TDP2–Dependent Non-Homologous End-Joining Protects against Topoisomerase II–Induced DNA Breaks and Genome Instability in Cells and In Vivo

Anticancer topoisomerase “poisons” exploit the break-and-rejoining mechanism of topoisomerase II (TOP2) to generate TOP2-linked DNA double-strand breaks (DSBs). This characteristic underlies the clinical efficacy of TOP2 poisons, but is also implicated in chromosomal translocations and genome instability associated with secondary, treatment-related, haematological malignancy. Despite this relevance for cancer therapy, the mechanistic aspects governing repair of TOP2-induced DSBs and the physiological consequences that absent or aberrant repair can have are still poorly understood. To address these deficits, we employed cells and mice lacking tyrosyl DNA phosphodiesterase 2 (TDP2), an enzyme that hydrolyses 5′-phosphotyrosyl bonds at TOP2-associated DSBs, and studied their response to TOP2 poisons. Our results demonstrate that TDP2 functions in non-homologous end-joining (NHEJ) and liberates DSB termini that are competent for ligation. Moreover, we show that the absence of TDP2 in cells impairs not only the capacity to repair TOP2-induced DSBs but also the accuracy of the process, thus compromising genome integrity. Most importantly, we find this TDP2-dependent NHEJ mechanism to be physiologically relevant, as Tdp2-deleted mice are sensitive to TOP2-induced damage, displaying marked lymphoid toxicity, severe intestinal damage, and increased genome instability in the bone marrow. Collectively, our data reveal TDP2-mediated error-free NHEJ as an efficient and accurate mechanism to repair TOP2-induced DSBs. Given the widespread use of TOP2 poisons in cancer chemotherapy, this raises the possibility of TDP2 being an important etiological factor in the response of tumours to this type of agent and in the development of treatment-related malignancy.


Introduction
The double-stranded helical structure of DNA creates topological problems in all processes that involve opening of the double helix and accessing the genetic information [1,2]. In particular, the transcription and duplication of DNA and its condensation into chromosomes generates knots and tangles that need to be resolved to avoid interference with diverse cellular processes and to ensure faithful chromosome segregation during mitosis. DNA topoisomerases are enzymes that introduce transient breaks in DNA to solve these topological problems. Type II topoisomerases, such as topoisomerase II in eukaryotes (TOP2) are essential homodimeric enzymes that relax, unknot and decatenate DNA molecules by catalyzing the passage of duplex DNA through a transient DNA double strand break (DSB) created by the enzyme [3]. Two isoforms of TOP2, a and b, exist in higher eukaryotes, with primary roles in replication and chromosome segregation and in transcription, respectively.
A key intermediate of TOP2 activity is the cleavage complex, in which each of two topoisomerase subunits is covalently linked to the 59terminus of an enzyme-generated DSB via a phosphodiester bond between the active-site tyrosine and the 59-phosphate. The cleavage complex is normally a very short-lived intermediate, because the topoisomerase rapidly re-ligates the DSB once DNA strand passage through the DSB has occurred. However, under certain circumstances, such as the presence of nearby DNA lesions, cleavage complexes can be stabilized resulting in an increased likelihood of collision with RNA or DNA polymerases [4]. Such collisions can convert cleavage complexes into potentially clastogenic or lethal DSBs that require cellular DNA repair pathways for their removal.
Cleavage complexes are the target of a widely used class of antitumor agents that 'poison' topoisomerase activity, thereby prolonging the half-life of the intermediate and increasing the possibility of DSB formation [4,5]. Thus, these drugs kill tumor cells by inducing high levels of TOP2-associated DSBs. Consequently, TOP2 poisons are commonly used antineoplastic drugs in the treatment of a broad range of tumor types including malignant lymphomas, sarcomas, leukemias, and lung, ovarian, breast and testicular cancers [5]. However, similar to other chemotherapeutic agents, TOP2-targeting drugs are only partially selective for tumour cells, resulting in unwanted toxicity in normal tissues and in therapy-associated chromosome translocations and secondary leukemias [6][7][8][9][10][11][12][13][14]. Moreover, some breakpoints in such translocations have actually been correlated with preferential sites of cleavage by TOP2 [13][14][15][16][17].
A characteristic feature of TOP2-induced DNA breaks is covalent attachment of the enzyme to 59 ends of the DNA, which must be removed by cellular end-processing enzymes if DSB repair is to occur [18]. Until recently, the only known mechanism for removal TOP2 peptide from DNA 59-termini in mammalian cells involved excision of the DNA fragment linked to the peptide using nucleases such as the MRN complex, CtIP or Artemis [19][20][21]. Recently, however, we identified a human 59-tyrosyl DNA phosphodiesterase (59-TDP) that can cleave 59-phosphotyrosyl bonds and thereby release TOP2 from DSB termini without the need to also remove DNA sequence [22]. Consequently, this enzyme, which was previously known as signalling protein and transcription cofactor TTRAP/EAPII [23,24], is now denoted tyrosyl DNA phopshodiesterase-2 (TDP2; Human Gene Nomeclature Organisation). Notably, consistent with its enzyme activity, TDP2 is required for cellular resistance to the anti-cancer TOP2 poison etoposide, but is not required for cellular resistance to ionizing radiation or methylmethane sulphonate [22,25]; agents that induce DNA damage independently of TOP2 activity.
Following DNA end processing, DSBs can be repaired either by homologous recombination (HR) or by non-homologous end joining (NHEJ) [26]. However, these pathways utilize fundamentally different mechanisms for rejoining DSBs and consequently differ in their accuracy. In particular, HR utilizes undamaged sister chromatids to replace any nucleotides removed from DNA termini during DNA end processing and consequently is normally 'error-free'. However, this process is available only during S phase or G2, when sister chromatids are available. In contrast, NHEJ is a 'cut-and-splice' process in which DSB termini are ligated together following DNA end processing without accurate replacement of missing nucleotides, and thus is potentially 'error-prone'.
Here, we employ avian and murine experimental models to show that TDP2/Tdp2 deletion results in hypersensitivity to a structurally diverse range of anti-cancer TOP2 poisons. Moreover, we present genetic, biochemical and cellular evidence for TDP2 functioning in a mechanism of NHEJ that protects genome integrity in response to TOP2-induced damage. Finally, we show that this TDP2 dependent pathway also operates in vivo, as, upon exposure to TOP2 poisons, it is required for normal adult mouse lymphopoiesis, intestinal mucosa homeostasis and the maintenance of genome stability in the bone marrow. Collectively, our results suggest that TDP2 defines an error-free mechanism of NHEJ in mammals, which is specialized in the repair of TOP2induced DSBs and reduces both tissue toxicity and genome instability in response to this particular type of DNA damage. These findings suggest the possibility of TDP2 being a significant etiological factor in the clinical tolerance and response to widely used TOP2 poisons.

