Oncogene (2009) 28, 3880–3891 & 2009 Macmillan Publishers Limited All rights reserved 0950-9232/09 $32.00 www.nature.com/onc ORIGINAL ARTICLE Anticancer DNA intercalators cause p53-dependent mitochondrial DNA re-modelling

N Ashley and J Poulton

Nuffield Department of Obstetrics and Gynaecology, University of Oxford, Level 3, Women’s Centre, John Radcliffe Hospital, Headington, Oxford, UK

Many anticancer drugs, such as doxorubicin (DXR), altering its molecular topology. Consequently, DNA intercalate into nuclear DNA of cancer cells, thereby polymerases and other DNA-related proteins are inhib- inhibiting their growth. However, it is not well understood ited and DNA replication is reduced, causing the death of how such drugs interact with mitochondrial DNA rapidly dividing cancer cells. Unfortunately, anthracy- (mtDNA). Using cell and molecular studies of cultured clines can cause severe cardiotoxicity (Singal and Iliskovic, cells, we show that DXR and other DNA intercalators, 1998), often manifesting months or even years after such as ethidium bromide, can rapidly intercalate into treatment (Steinherz and Steinherz, 1991). The reason for mtDNA within living cells, causing aggregation of this delayed toxicity remains unknown, but one possibility mtDNA and altering the distribution of nucleoid is that anthracyclines induce a lesion capable of damaging proteins. Remodelled nucleoids excluded DXR and the cell over a long period. A potential candidate for such maintained mtDNA synthesis, whereas non-remodelled a lesion is the mitochondrion, as cardiac-specific mito- nucleoids became heavily intercalated with DXR, which chondrial dysfunction is an early and specific feature of inhibited their replication, thus leading to mtDNA anthracycline cardiotoxicity (Jung and Reszka, 2001; depletion. Remodelling was accompanied by extensive Wallace, 2003; Conklin, 2005; Tokarska-Schlattner mitochondrial elongation or interconnection, and was et al., 2006), which can persist even after drug cessation suppressed in cells lacking mitofusin 1 and optic atrophy (Zhou et al., 2001). The mitochondria contain a 16.5-kb 1 (OPA1), the key proteins for mitochondrial fusion. In multicopy genome known as mitochondrial DNA contrast, remodelling was significantly increased by p53 or (mtDNA), and extensive mtDNA damage in the form ataxia telangiectasia mutated inhibition (ATM), indicat- of depletion and deletions has been found in the cardiac ing a link between nucleoid dynamics and the genomic tissue of DXR-treated patients (Lebrecht et al., 2005), as DNA damage response. Collectively, our results show that well as in rodent and cell culture models (Cullinane et al., DNA intercalators can trigger a common mitochondrial 2000; Lebrecht et al., 2004; Suliman et al., 2007). Using response, which likely contributes to the marked clinical the fluorescent DNA dye PicoGreen, which labels toxicity associated with these drugs. specifically nuclear and mtDNA within living cells, as Oncogene (2009) 28, 3880–3891; doi:10.1038/onc.2009.242; we have shown earlier (Ashley et al., 2005, 2008), we published online 17 August 2009 recently showed that anthracyclines are capable of rapidly penetrating into the mitochondria to directly intercalate Keywords: anthracyclines; mtDNA; mitochondria; into mtDNA (Ashley and Poulton, 2009). nucleoids; doxorubicin; mitofusin In vivo, mtDNA is arranged in punctate nucleoid structures (Iborra et al., 2004; Garrido et al.,2003;Legros et al., 2004), which consist of several copies of mtDNA bound to ‘nucleoid’ proteins, such as mitochondrial Introduction single-stranded DNA-binding protein (mtSSB), mito- chondrial transcription factor A (TFAM) and polymerase DNA-intercalating drugs form the cornerstone of many gamma (POLG). Although the functional significance of anticancer regimes. Prominent among them are the nucleoids remains obscure, yeast nucleoids can undergo anthracyclines, a group of compounds including doxo- structural remodelling in response to metabolic stress to rubicin (DXR) and daunorubicin (Hande, 1998). The protect mtDNA from damage (Chen et al., 2005b; Kucej DNA intercalators function by inserting themselves et al., 2008). The loss of the TFAM homologue and between the base pairs of the DNA double helix, thereby nucleoid protein Abf2p sensitizes yeast mtDNA to damage by the DNA intercalator ethidium bromide Correspondence: Dr N Ashley or J Poulton, Nuffield Department of (Chen et al., 2005b). Within human cells, the tumour Obstetrics and Gynaecology, University of Oxford, Level 3, Women’s suppressor p53 can translocate to the mitochondria and Centre, John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK. interact with POLG to protect mtDNA from damage by E-mails: [email protected] or ethidium bromide (Achanta et al., 2005). [email protected] Received 20 January 2009; revised 7 June 2009; accepted 8 July 2009; The mitochondria are dynamic organelles that can published online 17 August 2009 undergo rapid transient fission and fusion, mediated by Anthracyclines and mitochondrial DNA N Ashley and J Poulton 3881 pro-fusion proteins such as optic atrophy 1 (OPA1) methyl ester (TMRM) revealed that the mitochondria and mitofusins 1/2 (MFN1 and MFN2), and pro-fission within DXR-treated cells had undergone marked inter- proteins such as dynamin-related protein 1 (DRP1) linking, with highly interconnected mitochondria (Chen et al., 2005a). Recently, the anticancer DNA (Figure 1aiii and Supplementary Figure 1b). The intercalator actinomycin D has been shown to trigger remodelled nucleoids frequently colocalized with small OPA1 and MFN1-mediated stress-induced mitochon- mitochondrial swellings that were not commonly drial hyperfusion, which has a protective antiapoptotic observed in controls. Similar results were obtained for effect within cells (Tondera et al., 2009). Collectively, DXR-treated A549 cancer cells and H9C2 cardio- this evidence suggests that mitochondria and mtDNA myocytes (Supplementary Figures 1a and c), although may undergo a coordinated stress response in the H9C2s did not show mitochondrial interlinking (Sup- presence of DNA-intercalating drugs. plementary Figure 1d). In this study, we sought to analyse the effects of Anti-DNA and Mitotracker Red labelling (which anticancer DNA-intercalating drugs on mtDNA and detects mitochondria in fixed cells) showed that mitochondria. We show that anthracyclines can directly ethidium bromide also caused marked nucleoid remo- interact with mtDNA at clinically relevant concentra- delling accompanied by mitochondrial interlinking tions, producing major alterations of nucleoids and (Figure 1aiv). mitochondrial morphology and ultimately resulting in Quantification of an experiment similar to that shown the depletion of mtDNA. in Figure 1a showed that DXR could induce mitochon- drial interlinking and remodelling at doses below 1 mM. H9C2 rat cardiomyocytes were more sensitive at low Results concentrations of DXR than human fibroblasts, but the level of remodelling was unaffected by incubating cells Doxorubicin and ethidium bromide alter nucleoid and with the well-known reactive oxygen species scavenger mitochondrial morphology N-acetylcysteine (Figure 1b). DXR significantly reduced To determine the long-term effect of DXR on mtDNA, nucleoid numbers within individual fibroblasts we used PicoGreen fluorescence quenching to monitor (Figure 1c) and significantly increased the average DXR–DNA interactions within normal primary human nucleoid fluorescence (Figure 1d). fibroblasts, as described earlier (Ashley and Poulton, 2009). In untreated cells, nuclear and mtDNA were brightly labelled by PicoGreen (Figure 1ai), and Remodelled nucleoids are enriched with TFAM, mtSSB both signals were substantially quenched by addition and POLG of DXR, indicating the intercalation of DXR into Within nucleoids, mtDNA is complexed with ‘core’ nuclear and mtDNA. After 24 h of DXR exposure, nucleoid proteins, including mtSSB, POLG and PicoGreen fluorescence partially recovered in both the TFAM, and associated, perhaps indirectly, with tumer- nucleus and mitochondria, the latter containing a small ous imaginal disc 1 (Tid1) and ATAD3 (He et al., 2007; number of brightly labelled and grossly enlarged Bogenhagen et al., 2008). To determine whether DXR nucleoids. The average nucleoid diameter of these affected these proteins, we immunolabelled fibroblasts enlarged nucleoids was 1–3 mm, compared with treated with DXR or vehicle. In untreated cells, TFAM B0.8 mm of normal-sized nucleoids. Although normal- was concentrated into small dots (nucleoids) distributed sized nucleoids remained within the DXR-treated cells, evenly along mitochondria, but was heavily concen- they were often heavily quenched by DXR (t ¼ 24 h trated into large mitochondrial blobs after DXR shown by the arrow in the inset). Hereafter, we term exposure (Figure 2a). Large swathes of the mitochon- these DXR-induced nucleoid alterations as ‘nucleoid dria thus became depleted of detectable TFAM. The remodelling’. After DXR exposure, the surviving focal accumulations of TFAM colocalized with mito- cells, which were allowed to recover in drug-free chondrial swellings and with remodelled nucleoids medium for 72 h, showed a marked reduction in (detected using anti-DNA IgM1—not shown). Similar nucleoid PicoGreen labelling, despite recovery of the results were obtained using ethidium (not shown). Note nuclear signal, suggestive of mtDNA depletion, which that the strong red nuclear signal in DXR-treated was confirmed using quantitative real-time PCR cells is due to the auto-fluorescence of DXR intercalated (Figure 3d). Some remodelled nucleoids were still visible into nuclear DNA, and not because of Mitotracker. (shown by the arrow). Mitochondrial DXR was not visible. Western blot of Immunolabelling of DXR-treated fibroblasts with TFAM showed that the total amount of TFAM anti-DNA IgM (IgM1), capable of detecting nucleoids remained unaffected by DXR treatment (Supplementary (Legros et al., 2004), confirmed that DXR caused gross Figure 2d). enlargement of nucleoids (Figure 1aii). The enlarged The immunolabelling of POLG1 (Figure 2b) and remodelled nucleoids appeared to be composed of mtSSB (Supplementary Figure 2c) also showed marked aggregates of smaller nucleoids and there was a notice- reorganization into remodelled nucleoids in response to able clustering of nucleoids (shown by the arrows), DXR treatment, although it was rarely as marked as which was generally not seen in untreated cells. TFAM. Reduced camera exposure revealed that POLG- Colabelling of DXR-treated fibroblasts with PicoGreen enriched nucleoids consisted of small ‘dot-like’ struc- and the mitochondrial probe tetramethyl rhodamine tures (Figure 2b, bottom panels). Neither Tid1 or

