The nonsteroidal antiestrogen, tamoxifen, is used in women with breast cancer or at risk of developing it (Fisher et al., 2005). Tamoxifen has a 7-day half-life in humans (Lien et al., 1995). Due to faster metabolism, mice have to be treated with 200 mg/kg (0.54 mmol/kg) p.o. daily to reproduce the plasma levels of patients receiving 30 mg of tamoxifen twice daily (Robinson et al., 1991). Plasma tamoxifen levels progressively increase during the first days of tamoxifen administration (0.54 mmol/kg p.o. daily) to reach a plateau after 7 days in mice (Robinson et al., 1991). Tamoxifen is concentrated in tissues, particularly in liver (Robinson et al., 1991). At 7 days in mice, the hepatic tamoxifen content (215 μmol/kg) is 160 times the plasma concentration (1.34 μM) and 8-fold the skeletal muscle content (27 μmol/kg) (Robinson et al., 1991). In treated women, plasma tamoxifen concentrations range from nondetectable to 1.68 μM (Peyrade et al., 1996), and the hepatic concentration was 64 times the plasma concentration in one studied woman (Lien et al., 1991).

Tamoxifen is well tolerated but causes steatosis in 43% of recipients (Nishino et al., 2003). Steatohepatitis can develop, particularly in overweight women (Bruno et al., 2005), and can lead to cirrhosis (Oien et al., 1999). However, the mechanism for tamoxifen-induced steatosis is essentially unknown. Tamoxifen administration decreases fatty acid synthase (FAS) expression in rat liver (Lelliott et al., 2005), and tamoxifen both uncouples and inhibits mitochondrial respiration (Cardoso et al., 2001). However, effects on mitochondrial β-oxidation and mitochondrial DNA (mtDNA) have not been studied.

Tamoxifen is a cationic drug, and several such drugs are electrophoretically taken up by mitochondria to achieve high concentrations that inhibit both β-oxidation and respiration (Fromenty et al., 1990a,b; Berson et al., 1998). Tamoxifen is also a DNA-intercalating drug (Snyder and Brown, 2002), and such drugs can inhibit topoisomerases (Lin and Castora, 1991). Topoisomerase I monomers and topoisomerase II dimers transiently cut one and two strands of DNA, respectively, to allow the passage of another DNA strand or double helix, respectively, and then promptly reseal the DNA backbone (Wang, 2002). Both intercalation and the inhibition of topoisomerases by intercalating drugs can impair DNA replication and transcription (Li and Liu, 2001; Wang, 2002). Topoisomerases I and II are present in the mitochondria (Lin and Castora, 1991; Zhang et al., 2001), and the intercalating drug, tacrine, inhibits topoisomerases and depletes mtDNA in mice (Mansouri et al., 2003).

In the present study, we show that tamoxifen is electrophoretically concentrated within hepatic mitochondria, where it acutely impairs both β-oxidation and respiration. Tamoxifen also inhibits topoisomerases, decreases mtDNA replication, and depletes mtDNA in mice. Finally, tamoxifen inhibits hepatic triglycerides (TG) secretion. These combined effects could decrease fat removal from the liver and cause steatosis, despite a secondary down-regulation of hepatic FAS expression.

Materials and Methods

Materials and Chemicals. Tamoxifen citrate and [³H]tamoxifen (85 Ci/mmol) were purchased from Sigma Chemical Co. (St. Louis, MO) and GE Healthcare (Chalfont St. Giles, Buckinghamshire, UK), respectively. Topoisomerase I was from GE Healthcare, and topoisomerase II was from USB (Cleveland, OH). The pBR322 plasmid was purchased from Roche Diagnostics (Meylan, France). Antibodies against mouse cytochrome c oxidase subunit 1 (COX 1) and subunit 4 (COX 4) were from Molecular Probes (Eugene, OR). The rabbit polyclonal antibody against mouse FAS was from Santa Cruz Biotechnologies (Santa Cruz, CA). The antibody against medium-chain acyl-CoA dehydrogenase (MCAD) was donated by J. R. Wanders (Laboratory of Genetic Metabolic Diseases, Amsterdam, The Netherlands), and the anti-carnitine palmitoyl-transferase-I (CPT-I) antibody was donated by C. Prip-Buus (Institut National de la Santéet de la Recherche Médicale U567, Paris, France).

