Lipidomic profiling reveals protective function of fatty acid oxidation in cocaine-induced hepatotoxicity.

During cocaine-induced hepatotoxicity, lipid accumulation occurs prior to necrotic cell death in the liver. However, the exact influences of cocaine on the homeostasis of lipid metabolism remain largely unknown. In this study, the progression of subacute hepatotoxicity, including centrilobular necrosis in the liver and elevation of transaminase activity in serum, was observed in a three-day cocaine treatment, accompanying the disruption of triacylglycerol (TAG) turnover. Serum TAG level increased on day 1 of cocaine treatment but remained unchanged afterwards. In contrast, hepatic TAG level was elevated continuously during three days of cocaine treatment and was better correlated with the development of hepatotoxicity. Lipidomic analyses of serum and liver samples revealed time-dependent separation of the control and cocaine-treated mice in multivariate models, which was due to the accumulation of long-chain acylcarnitines together with the disturbances of many bioactive phospholipid species in the cocaine-treated mice. An in vitro function assay confirmed the progressive inhibition of mitochondrial fatty acid oxidation after the cocaine treatment. Cotreatment of fenofibrate significantly increased the expression of peroxisome proliferator-activated receptor α (PPARα)-targeted genes and the mitochondrial fatty acid oxidation activity in the cocaine-treated mice, resulting in the inhibition of cocaine-induced acylcarnitine accumulation and other hepatotoxic effects. Overall, the results from this lipidomics-guided study revealed that the inhibition of fatty acid oxidation plays an important role in cocaine-induced liver injury.

Cocaine, as one of the most widely abused psychological stimulants, is well known for causing addictive dependence and sudden death in its users ( 1 ). Besides its toxicity on the central nervous system (CNS) and the cardiovascular daily at a dose of 50 mg/kg by gavage. For cocaine and fenofibrate cotreatment, fenofi brate was administered 2 h earlier than cocaine on days 1, 2, and 3 of cocaine treatment. The control mice were treated with blank saline solution or CMC solution. Serum samples were collected by submandibular bleeding prior to treatment and at the end of each day of cocaine treatment, and urine samples were obtained by housing mice individually in metabolic cages for 24 h. Liver and other tissue samples were harvested after carbon dioxide euthanasia. All tissues samples were stored at Ϫ 80°C before further analysis, except for small aliquots of liver tissue for histology.

Assessment of cocaine-induced toxicity
Fresh live tissues were immediately fi xed, embedded, sectioned, and stained with hematoxylin and eosin (H&E) for general histo logy and with Oil Red O for lipid droplets. Cocaine-induced liver injury was evaluated by measuring serum alanine aminotransferase (ALT) and aspartate aminotransferase (AST) activities (Pointe Scientifi c, Canton, MI). Cocaine-elicited disruption of fatty acid metabolism was evaluated by measuring the levels of serum and hepatic triacylglycerol (TAG) using a colorimetric assay kit (Pointe Scientifi c). The status of lipid peroxidation was determined by thiobarbituric acid reactive substances (TBARS) assay using liver homogenates ( 29 ). Briefl y, 50 mg liver was homogenized in 500 µl of 10 mM phosphate buffered solution (pH = 7.4) containing 1% (v/v) Triton X-100. After mixing 100 µl of liver homogenate with 200 µl of ice-cold 10% trichloroacetic acid, 200 µl of supernatant was obtained by centrifuging at 2,200 g for 15 min at 4°C, and then incubated with an equal volume of 0.67% (w/v) thiobarbituric acid for 10 min in a boiling water bath. The concentration of malondialdehyde was determined at 532 nm using a Spectra-Max 250 spectrometer (Molecular Device, Sunnyvale, CA).

Preparation of serum and liver lipid extraction for LC-MS analysis
Deproteinization of serum was conducted by mixing one volume of serum with 19 vol of 66% aqueous acetonitrile and then centrifuging at 18,000 g for 10 min. Hepatic lipids were extracted from the liver based on the principle of Bligh and Dyer method ( 30 ). Liver samples (100 mg) were homogenized in 0.5 ml methanol and mixed with 0.5 ml chloroform and 0.4 ml water. Phase separation was achieved by 10 min centrifugation at 18,000 g . Lipid fraction in the chloroform phase was dried under nitrogen and then reconstituted in 0.5 ml n-butanol.