Oncogene Anthracyclines and mitochondrial DNA N Ashley and J Poulton 3882

Figure 1 Doxorubicin (DXR) and ethidium alter mitochondrial DNA (mtDNA) nucleoids and mitochondrial morphology. (ai) Time course of PicoGreen labelling of live primary human fibroblasts incubated with DXR (t ¼ hours of exposure). (aii) Anti-DNA (IgM1) labelling of mtDNA within DXR- (3.4 mM 24 h) and vehicle-treated fibroblasts. (aiii) PicoGreen/tetramethyl rhodamine methyl ester (TMRM) colabelling of DXR- and vehicle-treated fibroblasts (2 mM 24 h). (aiv) Anti-DNA (IgM1)/Mitotracker Red labelling of mtDNA/mitochondria within fibroblasts treated with vehicle or ethidium bromide (0.5 mg/ml 24 h). (b, upper panels) The percentage of DXR- and vehicle-treated fibroblasts showing mitochondrial interlinking after 24 h (n ¼ 200). (b, lower panels) The percentage of DXR- and vehicle-treated human fibroblasts and H9C2 rat cardiomyocytes showing nucleoid remodelling after 24 h, in the presence or absence of 100 mg/ml N-acetylcysteine (n ¼ 200). (c) Mean nucleoids per cell within fibroblasts treated with DXR (3.4 mM 24 h) or vehicle (n ¼ 10). (d) Average nucleoid fluorescence (a.u. ¼ arbitrary units) within fibroblasts treated with DXR/vehicle (3.4 mM 24 h) (n ¼ 100). *Po0.05 versus control. Error bars, þ s.d. Results are representative of three independent experiments. Size bars: 20 mm, except in (aii) (10 mm).

ATAD3 showed much relocalization in response to ingly, DXR treatment substantially reduced the DXR, and both remained evenly distributed within mtDNA-binding affinity of a second anti-DNA anti- mitochondria. This was also the case for cytochrome c body (IgM2) (Figure 2c), suggesting a structural (not shown) (Supplementary Figures 2a and b). Surpris- alteration to the mtDNA.