Animals and Treatments. Male Crl:CD-1(ICR)BR Swiss mice weighing 28 to 30 g were purchased from Charles River (Saint-Aubin-Lés-Elbeufs, France). Mice were fed ad libitum a standard diet (A04–10 biscuits; UAR, Villemoisson-sur-Orge, France) or were deprived of food for the last 24 h as indicated. Animals received humane care, and all experiments were performed according to national guidelines for the use of animals in biomedical research.

Mice received tamoxifen citrate (0.5 mmol/kg daily) and/or 0.9% NaCl by gastric intubation and were killed after 5, 12, or 28 days. Tamoxifen-treated mice consumed 21% less food than control mice, and their body weight only increased by 2% after 28 days, compared with 19% in control mice.

Polarographic Measurement of Oxygen Consumption and Mitochondrial Uptake of Tamoxifen. Liver mitochondria were prepared, and respiration was measured as described previously (Fromenty et al., 1990a). We measured the respiration uncoupled by 2,4-dinitrophenol (160 μM), the state 3 respiration stimulated by ADP (0.2 mM), and the basal (state 4) respiration after ADP exhaustion.

To assess uptake, we energized mitochondria (2 mg of protein/ml) with succinate (10 mM) and incubated them with [³H]tamoxifen (25 μM; 8.5 μCi/ml) at 30°C for 1 min. An aliquot was taken to measure proteins, and mitochondria were centrifuged for 2 min at 13,000g. Aliquots from the mitochondrial pellet and supernatant were extracted with ethanol and counted for ³H radioactivity. Mitochondrial drug concentrations were estimated assuming a mitochondrial volume of 1 μl per 1 mg of protein (McCormack et al., 1989; Fromenty et al., 1990a).

In Vitro Mitochondrial β-Oxidation, CPT-I Activity, and MCAD Activity. The β-oxidation of [U-¹⁴C]palmitic acid by mouse liver mitochondria was assessed as described previously (Fromenty et al., 1990b). The CPT-I-dependent release of CoA from palmitoyl-CoA (0.15 mM) in the presence of carnitine (2.5 mM) and KCN (1 mM) was assayed with the thiol-reactive, fluorescence-generating probe, 7-fluorobenz-2-oxa-1,3-diazole-4-sulfonamide, as described previously (Maisonneuve et al., 2004). MCAD activity was measured with octanoyl-CoA (50 μM) and ferricenium hexafluorophosphate (200 μM) as described previously (Berson et al., 1998).

Topoisomerase-Mediated DNA Relaxation. We used the pBR322 plasmid to assess topoisomerase-mediated DNA relaxation (Mansouri et al., 2003). The supercoiled plasmid (300 ng) was incubated for 30 min at 30°C with tamoxifen (0–100 μM) and 2.5 units of topoisomerase I or II under conditions specified by the manufacturers. DNA isoforms were resolved in 0.7% Seakem GTG agarose gels (FMC Bioproducts, Rockland, ME) without ethidium bromide. After staining with ethidium bromide (0.5 μg/ml), DNA bands were visualized by UV light.

[³H]Thymidine Incorporation into DNA. Twenty-four hours after the last dose of tamoxifen, mice received [³H]thymidine i.p. (0.03 μmol/kg, 0.75 mCi/kg). Two hours later, the liver was homogenized and centrifuged at 600g for 15 min. We used the nuclear pellet to isolate nuclear DNA (nDNA) and the supernatant to isolate mitochondria at 7700g (Mansouri et al., 2003).

nDNA and mtDNA were recovered with the phenol-chloroform method and counted for [³H]thymidine (Mansouri et al., 2003). nDNA and mtDNA Levels. Total hepatic DNA was isolated with QIAGEN Genomic-tip 100/G columns (QIAGEN, Valencia, CA) (Mansouri et al., 1999), and 300 ng was blotted onto Hybond-N⁺ nylon membranes (GE Healthcare) and hybridized with a 10.9-kb mtDNA probe (nucleotides 4964–15,896) generated by long polymerase chain reaction and labeled by random priming (Multiprime DNA labeling system; GE Healthcare) (Mansouri et al., 1999). Membranes were stripped and hybridized with a mouse C_ot-1 nDNA probe (Invitrogen, Cergy Pontoise, France). mtDNA and nDNA were assessed by densitometry analysis of autoradiographs (Mansouri et al., 1999).

Mitochondrial Complexes. Hepatic mitochondria were isolated, and submitochondrial particles were prepared by two cycles of freezing/thawing. Complex I was assayed by following the oxidation of NADH in the presence of decylubiquinone and antimycin A (Kwong and Sohal, 2000). Complex II was measured as the phenazine methosulfate-mediated reduction of 2,6-dichlorophenolindophenol in the presence of succinate, rotenone, and antimycin A. Complex IV was assessed by following the reduction of cytochrome c (Kwong and Sohal, 2000).