Results
TDP2 is required for cellular resistance to clinical TOP2 poisons and is the major 59-TDP activity in the mouse The discovery of TDP2 as the first 59-TDP activity raised the possibility of it being an important factor in the clinical response to TOP2 poisons [22,25]. Indeed, TDP2 deleted avian DT40 cells are hypersensitive to etoposide [22,25]. To address this question further, we examined the sensitivity of TDP2 2/2/2 cells to two additional, structurally diverse, TOP2 poisons. These drugs, denoted doxorubicin and amsacrine (m-AMSA), are employed widely during cancer therapy but in contrast to etoposide, 'poison' TOP2 by intercalating into DNA [5]. Nevertheless, similarly to etoposide, TDP2 2/2/2 cells displayed significant hypersensitivity to both doxorubicin and m-AMSA ( Figure 1A). Moreover, a functional TDP2 phosphodiesterase domain was required for cellular resistance to this type of drug, because expression of wildtype human TDP2 (hTDP2) rescued the sensitivity of TDP2 2/2/2 DT40 cells to m-AMSA, whereas hTDP2 D262A harbouring an inactivating mutation in the catalytic active site [5] did not ( Figure 1A). These results show that TDP2 is required for cellular resistance to a range clinically relevant and structurally diverse TOP2 poisons, and support our contention that this requirement reflects the 59-TDP activity of this enzyme.
To determine the impact of TDP2 on TOP2-induced DNA damage in mammals, and thus its possible relevance to anti-cancer therapy, we adopted a mouse model in which the first three exons of Tdp2, plus the 59-UTR, were deleted by Cre-mediated excision ( Figure 1B; see Materials and Methods). Mice homozygous for the deleted allele (Tdp2 flD , from here-on denoted Tdp2 D1-3 ) are viable, and so far we have not detected any abnormal pathology (unpublished observations). However, transformed Tdp2 D1-3 mouse embryonic fibloblasts (MEFs) were hypersensitive to etoposide ( Figure 1C, left, and Figure S1), but were not hypersensitive to DNA damage induced independently of TOP2 by c-irradiation ( Figure 1C, right).

Author Summary
DNA double-strand breaks (DSBs) are dangerous because they can lead to cellular death and tissue degeneration if not repaired, or to genome rearrangements, which are a common hallmark of cancer, if repaired incorrectly. Although required for all chromosomal transitions in cells, transient DNA cleavage by topoisomerase II (TOP2) is a potential endogenous source of DSBs, which are characteristic in that TOP2 remains covalently bound to the DNA termini. In addition, numerous chemotherapeutic regimes rely on compounds that ''poison'' TOP2 activity, stimulating the formation of DSBs that target tumour cells. However, these compounds also affect healthy tissue and confer undesirable side effects, including the stimulation of genome rearrangements that can trigger secondary malignancies (mainly acute leukemia). Identifying the factors that participate in the repair of TOP2induced DSBs and fully understanding their mechanism of action are therefore important for the design of chemotherapeutic regimes that are more effective and safer. Here we demonstrate that TDP2, a recently identified protein that can liberate DSB termini from blocked TOP2, functions as part of established cellular DSB repair processes and is required to safeguard genome integrity upon treatment with TOP2 poisons, both in cells and in mice. These results can therefore have important implications in cancer treatment.
Protein extracts from spleen, thymus, and bone marrow from wild type mice possess robust 59-TDP activity, but, importantly, this activity was absent in analogous protein extracts from Tdp2 D1-3 mice, confirming successful inactivation of the enzyme (Figure 2A). Cell extracts prepared from primary Tdp2 D1-3 MEFs also lacked detectable 59-TDP activity ( Figure 2B). This was true not only for blunt-ended DSB substrates, but also for DSB substrates harbouring a 4-bp 59-overhang ( Figure 2C), characteristic of TOP2-induced DSBs. Additionally, EDTA-mediated chelation of Mg 2+ , which is essential for TDP2 function, completely eliminates 59-TDP activity in wild type MEF extracts. These observations are significant because the related enzyme TDP1, whose activity is Mg 2+ independent, was recently reported to possess weak activity on this type of substrate [27]. Our data therefore suggest that TDP2 is the primary, if not only, source of 59-TDP activity in MEF extracts ( Figure 2C).