LC-MS analysis of serum and lipid extract
A 5 µl aliquot of diluted serum or liver lipid sample was injected into an Acquity UPLC system (Waters, Milford, MA) and separated by a gradient of mobile phase ranging from water to 95% aqueous acetonitrile containing 0.1% formic acid over a 10 min run. LC eluant was introduced into a SYNAPT QTOF mass spectrometer (Waters) for accurate mass measurement and ion counting. Capillary voltage and cone voltage for electrospray ionization (ESI) was maintained at 3 kV and 30 V for positive-mode detection, or at Ϫ 3 kV and Ϫ 35 V for negative-mode detection, respectively. Source temperature and desolvation temperature were set at 120°C and 350°C, respectively. Nitrogen was used as both cone gas (50 L/h) and desolvation gas (600 L/h), and argon as collision gas. For accurate mass measurement, the mass spectrometer was calibrated with sodium formate solution (range m/z 50-1,000) and monitored by the intermittent injection of the lock mass leucine enkephalin ( and survival leads to progressive necrosis and apoptosis of hepatocytes (18)(19)(20). Although previous studies have yielded multifaceted mechanistic information on cocaine-induced hepatotoxicity, our current understanding on this toxic event still remains insuffi cient. For example, the correlation between the metabolic events and the progression of cocaine-induced liver injury has not been extensively examined, even though the liver is the most important metabolic organ in the body. This situation is caused mainly by the limited analytical capacity and discovery power of the traditional targeted metabolite analysis approach. However, the developments in metabolomics and lipidomics have provided an alternative technical platform for examining the metabolic fl ux in a complex biomatrix through the combination of advanced analytical instrumentation and chemometric computation (21)(22)(23). The indiscriminant and untargeted nature of metabolomic and lipidomic analyses have led to the identifi cation of novel biomarkers that guided subsequent mechanistic investigations of diseases and chemical-induced toxicities (24)(25)(26). In fact, a recent metabolomics-based study has revealed the potential role of tryptophan and purine metabolism in cocaine addiction ( 27 ).
In this study, the infl uences of cocaine treatment on the chemical composition of mouse serum and liver were investigated by the LC-MS-based lipidomic analysis. Biomarkers originated from the cocaine-induced disruption of fatty acid metabolism were further characterized. Cotreatment of feno fi brate, an activator of peroxisome proliferator-activated receptor ␣ (PPAR ␣ ), was performed to investigate the role of fatty acid oxidation in cocaine-induced liver injury.

Animal treatment and sample collection
Male C57BL/6 mice, 10-12 weeks old, were used in this study ( 28 ). All animals were maintained in a University of Minnesota (UMN) animal facility under a standard 12 h light/12 h dark cycle with food and water ad libitum. Handling and treatment procedures were in accordance with animal study protocols approved by the UMN Animal Care and Use Committee.
For cocaine treatment, cocaine was dissolved in a saline solution and administered at a dose of 30 mg/kg by daily intraperitoneal injection for three consecutive days. For fenofi brate treatment, fenofi brate was suspended in a 0.5% CMC solution and administered

Gene expression analysis
Total RNA was extracted from the livers of C57BL/6 mice using TRIzol reagent. Quantitative real-time PCR (qPCR) was performed using cDNA generated from 1 mg total RNA with SuperScript III Reverse Transcriptase. Primers were designed using NCBI/Primer-BLAST (http://www.ncbi.nlm.nih.gov/tools/primer-blast/), and sequences are enlisted in supplementary Table I. qPCR reactions were carried out using SYBR green PCR master mix with a StepOnePlus system (Applied Biosystems). Gene expression levels were quantifi ed using the Comparative CT method and normalized to 18S rRNA.

Statistics
Experimental values are expressed as mean ± standard deviation (SD). Statistical analysis was performed with two-tailed Student t-tests for unpaired data, with P < 0.05 considered statistically signifi cant.