Oncogene Anthracyclines and mitochondrial DNA N Ashley and J Poulton 3883

Figure 2 Doxorubicin (DXR) alters the distribution of mitochondrial transcription factor A (TFAM), mitochondrial single- stranded DNA-binding protein (mtSSB) and polymerase gamma (POLG), and reduces mitochondrial DNA (mtDNA) binding by anti- DNA IgM2. (a) Anti-TFAM/Mitotracker labelling of fibroblasts incubated with vehicle/DXR (3.4 mM 24 h). The retained DXR fluorescence within the nuclei is indicated by ‘N’. (b) Anti-POLG/Mitotracker labelling of fibroblasts incubated with vehicle or DXR for 24 h (3.4 mM 24 h). 4’,6-Diamidino-2-phenylindole (DAPI) was used to stain the nuclei blue. The DXR-induced POLG accumulations are shown by the arrow. The bottom panels show a higher magnification of a POLG labelling cell acquired using reduced camera exposure. (c) Anti-DNA IgM2/Mitotracker labelling of fibroblasts incubated with vehicle/DXR (3.4 mM 24 h). The arrow shows a mitochondrial swelling indicating a remodelled nucleoid. Size bars: 20 mm. Results are representative of at least three independent experiments.

Oncogene Anthracyclines and mitochondrial DNA N Ashley and J Poulton 3884 In conclusion, remodelled nucleoids were enriched replication using a32P-dNTP labelling of permeabilized with the core nucleoid proteins TFAM, mtSSB and A549 cells (Emmerson et al., 2001), which undergo POLG, and had reduced affinity for certain anti-DNA remodelling and mitochondrial interlinking in response antibodies. to DXR (Supplementary Figure 1a). DXR substantially reduced the amount of mtDNA a32P-dCTP incorpora- tion relative to the total mtDNA mass (determined using DXR inhibits mtDNA replication but remodelled a total human mtDNA probe), indicating that DXR nucleoids remain replicatively active inhibited mtDNA replication (Figure 3a, left panel). As DXR inhibits polymerases (Hixon et al., 1981; Ellis Quantification of the bands showed a dose–response et al., 1987), we investigated the effect of DXR on DNA effect, with a >50% reduction of mtDNA synthesis

Figure 3 Doxorubicin (DXR) inhibits mitochondrial DNA (mtDNA) synthesis and causes its depletion. (a) Phosphor-screen image of a32P-dCTP incorporation of PVUII-digested mtDNA, representing newly synthesized mtDNA, isolated from DXR- and vehicle- treated A549 (3.4 mM DXR 24 h), and the corresponding signal of the same mtDNA fragments hybridized to a total human mtDNA probe, representing the total mtDNA mass per lane. The graph shows quantification of the relative mtDNA fragment intensities (mtDNA replication), on the basis of the mean of incorporated (I) versus probed (P) signal ratios (n ¼ 3). (b, left panel) Quantification of the average number of bromodeoxyuridine (brdU)-labelled (replicating) nucleoids of DXR- and vehicle-treated fibroblasts (3.4 mM DXR 24 h) (n ¼ 100), after labelling with brdU (10 mM 6 h), and immunodetection by anti-brdU. (b, right panel) Average nuclear brdU incorporation of S-phase cells from the same experiment (n ¼ 100). (c, upper panel) Image of brdU incorporation into the cellular DNA of fibroblasts used to generate data for (b). DAPI was used to stain nuclei and hence the S-phase nuclei appear white. (c, lower panel) Anti-mitochondrial transcription factor A (TFAM)/brdU co-labelling of DXR- and vehicle-treated fibroblasts (3.4 mM DXR 24 h). (d) The mtDNA to nuclear DNA ratios of fibroblasts treated with vehicle/DXR (1.7 mM) for up to 48 h, as determined by quantitative real-time PCR (n ¼ 3). Note that the drug was washed out after 24 h to reduce cellular toxicity. This had no effect on nucleoid remodelling or PicoGreen quenching (not shown). *Po0.05. Error bars þ s.d. Size bars: 20 mm. Results are representative of three independent experiments.

