitute. As the Lig1 homolog in yeast provides mitochondrial ligase function[12], we provided an MLS to murine Lig1 (Fig. 2a). GFP-tagged MtLig1, but not wild-type Lig1, was localized to mitochondria (Supplementary Fig. 3c), as expected. Stable Lig3 cell lines expressing MtLig1, but not wild-type Lig1, could be efficiently converted to Lig3 (35%; Fig. 2b). MtLig1 Lig3 clones were devoid of Lig3, and expressed instead a larger Lig1 protein due to the GFP tag (Fig. 3a). Thus, targeting Lig1 to mitochondria circumvents the viability requirement for Lig3, allowing the creation of Lig3 null cells. In this way, the DNA ligase repertoire of mammalian cells is converted to that of yeast.

Figure 3

Loss of Lig3 is not associated with sensitivity to several DNA damaging agents or with increased sister-chromatid exchange

(a) Western blot analysis showing the loss of endogenous Lig3 protein in Lig3 cells with the indicated transgenes.

(b-f) Sensitivity of Lig3 cell lines to the indicated DNA damaging agents was measured using colony formation assays. Brca1 and Ercc1 cells are only shown on graphs when they are sensitive. For each cell line and agent, n = 4 and error bars = SEM; error bars in some cases are smaller than the symbol.

(g) SCE analysis. The range of SCEs was between 5 and 21 per metaphase for each cell line. For cells expressing MtLig1, two transgenic cell lines were analyzed. The differences between the cell lines are not significant using a two-tailed unpaired t-test. Values are presented with 1 SD from the mean.

Given that ZnF and BRCT-truncated forms of Lig3 and MtLig1 could rescue Lig3 cells, we investigated their proficiency in mtDNA maintenance and repair. The mtDNA copy number was maintained as well (or better) in these cell lines as in wild-type Lig3-expressing cells (Fig. 2c), indicating that cells expressing these altered ligases are competent to replicate their mtDNA during continued passage. A long-range quantitative PCR assay[17] was performed to assess the mitochondrial base excision repair capacity of these cells in response to oxidative damage, and these altered ligases were similarly proficient in repairing mtDNA lesions compared to wild-type Lig3-expressing cells (Fig. 2d).

At 298 amino acids, Chlorella virus DNA ligase (ChVLig) is the smallest eukaryal ligase known, consisting solely of a catalytic core[3]. We expressed ChVLig and a modified form containing an MLS, MtChVLig (Fig. 2a), and found that expression of either allowed deletion of Lig3 from the mouse genome (Fig. 2b). It is conceivable that ChVLig contains an internal sequence that allows translocation into mitochondria[18].

Thus, a minimal ATP-dependent ligase, devoid of auxiliary domains, rescues the survival of Lig3 null mammalian cells. Further, Lig3 null cells rescued by MtChVLig were proficient at mtDNA maintenance (Fig. 2c) and repair (Fig. 2d).

Whereas ATP-dependent ligases are widespread, ligases that use NAD+ as a cofactor are usually restricted to bacteria[1]. E. coli DNA ligase, LigA, is NAD+-dependent and has a distinctive domain organization compared with mammalian ligases[1]. LigA and a modified form with an MLS (Fig. 2a) were expressed from transgenes, and like ChVLig, both forms were found to allow the survival of Lig3 null cells (Fig. 2b). Hence, there is no essential functional distinction between NAD+ and ATP-dependent ligases in the mammalian mitochondria, akin to swaps of NAD+ and ATP-dependent ligases performed in bacteria[19] and yeast[20].

We demonstrated that nuclear Lig3 is not required for cell survival, as MtLig1 Lig3 cells are null for Lig3. To impair nuclear localization of MtLig1, we removed the Lig1 nuclear localization signal, creating MtLig1-ΔNLS Lig3 cells (Fig. 2a,2b; Supplementary Fig. 3c). MtLig1-ΔNLS, like MtLig1, was expressed at a substantially lower level than endogenous Lig1 (Fig. 3a). As a complementary approach, we also created Lig3 cells expressing MtLig3-ΔBRCT-NES, whose interaction with XRCC1 is abrogated and which is excluded from the nucleus by addition of a potent nuclear export signal (NES)[21] (Fig. 2a,2b; Fig. 3a; Supplementary Fig. 3b).