Immunoblot Analysis of Selected Respiratory Chain Polypeptides and Lipid Metabolism Enzymes. Total liver proteins were separated by electrophoresis, blotted, and exposed first to antibodies against COX 1, COX 4, FAS, CPT-I, or MCAD, and then a secondary antibody conjugated with horseradish peroxidase (Bio-Rad, Hercules, CA). Blots were revealed with enhanced chemiluminescence reagents (GE Healthcare), stripped, and exposed to an anti-β-actin antibody.

In Vivo Formation of [¹⁴C]CO₂ from [U-¹⁴C]Palmitic Acid. Mice fasted for 24 h were given a tracer dose of [U-¹⁴C]palmitic acid (3.7 μCi/kg; 4 nmol/kg) by gastric intubation in 0.2 ml of corn oil and placed in a small plastic cage swept by an air flow of 50 ml/min (Fromenty et al., 1990b). The outflow was bubbled into a 60-ml mixture of ethanolamine and 2-methoxyethanol (30%/70%, v/v) to trap the exhaled [¹⁴C]CO₂ (Fromenty et al., 1990b).

Hepatic Secretion of TG and Apo B. Plasma TG and Apo B were determined with commercial kits (Biotrol A 01549, Chennevières-les-Louvres, France and Sigma kit no. 357, respectively) 4 h after the administration of Triton WR-1339, which inhibits lipoprotein lipase activity and blocks the peripheral removal of TG (Lettéron et al., 2003). In 24-h-fasted animals (with no intestinal fat absorption), the increase in TG mostly reflects hepatic TG secretion (Lettéron et al., 2003).

Plasma Chemistry, Cytokines, Alanine Aminotransferase Activity, Caspase Activities, and Assessment of nDNA Fragmentation. Plasma β-hydroxybutyrate, glucose, TG, cholesterol, tumor necrosis factor (TNF)-α, transforming growth factor (TGF)-β, and alanine aminotransferase (ALT) activity were measured with commercial kits.

Livers were minced and homogenized in 1 mM EDTA, 50 mM HEPES, 0.1% 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate, 5 mM dithiothreitol, 4 mg/ml leupeptin, and 4 mg/ml pepstatin, pH 7.4. After centrifugation at 14,000g, caspase-3 activity was measured in the supernatant with a fluorescent assay kit (Biomol, Plymouth Meeting, PA). Caspase-9 activity was measured with 25 μM Ac-LEHD-AFC (Calbiochem, Darmstadt, Germany).

Hepatic DNA (5 μg) was loaded on 2% agarose gels (Mansouri et al., 1999). After electrophoresis for 4 h at 50 V, DNA was transferred to a nylon membrane (Hybond-N⁺) and hybridized with a mouse Cot-1 nDNA probe (Mansouri et al., 1999).

Hepatic TG and Cholesterol, Liver Morphology, Terminal Deoxynucleotidyl Transferase dUTP Nick-End Labeling Assay, and Liver Ultrastructure. Liver TG levels were determined as described previously (Fromenty et al., 1990b). Liver cholesterol was measured using the Cholesterol/Cholesteryl Ester Quantitation kit (BioVision, Mountain View, CA) following the manufacturer's recommendations.

A liver fragment for light microscopy was fixed with 10% formalin in phosphate-buffered saline, dehydrated in an alcoholic series, placed in toluene baths, and embedded in paraffin. Three-micrometer-thick sections were prepared for hematoxylin-eosin staining. In situ detection of apoptosis was performed with the terminal deoxynucleotidyl transferase dUTP nick-end labeling assay using the Apo-Tag peroxidase in situ apoptosis detection kit (Chemicon International, Ltd., Hampshire, UK).

Another liver fragment was cut into 1-mm³ blocks and fixed in phosphate-buffered saline with 2.5% glutaraldehyde. After dehydration in an alcoholic series, blocks were placed in propylene oxide solutions and embedded in epoxy resin. Semithin sections were stained with toluidine blue and examined on light microscopy for orientation. Ultrathin sections were stained with uranyl acetate and lead citrate and examined with a JEOL 1010 electron microscope (Tokyo, Japan).