Oncogene Anthracyclines and mitochondrial DNA N Ashley and J Poulton 3885 after exposure to 3.4 mM DXR (Figure 3a, right panel). (Supplementary Figure 3aiv). DXR did not significantly We also examined DNA synthesis in DXR-treated compromise the mitochondrial function of fibroblasts or fibroblasts using bromodeoxyuridine (brdU) labelling H9C2 cells, as determined by adenosine triphosphate (Magnusson et al., 2003). The DXR- and vehicle-treated generation, cellular oxygen consumption, cytochrome fibroblasts were incubated with brdU, and brdU- c oxidase histochemistry and JC-1 fluorescence labelled nuclear and mtDNA were visualized using (Supplementary Figure 3b). anti-brdU antibody. Measurements of the brdU inten- sity of S-phase nuclei and the number of brdU-labelled DXR-induced nucleoid remodelling is mediated by OPA1 nucleoids showed that DXR strongly inhibited nuclear and mitofusin 1, but not by mitofusin 2 synthesis, and sharply reduced the number of replicating We next determined the influence of mitochondrial nucleoids (Figure 3b). Most of the replicating nucleoids fusion on DXR-induced remodelling by comparing in DXR-treated cells were enlarged, that is, remodelled wild-type mouse embryonic fibroblasts (MEFs) with (Figure 3c, upper panels). As the earlier anti-DNA mitochondrial fusion-deficient MEFs harbouring null labelling had shown that numerous non-remodelled mutations in either MFN1 or MFN2 (Chen et al., 2003). nucleoids remained within DXR-treated cells These MFN mutants contained fragmented mitochon- (Figure 1aii), we determined whether DXR had pre- dria (Supplementary Figure 4a). Anti-DNA (IgM1) ferentially inhibited their replication, by double labelling labelling indicated that although DXR induced sub- DXR-treated cells with brdU and anti-TFAM stantial nucleoid remodelling within wild-type and (Figure 3c, lower panels). Within untreated cells most MFN2 null MEFs, remodelling was almost totally of the TFAM-positive nucleoids showed brdU labelling, suppressed within MFN1 null MEFs (Figure 4a). Next but within DXR-treated cells only the largest remo- we determined whether OPA1 was also involved in delled nucleoids showed brdU incorporation compar- nucleoid remodelling, by ablating OPA1 levels in human able to or greater than that of untreated nucleoids. fibroblasts using small interfering (si)RNA and con- A large number of non- or partially remodelled firming by western blot (Supplementary Figure 4b). nucleoids labelled by TFAM showed little or no brdU DXR nucleoid remodelling, as detected by PicoGreen, label (examples shown by the arrows in the insets). was inhibited within the OPA1 siRNA-transfected To determine whether DXR caused mtDNA deple- fibroblasts, but not in scramble siRNA-transfected cells. tion, we measured the total mtDNA copy number using Quantification showed that remodelling was almost quantitative real-time PCR (Figure 3d). The mtDNA totally suppressed by loss of OPA1 and MFN1, but levels were unaffected after a 24-h drug exposure, but not by loss of MFN2 (Figure 4a, graph). showed a significant depletion of mtDNA after 48 h To determine whether the suppression of nucleoid when compared with controls. The DXR was removed remodelling within MFN1 null cells led to increased after 24 h to reduce cellular toxicity, but this did not sensitivity of mtDNA to DXR, we pulse-labelled wild- alter the level of PicoGreen quenching (not shown), type MFN1 and MFN2 MEFs with brdU after a 24-h indicating that DXR remained intercalated. treatment with DXR (1.7 mM). As shown in Figure 4d, Thus, DXR preferentially inhibits mtDNA replica- mtDNA synthesis was markedly suppressed within tion of non-remodelled nucleoids, causing mtDNA MFN1 null cells, yet was well maintained by MFN2 depletion. and wild-type cells. To determine whether suppression of mitochondrial fission could alter nucleoids, we ablated DRP1 using Nucleoid remodelling is not secondary to apoptosis and siRNA. PicoGreen–tetramethyl rhodamine methyl ester does not alter short-term mitochondrial function co-staining showed that prolonged knockdown of We next determined whether remodelling preceded cell DRP1 in fibroblasts led to the formation of hyperfused death or compromised mitochondrial function. Using mitochondria and large mitochondrial swellings that terminal deoxynucleotidyl transferase-mediated dUTP colocalized with apparently remodelled nucleoids nick end labelling, we determined the cellular apoptotic (Figure 4c). index of DXR/vehicle-treated cells (Supplementary Collectively, the data above suggest that the nucleoid Figure 3ai). After a 24- to 48-h DXR exposure remodelling induced by DXR is mediated by MFN1 and (3.4 mM), 10% of cells were apoptotic, which was well o OPA1, but not by MFN2, and that suppression of below the number of cells showing remodelled nucleoids remodelling renders mtDNA replication more suscep- or increased mitochondrial interconnection. Calcein/ tible to DXR inhibition. Furthermore, nucleoid remo- propidium labelling showed that 5% of the attached o delling can be induced in the absence of DXR by cell population were necrotic (Supplementary Figure inhibition of mitochondrial fission. 3aii). 4’,6-Diamidino-2-phenylindole/cytochrome c im- munolabelling showed no differences between vehicle- and DXR-treated cells, unlike cells treated with hydro- p53, ATM and POLG suppress DXR-induced remodelling gen peroxide, which showed nuclear located cytochrome As the DNA damage induced by DXR can trigger c and chromosomal condensation (very bright 4’,6- upregulation of p53 and its translocation to the diamidino-2-phenylindole signal), the hallmarks of mitochondria, where it is involved in mtDNA protection apoptosis (Supplementary Figure 3aiii). Remodelling (Nithipongvanitch et al., 2007), we determined whether did not occur during staurosporine-induced apoptosis altering the p53 activity altered nucleoid remodelling.

Oncogene Anthracyclines and mitochondrial DNA N Ashley and J Poulton 3886

Figure 4 Influence of mitofusin 1/2 (MFN1/2), optic atrophy 1 (OPA1) and dynamin-related protein 1 (DRP1) on doxorubicin (DXR)-induced nucleoid remodelling. (a, left panel) Anti-DNA IgM1 labelling of DXR- and vehicle-treated wild-type, MFN1 null and MFN2 null mouse embryonic fibroblasts (MEFs) (7.2 mM DXR 24 h). (a, right panel) PicoGreen labelling of fibroblasts transfected with either OPA1 targeting small interfering RNA (siRNA) (48 h) or a scramble siRNA control, after treatment with DXR (7.2 mM 24 h). (a, graph) Quantification of the percentage of cells with DXR-induced remodelled nucleoids (n ¼ 200). (b) Representative bromodeoxyuridine (brdU) labelling (20 mM 24 h) of wild-type, MFN1 and MFN2 null cells, treated with vehicle/DXR (3.4 mM 24 h). (c) PicoGreen/tetramethyl rhodamine methyl ester (TMRM) labelling of DRP1/scramble siRNA-treated cells (6 days). Bars 20 mm. Results, þ s.d. Results are representative of three independent experiments.

In DXR-treated fibroblasts, p53 was upregulated and DXR is mediated mainly by ATM (Kurz et al., 2004), this could be inhibited by p53 siRNA (Figure 5a). Using we compared DXR remodelling in cells treated with PicoGreen, we found that DXR remodelling was either vehicle or the ATM inhibitor KU55933. ATM substantially increased by p53 knockdown (Figure 5b). inhibition substantially increased the level of DXR Chemical inhibition of p53 by pifithrin-a also increased remodelling (Figure 5b). remodelling, confirming the specificity of the result. To determine whether the transcriptional activity of Remodelled nucleoids were larger and brighter in the p53 was necessary for its influence on remodelling, p53-inhibited cells than in scramble or vehicle-treated we treated RH30 cells with DXR in the presence or controls (Figure 5b, lower panels). As p53 activation by absence of pifithrin-a, and counted the proportion of

Oncogene Anthracyclines and mitochondrial DNA N Ashley and J Poulton 3887

Figure 5 p53, ataxia telangiectasia mutated (ATM) and polymerase gamma (POLG) suppress doxorubicin (DXR)-induced nucleoid remodelling. (a, upper panel) Anti-p53 immunolabelling of fibroblasts treated with vehicle or DXR (1.7 mM for 24 h) or transfected with p53/control after exposure to 3.4 mM DXR for 24 h. (a, lower panel) Immunoblot of an experiment similar to (a). (b, upper panels) Quantification of nucleoid remodelling of DXR-treated (3.4 mM 24 h) fibroblasts in which p53 was inhibited by small interfering RNA (siRNA) or pifithrin-a (10 mg/ml), or ATM inhibited by KU55933 (10 mM)(n ¼ 20). (b, lower panels) Typical PicoGreen staining of cells treated with DXR in the presence of vehicle or pifithrin-a.(c) Effect of vehicle or pifithrin-a (10 mg/ml) on DXR (3.4 mM 24 h)-induced nucleoid remodelling in RH30 cells expressing a p53 transcription mutant (n ¼ 20). (d) Nucleoid remodelling exhibited by wild-type or POLG exonuclease null mouse embryonic fibroblasts (MEFs) treated with DXR (3.4 mM 24 h) (n ¼ 100). Results, þ s.d. *Po0.05. Bars, 20 mm. Results are representative of three independent experiments.