TDP2 creates ligatable DSBs and functions in NHEJ
Based on the mechanism of TOP2 cleavage, we anticipated that TDP2 activity would reconstitute 'clean' DSBs (59 phosphate and 39 hydroxyl termini) with 4-bp overhangs, which would be an ideal substrate for ligation by NHEJ. Interestingly, these ligation events would accurately preserve the DNA sequence, suggesting the possibility of an error-free NHEJ mechanism that specifically acts on TOP2-induced DSBs. To test this hypothesis, we examined whether TDP2 action at DSBs typical of those induced by TOP2 creates termini that can be ligated by T4 DNA ligase. Indeed, inclusion of T4 DNA ligase in reactions containing wild type MEF extract resulted in the additional appearance of a product of 46-nt, indicative of the completion of DSB repair by DNA ligation. However, this product was not observed if reactions contained cell extract from Tdp2 D1-3 MEFs, confirming that DNA ligation was dependent on TDP2 activity ( Figure 2D). Interestingly, the length of the product is consistent with a ligation event in which DNA sequence is preserved. To analyse ligation events directly catalysed by cell extracts, we generated linear plasmids harbouring 59 phosphate or 59 phosphotyrosine ends by PCR amplification with the corresponding modified primers. The incubation of these substrates with NHEJ-competent nuclear extracts [28] results in plasmid circularization events that can be scored as colonies following bacterial transformation. As can be seen in Figure 2E, nuclear extracts from Tdp2 D1-3 MEFs efficiently circularized linear plasmids with 59 phosphate ends but not linear plasmids harbouring 59-phosphotyrosine. This difference was lost upon addition of recombinant TDP2 to the reaction, confirming the TDP2-dependent nature of the repair reaction. Collectively, our data suggest that TDP2 activity facilitates NHEJ of 59 tyrosineblocked ends by generating DSBs with ligatable termini, consistent with our hypothesis that this enzyme can support error-free NHEJ of TOP2-induced DNA damage.
To genetically test whether TDP2 functions indeed during NHEJ, we generated TDP2 2/2/2 DT40 cells harboring a targeted deletion of Ku70, a core component of the NHEJ pathway ( Figure  S2). Whilst both TDP2 2/2/2 and KU70 2/2 cells were hypersensitive to etoposide, cells in which both genes were deleted (TDP2 2/2/2 /KU70 2/2 ) were no more hypersensitive than cells in which Ku70 alone was deleted ( Figure 3A). In contrast to this epistatic relationship with a core NHEJ factor, transient knockdown of TDP2 further enhances etoposide sensitivity of HR defective (BRCA2 mutated) human fibroblasts ( Figure 3B). Based on these genetic relationships, we conclude that TDP2 functions in a NHEJ-mediated and HR-independent pathway for the repair of TOP2-induced DSBs.
To further assign a role for TDP2 in the NHEJ pathway for DSB repair, we measured DSB repair rates in primary Tdp2 D1-3 MEFs by immunodetection of cH2AX, a phosphorylated derivative of histone H2AX that arises at sites of chromosomal DSBs [29]. We measured DSB repair in specific phases of the cell cycle, because whilst NHEJ is operative throughout, HR-mediated DSB repair is operative only in S/ G2 [30]. Notably, DSB repair rates were markedly reduced in Tdp2 D1-3 MEFs following etoposide treatment, both in G0/G1 ( Figure 3C) and G2 ( Figure 3D), consistent with TDP2 functioning, as NHEJ, independently of cell cycle. These results were not specific to murine cells, since similar results were observed in TDP2-depleted human A549 cells ( Figure S4). In contrast to treatment with etoposide, the rate of DSB repair was normal in Tdp2 D1-3 MEFs following cirradiation, consistent with a role for TDP2 specifically at TOP2induced DSBs ( Figure 3E). Collectively, these data demonstrate that TDP2 is required in mammalian cells for rapid repair of TOP2induced DSBs by NHEJ, and for cellular resistance to these lesions. TDP2 promotes genome stability following TOP2induced DNA damage We hypothesized that this TDP2-mediated error-free NHEJ mechanism would be important to maintain genome integrity upon exposure to TOP2 poisons. To address this possibility, we quantified the frequency of micronuclei (MN), nucleoplasmic bridges (NB), and chromosomal aberrations following etoposide treatment. These events constitute well-established indicators of genome instability caused by misrepair of DSBs in which acentric, dicentric and aberrant chromosomes or chromosome fragments can be formed. As expected, etoposide increased the number of micronuclei and nucleoplasmic bridges in both transformed Tdp2 +/+ and Tdp2 D1-3 MEFs, but this increase was significantly higher (up to three-fold) in Tdp2 D1-3 cells ( Figure 4A). Primary Tdp2 D1-3 MEFs at low passage (P3-4) similarly displayed elevated levels of micronuclei and nucleoplasmic bridges following etoposide treatment, compared to wild type primary MEFs ( Figure 4B), although in the case of nucleoplasmic bridges the low number of cells displaying these structures prevented the difference from reaching statistical significance.
An additional indicator of genome instability is elevated frequencies of chromosome aberrations. Consequently, we quantified the frequency of chromosome breaks and exchanges in metaphase spreads of transformed Tdp2 +/+ and Tdp2 D1-3 MEFs. In agreement with the increased cell cycle arrest of TDP2 2/2/2 DT40 cells in G2 following etoposide treatment [25], we noted an etoposide-dependent reduction in metaphase cells that was particularly severe in Tdp2 D1-3 MEFs (unpublished observations). However, of those metaphases observed and scored, both chromosome exchanges and breaks were significantly higher (2 to 5-fold) in Tdp2 D1-3 MEFs than in Tdp2 +/+ MEFs ( Figure 4C). A similar increase in these events in Tdp2 D1-3 MEFs, compared to wild type cells, was observed if low-passage primary MEFs were employed, ruling out the possibility that the elevated genome instability in Tdp2 D1-3 MEFs was an artefact of cellular transformation ( Figure 4D). In the latter case, etoposide treatment almost ablated the appearance of mitotic cells in populations of both wild type and Tdp2 D1-3 MEFs, necessitating the use of caffeine to prevent G2 arrest. Taken together these results demonstrate that loss of TDP2 results in increased genome instability following TOP2-induced DNA strand breakage.