Statistical Analysis. Data were analyzed with one-way analysis of variance and Fischer's tests. The significance level was set at P < 0.05. Results in mice treated with saline for 5, 12, or 28 days did not differ from each other. For clarity, they were pooled into a single control group in the table and figures showing results at different treatment times.

Results

Tamoxifen Is Electrophoretically Taken up by Mitochondria. We isolated mitochondria, energized them with succinate, and added [³H]tamoxifen (25 μM) (Table 1). Tamoxifen was concentrated 40-fold in mitochondria, reaching a concentration of 640 μM compared with a residual concentration of 16 μM in the medium. The electrophoretic uptake of tamoxifen was decreased by valinomycin (70 nM), a potassium ionophore, which lowers the mitochondrial membrane potential (Teplova et al., 2006).

Tamoxifen Impairs Mitochondrial Respiration and Fatty Acid β-Oxidation. At 50 μM, tamoxifen increased basal (state 4) respiration (Supplemental Table S1), indicating an uncoupling effect. At 150 μM, tamoxifen no longer stimulated basal respiration but inhibited the high respiratory rate triggered by the uncoupler, 2,4-dinitrophenol (Supplemental Table S1), suggesting that tamoxifen impairs electron flow in the respiratory chain. At an intermediate concentration (100 μM), tamoxifen inhibited ADP-stimulated respiration but did not significantly inhibit 2,4-dinitrophenol-stimulated respiration. The more severe impairment of ADP-stimulated respiration may suggest additional inhibitory effect(s) on the adenine nucleotide translocator and/or ATP synthase. Both are involved in ADP-stimulated respiration but not in 2,4-dinitrophenol-uncoupled respiration.

We next assessed the β-oxidation of [U-¹⁴C]palmitic acid by isolated mitochondria (Fig. 1A). β-Oxidation was decreased by 44, 70, 92, and 97% by 100, 150, 200, and 250 μM tamoxifen, respectively.

A severe impairment of mitochondrial respiration can decrease the reoxidation of FADH₂ and NADH into FAD and NAD⁺, thus inhibiting fatty acid β-oxidation, which requires these oxidized cofactors (Fromenty and Pessayre, 1995). To assess whether tamoxifen inhibits β-oxidation directly or only as a consequence of the inhibition of respiration, we added ferricenium hexafluorophosphate and acetoacetate to the β-oxidation medium (Fig. 1B). Ferricenium accepts the electrons coming from the enzyme-bound FADH₂ in acyl-CoA dehydrogenases (Lettéron et al., 1997), whereas acetoacetate reoxidizes NADH into NAD⁺, which is required by 3-hydroxyl-acyl-CoA dehydrogenases (Fromenty and Pessayre, 1995). Although acetoacetate and ferricenium in combination partly relieved the inhibition of mitochondrial β-oxidation that was caused by rotenone or antimycin A, they did not relieve the inhibition caused by tamoxifen (250 μM) (Fig. 1B), suggesting that tamoxifen directly inhibits the β-oxidation process.

CPT-I partly controls the mitochondrial β-oxidation rate of long-chain fatty acids by modulating their entry into mitochondria (Fromenty and Pessayre, 1995). Tamoxifen (50, 100, 200, or 250 μM) inhibited CPT-I activity by 38, 43, 58, and 64%, respectively (Fig. 1C).

To determine whether tamoxifen also impairs in vivo fatty acid oxidation, we measured [¹⁴C]CO₂ exhalation after a tracer dose of [U-¹⁴C]palmitic acid (Fig. 1D). [¹⁴C]CO₂ exhalation was unchanged in mice treated with a single dose of 0.1 mmol/kg tamoxifen but was decreased by 49 and 46% after 0.5 and 1 mmol/kg, respectively.

Tamoxifen Also Inhibits Hepatic TG Secretion in Vivo. To assess hepatic TG secretion, we measured the increase in plasma TG 4 h after administration of Triton WR-1339, which blocks peripheral TG utilization (Lettéron et al., 2003). Tamoxifen (1 mmol/kg) inhibited the increase in TG by 68% but did not significantly blunt the increase in Apo B (Fig. 2), suggesting that tamoxifen impairs Apo B lipidation.

Tamoxifen Inhibits Topoisomerases, Impairs mtDNASynthesis, and Depletes mtDNA. We used the supercoiled pBR322 plasmid to assess the effects of tamoxifen on topoisomerase activities (Fig. 3). Tamoxifen (20–100 μM) inhibited DNA relaxation mediated by topoisomerases I and II.