Time course of cocaine-induced liver injury
To characterize the development of cocaine-induced hepatotoxicity, C57BL/6 mice were treated with 30 mg/kg of cocaine by intraperitoneal injection for three consecutive days. Microscopic examination of liver histology by H&E and Oil Red O staining revealed the progression of cocaineinduced liver injury ( Fig. 1A and supplementary Fig. I-A). On day 1, hepatocytes were slightly swelling with increased vacuoles, but no clear cell death was observed. On day 2, hepatocytes in the centrilobular area appeared to be ballooning and started to lose nuclei, and enlarged lipid droplets became apparent. On day 3, necrotic hepatocytes were prevalent around the central vein along with infl ammatory infi ltration, and the development of microvesicular steatosis was evident in the surrounding cells ( Fig. 1A and supplementary Fig. I-A). Consistent with the histological observation, serum ALT and AST activity remained unchanged on day 1 but dramatically increased on day 2 and 3 of cocaine treatment ( Fig. 1B and supplementary Fig. I-B), which suggests that cocaine-induced damage on hepatocytes at this dose (30 mg/kg) is a subacute toxic event instead of an acute one. Based on the appearance of lipid vacuoles and vesicles in cocaine-treated liver, the degree of lipid accumulation was evaluated by measuring TAG in the liver and serum. Both liver and serum TAG levels increased signifi cantly after just one dose of cocaine exposure. However, the serum TAG level remained largely unchanged after day 1, whereas the hepatic TAG level further increased dramatically on day 3 of treatment ( Fig. 1C, D ). Similar to the change in hepatic TAG level, the hepatosomatic index (ratio of liver weight/body weight) was also elevated on days 1 and 3 of cocaine treatment ( Fig. 1E ). Lipid peroxidation was examined by measuring TBARS content in the liver. The TBARS level peaked rapidly on day 1 of cocaine treatment, but it was not further increased afterwards ( Fig. 1F ). Overall, among all measured biochemical parameters, the increase of hepatic TAG on day 3 of cocaine treatment had the best correlation with the development of necrosis in the liver and the release of transaminase from the liver.

Chemometric analysis and biomarker identifi cation
Chromatographic and spectral data of serum samples were deconvoluted by MarkerLynx software (Waters). A multivariate data matrix containing information on sample identity, ion identity [retention time (RT) and m/z ], and ion abundance was generated through centroiding, deisotoping, fi ltering, peak recognition, and integration. The intensity of each ion was calculated by normalizing the single ion counts (SIC) versus the total ion counts (TIC) in the whole chromatogram. The data matrix was further exported into SIMCA-P+ software (Umetrics, Kinnelon, NJ) and transformed by mean-centering and Pareto scaling. Based on the complexity and quality of the data, either unsupervised or supervised multivariate data analysis (MDA), including principal components analysis (PCA) and projection to latent structures-discriminant analysis (PLS-DA), were adopted to analyze the serum and liver lipid data from control and cocaine-treated C57BL/6 mice. Major latent variables in the data matrix were described in a scores scatter plot of defi ned multivariate model. Potential biomarkers were identifi ed by analyzing ions contributing to the principal components and to the separation of sample groups in a S-loadings plot of orthogonal PLS (OPLS) discriminant analysis ( 22,31

Quantitation of palmitoylcarnitine in serum and liver
Palmitoylcarnitine (PalC) in serum and liver lipid extracts was quantifi ed by accurate mass-based ion extraction chromatograms. Stable isotope-labeled PalC ([ 13 C 4 ]PalC) was used as the internal standard. PalC concentrations were determined by calculating the ratio between the peak area of PalC and the peak area of [ 13 C 4 ]PalC and fi tting with a standard curve with a linear range from 10 nM to 1 µM ( r = 0.99) using QuanLynx software (Waters).

In vitro assay of hepatic mitochondrial ␤ -oxidation activity
Hepatic mitochondrial ␤ -oxidation activity was evaluated by the rate of PalC utilization ( 32 ). Mitochondria were isolated from the liver by differential centrifugation ( 33 ). Incubations of 20 µM PalC with 0.5 mg/kg liver mitochondria homogenate were then carried out in a Tris-HCl buffer containing 0.05% Triton X-100, 10 mM MgCl 2 , and 1 mM CoASH in a fi nal volume of 200 µl for 15 min at 37°C. The reactions were terminated by adding 400 µl acetonitrile. The rate of PalC utilization was calculated based on the LC-MS quantitation of PalC concentration after the incubation.