To assess the nuclear role of Lig3, we tested the sensitivity of Lig3 null (Lig3; MtLig1-ΔNLS) and nuclear Lig3-deficient (Lig3; MtLig3-ΔBRCT-NES) cells to a variety of DNA damaging agents. XRCC1-deficient cells are highly sensitive to alkylating agents like methyl methanesulfonate (MMS)[4,5,6]. If the interaction of XRCC1 with Lig3 is relevant to base excision repair, cells without nuclear Lig3 would also be expected to be sensitive to MMS; however, we found that these cells were no more sensitive than transgenic cells expressing wild-type Lig3 (Fig. 3b) or the parental cells (Supplementary Fig. 5). XRCC1-deficient cells are also sensitive to agents which produce DNA single and double-strand breaks, including hydrogen peroxide and ionizing radiation[4,5,6], and to ultraviolet radiation[22]. By contrast, we found that cells without nuclear Lig3 were not any more sensitive to these agents than control cells (Fig. 3c-e, Supplementary Fig. 5). Thus, Lig3 appears to be dispensable for nuclear DNA damage repair that requires XRCC1. Finally, we tested sensitivity to Parp inhibiton, which causes the accumulation of single-strand breaks[23], and found that nuclear Lig3 was also not required for resistance of cells to Parp inhibiton (Fig. 3f).

As the ZnF domain of Lig3 has been reported to be critical for its intermolecular ligation activity[16], we also investigated whether deletion of this domain in the context of an otherwise wild-type Lig3 would impair resistance of cells to ionizing radiation. As with the other mutants, Lig3-ΔZnF Lig3 cells were no more sensitive than control cells (Fig. 3e).

XRCC1-deficient cells are notable for their high rate of spontaneous sister-chromatid exchange (SCE): both mouse and hamster XRCC1 mutants have ~10-fold higher SCE levels than control cells[4,5]. We examined spontaneous SCEs in MtLig1 Lig3 cells and found levels similar to control cells (Fig. 3g). Thus, the high level of SCEs found with XRCC1 deficiency is not recapitulated with Lig3 deficiency.

The lack of Lig3 in many model organisms has limited their use to study its function. In mouse, disruption of any of the DNA ligase genes leads to embryonic death, but the most severe phenotype occurs with Lig3 disruption[2,24,25]. Lig1 has been considered to be the replicative ligase[1,26], but the earlier death associated with Lig3 disruption, together with the inability to obtain viable Lig3 null cells, left open the possibility that Lig3 could have a critical role in nuclear DNA metabolism. The generation of viable and healthy Lig3 null cells by providing a mitochondrial ligase conclusively rules out an essential role for Lig3 in the nucleus.

The well-documented interaction between Lig3 and XRCC1 had suggested that Lig3 would be critical for the same nuclear DNA repair pathways as XRCC1, similar to the Lig4-XRCC4 complex in DSB repair[1]. However, the lack of sensitivity of Lig3 null cells to the spectrum of DNA damaging agents that sensitize XRCC1-deficient cells, as shown here in proliferating cells and in the accompanying report in quiescent cells[27], together with a normal SCE frequency, provides strong evidence that Lig3 is not required for XRCC1-dependent nuclear DNA repair, pointing instead to a role for Lig1.

Our results demonstrate instead that Lig3 is an essential gene because of its requirement in mitochondria. However, Lig3 can be replaced in mitochondria with Lig1, the mitochondrial ligase in lower eukaryotes, with an algal viral ligase consisting solely of a catalytic core, and even the NAD+-dependent E. coli LigA. Thus, these results attest to the requirement for a functional DNA ligase, which trumps even co-factor specificity. Why vertebrates developed a requirement for Lig3 is uncertain, but given our results, the additional domains found in Lig3 do not appear to be essential for mitochondrial function, including mtDNA maintenance or repair of oxidative damage. These results underscore a surprising plasticity that mammalian cells have in their mitochondrial DNA ligase requirement.

Methods Summary

Cell culture

To construct stable cell lines expressing various DNA ligases, 5 × 106 Lig3 cells were electroporated with 12 μg ligase expression vector at 800 V, 3 μF. Hygromycin resistant clones were picked after incubation for 10 days in 150 μg/μl hygromycin. Initial screening for exogenous ligase expression was performed by RT-PCR using specific primers, followed by Western blotting. For deletion of the endogenous Lig3 allele, 5 × 106 cells were electroporated with 5 μg Cre recombinase vector at 250 V, 950 μF. Cells were plated based on a serial dilution. After 7 days, colonies were picked and expanded, and then replica plated into two 96-well plates. One plate was cultured with 200 μg/μl G418, whereas the other plate was cultured in normal media. Clones that did not grow in G418, but grew in normal media, have converted the Lig3 allele to a Lig3 allele. The genotype of these clones was confirmed by PCR.