Intercalating topoisomerase inhibitors can inhibit DNA replication (Li and Liu, 2001; Wang, 2002; Mansouri et al., 2003). To assess DNA synthesis, we treated mice with tamoxifen for 12 or 28 days, injected [³H]thymidine 24 h after the last dose, and killed the animals 2 h later (Fig. 4A). [³H]Thymidine incorporation into mtDNA was decreased by 32% at 12 days and by 62% at 28 days. In contrast, [³H]thymidine incorporation into nDNA was unchanged at 12 days and was decreased by only 25% at 28 days.

Tamoxifen administration progressively decreased hepatic mtDNA level without affecting nDNA content (Fig. 4B). We used the mtDNA/nDNA hybridization ratio to assess mtDNA changes (Fig. 4C). This ratio was unchanged after 5 days of treatment but was decreased by 31% at 12 days and by 60% at 28 days (Fig. 4C).

Tamoxifen Administration Decreases mtDNA-Encoded Respiratory Complexes and Sensitizes Mitochondria to the Inhibitory Effects of Tamoxifen. Although the mtDNA-encoded COX 1 polypeptide was decreased by 66% in mice treated with tamoxifen for 28 days, the nDNA-encoded COX 4 was unchanged (Fig. 5). To assess the activities of mitochondrial complexes, we isolated mouse liver mitochondria and prepared submitochondrial particles (Table 2). The washings and dilutions involved in the isolation of mitochondria can remove drugs present in the liver of treated animals at the time of sacrifice, thus suppressing their in vivo inhibitory effects on mitochondrial respiration and/or β-oxidation (Fromenty and Pessayre, 1995). Indeed, the activities of mitochondrial complexes were unchanged in the submitochondrial particles isolated from mice killed 24 h after the last dose of tamoxifen, 0.5 mmol/kg daily for 7 days (Table 2). In contrast, the activities of complexes I and IV were decreased by 32 and 20%, respectively, in mice treated for 28 days (Table 2). The deterioration occurring between days 7 and 28 is unlikely to be due to tamoxifen accumulation because plasma concentrations already reach a plateau at day 7 (Robinson et al., 1991). Instead, the impairment of complexes I and IV after 28 days, without significant impairment of complex I, may be due to progressive mtDNA depletion. Although complex II is encoded by nDNA, some of the polypeptides of complexes I and IV are mtDNA-encoded (Fromenty and Pessayre, 1995).

The activity of mitochondrial complexes has to be severely inhibited (beyond a 70–80% threshold) for mitochondrial respiration to decrease (Rossignol et al., 1999). In keeping with the moderately impaired activities of complexes I and IV, respiration was unchanged in the mitochondria isolated from mice treated with tamoxifen for 28 days (Fig. 6). As already mentioned, the isolation of mitochondria can remove drugs present in the liver at the time of sacrifice so that their in vivo inhibitory effects on mitochondrial function are no longer observed ex vivo (Fromenty and Pessayre, 1995). To reproduce the in vivo situation, we added tamoxifen to some incubation flasks (Fig. 6). Combined with the depletion of mtDNA-encoded respiratory polypeptides, the acute inhibitory effects of tamoxifen may then hamper enough the flow of electrons within the respiratory chain to significantly decrease mitochondrial respiration. Indeed, adding a small concentration of tamoxifen (15 μM) to the mitochondria of mice treated with tamoxifen for 28 days markedly inhibited the respiration that was supported by malate and glutamate, which donate electrons to the partially mtDNA-encoded complex I (Fig. 6). In contrast, in mice treated for only 7 days, tamoxifen (15 μM) failed to inhibit respiration (Fig. 6), probably due to the lack of significant mtDNA depletion at that time. Collectively, these data indicate that the depletion of mtDNA-encoded polypeptides after prolonged tamoxifen administration sensitizes mitochondria to the acute inhibitory effects of a small concentration of tamoxifen on mitochondrial respiration.

Western Blot Analysis of Selected Proteins Involved in Fatty Acid Metabolism and Effects on ex Vivo MCAD Activity. The expression of FAS and CPT-I were similar in control mice and mice treated with tamoxifen for 5 or 12 days (Fig. 7). However, at 28 days, FAS protein expression decreased by 52% in treated mice, and CPT-I expression increased by 22%. MCAD protein expression did not change, whatever the duration of treatment (Fig. 7). We also determined the effect of tamoxifen administration on ex vivo MCAD activity. MCAD activity (mean ± S.E.M. for five mice) was not significantly decreased in mice treated with tamoxifen (0.5 mmol/kg daily) for 28 days (18 ± 4 nmol/min/mg protein in tamoxifen-treated mice versus 26 ± 2 in control mice).