cells showing strong nucleoid remodelling. RH30 cells In summary, the data above indicate that p53/ATM express a mutant p53 protein that is unable to initiate and POLG exonuclease activity suppresses DXR transcription of p53-responsive genes (Felix et al., 1992; nucleoid remodelling. McKenzie et al., 2002). The amount of remodelling was increased by pifithrin-a treatment within RH30, suggest- ing that p53 transcriptional activity was not required for DNA intercalators induce nucleoid remodelling the suppression of nucleoid remodelling (Figure 5c). To determine the forms of DNA damage that are To determine the influence of POLG exonuclease capable of triggering nucleoid remodelling, we tested (proofreading) activity on DXR nucleoid remodelling, several drugs with known DNA-damaging effects for we compared POLG wild-type and POLG exonuclease their ability to induce remodelling, which was detected À/À MEFs derived from a transgenetic knockin mouse by either anti-DNA or PicoGreen labelling (Table 1). (Trifunovic et al., 2004). POLG exonuclease À/À MEFs Certain anticancer drugs were found to induce varying exhibited significantly more DXR-induced nucleoid levels of remodelling, including several anthracyclines, remodelling than identically treated wild-type controls actinomycin D and ditercalinium. All compounds that (Figure 5d). induced remodelling were known DNA intercalators.

Oncogene Anthracyclines and mitochondrial DNA N Ashley and J Poulton 3888 Table 1 Compounds tested for ability to induce nucleoid remodelling in fibroblasts Drug/process Tested concentration Physiological effect Nucleoid remodelling inducer

Doxorubin (anthracycline) 0.7–7.5 mM DNA intercalator, II inhibitora Yes Daunorubicin (anthracycline) 0.5–3.4 mM DNA intercalator, topoisomerase II inhibitora Yes Epirubicin (anthracycline) 0.5–3.4 mM DNA intercalator, topoisomerase II inhibitora Yes Idarubicin (anthracycline) 0.5–3.4 mM DNA intercalator, topoisomerase II inhibitora Yes Aclarubicin (anthracycline) 1.7–3.4 mM DNA intercalator, topoisomerase II inhibitorb Yes Ethidium bromide 0.5–1 mg/ml DNA intercalator Yes Actinomycin D 5–10 mM DNA intercalator, topoisomerase inhibitora Yes Ditercalinium 0.5–1 mg/ml DNA intercalator, topoisomerase inhibitorb Yes Etoposide 10–20 mM Topoisomerase II inhibitorb No Ciprofloxacin 10–50 mM Topoisomerase II inhibitora No Sobuzoxane 50–100 mM Topoisomerase II inhibitorb No Camptothecin 10–20 mM Topoisomerase I inhibitor No ICRF-193 50–100 mM Topoisomerase II inhibitorb No Cisplatin 10–50 mM DNA cross-linker No Bleomycin 50 mg/ml DNA damager No Hydroxyurea 0.5–1 mM Ribonucleotide reductase inhibitor No Menadione 5–10 mM Mitochondrial ROS generator No Bromodeoxyuridine 10–20 mM Nucleoside analogue No Dideoxycytidine 5–15 mM Nucleoside analogue No g-Irradiation 3000 rad DNA damager No Chloramphenicol 25–50 mg/ml Mitochondrial translation inhibitor No Xanthine/xanthine oxidase 50 U ROS generator No H2O2 200–500 mM ROS generator No

Abbreviation: ROS, reactive oxygen species. Cells were incubated with a medium containing compounds at the specified concentrations for 24 h, and nucleoid remodelling was detected using PicoGreen. aPoison. bCatalytic inhibitor.

Remodelling was not observed with specific inhibitors ‘nucleoid remodelling’. Although many non-remodelled of , reactive oxygen species generators, nucleoids remained, most of them failed to recover their g-radiation, DNA alkylating agents or nucleoside fluorescence, indicating that they lacked the ability to analogues, indicating that DNA intercalation and not exclude DXR, unlike their remodelled counterparts. other forms of DNA damage triggered nucleoid Similar results were found within DXR-treated H9C2 remodelling. cardiomyocytes and A549 cancer cells. Anti-DNA labelling confirmed that DXR-exposed nucleoids were abnormal, and also showed that ethidium also induced Discussion nucleoid alterations. As these changes largely preceded DXR-induced cell death, were not induced by staur- This study addresses the hypothesis that anticancer osporine and were induced by nontoxic doses of DNA-intercalating drugs can directly affect mtDNA. ethidium, we conclude that remodelling was not We have earlier shown that such drugs can rapidly secondary to apoptosis. intercalate into the mtDNA of living cells (Ashley and TFAM and, to a lesser extent, even mtSSB and POLG Poulton, 2009). Here we show how these drugs became strongly enriched into remodelled nucleoids. In progressively alter the structure and arrangement of contrast, nucleoid proteins that were not believed to be nucleoids and mitochondria, by causing nucleoid directly bound to mtDNA, such as ATAD3 and aggregation and mitochondrial interlinking. These tumorous imaginal disc 1, showed much less alteration effects were influenced by mitochondrial dynamics and in their mitochondrial distribution. The mtDNA bind- ATM/p53 activation, thereby showing that mitochon- ing by another anti-DNA IgM (IgM2) capable of drial nucleoids are linked to the genomic DNA damage detecting mtDNA was markedly reduced by DXR, response. suggesting that nucleoid structure was altered so as to Within fibroblasts exposed to DXR, initial strong reduce its binding affinity. As IgM2 and IgM1 were fluorescent quenching of PicoGreen by DXR showed raised against differing cellular epitopes, remodelling that substantial intercalation occurred rapidly within may alter the access of the antibodies to one but not the nuclear and mtDNA (Figure 1a), consistent with earlier other epitope. a32P-dCTP radiolabelling and brdU findings (Ashley and Poulton, 2009). However, after incorporation showed that mtDNA synthesis was several hours, mtDNA fluorescence started to recover significantly reduced by DXR (Figure 3a) and DXR- and was accompanied by the formation of giant, exposed cells consequently became depleted of mtDNA abnormally bright nucleoids, a process we describe as (Figure 3d).