Loss of TDP2 results in elevated homologous recombination
The above results demonstrate increased genome instability in Tdp2 D1-3 MEFs, consistent with a role for TDP2 in error-free NHEJ-mediated repair of TOP2-induced DSBs. In this scenario, we considered the possibility that loss of TDP2 might also result in channelling of DSB repair towards HR. To address this question, we analyzed the formation of RAD51 foci, a well-established indicator of repair by HR. Following treatment with etoposide, the average number of Rad51 foci per cell was ,3-fold higher in Tdp2 D1-3 than in wild-type MEFs ( Figure 5A), in agreement with an increase in the use of HR to repair TOP2-induced DSBs when TDP2 is not present. Furthermore, we compared the frequency of etoposide-induced sister chromatid exchanges (SCEs), a molecular hallmark of HR [31], in wild type and Tdp2 D1-3 MEFs ( Figure 5B). Notably, SCE levels increased substantially in transformed MEFs following acute etoposide exposure, being significantly higher in Tdp2 D1-3 cells at two etoposide concentrations tested (1 and 2.5 mM). These data confirm that, upon etoposide treatment, the frequency of HR is elevated in Tdp2 D1-3 MEFs, consistent with TDP2 functioning in NHEJ.
Elevated hypersensitivity to TOP2-induced DNA damage in Tdp2 D1-3 mice To address the relevance of TDP2-mediated repair of TOP2induced DSBs in vivo, we compared the impact of etoposide on adult (8 wk) wild type and Tdp2 D1-3 mice. A single intraperitonal injection of etoposide (75 mg/kg) caused a decrease in body weight in the initial 4 days post-treatment both in wild type and Tdp2 D1-3 animals ( Figure 6A). However, whereas Tdp2 +/+ mice exhibited relatively mild and transient weight loss, Tdp2 D1-3 littermates lost weight progressively and were sacrificed at day 6 to prevent suffering. No differences in body weight were observed between mock-treated (with DMSO) wild type and Tdp2 D1-3 mice. Histopathological analysis of Tdp2 D1-3 mice sacrificed 6 days after etoposide treatment revealed marked villous atrophy in the small intestinal mucosa as the likely cause of the drastic weight loss ( Figure 6B). This was not observed in either wild-type and/or DMSO treated animals (data not shown), suggesting a protective role for TDP2 against adverse effects of etoposide in vivo.