Profi ling cocaine metabolism in mouse
The profi le of cocaine metabolism was determined by LC-MS analysis of cocaine metabolites in the urine. Details of the identifi cation and structural elucidation of cocaine metabolite will be described in a separate report. Briefl y, the chromatographic peaks of all identifi ed urinary cocaine metabolites were identifi ed, and their peak areas were determined using accurate mass measurement-based MetaboLynx software (Waters) ( 34 ). The profi les of urinary cocaine metabolites during three-day cocaine treatment were compared by calculating the percentage of the peak area of each single metabolite in the pooled total peak area (area % ± SD) of cocaine metabolites.
correlating to the development of cocaine-induced liver injury. Serum ions (I-XXII) contributing to the classifi cation of cocaine and control groups were further identifi ed in a loadings S-plot ( Fig. 2B ). The majority of identifi ed biomarkers belong to the subfamilies of polar lipids, which are phosphatidylcholines (PC: III-XI, XIV, XXI), lysopho sphatidylcholines (LPC: I, II, XII, XIII), phosphatidylethanolamine (PE: XXII), lysophosphatidylethanolamines (LPE: XVIII-XX), and acylcarnitines (XV-XVII) ( Table 1 ). These identifi cations were made through accurate mass measurement; elemental composition analysis; MS/MS fragmentation; chromatographic and spectroscopic comparison with authentic standards; or published data in metabolomic and lipidomic databases. For example, palmitoylcarnitine (PalC, XVI) was identifi ed through chromatographic and spectral comparisons with the authentic stand ard ( Fig. 2C, D ), whereas the structural moiety and fatty acid content of PC and PE molecules were elucidated through the interrogation of MS/MS spectra, even though the positions of acyl group in the glycerol backbone was The stepwise changes in hepatic TAG level suggested that different metabolic events and mechanisms might contribute to the histological phenotypes and steatosis on days 1 and 3 of cocaine treatment.

Metabolomic investigation of cocaine-induced metabolic events
To further explore the metabolic events associated with subacute hepatotoxic effects of cocaine, especially the changes in the lipidome, serum samples and liver lipid extracts were examined through LC-MS-based lipidomic analysis. After the chromatographic and mass spectral data acquired from LC-MS analysis were processed by multivariate ana lysis, a two-component model represented by a scores scatter plot ( Fig. 2A ) was constructed to illustrate the relationship among sample groups. The distinctive separation of serum samples collected at the different time points of cocaine treatment ( Fig. 2A ) indicated that cocaine dramatically altered the chemical composition of serum in a time-dependent pattern, suggesting the existence of serum metabolites analysis, and the lipid species contributing to the separation were similar to those identifi ed through serum lipid analysis (supplementary Fig. III-A, B).

Infl uence of cocaine treatment on mitochondrial fatty acid oxidation in the liver
Among all identifi ed changes in the lipidome ( Figs. 2 and 3 and Table 1 ), the increase in the relative abundance of long-chain acylcarnitines is highly relevant to cocaineinduced hepatic steatosis due to their identity as specifi c substrates of mitochondrial ␤ -oxidation. To validate this observation, the levels of PalC (XVI) in both serum and liver during the three-day cocaine treatment were quantifi ed. The results showed that PalC in serum and liver remained unchanged on day 1 of cocaine treatment, was signifi cantly elevated on day 2, and was dramatically increased on day 3 ( Fig. 4A , B ), indicating that acylcarnitine accumulated in the liver and was subsequently released into serum after cocaine treatment. The inhibition of fatty acid oxidation, as suggested by the accumulation of PalC, was further confi rmed by the in vitro mitochondrial not resolved (supplementary Fig. II-A, B). Examination of their relative abundances in the detected MS signals of serum revealed several clear trends of changes in serum lipidome, which included the following. First, dramatic increase of long-chain acylcarnitines on days 2 and 3 of cocaine treatment, including PalC (XVI), oleoylcarnitine (OC; XVII), and stearoylcarnitine (SC; XV) ( Fig. 3A ). Second, the increase of LPC(16:0) (XII) and LPC(18:2) (XIII) but the decrease of LPC(20:4) (I) and LPC(20:3) (II) ( Fig. 3B ), suggesting that the elongation and desaturation of fatty acids in vivo were potentially reduced after cocaine treatment. This is further supported by the observation of the increase of PC(16:0/18:2) (XI) and PC(16:0/16:0) (XXI), which only contain 16-and 18-carbon fatty acids, and the decrease of PC(18:0/20:3) (III) and PC(18:1/22:6) (VI), which contain 20-and 22-carbon PUFAs ( Fig. 3C ). Third, the signifi cant increase of LPE species on day 3 of cocaine treatment, including LPE(18:2) (XVIII), LPE(22:6) (XIX), and LPE(20:4) (XX) ( Fig. 3D ). Furthermore, the time-dependent separation among the control and cocainetreated liver lipid extracts was observed following lipidomic (supplementary Fig. VI-B, C) induced by three-day cocaine treatment.