Oncogene Anthracyclines and mitochondrial DNA N Ashley and J Poulton 3889 Remodelling was closely linked to mitochondrial within remodelled nucleoids shows that they were more dynamics as remodelling was accompanied by marked efficient at excluding DXR than non-remodelled nu- mitochondrial interlinking in certain cell types. Further- cleoids. Second, brdU/TFAM colabelling showed that more, remodelling required the mitochondrial pro- mtDNA synthesis was better maintained by large fusion proteins MFN1 and OPA1 (Figure 4), suggesting remodelled nucleoids than by smaller or non-remodelled that remodelling requires the aggregation of nucleoids nucleoids (Figure 3c). Third, mtDNA synthesis in the from separate mitochondria. Remodelling did not presence of DXR was markedly reduced within MFN1 require MFN2, which has a lower pro-fusion activity null cells in which remodelling was suppressed, when than that of MFN1 and is not essential for fusion compared with that of MFN2 null or wild-type cells (Cipolat et al., 2004; Chen et al., 2005a; Tondera et al., (Figure 4a). This suggests that reducing nucleoid 2009). The DXR-mediated mitochondrial interlinking remodelling renders mtDNA more susceptible to DXR may represent stress-induced mitochondrial hyperfu- by, for example, increasing DXR intercalation. Finally, sion, which has recently been shown to be a pro-survival mitochondrial function was well maintained within response of cells treated with toxic agents (Tondera DXR-treated cells (Supplementary Figure 3), indicating et al., 2009). A noteworthy finding is that actinomycin that remodelling did not compromise short-term mito- D, which can induce stress-induced mitochondrial chondrial function, as might be expected if the process hyperfusion, also triggers nucleoid remodelling (Supple- were pathogenic. mentary Figure 2c), suggesting that both phenomena are This study and our earlier work (Ashley and Poulton, linked. Strikingly, both stress-induced mitochondrial 2009) show that mtDNA is a direct target for several hyperfusion and nucleoid remodelling similarly require clinically useful anticancer DNA drugs. This is consis- OPA1 and MFN1, but not MFN2 activity. Mitochon- tent with the extensive mtDNA damage associated with drial fission may also have a function in nucleoid human DXR pathology (Lebrecht et al., 2005). Ethi- remodelling, as prolonged inhibition of fission through dium bromide, which had a similar effect on mtDNA as knockdown of DRP1 caused the formation of highly DXR, can trigger long-term mitochondrial dysfunction interlinked mitochondria and enlarged nucleoids that by damaging mtDNA (von Wurmb-Schwark et al., colocalized with mitochondrial swellings (Figure 4c). 2006). Thus, in vivo, DXR intercalation into cardiac However, immunoblotting failed to detect any DXR- mtDNA and subsequent damage may contribute to the induced alterations of the expression of OPA1, DRP1, longer-term tissue dysfunction characteristic of DXR MFN2 or MFN2 (N Ashley, unpublished observations). pathology. The nucleoid remodelling was specific to DNA intercalators and not to other forms of damage, such as reactive oxygen species (Table 1). The DNA damage Materials and methods induced by DXR triggered the accumulation of p53 within the nuclei of treated cells, and inhibition of p53, Materials or its upstream activator ATM, increased DXR All chemicals were from Sigma-Aldrich Company Ltd nucleoid remodelling (Figure 5b), indicating that p53 (Poole, UK), Axxora (UK) Ltd (Nottingham, UK) or repressed remodelling. As this was apparently not Invitrogen Ltd (Paisley, UK). Antibodies to b-actin, cyto- because of the p53’s role as a transcription factor chrome c, p53, POLG and Tid1 were from Fisher Scientific (Figure 5c), it might well be a result of its direct UK Ltd (Loughborough, UK); anti-TFAM was raised in our involvement in mtDNA protection. p53 can interact own laboratory. Anti-DNA (IgM1) was from PROGEN Biotechnick GmbH (Heidelberg, Germany), or a kind gift of with POLG (Achanta et al., 2005), and is important for Peter Cook (IgM2). Anti-brdU was from Roche Diagnostics protecting mtDNA from damage by ethidium bromide Ltd (Burgess Hill, UK). Anti-mtSSB, anti-ATAD3 and (Achanta et al., 2005) and DXR (Nithipongvanitch ditercalinium were generous gifts from Professor Zeviani, Dr et al., 2007). The POLG exonuclease null MEFs I Holt and Professor BP Roques, respectively. MFN1/2 (Chen exhibited markedly increased remodelling when com- et al., 2003) and POLG exonuclease null (Trifunovic et al., pared with their wild-type counterparts (Figure 5d), 2004) MEFs were generous gifts from Dr D Chan and Dr A indicating that the proofreading exonuclease activity of Trifunovic, respectively. POLG also suppresses nucleoid remodelling. As p53 enhances the exonuclease activity of POLG (Bakha- Cell culture and drug treatment nashvili et al., 2008, 2009), it may exert its influence on Mycoplasma-free normal human fibroblasts, H9C2 rat cardi- nucleoid remodelling through its enhancement of POLG omyocytes, RH30 and A549 cells were cultured as described in exonuclease activity. an earlier study (Ashley et al., 2005). For drug treatments, cells B 6 2 In yeast, nucleoids can undergo structural alterations were seeded at 2 Â 10 per cm in dishes and allowed to attach overnight before exposure to 1 ml of drug/vehicle to protect mtDNA from damage during metabolic medium per cm2 of vessel surface area. stress, possibly by allowing mtDNA to adopt a compact state (Chen et al., 2005b; Kucej et al., 2008). It is Cytochemical labelling therefore conceivable that the ‘nucleoid remodelling’, The immunocytochemical labelling, histochemical staining of which we have characterized within mammalian cells, cytochrome c oxidase activity, brdU labelling of newly helps to protect mtDNA from damaging intercalating synthesized DNA and PicoGreen labelling were performed as agents. Several lines of evidence support this theory. described in an earlier study (Ashley et al., 2005). Tetramethyl First, the absence of PicoGreen quenching by DXR rhodamine methyl ester co-staining was carried out using

Oncogene Anthracyclines and mitochondrial DNA N Ashley and J Poulton 3890 30 nM for 20 min. Cells were visualized using a Leica DMI50 CellTiter-Glo assay (Promega Corporation, Madison, WI, microscope (Leica Microsystems (UK) Ltd, Milton Keynes, USA) as per the manufacturer’s instructions, using a Turner UK) fitted with a Hamamatsu ORCA-II camera (Hamamatsu Biosystems luminometer (Turner BioSystems, Inc., Sunnyvale, Photonics UK Ltd, Welwyn Garden City, UK). CA, USA).