TOP2-induced DNA damage results in increased lymphoid toxicity in Tdp2 D1-3 mice
In addition to severe intestinal damage, etoposide administration resulted in elevated splenic and thymic atrophy in Tdp2 D1-3 mice, compared to wild type mice ( Figure 6C), consistent with the known hypersensitivity of these organs to this drug [32]. Histological analysis of these tissues revealed a marked reduction in the cellular content in Tdp2 D1-3 animals ( Figure 6C, right, note the low density of dark-stained nuclei). In light of these results, we analysed B-cell and T-cell maturation in wild type and Tdp2 D1-3 mice ( Figure 6D and Figure S5). In the case of B-cell precursors in bone marrow, treatment with etoposide resulted in a decrease of 30-50% in the fraction of cells that were CD43 + B220 + progenitors (Pro-B cells) and a decrease of .95% in the fraction of cells that were CD43 2 B220 low (Pre-B cells) or CD43 2 B220 high (immature B cells) precursors. In all cases the reduction in B-cell precursors was greater in Tdp2 D1-3 mice, but the differences were not statistically significant at the administered dose. In contrast, in the case of T-cell maturation, whereas etoposide treatment reduced the fraction of CD4 + CD8 + immature T cells by 30-40% in wild type mice, these cells were almost completely eliminated in Tdp2 D1-3 mice ( Figure 6A, bottom right). No effect was observed in CD11b/Mac-1 + myeloid cells in the bone marrow ( Figure S6). Taken together, these results suggest that loss of TDP2 increases cellular attrition in the lymphoid system, particularly in the T-cell lineage, in response to TOP2-induced DNA damage.
Tdp2 D1-3 mice display increased TOP2-induced genome instability in bone marrow A major side-effect of cancer therapy employing TOP2 poisons is secondary hematological malignancy, and in particular acute leukemia, resulting most likely from error prone/erroneous repair of TOP2-induced DSBs and chromosome translocations [4,7]. Given our findings that TDP2 limits genome rearrangements induced by etoposide in cells, we examined whether TDP2 also promotes genome stability in bone marrow in vivo. We quantified the fraction of micronucleated polychromatic erythrocytes (PCEs) in bone marrow smears from Tdp2 D1-3 and Tdp2 +/+ mice 24 hour after intraperitoneal injection of etoposide (1 mg/kg). The rodent erythrocyte micronucleus test is a standard procedure to detect cytogenetic damage in toxicological studies and is based on the detection of micronuclei in erythrocyte precursors (Hayashi et al 1994). As expected, etoposide increased the fraction of PCEs that were micronucleated in both wild type and Tdp2 D1-3 animals  Figure 7). However, this increase was ,2-fold higher in Tdp2 D1-3 mice than in wild type mice, suggesting that TDP2 protects heamatopoietic cells from genome instability induced by anticancer TOP2 poisons.

TDP2 is the major 59-tyrosyl DNA phosphodiesterase in mammals
In the current study we observe that Tdp2 deletion ablates detectable 59-TDP activity in different mouse tissues and MEFs, consistent with our previous observations in DT40 cells [25]. It is worth noting that other roles have been assigned to this protein, in other cellular processes such as signal transduction and transcriptional regulation [33]. So far, however, we have been unable to detect any spontaneous phenotype caused by TDP2 loss, either at cellular level or in vivo, while dramatic effects are observed upon etoposide treatment. This suggests that the most important function of TDP2, following Top2 induced DNA damage at least, is related to the 59 TDP activity of this enzyme. Additionally, our data suggest that alternative, TDP2-independent, mechanisms of DSB repair are sufficient to cope with the endogenous level of TOP2 damage arising during normal mouse development and life.
A role for human TDP1 in repairing TOP2-induced DSBs was recently suggested by a weak 59-TDP activity of human recombinant protein on DSBs possessing 4-bp 59-overhangs, and on a mild sensitivity of TDP1 2/2 DT40 cells to etoposide [27]. This is also consistent with the increased resistance to etoposide reported in cells highly overexpressing TDP1 [34], and with the reported 59-TDP activity of Tdp1 in Saccharomyces cerevisiae [35]. However, while our standard activity assays employs DSBs with blunt-ended 59-phosphotyrosyl termini, in the current study we similarly failed to detect residual 59-TDP activity in Tdp2 D1-3 MEF extracts on DSB substrates with 4-bp 59-overhangs ( Figure 2C). In addition, in our hands, TDP1 2/2 DT40 cells are not hypersensitive to etoposide, and deletion of TDP1 in TDP2 2/2/2 DT40 cells does not increase sensitivity to etoposide above that observed by TDP2 deletion alone [36]. Consequently, we conclude that TDP2 is the major if not only 59-TDP activity in mammals (as in DT40 chicken cells), at physiologically relevant enzyme concentrations at least.

TDP2 is required for survival and efficient repair upon induction of TOP2-mediated DSBs in mammals
We have shown that Tdp2-deleted mouse cells are hypersensitive to TOP2-induced DNA damage, but not to ionizing radiation, in agreement with previous results with TDP2 2/2/2 DT40 cells [25]. Moreover, we demonstrate that this hypersensitivity correlates with a defect in the repair of etoposide-induced DSBs, as measured by immunostaining for sites of cH2AX, which suggests that TDP2-mediated repair promotes tolerance to TOP2-induced DNA damage in mammalian cells. Remarkably, we observed that TDP2 is required for resistance to TOP2-induced DNA damage not only at the cellular level, but also at the whole-organism level. Indeed, etoposide administration in Tdp2 D1-3 mice resulted in both increased mortality due to intestinal damage and in elevated  toxicity in lymphoid tissue, established in vivo targets of etoposide [32]. TDP2 is therefore a critical factor in the cellular and physiological response to TOP2 poisons.