Cocaine metabolism during cocaine and fenofi brate treatments
As a versatile transcription factor, PPAR ␣ not only regulates enzymes and transporters important in lipid meta bolism but also can affect many xenobiotic-metabolizing enzymes (36,37 ). Because cocaine-induced hepatotoxicity is initiated by cocaine metabolism, especially the production of reactive cocaine metabolites, the effects of fenofibrate on cocaine metabolism were evaluated through LC-MS-based profi ling of major urinary cocaine metabolites on days 1 and 3 of the treatments. The resulted showed that the levels of two reactive N -demethylation metabolites, benzoylnorecgonine (BNE) and norcocaine (NCOC), on day 3 were higher than their levels on day 1, whereas an opposite trend was observed for benzoylecogonine (BE) and N -hydroxybenzoylecogonine (OHBE), two nontoxic metabolites formed by the hydrolysis of cocaine ( Fig. 6 ). These observations suggest that the bioactivation of cocaine was induced by itself. Compared with observed time-dependent differences, fenofi brate did not affect cocaine metabolism, because the cocaine metabolite profi le in the cotreatment group was comparable to that in the cocaine-alone group on both day 1 and day 3 ( Fig. 6 ). Therefore, the protective effects of fenofi brate on cocaine-induced liver injury were not due to the modifi cation of cocaine metabolism.

Effects of fenofi brate on cocaine-induced changes in the lipidome
The lipidomic changes associated with the protective effects of fenofi brate were examined through LC-MS ␤ -oxidation functional assay. The utilization of PalC by mitochondria was signifi cantly reduced on day 3 compared with other time points ( Fig. 4C ).

Protective effects of fenofi brate on cocaine-induced liver injury
Fatty acid oxidation is mainly regulated by transcriptional factor PPAR ␣ ( 35 ). To examine the role of fatty acid oxidation in cocaine-induced hepatotoxicity, fenofi brate, a PPAR ␣ agonist, was fed to the mice on all three days of cocaine treatment. The results of H&E and Oil Red O staining clearly showed that fenofi brate cotreatment was highly protective against cocaine-induced histological changes in the centrilobular region of the liver, as evidenced by the lack of necrotic cells and microvesicular steatosis in the liver treated with both cocaine and fenofi brate ( Fig. 5A and supplementary Fig. IV-A). Consistent with the histological fi ndings, biochemical indices of cotreated mice, including ALT, AST, hepatosomatic index, and serum and hepatic TAG, remained in normal levels in contrast to the dramatic changes in the cocaine-alone mice ( Fig. 5B-E and supplementary Fig. IV-B). This protective effect of fenofi brate was also confi rmed by observing the decrease of cocaine-induced TBARS level in the liver ( Fig. 5F ). Therefore, concurrent fenofi brate treatment was highly effective in rescuing the liver from cocaineinduced steatosis and other histopathological changes. Moreover, the effects of treating with fenofi brate one day and two days after starting cocaine treatment were examined (supplementary Fig. VI-A). Even though not as effective as the concurrent cotreatment, the postcocaine administrations of fenofi brate were still able to partially and timedependently reduce the elevation of serum aminotransferases The process of identifying serum metabolites responsive to cocaine treatment is described in Experimental Procedures. The most prominent markers revealed by the S-plot of OPLS model (labeled in Fig. 2B ) were further investigated by accurate mass measurement, MS/MS fragmentation, and structural elucidation. Retention time (RT) of each marker was its elution time from a 10 min run in a C 18 UPLC column. Fatty acid contents of phospholipids were determined by examining their MS/MS pattern, but their positions ( sn -1 and sn -2) in PCs and PEs was not resolved. ↑ , increased; ↓ , decreased.
in an S-plot and labeled according to Table 1 ( Fig. 7B ). Examination of their relative abundances in serum revealed the protective effects of fenofi brate against the cocaineinduced alterations in the lipidome. For example, fenofi brate completely abolished the cocaine-induced acylcarnitine accumulation ( Fig. 7C ), and it also partially reversed the change in several LPC, PC, and LPE species ( Fig. 7D ). Furthermore, analyzing the liver lipids revealed similar separation among sample groups (supplementary Fig. V-A). Lipid species contributing to the separation were overlapping with ones identifi ed in serum analysis (supplementary Fig. V-B).
analysis of serum samples from fenofi brate-cocaine, cocainealone, fenofi brate-alone, and control treatments. The results showed distinctive separation of four sample groups in the scores scatter plot of a three component PLS-DA model ( Fig. 7A ). Besides expected differences between control and cocaine groups, the separation of the cotreatment group from the cocaine-alone group was also apparent, suggesting fenofi brate signifi cantly altered the lipid composition of cocaine-treated mice. The serum ions contributing to the distinction of the cocaine group from the cotreatment and control groups were further characterized  alone increased the expression of PPAR ␣ -targeted genes ( Fig. 9A -C ). The trend of upregulating PPAR ␣ activity was also observed from the cocaine-alone treatment ( Fig. 9 ). The cotreatment of fenofi brate and cocaine led to the highest levels of gene expression of Cpt1a , Cpt2 , and Acot1 among all treatments ( Fig. 9 ). This correlation between the activation of PPAR ␣ -targeted genes and the recovery of mitochondrial ␤ -oxidation activity ( Fig. 8C ) suggested that the protective effect of fenofi brate against cocaineinduced lipid accumulation and hepatotoxicity was at least partially due to the PPAR ␣ -mediated upregulation of fatty acid oxidation. Furthermore, the results from the postcocaine administration of fenofi brate indicated that early activation of PPAR ␣ -targeted genes by fenofi brate (supplementary Fig. VI-F) was positively correlated with the protection against cocaine-induced liver injury.