Western Blotting

Whole cell extracts were prepared with Nonidet-P40 buffer and were run on a 7.5% (w/v) Tris-HCl SDS page gel, blotted, and then probed with Lig3 antibody clone 7 (BD Transduction Labs), which recognizes both the human and mouse Lig3 proteins, or Lig1 antibody N-13 (Santa Cruz). α-tubulin (Sigma) was used as a loading control.

Full methods accompany this paper.

Full Methods

DNA constructs

A vector containing wild-type human Lig3 cDNA (with both mitochondrial and nuclear translation initiation sites), a gift from Dr. K.W. Caldecott (University of Sussex, Brighton, UK), was digested with NheI and XbaI and subcloned into the NheI site of pCAGGS. As the cDNA contained a 51 bp linker located before the nuclear translation initiation site, it was modified by site directed mutagenesis to remove the linker, with the primers: 5’-GTGGCCCCTGTGAGATGGCTGAGCAACG-3’ and 5’-CGTTGCTCAGCCATCTCACAGGGGCCAC-3’, to restore an unmodified Lig3 sequence, creating pCAGGS-Lig3. A Pgk-hygromycin resistance gene was inserted at the PsiI site to create pCAGGS-Lig3-hyg. MtLig3 was generated by using site directed mutagenesis to generate a M88T mutation in pCAGGS-Lig3-hyg using the primers: 5’GAGAGGCCCCTGTGAGACCGCTGAGCA-3’ and 5’GAGAGGATCCCTAGCAGGGAGCTACCAGTCTC-3’. For NucLig3, amino acids 1 to 87 were deleted by introducing NotI and BamHI sites into pCAGGS-Lig3-hyg via PCR using the primers: 5’-GCATGCGGCCGCCTGTGAGATGGCTGAGCAACGGT-3’, 5’-GGATGGATCCCTAGCAGGGAGCTACCAGTC-3’. For the ΔBRCT mutation, amino acids 934 to 1009 were deleted by introducing NheI and MfeI sites via PCR using the primers: 5’-GGCCGCTAGCGGGCAGCTATATGTCTTTGGCTTTCAAGAT-3’ and 5’-GAGACAATTGTTACTATACCTTTGTTTGGCACAGCGTC-3’. The ΔZnF mutation was generated by deleting amino acids 89 to 258 using site directed mutagenesis with primers 5’-TGGCCCCTGTGAGATGAAGGACTGTCTGCTAC-3’ and 5’-GTAGCAGACAGTCCTTCATCTCACAGGGGCCA-3’. For GFP tagging of the Mt-tagged constructs, SacII and AgeI sites were introduced and stop codons of the full length or ΔBRCT proteins were converted into alanine codons by PCR and cloned in frame into SacII and AgeI sites of pEGFP-N1 (Clontech). PCR primers for full length were 5’-ACGGTACCGCGGCAGCTATATGTCTTTGG-3’and 5’- ACGGTACCGCGGCAGCTATATGTCTTTGG-3’, and for ΔBRCT were 5’- ACGGTACCGCGGCAGCTATATGTCTTTGG-3’ and 5’- GGCGACCGGTGGTACCTTTGTTTGGCACAGCG-3’.

For other constructs with GFP fusions (NucLig3, Lig3, ΔZnF and K508V), plasmids were digested with PmlI and ligated into the vector backbone of MtLig3-GFP using the same enzyme. The MAPKK nuclear export signal (NES)[21] was fused to the C terminus of GFP via PCR using the primers 5’-GCCCCCTCAGCCAGTACCAAGAA-3’ and 5’GGCCAATTGGCCTTATTACTGCTGCTCGTCCAGCTCCAGCTCCTCCAGCTT CTTTTGGAGGTCCACGAGATTCTTGTACAGCTCGTCCAT-3’. Mouse Lig1 cDNA (Invitrogen) was amplified with primers introducing KpnI and AgeI sites and changing a stop codon into an alanine codon; this fragment was cloned in frame into the KpnI and AgeI sites of pEGFP-N1. The Lig3 MLS was amplified with the following primers 5’-GGCGAATTCTATATGTCTTTGGCTTTCAAGATCTTCTTTC-3’and 5’-ATTGGTACCCCTCACAGGGGCCACTGCAG-3’ and cloned into the EcoRI and KpnI sites of Lig1-GFP-hyg vector. The Chlorella virus DNA ligase coding region was amplified with ChV-NheI and ChV-MfeI primers and cloned into the NheI and MfeI sites of the pCAGGS-Lig1-Hyg vector. ChV-NheI: 5’-GCCGCTAGCACCATGGCAATCACAAAGCCATT-3’, ChV-MfeI: 5’-GCCCAATTGTTAACGGTCTTCCTCGTGAC-3’. The Escherichia coli DNA ligase coding region was amplified with LigA-NheI and LigA-MfeI primers and cloned into the NheI and MfeI sites of the pCAGGS-Lig1-Hyg vector. LigA-NheI: 5’-GCCGCTAGCACCATGGAATCAATCGAACAACAA -3’, LigA-MfeI: 5’-GCCCAATTGTCAGCTACCCAGCAAACG -3’.