TDP2 functions in NHEJ and protects genome integrity
One important result of our study was to uncover the relationship between TDP2 and the major DSB-repair pathways, NHEJ and HR. We have shown that TDP2 can convert DSBs with 59-phosphotyrosyl termini into DSBs that are directly ligatable, and might thus be of particular utility in facilitating an error-free NHEJ pathway for repair of TOP2-induced DSBs. Several of our observations support the idea that TDP2 is a component of NHEJ. First, the contribution of TDP2 to cellular resistance to TOP2 induced DNA damage is dependent on the NHEJ machinery and independent on HR, as, with regards to etoposide sensitivity, KU70 is epistatic over TDP2 deletion in DT40 cells while an additive effect is observed when TDP2 is depleted in BRCA2-deficient human fibroblasts. Second, loss of TDP2 results in a DSB repair defect not only in G2 but also in G0/G1, cell cycle stages in which NHEJ is the main if not only DSB repair mechanism available [26,30,37,38]. Third, Tdp2 D1-3 MEFs exhibit increased levels of HR-mediated DSB repair, as measured by elevated frequencies of RAD51 foci and sister chromatid exchange in response to etoposide treatment, which is a phenotype observed in other cell lines in which NHEJ is defective [39][40][41]. Additionally, we have been unable to generate DT40 cells in which both TDP2 and XRCC3 are deleted, suggesting that loss of both TDP2 and HR-mediated DSB repair is cell lethal (unpublished observations).
Whilst the above observations argue strongly that TDP2 is a component of NHEJ, it is important to note that TDPindependent NHEJ mechanisms to process TOP2-linked termini most likely also exist and employ nucleases such as MRN complex, CtIP or Artemis [5,[18][19][20][21]. This explains why KU70 2/2 DT40 cells exhibit much greater hypersensitivity to etoposide than TDP2 2/2/2 DT40 cells, and why Tdp2 D1-3 MEFs still repair a significant fraction of etoposide-induced DSBs in G0/G1 (when NHEJ is the only DSB repair pathway available). Whilst nucleasemediated NHEJ can support cell survival in response to TOP2induced DNA damage, they most likely do so at the expense of increased genetic instability. This is because the removal of sequence from 4-bp complementary 59-overhang during NHEJ will, on the one hand, likely result in chromosome deletions, and on the other hand, increase the propensity for DSB misjoining and chromosome translocation. In contrast, HR provides an error-free pathway to repair TOP2-induced DSBs that have been processed by nucleases, by restoring any missing DNA sequence from and intact sister chromatid in S and G2 [30,37,42]. In this scenario, the increased etoposide-induced genome instability in Tdp2 D1-3 mice, both in cultured cells from these animals and in bone marrow in vivo, likely reflects the use of TDP2-independent NHEJ in cellular contexts in which HR-mediated DSB repair is unavailable (e.g. in cells in G0/G1), or is saturated by the number of etoposideinduced DSBs.
In summary, based on these and our previously published data, we suggest that TDP2 defines a novel error-free NHEJ subpathway that converts TOP2-linked 59-termini into ligatable DNA termini. We suggest that this may be particularly important during G1 and in post-mitotic cells, which lack HR-mediated repair, and thus in which it may be the only mechanism for error-free DSB repair of TOP2-induced DSBs (Figure 8).

TDP2 and cancer therapy
The results presented here can have important implications in the treatment of cancer. Given the widespread use of TOP2 poisons in cancer therapy, and the observed hypersensitivity to TOP2 poisons of cells lacking TDP2, our findings suggest that TDP2 could affect the response of tumour cells to chemotherapy. In this context, TDP2 expression is reportedly elevated in the majority of non-small cell lung cancer cells [43], and mutant-p53dependent over-expression of TDP2 has been implicated in cellular resistance to etoposide in lung cancer cells [44]. TDP2 might therefore be a valid target for overcoming tumour resistance to TOP2 poisons and/or a useful predictive biomarker for clinical response to these agents.
In addition, our toxicity assays in mice and the increased genome instability in cells and in mouse bone marrow correlate well with known side effects of treatment with TOP2 poisons during cancer therapy. This raises the possibility that heterogeneity in expression levels or activity of TDP2 could be an Macroscopic (left) and histological (right) representative image of spleen and thymus from wild-type and Tdp2 D1-3 animals 6 days after treatment. Average weight of these organs 6 s.e.m. and statistical significance by Two-way ANOVA with Bonferroni post-test is shown (centre). D. important etiological factor both in the toxicity that accompanies chemotherapy involving TOP2 poisons [45] and on the incidence of treatment-related hematological malignancy, typically acute leukemia occurring in a relatively high proportion of patients [4,7,8]. Like other acute leukemias, therapy-related malignancies are linked to specific translocations that result in the expression of fusion proteins and contribute in some way to disease development. Intriguingly, in some cases, these translocations map to regions of preferential TOP2 cleavage, supporting a model in which the translocations arise via erroneous repair of TOP2induced DSBs. These translocations are also surprisingly similar to those found in infant leukemia [46], suggesting that erroneous repair of TOP2-induced DSBs may also be a source of primary malignancy. Consistent with this idea, TOP2-induced DSBs are implicated in translocations commonly associated with prostate cancer [47]. In the light of our findings, it is tempting to speculate that TDP2 activity reduces the likelihood of oncogenic translocations, by ensuring rapid and accurate repair of TOP2-induced DSBs. It is possible, however, that TDP2 might occasionally promote a translocation, by liberating a DSB that engages in erroneous DNA ligation, as might be the case in some extremely conservative rearrangements that have been reported [12,13].