Effects of Tamoxifen on Plasma β-Hydroxybutyrate, Glucose, Cholesterol, TG, TNF-α, TGF-β, and ALT Activity. The β-oxidation of fatty acids generates ketone bodies, the most abundant of which is β-hydroxybutyrate (Fromenty and Pessayre, 1995). In fed mice treated with tamoxifen for 28 days, plasma β-hydroxybutyrate was decreased by 64% (Table 3), thus confirming the decreased β-oxidation. Plasma glucose was unchanged, whereas plasma cholesterol was decreased by 27% (Table 3).

We also measured plasma β-hydroxybutyrate, TG, TNF-α, and TGF-β levels and ALT activity in mice treated for 28 days and fasted for the last 24 h (Table 3). Plasma β-hydroxybutyrate was decreased by 25% and TG by 55% in tamoxifen-treated mice (Table 3). Plasma TNF-α was decreased by 43%, whereas TGF-β was unchanged. Plasma ALT activity was increased by 87% in treated mice (Table 3).

Tamoxifen Triggers Hepatic Steatosis and Ultrastructural Mitochondrial Abnormalities but No Apoptosis. Liver histology was evaluated in control mice (n = 11) and in mice treated with tamoxifen for 5 (n = 4), 12 (n = 4), or 28 (n = 9) days. No steatosis, or minimal steatosis, was observed in control mice (Fig. 8A) or mice treated for 5 days (Fig. 8B). Three of the four mice treated for 12 days had mild periportal steatosis (Fig. 8C). Eight of the nine mice treated for 28 days exhibited moderate, periportal, macrovacuolar, and microvesicular steatosis, as well as ballooning of periportal hepatocytes, without apoptosis. One of these animals also exhibited a few necrotic hepatocytes, without inflammation and without apoptosis (Fig. 8D). Although moderate steatosis was observed in most mice treated with tamoxifen for 28 days (Fig. 8D), TG and cholesterol contents measured in total liver extracts were not different in six control mice and five mice treated with tamoxifen for 28 days (12 ± 1mg of TG/g liver in control mice versus 13 ± 3 in tamoxifen-treated mice; 2.7 ± 0.7 μmol of cholesterol/g liver in control mice versus 2.2 ± 0.2 μmol in tamoxifen-treated mice). This discrepancy is thought to be due to the fact that treated mice only exhibited moderate steatosis, which was selectively localized to the periportal region of the liver (Fig. 8D), whereas TG and cholesterol were determined in whole-liver extracts. Electron microscopy at 28 days showed small lipid droplets (Fig. 9, B and C) and elongated, hypertrophied, or swollen mitochondria with dilatation of cristae (Fig. 9, E and F).

Hepatic caspase-9 activity was unchanged in mice treated for 28 days (Table 4). Although caspase-3 was increased, this increase was mild compared with diverse apoptosis models. Even with the sensitive Cot-1 hybridization technique, we could not detect nDNA fragmentation in any of 15 mice treated with tamoxifen (0.5 mmol/kg) for 28 days. The terminal deoxynucleotidyl transferase dUTP nick-end labeling assay was negative in four control mice and five mice treated with tamoxifen for 28 days (data not shown).

Discussion

Although tamoxifen administration causes hepatic steatosis in 43% of recipients (Nishino et al., 2003), the mechanisms for fat accumulation have remained elusive. This work demonstrates that tamoxifen is electrophoretically taken up by mitochondria, where it targets not only respiration, as already reported (Cardoso et al., 2001), but also β-oxidation and mtDNA.

In the acidic intermembrane space of mitochondria, the weak base, tamoxifen (pK_a, 8.5), is protonated. Like other cationic amphiphilic drugs (Fromenty et al., 1990a,b; Berson et al., 1998), tamoxifen is electrophoretically transported into the mitochondrial matrix, thus achieving high intramitochondrial concentrations (Table 1). In the alkaline matrix, protonated tamoxifen may then partly dissociate into a proton and uncharged tamoxifen. This protonophoric effect may decrease the mitochondrial membrane potential (Cardoso et al., 2001), thus unleashing the flow of electrons in the respiratory chain and increasing basal mitochondrial respiration (Supplemental Table S1). Although tamoxifen increases basal respiration, it decreases the high respiratory rates triggered by ADP or 2,4-dinitrophenol (Supplemental Table S1). Tamoxifen partitions into biomembranes and decreases membrane fluidity (Custodio et al., 1993). It could thus impair the interactions between respiratory chain components and the diffusional mobility of membrane proteins (Custodio et al., 1993), thus forbidding high rates of electron flow and oxidative phosphorylation.