Propidium iodide/calcein-AM determination of viability Radiolabelling of mtDNA synthesis 32 Cells were incubated with 1 mM calcein-AM and 0.5 mg/ml The a P-dCTP labelling of newly synthesized mtDNA in propidium iodide for 10 min and were analysed by microscopy. permeabilized cells was performed as described in an earlier study (Emmerson et al., 2001; Ashley et al., 2007). Terminal deoxynucleotidyl transferase-mediated dUTP nick end labelling Small interfering RNA transfections Detection of methanol-fixed apoptotic cells was done using a For OPA1/p53/DRP1 knockdowns, we used the previously kit from Takara Bio Inc. (Shigo, Japan). validated siRNA sequences, 50-GTTATCAGTCTGAGCCA GGdTdT-30 for OPA1 (Olichon et al., 2003), 50-GCAUGAA CCGGAGGCCCAUdTdT-30 for p53 (Martinez et al., 2002) Digital image analysis and 50-GCAGAAGAATGGGGTAAATdTdT-30 for DRP1 The SimplePCI 6 imaging software (Hamamatsu Photonics (Griparic et al., 2007). Qiagen (Qiagen, Crawley, UK) AllStars UK Ltd) was used for image capture. Image analysis was siRNA was used as a negative control and 100 nM siRNA was performed on unprocessed, 8-bit digital images using ImageJ transfected using Dharmacon Dharmafect (Fisher Scientific (http://rsbweb.nih.gov/ij/). Fluorescent intensities (pixel grey UK Ltd), as per the manufacturer’s protocols. values) of defined regions of interest (nuclei or nucleoids), corrected for cellular background, were determined as described (Abramoff et al., 2004). Cell counts were done by Immunoblotting manually counting cells from randomly acquired images. Immunoblotting was performed as described in an earlier study (Poulton et al., 1994). Quantitative real-time PCR The relative ratio of mtDNA to nuclear DNA was determined Statistical analysis as described in an earlier study (Ashley et al., 2005), except Statistical analysis was calculated using Microsoft Excel or that the nuclear probe was labelled with Vic at the 50 end and SPSS. Significance was calculated using Student’s t-test. amplifications were done simultaneously using a GeneAmp 7700 sequence detection system (Applied Biosystems Inc, Foster City, CA, USA). Conflict of interest

Oxygen consumption The authors declare no conflict of interest. The cells were harvested, counted with a haemocytometer and resuspended in Dulbecco’s modied Eagle’s medium without serum or glucose, at 5 Â 106 cells/ml. Oxygen consumption was Acknowledgements performed on whole cells at 37 1C, using a Hansatech Clarke- type oxygen electrode (Hansatech Ltd, King’s Lynn, UK). We thank I Sargent for providing the equipment; E Brampton, M Zeviani, I Holt, B Roques, A Trifunovic, P Cook and Adenosine triphosphate measurements D Chan for materials; and K Morten and F Brook for The cells were harvested and counted using a haemocytometer. technical support. This work was funded by the MRC/ Adenosine triphosphate was measured using a Promega Wellcome Trust.

References

Abramoff MD, Magelhaes PJ, Ram SJ. (2004). Image processing with Ashley N, Poulton J. (2009). Mitochondrial DNA is a direct target of ImageJ. Biophotonics Int 11: 36–42. anti-cancer anthracycline drugs. Biochem Biophys Res Commun 378: Achanta G, Sasaki R, Feng L, Carew JS, Lu W, Pelicano H et al. 450–455. (2005). Novel role of p53 in maintaining mitochondrial genetic Bakhanashvili M, Grinberg S, Bonda E, Rahav G. (2009). Excision stability through interaction with DNA Pol gamma. EMBO J 24: of nucleoside analogs in mitochondria by p53 protein. AIDS 27: 3482–3492. 779–788. Ashley N, Adams S, Slama A, Zeviani M, Suomalainen A, Andreu AL Bakhanashvili M, Grinberg S, Bonda E, Simon AJ, Moshitch- et al. (2007). Defects in maintenance of mitochondrial DNA are Moshkovitz S, Rahav G. (2008). p53 in mitochondria enhances associated with intramitochondrial imbalances. Hum the accuracy of DNA synthesis. Cell Death Differ 15: 1865–1874. Mol Genet 16: 1400–1411. Bogenhagen DF, Rousseau D, Burke S. (2008). The layered Ashley N, Harris D, Poulton J. (2005). Detection of mitochondrial structure of human mitochondrial DNA nucleoids. J Biol Chem DNA depletion in living human cells using PicoGreen staining. Exp 283: 3665–3675. Cell Res 303: 432–446. Chen H, Chomyn A, Chan DC. (2005a). Disruption of fusion results in Ashley N, O’Rourke A, Smith C, Adams S, Gowda V, Zeviani M et al. mitochondrial heterogeneity and dysfunction. J Biol Chem 280: (2008). Depletion of mitochondrial DNA in fibroblast cultures from 26185–26192. patients with POLG1 mutations is a consequence of catalytic Chen H, Detmer SA, Ewald AJ, Griffin EE, Fraser SE, Chan DC. mutations. Hum Mol Genet 17: 2496–2506. (2003). Mitofusins Mfn1 and Mfn2 coordinately regulate mitochon-