Conclusions
We have shown that TDP2 protects mouse cells from the cytotoxic and clastogenic effects of TOP2 poisons, most likely by functioning in error-free pathway for NHEJ. These results have important implications in the treatment of cancer. For example, development of small molecule inhibitors for TDP2 may provide a way of sensitizing particular types of tumor to chemotherapy, though precaution is necessary to consider the possible consequences of TDP2 inhibition on normal cells and on the generation of secondary malignancies.

Ethics statement
All animal procedures were performed in accordance with European Union legislation and with the approval of the Ethical Committee for Animal Experimentation of the University of Leuven and the University of Seville, respectively.
Primary MEFs were isolated from littermate embryos at day 13 p.c. and cultured at 37uC, 5% CO 2 , 3% O 2 in Dubelcco's Modified Eagle's Medium (DMEM) supplemented with penicillin, streptomycin, 10% FCS and non-essential aminoacids. All experiments were carried out between P2 and P4. MEFs were transformed by retroviral delivery of T121, a fragment of the SV40 large T antigen that antagonizes the three Rb family members but not p53 [50]. Transformed MEFs were maintained at 37uC, 5% CO 2 in DMEM supplemented with penicillin, streptomycin and 10% FCS.

Generation of Tdp2 conditional and knockout mice
A targeting construct was generated for Tdp2 in which the first three exons were flanked by loxP sites, followed by an FRT-and loxP-flanked neomycin-resistance (neo) cassette. These three exons encode for the N-terminal half of TDP2 and contain mapped interaction domains for e.g. TDP2 itself, CD40 and TRAF6 [23]. The Tdp2 flEx1-3,neo targeting construct was electroporated in E14 (129Ola) ES cells and correctly recombined ES cell clones were confirmed by Southern blot analysis. The functionality of the loxP sites was shown in vitro by electroporation of a correctly targeted ES cell clone with a Cre-expressing vector. Several correctly targeted ES cell clones were used for aggregation with CD1 morulae and transferred into pseudo-pregnant recipient females to Alternatively, nucleolytic attack on the DNA backbone can also remove the protein adduct from the DSB but genetic information is lost from the ends. Accurate repair of this break would therefore need HR to copy the missing information from the sister chromatid, while NHEJ would result in error-prone repair. doi:10.1371/journal.pgen.1003226.g008 obtain chimaeric mice. Three chimaeric males produced heterozygous offspring after breeding with CD1 wild-type females. The obtained offspring was genotyped with both a loxP-specific and a neo-specific PCR. Intercrosses between Tdp2 flEx1-3,neo/+ mice led to the generation of homozygous floxed Tdp2 mice which were viable and fertile. To delete the critical exons we crossed the heterozygous Tdp2 mice with an EIIa-Cre mouse (Adenovirus EIIa-promoter driven Cre) and obtained Tdp2 flD/+ mice. Intercrosses of the latter mice resulted in viable homozygous knockout mice (from now on denoted Tdp2 D1-3 ) at the normal 25% Mendelian distribution. Southern blot analysis confirmed the complete recombination of the loxP-flanked sequences in the homozygous mice and hence the generation of Tdp2 knockout mice.

Clonogenic survival assays
To determine sensitivity in DT40, cells were plated in 5 ml of medium containing 1.5% (by weight) methylcellulose (Sigma) in 6well plates at 50, 500, and 5000 cells/well per treatment condition. Media also contained the indicated concentration of doxorubicin (Sigma), mAMSA (Sigma) or etoposide (Sigma). In all experiments, cells were incubated for 7-11 days and visible colonies were counted.
Survival assays in MEFs were carried out seeding 2000 cells in 100 mm dishes, in duplicate for each experimental condition. After 6 hours, cells were irradiated or treated with the indicated concentrations of etoposide for 3 hours, washed with PBS and fresh medium was added. Cells were incubated for 10-14 days and fixed and stained for colony scoring in Crystal violet solution (0.5% Crystal violet in 20% ethanol). The surviving fraction at each dose was calculated by dividing the average number of visible colonies in treated versus untreated dishes.
Human fibroblasts were transfected with non-targeting Negative Control and TDP2 smartpool siRNAs (Thermo Scientific) using HyperFect transfection reagent (Invitrogen). Cells were transfected twice in two consecutive days and used for survival 48 hours after second transfection. Other details as described above.