The present study shows for the first time that tamoxifen directly inhibits the mitochondrial β-oxidation of fatty acids, in part by inhibiting CPT-I (Fig. 1). CPT-I allows the entry of long-chain fatty acids into mitochondria and thus partly controls the mitochondrial β-oxidation rate (Fromenty and Pessayre, 1995). CPT-1 activity was significantly inhibited by tamoxifen in a concentration-dependent manner (Fig. 1C). Yet, the level of CPT-1 protein was slightly but significantly increased at 28 days, although not at 5 or 12 days (Fig. 7). The secondary up-regulation of CPT-1 may mitigate the inhibitory effect of tamoxifen on CPT-1. Like the secondary down-regulation of hepatic FAS expression (discussed further on), the secondary up-regulation of CPT-I may be an adaptive response tending to prevent further fat accretion.

The present work also identifies for the first time mtDNA as an important target in tamoxifen-induced mitochondrial toxicity (Fig. 4). Tamoxifen intercalates between DNA bases (Snyder and Brown, 2002), and several DNA-intercalating drugs inhibit topoisomerases (Lin and Castora, 1991; Mansouri et al., 2003). At concentrations (20–100 μM) in the range of those present in the liver of treated mice and some treated women (Lien et al., 1991; Robinson et al., 1991; Peyrade et al., 1996), tamoxifen inhibited the relaxation of a supercoiled plasmid by topoisomerases I or II (Fig. 3). Like this plasmid, mtDNA is a supercoiled molecule (Mansouri et al., 1999), and mitochondria contain both topoisomerase I and a bacterial-like topoisomerase II activity (Lin and Castora, 1991; Zhang et al., 2001).

Topoisomerase inhibition by tamoxifen should keep mtDNA in its supercoiled form, which, together with intercalation, may hamper the progression of DNA polymerase γ during mtDNA replication (Nelson et al., 1984; Mansouri et al., 2003). Indeed, tamoxifen decreased [³H]thymidine incorporation into mtDNA, with limited effects on nDNA (Fig. 4A), probably because mtDNA is exposed to higher tamoxifen concentrations than nDNA. Liver mitochondria may be particularly targeted because of the higher tamoxifen concentrations in the liver than other organs (Robinson et al., 1991).

Even in tissues whose cells do not divide or rarely divide, such as the liver, there is a basal turnover of mitochondria. Therefore, drugs that decrease mtDNA synthesis can progressively deplete mtDNA (Rowe et al., 2001; Mansouri et al., 2003). Indeed, tamoxifen administration progressively decreased mtDNA levels, down to 40% of control levels after 28 days of treatment (Fig. 4C).

mtDNA depletion may decrease the synthesis of mtDNA-encoded respiratory chain polypeptides (Fromenty and Pessayre, 1995). After 28 days of tamoxifen administration, the mtDNA-encoded COX 1 polypeptide was decreased by 66% (Fig. 5), and the activities of complexes I and IV were decreased by 32 and 20%, respectively (Table 2). However, the activity of a respiratory complex has to be severely inhibited (in excess of a 70–80% inhibition threshold) for mitochondrial respiration to be impaired (Rossignol et al., 1999). Once this threshold is crossed, any further inhibition of the complex then causes a precipitous decline in respiration (Rossignol et al., 1999). In mice treated with tamoxifen for 28 days, the depletion of mtDNA-encoded polypeptides alone did not impair respiration (Fig. 6). However, combined with the acute inhibitory effects of a small concentration of tamoxifen (15 μM), it impaired the flow of electrons sufficiently to decrease respiration (Fig. 6). Severe impairment of mitochondrial respiration can decrease the reoxidation of NADH into the NAD⁺, which is required to sustain β-oxidation (Fromenty and Pessayre, 1995). This added effect could further decrease a β-oxidation rate already hampered by the acute inhibitory effects of tamoxifen on β-oxidation. This could explain why steatosis was absent at 5 days, mild at 12 days, and more prominent at 28 days (Fig. 8).

Tamoxifen also decreased hepatic TG secretion (Fig. 2). Although the mechanism of this effect remains to be ascertained, we previously reported that several steatogenic drugs inhibited not only fatty acid oxidation but also microsomal triglyceride transfer protein activity and very low-density lipoprotein secretion, thus concomitantly decreasing the two pathways removing fat from the liver (Lettéron et al., 2003).