DISCUSSION
Mechanisms of cocaine-induced liver injury have been studied for over three decades due to the stature of cocaine as one of the most harmful substances in modern society (1) . Following the identifi cation of reactive cocaine

Effects of fenofi brate on cocaine-induced inhibition of mitochondrial ␤ -oxidation function
Quantitative analysis of serum and hepatic PalC levels from the concurrent cotreatment and postcocaine treatments of fenofi brate was performed to validate the observations from the lipidomic analysis. The result showed that the concurrent cotreatment of fenofi brate with cocaine prevented the cocaine-induced increase of PalC in both serum and liver ( Fig. 8A , B ), and the postcocaine treatment of fenofi brate also partially and time-dependently reversed PalC accumulation (supplementary Fig. VI-D, E). The protective effect of fenofi brate was likely due to the preservation of mitochondrial ␤ -oxidation function in cocaine-treated mice, as the mitochondrial utilization of PalC was sustained in the cotreated mice ( Fig. 8C ).

Effects of fenofi brate on PPAR ␣ -targeted genes
The infl uence of fenofi brate and cocaine on PPAR ␣targeted genes was further examined by measuring the expression of carnitine palmitoyltransferase 1a ( Cpt1a ), carni tine palmitoyltransferase 2 ( Cpt2 ), and acyl-CoA thioesterase 1 ( Acot1 ), three genes involving in mitochondrial ␤ -oxidation and acyl-CoA turnover. As expected, fenofi brate treatment Values were presented as mean ± SD (n = 3-4). * P < 0.05 and ** P < 0.01 indicate statistical signifi cance between control and marked samples. 1 ( Fig. 1A, B and supplementary Fig. I-A, B). Therefore, examining the causes underlying the increase of hepatic TAG level might provide insights on cocaine-induced hepatotoxicity. Hepatic TAG level is regulated by multiple processes in TAG turnover in the liver, which include import and biosynthesis of fatty acids, TAG synthesis, TAG hydrolysis, fatty acid utilization (mainly oxidation), and TAG secretion. In this study, metabolomic analysis revealed that long-chain acylcarnitine levels in serum and liver were well correlated to the progression of hepatotoxicity ( Fig. 4A, B and Table 1 ), and the acylcarnitine accumulation in vivo was confi rmed by observing reduced PalC utilization by the mitochondria from cocaine-treated liver ( Fig. 4C ), indicating that cocaine treatment caused significant compromise of mitochondrial ␤ -oxidation function. Because of the importance of fatty acid oxidation in TAG turnover and energy metabolism, these results revealed a contributing mechanism for the development of microvesicular steatosis and liver toxicity in subacute cocaine treatment.
Mitochondrial dysfunction as a key component in cocaine toxicity has been investigated previously. Supporting evidence includes the observations of morphological alterations of hepatic mitochondria in mice after cocaine treatment and a dose-and time-dependent decrease of mitochondrial membrane potential ( 16,51 ). A recent study using purifi ed rat heart mitochondria also reported that cocaine inhibited octanoylcarnitine-mediated respiration ( 52 ). Identifi cation of cocaine-elicited inhibition of hepatic ␤ -oxidation in vivo in this study further underscores the importance of mitochondria in cocaine-induced hepatotoxicity. Besides cocaine, the inhibition of mitochondrial fatty acid oxidation has been identifi ed as a major mechanism of the hepatotoxicity induced by many therapeutic agents and environmental toxicants, including acetaminophen ( 53 ), aspirin, and hypoglycin ( 54 ). The direct consequences of suppressed fatty acid oxidation function are the disruption of energy metabolism and the accumulation of TAG and fatty acids. These changes can further impact other metabolic events, as well as the susceptibility of hepatocytes to oxidative stress and chemical insults. In fact, the decrease of ATP content, which refl ects the disruption of energy metabolism, and the increase of lipid peroxidation products, which represents oxidative stress, have been observed previously in cocaine-treated mouse liver ( 55 ). With regard to the metabolic events associated with decreased fatty acid oxidation, one interesting observation in this study was the lack of change in serum TAG level on day 3 of cocaine treatment in contrast to the dramatic change in hepatic TAG level ( Fig. 1C, D ), since fatty acids not used for oxidative metabolism in the liver were usually incorporated into TAG and then secreted into the blood via very low density lipoprotein (VLDL) ( 56 ). It has been shown that microsomal TAG transfer protein (MTP), which is responsible for assembling VLDL, is inhibited by other known inhibitors of mitochondrial ␤ -oxidation ( 57 ). Whether this event contributes to cocaine-induced TAG accumulation and toxicity in the liver requires further investigation. metabolites, such as N -hydroxynorcocaine ( 38 ) and norcocaine nitroxide ( 39 ), the majority of mechanistic studies have focused on the bioactivation of cocaine as well as subsequent covalent protein binding and disruption of redox homeostasis, including glutathione depletion and lipid peroxidation. Results from these studies have clearly indicated that metabolic activation and oxidative stress are indispensable components of cocaine toxicity (11)(12)(13). However, the adverse effects of cocaine exposure are not limited to the direct interaction with cytoprotective and cytotoxic pathways. As a potent uptake inhibitor of catecholamines, cocaine, especially cocaine abuse, can dramatically increase norepinephrine and epinephrine levels in synapses and sympathetic nerve endings ( 40 ). Because adrenergic receptors, the physiological targets of catecholamines, are actively involved in the regulation of carbohydrate, amino acid, and lipid metabolism (41)(42)(43)(44), cocaine is expected to signifi cantly affect intermediary metabolism. Observed pathophysiological events, including the increase of lactate and basal metabolism rate ( 45,46 ), as well as the elevation of cardiac output and blood pressure ( 48 ), support this assumption. However, the exact effects of cocaine exposure on intermediary metabolism remain largely unknown. In addition, the role of intermediary metabolism in cocaine-induced hepatotoxicity has not been explored.
Cocaine-induced fatty liver in animal and human has been reported previously (48)(49)(50). In this study, a new feature of fatty liver development in mouse following subacute treatment of cocaine was identifi ed, i.e., hepatic TAG level increased on days 1 and 3 of cocaine treatment, respectively, whereas serum TAG level increased on day 1 but remained largely unchanged thereafter ( Fig. 1C, D ). Compared with serum TAG level, the two-stage induction of hepatic TAG was better correlated to the results of histological and biochemical analyses, which showed that apparent development of microvesicular steatosis and dramatic increase of transaminase activity mainly occurred after day Fig. 6. Infl uence of repeated cocaine exposure and fenofi brate on cocaine metabolism. The procedure of determining the profi le of urinary cocaine metabolites is described in Experimental Procedures. Relative abundances of cocaine and eight major urinary metabolites on days 1 and 3 of cocaine-alone treatment (30 mg/kg) and fenofi brate (50 mg/kg)-cocaine cotreatment were compared. Values were presented as mean ± SD (n = 3-4). * P < 0.05 indicates statistical signifi cance between days 1 and 3.
The observation of protective effects of fenofi brate cotreatment in this study further confi rmed the importance of fatty acid oxidation in cocaine-induced hepatotoxicity ( Fig. 5 ). As an agonist of nuclear receptor PPAR ␣ , fenofi brate induced the expression of PPAR ␣ -targeted genes ( Fig. 9 ). Concurrent administration of fenofi brate with cocaine treatment abolished the adverse effects of cocaine on TAG turnover, lipid peroxidation, acylcarnitine level, and ␤ -oxidation activity ( Figs. 5, 7, and 8 ) without affecting cocaine metabolism ( Fig. 6 ), which suggests that the regulation of PPAR ␣ -mediated pathways, especially fatty acid metabolism, is likely the main mechanism underlying the protective effects of fenofi brate. Fenofi brate and other fi brate compounds have been widely used to treat hypertriglyceridemia in humans. Recently, two clinical studies have reported high TAG levels in the blood and heart of cocaine-dependent human subjects ( 59,60 ). Based on the protective effects of fenofi brate against cocaine-induced hepatotoxicity in this study and the proven Although the exact mitochondrial targets of cocaine in mouse liver was not revealed by this study, a potential cause behind the observation of ␤ -oxidation inhibition