Cytogenetic analysis
Micronuclei and nucleoplasmic bridges were analysed in transformed and low passage primary MEFs previously seeded onto coverslips. Following treatment, cytochalasin B (Sigma) was added at 4 mg/ml to transformed but not to primary MEFs. 22 h (transformed) or 30 h (primary) post-treatment, cells were fixed and subject to DAPI staining as described above. In transformed cells only binucleated cells were scored, which was confirmed by visualization of the cytoplasm with anti Tubulin immunofluorescence (performed as described above).
Chromosomal aberrations were scored in Giemsa stained metaphase spreads. Following treatment, recovery in fresh medium was allowed for 2 h (transformed MEFs) or 4 h (primary MEFs) and demecolcine (Sigma) was added at a final concentration of 0.2 mg/ml. Caffeine (Sigma) was also added at a final concentration of 0.1 mg/ml but only to primary cells. 4 h later cells were collected by trypsinisation, subject to hypotonic shock for 1 hour at 37uC in 0.3 M sodium citrate and fixed in 3:1 methanol:acetic acid solution. Cells were dropped onto acetic acid humidified slides and stained 20 minutes in Giemsa-modified (Sigma) solution (5% v/v in H 2 O).
For SCEs 10 mM BrdU (Sigma) was added to the medium for two complete cycles (approximately 48 hours) before collection. Drug treatment was applied for 30 minutes 6-8 hours before cell collection. Metaphase spreads were obtained as described above. Before Giemsa staining, slides were incubated in Hoescht 33258 solution (10 mg/ml) for 20 minutes, exposed to UV light (355 nm) for 1 hour and washed for 1 hour at 60uC in 206 SCC.

Animal maintenance
The mouse colony was maintained in an outbred 129Ola, CD1 and C57BL/6 background under standard housing conditions, at 2161uC with a photoperiod of 12:12 h (lights on at 8:00). They were housed in isolated cages with controlled ventilation trough HEPA-filters and were in flow cabins. Sterile food pellets and water were available ad libitum. Breeding pairs between heterozygotes (Tdp2 +/flD 6Tdp2 +/flD ) were set to obtain wild-type (Tdp2 +/+ ) and knock-out (Tdp2 D1-3 ) littermates for analysis. Mice were genotyped with Phire Animal Tissue Direct PCR Kit (Thermo) following manufacturer instructions and using primers 59-CCTTCATTACTTCTCGTAGGTTCTGGGTC-39, 59-AC-CCGCTCTTCACGCTGCTTCC-39 and 59-TACACCGTGC-CATAATGACCAAC-39. This results in amplification of a 429 bp fragment from the wild-type allele or 561 bp fragment from the mutant allele.
In vivo etoposide sensitivity At 8 weeks of age, Tdp2 +/+ and Tdp2 D1-3 mice underwent intraperitoneal injection with 3 ml/g of body weight of either DMSO (vehicle control) or etoposide at 25 mg/ml in DMSO for a final dose of 75 mg/kg. Weight and general health status was monitored daily from the day of injection (inclusive). 6 days posttreatment mice were sacrificed by cervical dislocation and dissected for histopathological analysis. Weight of spleen and thymus was recorded prior to their histological or cell content analysis. Bone marrow (BM) from femurs and tibias of each mouse was also obtained.

Micronuclei analysis in vivo
At 8 weeks of age, Tdp2 +/+ and Tdp2 D1-3 mice underwent intraperitoneal injection with 2.5 ml/g of body weight of either 10% DMSO (vehicle control) or etoposide at 400 mg/ml in 10% DMSO for a final dose of 1 mg/kg. Mice were sacrificed by cervical dislocation 24 h after injection and BM from one femur and tibia was extracted and homogenized in 3 ml FBS. Cellular content was concentrated in 150 ml FBS by centrifugation and smears were prepared on glass slides. Following 5 min fixation in methanol, slides were stained 30 min in Giemsa-modified (Sigma) solution (5% v/v in 100 mM Tris-HCl pH 6.8) and visualized under the microscope. 2000 polychromatic erythrocytes (PCE) were scored for the presence of micronuclei (MN-PCE) in each slide. Figure S1 TDP2 promotes survival following TOP2-induced DSBs in mammalian cells. Clonogenic survival of wild-type and Tdp2D1-3 transformed MEFs after continuous exposure to the indicated concentrations of etoposide. Average 6 s.e.m. of three independent experiments and statistical significance by Two-way ANOVA test with Bonferroni post-test is shown. (TIF) Figure S2 Targeted deletion of KU70 in DT40 cells. Southernblot analysis of EcoRI-digested DNA from wild-type (+/+), heterozygous (+/2) and knock-out (2/2) DT40 cells in TDP2 +/ +/+ and TDP2 2/2/2 background. A probe hybridizing to a region of the KU70 locus not contained in the deletion construct was used. The 5.5-kb (wild-type) and 2.8-kb (deleted) expected bands are indicated. Note that two clones were selected for further analysis. Other details as in ''A''. Average 6 s.e.m. of at least three independent experiments is shown. Statistical significance by Twoway ANOVA test with Bonferroni post-test is indicated. TDP2 depletion was performed as previously described [22]. (TIF) Figure S5 The absence of TDP2 sensitizes lymphocyte precursors to etoposide treatment in vivo. FACS analysis of B-cells in bone marrow (A) and T-cells in thymus (B) in wild-type and Tdp2 D1-3 animals 6 days after treatment with 75 mg/kg etoposide or vehicle control (DMSO). Pro-B cell (CD43 + B220 + ), Pre-B cell (CD43 2 B220 low ) immature B cell (CD43 2 B220 high ) and immature T cell (CD4 + CD8 + ) populations are indicated (rectangles). (TIF)