RT-PCR

Hygromycin resistant clones were screened by RT-PCR using primers specific to human Lig3. A primer pair was used with the forward primer to the pCAGGS backbone and the reverse primer to exon 3 of human Lig3, resulting in a size difference for mitochondrial and nuclear forms (Supplementary Fig. 3): pCAGRTfw 5’-CAACGTGCTGGTTATTGTGC-3’, hLig3Rv 5’-ACAGCTTTCTTCTTTGGTGTACCT-3’. A similar strategy was used for Lig1, with primers pCAGRTfw and Lig1RT_RV (5’-ACCGCTGAGCAACGGTTCT-3’), for Chlorella virus DNA ligase, with primers pCAGRTfw and chlRTPCR-RV1 (5’-CAGCACTTGTGGTGTCTTGAA-3’) and, for Escherichia coli DNA ligase, with primers pCAGRTfw and LigARTPCR-RV1 (5’-CCTGCACACGTTTGTTGAAA -3’). RNA was isolated using RNeasy Mini Kit (Qiagen) and cDNA was generated by SuperScript III First-strand Synthesis system (Invitrogen).

Genotyping

Genomic DNA was isolated using the Genelute Mammalian Genomic DNA Miniprep Kit (Sigma). Each primer was named for the location on the genomic DNA (e.g., Int5-6Fw means that the primer is at the intron between exons 5 and 6). Primer pairs used for genotyping are as follows: Exon 5Fw and Neo2Rv (primer pair a in Fig. 1a): 5’-GGCTTTCACGGTGATGTGTA-3’ and 5’-TCTGGATTCATCGACTGTGG-3’, using an annealing temperature of 62°C; Int5-6Fw and Int16-17Rv (primer pair b in Fig. 1a): 5’-CGGGTGTAGGGAGGTCATAA-3’, 5’-GAAGGAAGAGGTCTCCAGCA-3’, using an annealing temperature of 62°C; Int10-11Fw and Int11-12Rv (primer pair c in Fig. 1a): 5’-CACTAAACGTGGCAGAGCAA-3’, 5’-CCAGCCCAGACTACAGCTTC-3’, using anannealing temperature of 62°C; Int5-6Fw2 and Int5-6Rv (Supplementary Fig. 1d): 5’-GCCAAGTGTGAATATACAGC-3’ and 5’-CAGGGAGCTTGGGACGGATGC-3’, using an annealing temperature of 64°C; Int5-6 and Int16-17(Supplementary Fig. 1d): 5’-CGGGTGTAGGGAGGTCATAA-3’ and 5’-GAAGGAAGAGGTCTCCAGCA-3’, using an annealing temperature of 64°C.

Microscopy

The subcellular localization of the various GFP fusion constructs was checked by Mitotracker Red CMXRos (Invitrogen) and Hoechst 33342 (Invitrogen) to stain mitochondria and nucleus, respectively. DNA constructs were transiently transfected with Lipofectamine 2000 (Invitrogen). After incubating cells with Opti-MEM (Invitrogen) containing 10 nM Mitotracker Red CMXRos and 2.5 μM Hoechst 33342 for 20 min at 37°C, cells were monitored with a Zeiss LSM 510 META laser scanning microscope.