Finally, as previously reported (Lelliott et al., 2005), tamoxifen administration decreased hepatic FAS expression (Fig. 7). This effect is not due to the antiestrogenic effect of tamoxifen, which should have the opposite effect (D'Eon et al., 2005). Although an agonistic effect is possible, the late down-regulation of FAS (at 28 but not at 12 days) rather suggests an adaptive response to tamoxifen-induced steatosis. Indeed, in a previous study, a high-fat diet increased hepatic TG 2.5-fold and decreased hepatic FAS mRNA 2.5-fold in mice (Kim et al., 2004). In another study, an Apo B antisense oligonucleotide administered twice weekly for 6 weeks doubled the mean hepatic TG content (albeit not significantly) and decreased hepatic FAS protein expression (Crooke et al., 2005). Finally, a diet deficient in methionine and choline triggered hepatic steatosis and severely down-regulated hepatic FAS mRNA expression (Tomita et al., 2006). It is difficult to predict what may be the consequences of this secondary down-regulation of FAS on the outcome of tamoxifen-mediated steatosis. On the one hand, the secondary down-regulation of FAS may increase hepatic malonyl-CoA and inhibit CPT-I activity (Lelliott et al., 2005), which could further decrease hepatic fat oxidation. On the other hand, by decreasing hepatic fatty acid synthesis, FAS down-regulation may prevent further fat accretion. A new steady state may thus be achieved, whereby the decreased hepatic fatty acid synthesis may now equilibrate the decreased hepatic fatty acid oxidation and decreased TG secretion, resulting in stable and moderate periportal steatosis. In a previous study, the predominance of steatosis in periportal hepatocytes after valproate administration has been ascribed to a lower rate of oleate oxidation into ketone bodies in periportal than in pericentral hepatocytes and also to a relatively more severe inhibition of ketogenesis by valproate in periportal than in central hepatocytes (Olson et al., 1986).

Although TGF-β was unchanged, plasma TNF-α was decreased by 43% in tamoxifen-treated mice (Table 3). A possible mechanism could be that tamoxifen is not a pure antiestrogenic compound but also displays some estrogenic activity. Estrogens have been shown to decrease TNF-α gene expression by blocking JNK activity and the expression of c-Jun and JunD (Srivastava et al., 1999).

Although hepatocyte ballooning was observed in tamoxifen-treated mice, inflammation, apoptosis, or necrosis were absent, with the exception of a single mouse, which had mild liver cell necrosis. Likewise, most patients treated with tamoxifen may only have steatosis (Nishino et al., 2003). Tamoxifen-associated steatohepatitis seems to occur mostly in obese persons (Bruno et al., 2005). Although tamoxifen decreases the removal of fat from the liver, obesity leads to insulin resistance in muscle, thus increasing plasma glucose and/or insulin, which induce hepatic FAS expression and fatty acid synthesis (Pessayre and Fromenty, 2005). Due to different targets, tamoxifen and obesity in combination could cause more steatosis than tamoxifen alone or obesity alone. Furthermore, the hepatic steatosis associated with obesity can lead to mitochondrial lesions, mtDNA depletion, and decreased activities of respiratory chain complexes (Pessayre and Fromenty, 2005). Combined with the effects of tamoxifen, the mitochondrial effects of obesity could severely hamper mitochondrial respiration, possibly leading to reactive oxygen species formation, cytokine induction, and steatohepatitis lesions (Pessayre and Fromenty, 2005).

The present study and a few others have identified mtDNA as a new target for drug-induced hepatotoxicity (Rowe et al., 2001). Rapid mtDNA depletion occurs after high doses of alcohol or paracetamol, which trigger mtDNA degradation (Mansouri et al., 1999; Cover et al., 2005), whereas progressive mtDNA depletion occurs during prolonged treatments with tacrine or nucleoside reverse transcriptase inhibitors, which impair mtDNA replication (Gaou et al., 2001; Mansouri et al., 2003).

In conclusion, tamoxifen inhibits hepatic TG secretion and concentrates within mitochondria, where it inhibits β-oxidation and respiration. Tamoxifen also inhibits topoisomerases and impairs mtDNA synthesis. After 4 weeks of treatment with a dose reproducing human plasma concentrations, tamoxifen depletes mtDNA in mice. These combined effects may decrease fat removal from the liver and cause steatosis, despite a secondary down-regulation of hepatic FAS expression.