on day 3, but not day 1, of cocaine treatment, was revealed by the profi ling of urinary cocaine metabolites, which showed that two N -oxidative metabolites, norcocaine and benzoylnorecgonine, signifi cantly increased on day 3, whereas the level of benzoylecognine, a hydrolytic product of cocaine, decreased signifi cantly ( Fig. 6 ). It is known that oxidative metabolites of cocaine are responsible for cocaine-induced depression of mitochondrial function and generation of reactive oxygen species ( 14,58 ). Therefore, the correlation of their urinary levels with acylcarnitine levels in serum and liver implied that the increased oxidative metabolism of cocaine might contribute to the cocaine-induced impairment of mitochondrial ␤ -oxidation. Examining the effect of individual cocaine metabolites on acylcarnitine level in future studies will reveal their separate contributions to cocaine-induced lipotoxicity.  Fig. 2B were labeled, and their chemical identities are listed in Table 1 . (C) Relative abundance of acylcarnitines in the serum. (D) Relative abundance of LPC, PC, and LPE in the serum. Values were presented as mean ± SD (n = 3-4). * P < 0.05 and ** P < 0.01 indicate statistical signifi cance between control and marked samples. related to these changes is the similarity to the reported lipidomic changes induced by ethanol exposure ( 61,62 ), such as the decrease of PUFA-comprised lipid species and the imbalance between LPC and LPE species ( Fig. 3B, D and Table 1 ). Another feature of these observed changes in the lipidome is their potential relevance to the functions of PPAR ␣ , such as the elongation and desaturation of fatty acids ( 63 ). Whether these changes in the lipidome are common events in chemical-induced hepatotoxicity and how PPAR ␣ contributes to these changes require further study.
Overall, the combination of LC-MS-based lipidomics, animal modeling, and biochemical analysis in this study enabled the characterization of cocaine-induced dramatic changes in lipid metabolism. The inhibition of mitochondrial fatty acid oxidation was identifi ed as a contributing mechanism of cocaine-induced hepatotoxicity, and the protective effects of fenofi brate suggest a potential strategy in reducing the hepatic side effects of cocaine in clinical settings.
The authors thank the National Institute on Drug Abuse (NIDA) Research Resources Drug Supply Program for providing the cocaine and cocaine metabolites used in this study.
benefi cial effects of fi brates on dyslipidemia and cardiovascular diseases, it is reasonable to postulate that fi brates might offer hepato-and cardioprotective effects on cocaine users through metabolic modulation of mitochondrial functions. In fact, this potential value of fi brate in the protection against preexisting cocaine toxicity in humans was indicated by the observed partial protective effects from postcocaine administration of fenofi brate in this study (supplementary Fig. VI).
Besides identifying long-chain acylcarnitines as the biomarkers of cocaine-induced hepatotoxicity, LC-MSbased metabolomic and lipidomic analyses of serum and liver in this study revealed dramatic changes in the levels of major phospholipid species, including LPCs, PCs, LPEs, and PEs, as well as the increase of saturated fatty acids and the decrease of PUFAs in the acyl composition of phospholipids ( Table 1 ). Examining the pattern of these time-dependent changes indicated that some lipid species (I, III, VI, XIII, XI, XVIII) were affected immediately after cocaine exposure, whereas others (II, XII, XIX-XXI) were better correlated to the appearance of liver injury ( Fig. 3 ). More importantly, the majority of these changes, together with liver toxicity, were reversed by fenofi brate ( Fig. 7C ), which suggests their relevance to cocaine-induced toxicities. One interesting observation The procedure of in vitro ␤ -oxidation functional assay is described in Experimental Procedures. The utilization of PalC was interpreted as micromoles of PalC that was translocated from the cytosol to the mitochondria of hepatocytes per gram of liver protein per minute. Values were presented as mean ± SD (n = 3-4). * P < 0.05 indicates statistical signifi cance between control and marked samples. Values were presented as mean ± SD (n = 3-4). * P < 0.05 and ** P < 0.01 indicate statistical signifi cance between control and marked samples.