qPCR mtDNA repair assay

1 × 106 ES cells with the indicated genotypes were plated on 6-cm2 plates. After 16 h, cells were cultured with 6.25 ml, 175 μM hydrogen peroxide for 15 min and then cultured with conditioned medium for 1.5 h. mtDNA copy number and mtDNA repair were determined by a long-range quantitative PCR assay[17]. Basically, DNA was extracted from pellets of 1×10ˆ6 cells with the DNeasy Blood and Tissue kit (Qiagen) by a QIAcube automated DNA extraction robot (Qiagen). Initial DNA concentration was measured using Picogreen dsDNA binding agent (Invitrogen) and a DNA standard curve. Total mouse genomic DNA at an approximate final concentration of 4.5 ng/μL was then digested with with HaeII (New England Biosciences) for 1 hr at 37°C in a reaction mixture containing1X NEBuffer 4, 1X BSA, and 20 U undiluted HaeII enzyme. HaeII linearizes the mouse mtDNA by digesting once (2604) near the D-loop region. Linearization of mitochondrial DNA ensures efficient amplification and allows accurate determination of mtDNA copy number. After digestion, DNA concentration was measured with Picogreen and an appropriate volume was directly removed from the digest to use for qPCR, with less than 5% variability in DNA concentration between samples.

The qPCR reaction was performed with the GeneAmp XL PCR kit (Applied Biosystems) as follows: 10-15 ng total DNA, in a reaction mix of 50 μL, with 1X buffer, 100 ng/μL BSA, dNTPs at 200 μM each nt, 1.2 mM MgO(Ac)2, and 20 pmoles for each of two primers. Primer pairs for a 10 kb fragment of mtDNA (long) and for a 117 bp fragment of mtDNA (short) were used for calculating mtDNA damage and mtDNA copy number, respectively. Primer sequences are as described previously[17]. DNA polymerase was added at a concentration of 1 U/reaction. A 50% control and a “no template” blank were used to ensure that the assay was within quantitative range and free of contamination, respectively. PCR products were quantified using fluorescence-blank measurements from the Picogreen dsDNA binding agent. PCR products from the long fragment were normalized to the short fragment to account for the effect of differing mtDNA copy number on amplification of the long fragment.

Sister chromatid exchange

5 × 106 ES cells with the indicated genotypes were plated on 10-cm2 plates. After 24 h, cells were cultured with 10 μM bromodeoxyuridine for 20 h (approximately two cell cycle periods) and pulsed with 0.03 μg/ml Colcemid for the final 30 min. The cells were collected by centrifugation and exposed to 0.075 mM KCl hypotonic solution for 30 min at 25°C. The cells were washed twice with the fixative (methanol:acetic acid = 3:1) and suspended in a small volume of the fixative. The cell suspension was dropped onto ice-cold glass slides and air-dried at 60°C for 2 h. Two days later, slides were incubated with 1 μg/ml Hoechst 33258 in Sorensen’s phosphate buffer (38 mM KH2PO4, 60mM Na2HPO4, pH 7.0) for 10 min, rinsed with 2X SSC buffer (300 mM NaCl, 30 mM Na3C6H5O7, pH 7.0) and then overlaid with coverslips. Slides were exposed to black light (λ = 352 nm) at a distance of 1 cm for 20 min. After removal of coverslips, the slides were incubated in 2X SSC at 60°C for 2 h and then stained with 4% (v/v) Giemsa solution in Sorensen’s buffer for 10 min, rinsed in water, and air-dried. A two-tailed unpaired t-test was used to analyze the data.

Drug Sensitivity assays

2 × 103 ES-cells per well were seeded in 24-well plates in duplicate. After 24 h, cells were incubated with various drugs at the indicated concentrations (Fig. 3a) for 24 h in ES cell medium, except hydrogen peroxide which was for 1 h. For irradiation, plates were exposed to the X-ray source from an X-RAD 225C apparatus at a rate of 687 cGy/minute. Six days later cells were fixed with a solution of 12.5% (v/v) acetic acid, 12.5% (v/v) methanol for 15 min and then stained with 1% (w/v) crystal violet. Afterwards, stained cells were treated with 0.1% SDS in methanol and the absorbance was measured at 595 nm. Each point in the plots was the average of 2 experiments where each experiment had a duplicate and is a percentage of the absorbance from untreated ES cells. For Lig3 null cells expressing the Lig3 ΔZNF-GFP, MtLig3-ΔBRCT-GFP-NES and MtLig1 GFP transgenes, two independent null clones were used for each. For the colony formation assays (Fig. 3b), 2 × 103 ES cells were plated in 10-cm2 plates and exposed to ionizing radiation or ultraviolet radiation (UVC). Eight days later, surviving colonies were fixed with methanol and stained with 3% (v/v) Giemsa.