Oncogene Anthracyclines and mitochondrial DNA N Ashley and J Poulton 3891 drial fusion and are essential for embryonic development. J Cell Biol Legros F, Malka F, Frachon P, Lombes A, Rojo M. (2004). 160: 189–200. Organization and dynamics of human mitochondrial DNA. J Cell Chen XJ, Wang X, Kaufman BA, Butow RA. (2005b). Aconitase Sci 117: 2653–2662. couples metabolic regulation to mitochondrial DNA maintenance. Magnusson J, Orth M, Lestienne P, Taanman JW. (2003). Replication Science 307: 714–717. of mitochondrial DNA occurs throughout the mitochondria of Cipolat S, Martins de Brito O, Dal Zilio B, Scorrano L. (2004). OPA1 cultured human cells. Exp Cell Res 289: 133–142. requires mitofusin 1 to promote mitochondrial fusion. Proc Natl Martinez LA, Naguibneva I, Lehrmann H, Vervisch A, Tchenio T, Acad Sci USA 101: 15927–15932. Lozano G et al. (2002). Synthetic small inhibiting RNAs: efficient Conklin KA. (2005). Coenzyme q10 for prevention of anthracycline- tools to inactivate oncogenic mutations and restore p53 pathways. induced cardiotoxicity. Integr Cancer Ther 4: 110–130. Proc Natl Acad Sci USA 99: 14849–14854. Cullinane C, Cutts SM, Panousis C, Phillips DR. (2000). Interstrand McKenzie PP, McPake CR, Ashford AA, Vanin EF, Harris LC. cross-linking by adriamycin in nuclear and mitochondrial DNA of (2002). MDM2 does not influence p53-mediated sensitivity to DNA- MCF-7 cells. Nucleic Acids Res 28: 1019–1025. damaging drugs. Mol Cancer Ther 1: 1097–1104. Ellis CN, Ellis MB, Blakemore WS. (1987). Effect of adriamycin on Nithipongvanitch R, Ittarat W, Velez JM, Zhao R, St Clair DK, heart mitochondrial DNA. Biochem J 245: 309–312. Oberley TD. (2007). Evidence for p53 as guardian of the Emmerson CF, Brown GK, Poulton J. (2001). Synthesis of mitochon- cardiomyocyte mitochondrial genome following acute adriamycin drial DNA in permeabilised human cultured cells. Nucleic Acids Res treatment. J Histochem Cytochem 55: 629–639. 29: E1. Olichon A, Baricault L, Gas N, Guillou E, Valette A, Belenguer P Felix CA, Kappel CC, Mitsudomi T, Nau MM, Tsokos M, Crouch et al. (2003). Loss of OPA1 perturbates the mitochondrial inner GD et al. (1992). Frequency and diversity of p53 mutations in membrane structure and integrity, leading to cytochrome c release childhood rhabdomyosarcoma. Cancer Res 52: 2243–2247. and apoptosis. J Biol Chem 278: 7743–7746. Garrido N, Griparic L, Jokitalo E, Wartiovaara J, van der Bliek AM, Poulton J, Morten K, Freeman-Emmerson C, Potter C, Sewry C, Spelbrink JN. (2003). Composition and dynamics of human Dubowitz V et al. (1994). Deficiency of the human mitochondrial mitochondrial nucleoids. Mol Biol Cell 14: 1583–1596. transcription factor h-mtTFA in infantile mitochondrial Griparic L, Kanazawa T, van der Bliek AM. (2007). Regulation of the myopathy is associated with mtDNA depletion. Hum Mol Genet mitochondrial dynamin-like protein Opa1 by proteolytic cleavage. 3: 1763–1769. J Cell Biol 178: 757–764. Singal PK, Iliskovic N. (1998). Doxorubicin-induced cardiomyopathy. Hande KR. (1998). Clinical applications of anticancer drugs targeted N Engl J Med 339: 900–905. to topoisomerase II. Biochim Biophys Acta 1400: 173–184. Steinherz L, Steinherz P. (1991). Delayed cardiac toxicity from He J, Mao CC, Reyes A, Sembongi H, Di Re M, Granycome C et al anthracycline therapy. Pediatrician 18: 49–52. (2007). The AAA+ protein ATAD3 has displacement loop binding Suliman HB, Carraway MS, Ali AS, Reynolds CM, Welty-Wolf KE, properties and is involved in mitochondrial nucleoid organization. Piantadosi CA. (2007). The CO/HO system reverses inhibition of J Cell Biol 176: 141–146. mitochondrial biogenesis and prevents murine doxorubicin cardio- Hixon SC, Ellis CN, Daugherty JP. (1981). Heart mitochondrial DNA myopathy. J Clin Invest 117: 3730–3741. synthesis: preferential inhibition by adriamycin. J Mol Cell Cardiol Tokarska-Schlattner M, Wallimann T, Schlattner U. (2006). Altera- 13: 855–860. tions in myocardial energy metabolism induced by the anti-cancer Iborra FJ, Kimura H, Cook PR. (2004). The functional organization drug doxorubicin. C R Biol 329: 657–668. of mitochondrial genomes in human cells. BMC Biol 2:9. Tondera D, Grandemange S, Jourdain A, Karbowski M, Mattenber- Jung K, Reszka R. (2001). Mitochondria as subcellular targets for ger Y, Herzig S et al. (2009). SLP-2 is required for stress-induced clinically useful anthracyclines. Adv Drug Deliv Rev 49: 87–105. mitochondrial hyperfusion. EMBO J 28: 1589–1600. Kucej M, Kucejova B, Subramanian R, Chen XJ, Butow RA. (2008). Trifunovic A, Wredenberg A, Falkenberg M, Spelbrink JN, Mitochondrial nucleoids undergo remodeling in response to Rovio AT, Bruder CE et al. (2004). Premature ageing in mice metabolic cues. J Cell Sci 121: 1861–1868. expressing defective mitochondrial DNA polymerase. Nature 429: Kurz EU, Douglas P, Lees-Miller SP. (2004). Doxorubicin activates 417–423. ATM-dependent phosphorylation of multiple downstream targets von Wurmb-Schwark N, Cavelier L, Cortopassi GA. (2006). in part through the generation of reactive oxygen species. J Biol A low dose of ethidium bromide leads to an increase of total Chem 279: 53272–53281. mitochondrial DNA while higher concentrations induce the Lebrecht D, Kokkori A, Ketelsen UP, Setzer B, Walker UA. (2005). mtDNA 4997 deletion in a human neuronal cell line. Mutat Res Tissue-specific mtDNA lesions and radical-associated mitochon- 596: 57–63. drial dysfunction in human hearts exposed to doxorubicin. J Pathol Wallace KB. (2003). Doxorubicin-induced cardiac mitochondriono- 207: 436–444. pathy. Pharmacol Toxicol 93: 105–115. Lebrecht D, Setzer B, Rohrbach R, Walker UA. (2004). Mitochondrial Zhou S, Palmeira CM, Wallace KB. (2001). Doxorubicin-induced DNA and its respiratory chain products are defective in doxorubicin persistent oxidative stress to cardiac myocytes. Toxicol Lett 121: nephrosis. Nephrol Dial Transplant 19: 329–336. 151–157.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene