Targeting of Splice Variants of Human Cytochrome P450 2C8 (CYP2C8) to Mitochondria and Their Role in Arachidonic Acid Metabolism and Respiratory Dysfunction*

Background: Human CYP2C8 is involved in the metabolism of >20% of drugs on the market. Results: Both full-length (WT) CYP2C8 and splice variant 3 are bimodally targeted to mitochondria. Conclusion: Mitochondrial CYP2C8 metabolizes paclitaxel and arachidonic acid (ω-hydroxylation). Significance: Mitochondrial CYP2C8 is likely to play role in ischemic injury and oxidative stress. In this study, we found that the full-length CYP2C8 (WT CYP2C8) and N-terminal truncated splice variant 3 (∼44-kDa mass) are localized in mitochondria in addition to the endoplasmic reticulum. Analysis of human livers showed that the mitochondrial levels of these two forms varied markedly. Molecular modeling based on the x-ray crystal structure coordinates of CYP2D6 and CYP2C8 showed that despite lacking the N-terminal 102 residues variant 3 possessed nearly complete substrate binding and heme binding pockets. Stable expression of cDNAs in HepG2 cells showed that the WT protein is mostly targeted to the endoplasmic reticulum and at low levels to mitochondria, whereas variant 3 is primarily targeted to mitochondria and at low levels to the endoplasmic reticulum. Enzyme reconstitution experiments showed that both microsomal and mitochondrial WT CYP2C8 efficiently catalyzed paclitaxel 6-hydroxylation. However, mitochondrial variant 3 was unable to catalyze this reaction possibly because of its inability to stabilize the large 854-Da substrate. Conversely, mitochondrial variant 3 catalyzed the metabolism of arachidonic acid into 8,9-, 11,12-, and 14,15-epoxyeicosatrienoic acids and 20-hydroxyeicosatetraenoic acid when reconstituted with adrenodoxin and adrenodoxin reductase. HepG2 cells stably expressing variant 3 generated higher levels of reactive oxygen species and showed a higher level of mitochondrial respiratory dysfunction. This study suggests that mitochondrially targeted variant 3 CYP2C8 may contribute to oxidative stress in various tissues.

In this study, we found that the full-length CYP2C8 (WT CYP2C8) and N-terminal truncated splice variant 3 (ϳ44-kDa mass) are localized in mitochondria in addition to the endoplasmic reticulum. Analysis of human livers showed that the mitochondrial levels of these two forms varied markedly. Molecular modeling based on the x-ray crystal structure coordinates of CYP2D6 and CYP2C8 showed that despite lacking the N-terminal 102 residues variant 3 possessed nearly complete substrate binding and heme binding pockets. Stable expression of cDNAs in HepG2 cells showed that the WT protein is mostly targeted to the endoplasmic reticulum and at low levels to mitochondria, whereas variant 3 is primarily targeted to mitochondria and at low levels to the endoplasmic reticulum. Enzyme reconstitution experiments showed that both microsomal and mitochondrial WT CYP2C8 efficiently catalyzed paclitaxel 6-hydroxylation. However, mitochondrial variant 3 was unable to catalyze this reaction possibly because of its inability to stabilize the large 854-Da substrate. Conversely, mitochondrial variant 3 catalyzed the metabolism of arachidonic acid into 8,9-, 11,12-, and 14,15-epoxyeicosatrienoic acids and 20-hydroxyeicosatetraenoic acid when reconstituted with adrenodoxin and adrenodoxin reductase. HepG2 cells stably expressing variant 3 generated higher levels of reactive oxygen species and showed a higher level of mitochondrial respiratory dysfunction. This study suggests that mitochondrially targeted variant 3 CYP2C8 may contribute to oxidative stress in various tissues.
Cytochrome P450 2C8 (CYP2C8; 2 EC 1.14.14.1) is constitutively expressed in the liver and only represents about 5-7% of the total tissue P450 pool (1), but it is involved in the metabolism of a relatively large number of drugs used in human medicine. CYP2C8 is also expressed in many extrahepatic tissues, such as brain, mammary gland, ovary, heart, endothelium, kidneys, adrenal gland, breast, gastrointestinal tract, and tonsils (2). It is regarded as an important biomarker in breast cancers because of high levels of expression in mammary tumors (3)(4)(5)(6). Presently, at least 15 allelic variants of CYP2C8 have been described (The Human Cytochrome P450 (CYP) Allele Nomenclature Database). These genetic polymorphisms are believed to be the main cause of interindividual differences in enzyme activity, an important factor in clinical interventions. Ethnic background is an important factor in the altered catalytic activity of the enzyme (7)(8)(9). Two CYP2C8 variants, CYP2C8*2 and CYP2C8*3, have been reported that exhibit reduced activity compared with the normal/"wild-type" form (CYP2C8*1). The CYP2C8*2 variant has been reported to occur in Ͻ1% of Caucasians and about 15% of African-Americans. However, the CYP2C8*3 variant is rare in African-Americans but has been reported in as many as 20% of Caucasians (7).
CYP2C8 accounts for the metabolism of Ͼ20% of the drugs used in human medicine (10) and is the principal enzyme responsible for the metabolism of the anticancer drug paclitaxel (Taxol) and more than 60 clinically important therapeutic drugs, including a wide spectrum of antimalarial, antidiabetic, anti-inflammatory, and anticancer agents (e.g. amiodarone, cabazitaxel, carbamazepine, cerivastatin, chloroquine, diclofenac, fluvastatin, ibuprofen, pioglitazone, rosiglitazone, repaglinide, and treprostinil) (10 -12). In addition to the metabolism of the wide spectrum of drugs listed above, along with CYP2J2, CYP2C8 is also implicated in myocardial function. Both of these CYP epoxygenases catalyze the oxidation of arachidonic acid to four regioisomeric epoxyeicosatrienoic acids (EETs) (12,13), compounds that induce vasodilation and are mitogenic, antiapoptotic, and anti-inflammatory in endothelial cells (14 -16). These metabolites are thought to promote cardiac recovery against ischemia-reperfusion injury. However, members of the CYP2C subfamily also catalyze the epoxidation of linoleic acid to products that can have cytotoxic vasoconstrictive effects (14 -16). Indeed, transgenic overexpression of CYP2C8 in cardiomyocytes increased necrosis in a Langendorff in vitro heart perfusion system (17)(18)(19) and an associated increase in the production of the leukotoxin diols 9,10-and 12,13-dihydroxyoctadecenoic acid and ROS. Furthermore, trimethoprim, a selective inhibitor of human CYP2C8, improved left ventricular developed pressure recovery and reduced infarct size after ischemia-reperfusion in isolated CYP2C8overexpressing mouse heart preparations (20). Notably, infarct size was also reduced to control levels by the antioxidants N-acetylcysteine and superoxide dismutase, suggesting a role for ROS in CYP2C8-mediated susceptibility to ischemia-reperfusion injury. However, the precise mechanism of the cardiomodulatory function of CYP2C8 remains unclear.
Several studies from our and other laboratories have demonstrated that various cytochrome P450 members of Family 1, 2, and 3 CYPs are associated with mitochondria in addition to the endoplasmic reticulum (ER) (21)(22)(23). Studies from our laboratory have described the mechanisms of bimodal targeting of CYP1A1, -1B1, -2B1, -2E1, and -2D6 to mitochondria and ER (24 -32). The bimodal targeting of these proteins is driven by their N-terminal "chimeric signals" resulting in their dual targeting properties to both ER and mitochondria. CYPs localized in the mitochondrial compartment catalyze drug metabolism, efficiently interacting with mitochondrial adrenodoxin (Adx) and adrenodoxin reductase (AdxR) (25,26,28). In the case of CYP2B1, -2D6, and -2E1, activation of cryptic mitochondrial targeting signals by protein kinase A-or protein kinase C-mediated phosphorylation is required at sites immediately flanking the targeting signal and/or membrane-anchoring regions (26 -30). However, activation by endoproteolytic cleavage by a cytosolic endoprotease is required in the cases of CYP1A1 and CYP1B1, exposing the mitochondrial signal (24,31,32). The truncated forms of CYP1A1 and CYP1B1 exhibit altered substrate specificity compared with the microsomal counterparts (24,31,32). However, it remains unclear whether CYP2C8 is also targeted to mitochondria and whether the mitochondrial form plays a role in drug metabolism and drug-induced toxicity.
Most studies on genetic polymorphic forms of CYP2C8 are focused on the mutations targeted to the catalytic domain of the enzyme (7,(11)(12)(13). First, analysis of human liver samples in this study showed that, similar to some other Family 1 and 2 CYPs, CYP2C8 can also be localized in the mitochondria. Second, a nearly full-length form (comparable in size with the microsomal CYP2C8; designated as Var_1) and a truncated form (44 kDa; designated as Var_3) were detected in mitochondria. Notably, despite missing the critical Arg-97 residue involved in binding to the heme group in the full-length enzyme (33), the N-terminal truncated form is capable of binding to heme and catalyzed the metabolism of some but not all CYP2C8 substrates. In this study, we show that three major splice variants of CYP2C8 expressed in human liver exhibit remarkable differences with respect to relative levels of microsomal (ER) and mitochondrial targeting, metabolism of paclitaxel and arachidonic acid, and induction of cellular oxidative stress.

Isolation of Mitochondria and Microsomes from Frozen
Human Liver Samples-Human liver samples were obtained through Tennessee Donor Services (Nashville, TN) and used in accordance with Vanderbilt Institutional Board guidelines. Mitochondria and microsomes were isolated from human liver samples by using a modification of a method described previously (29,30). Briefly, livers were washed in ice-cold saline and homogenized in 10 volumes of sucrose-mannitol buffer (20 mM potassium HEPES buffer (pH 7.5) containing 70 mM sucrose, 220 mM mannitol, 2 mM EDTA, and 0.5 mg/ml BSA). Mitochondrial and microsomal fractions were isolated from the homogenates using a differential centrifugation method (24,34). Mitochondria were pelleted at 8,000 ϫ g for 15 min. Crude mitochondria were washed twice with the above buffer and pelleted through a 0.8 M sucrose layer at 14,000 ϫ g for 30 min, and the mitochondrial pellet was washed twice with sucrosemannitol buffer. Mitoplasts were prepared by treatment with digitonin (75 g/mg protein; Calbiochem) at 4°C for 10 min. The resulting mitoplast pellet was washed twice with sucrosemannitol buffer. Microsomes were isolated from the postmitochondrial supernatant by centrifugation at 100,000 ϫ g for 60 min at 4°C. All subcellular preparations were resuspended in 50 mM potassium phosphate buffer (pH 7.5) containing 20% glycerol (v/v), 0.1 mM EDTA, 0.1 mM dithiothreitol, and 0.1 mM phenylmethanesulfonyl fluoride and frozen at Ϫ80°C.
Characterization of Human CYP2C8 Variants-Following the CYP nomenclature of Nelson et al. (35), the three different molecular forms characterized in this study are: WT CYP2C8 containing all nine exons (hereafter called Var_1), differentially spliced ⌬1ade2a (Var_2), and differentially spliced 1n⌬1a⌬2a (Var_3). The nucleotide sequences of the WT CYP2C8 (GenBank TM accession number NM_000770.3) and two reported transcripts Var_2 (GenBank accession number NM_ 001198853.1) and transcript Var_3 (GenBank accession number NM_001198854.1) were aligned by a Clustal format alignment, and common 5Ј and 3Ј internal primers were generated so that all three forms of CYP2C8 (Var_1, Var_2, and Var_3) were amplified in one single PCR. Total RNA was isolated from human liver samples using TRIzol reagent in accordance with the manufacturer's instructions (Invitrogen). Total RNA (5 g) was reverse transcribed using a High Capacity cDNA Archive kit (Applied Biosystems, Carlsbad, CA), and 100 ng was used for PCR amplification using the sense 5Ј-GTCCTGGTGCTGTG-TCTCTCTTTTAT-3Ј and antisense 5Ј-GAAACGCCGGAT-CTCCTTCCATC-3Ј primers from the region conserved in all three mRNAs. The amplicons were resolved by electrophoresis on 1.5% agarose gels and cloned in a TOPO vector using a TOPO TA cloning kit (Invitrogen), and the nucleotide sequences of all three cDNA amplicons were confirmed by nucleotide sequence analysis.
Molecular Modeling of CYP2C8 -The structure of human CYP2C8 Var_3 was modeled using the online protein structure prediction server Phyre 2 (36). The best model was aligned with the available CYP2C8 structure complexed with felodipine (Protein Data Bank code 2NNJ) using the PyMOL Molecular Graphics System (Version 1.5.0.4, Schrödinger). The putative heme-binding residues of WTCY2C8 and Var_3 were predicted based on the known structures of CYP2C5 and CYP2C8 (33,37,38).
Construction of WT and Variant CYP2C8 cDNAs-The ORF clone of human CYP2C8 (RG204605) was purchased from Origene Technologies, Rockville, MD. Var_2 and Var_3 reading frames were amplified and cloned in the same pCMV6-AC-GFP vector (Precision Shuttle mammalian vector with C-terminal truncated GFP tag) in SgfI and MluI restriction sites. For stable cell expression, wild-type CYP2C8 (Var_1), Var_2, and Var_3 cDNAs were cloned in pGP-lenti viral plasmid (Gene-Pass, Nashville, TN) in BamHI and AvrII restriction sites. An internal BamHI site was conservatively mutated using a QuikChange II XL site-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA). PCR amplification was carried out using a GeneAmp XL PCR kit (Applied Biosystems/Roche Applied Science) following the manufacturer's suggested protocol.
In Vitro Transport of 35 S-Labeled Protein into Isolated Mitochondria-The three cDNA constructs (in pCMV6-AC-GFP vectors) were used as templates in T7 polymerase-coupled rabbit reticulocyte lysate transcription-translation systems (Promega, Madison, WI) in the presence of [ 35 S]Met as described previously (24). Import of 35 S-labeled translation products in rabbit reticulocyte lysate was carried out as described by Gasser et al. (39) and modified by Bhat and Avadhani (40) and Addya et al. (24) using freshly isolated rat liver mitochondria. For some control experiments, mitochondria were preincubated with carbonyl cyanide m-chlorophenylhydrazone (50 M; Sigma-Aldrich) or oligomycin (50 M; Sigma-Aldrich) at 25°C for 20 min prior to initiating the import reaction. After import, trypsin digestion of mitochondria was performed for 20 min on ice (150 g of trypsin/mg of mitochondrial protein). Control mitochondria were incubated similarly without added trypsin. Soybean trypsin inhibitor (1.5 mg/mg of protein) was added to all samples to terminate the reactions. Mitochondria from both trypsin-treated and untreated samples were reisolated by pelleting through 0.8 M sucrose, and the proteins were subjected to SDS-PAGE followed by fluorography.
Transient Transfection and Protease Protection Assay-Transfection of COS-7 cells with cDNA constructs (6 g/100mm 2 plate) was carried out using the lipophilic reagent FuGENE 6 (Roche Applied Sciences). Cells were harvested 48 h post-transfection and used for isolating subcellular fractionations. Mitochondria and microsomes were suspended at 10 mg/ml in 50 mM potassium phosphate buffer (pH 7.4) containing 20% (v/v) glycerol, 0.1 mM EDTA, and 0.1 mM dithiothreitol and subjected to trypsin digestion (75 g/mg of protein) for 20 min on ice. The reaction was stopped by adding a 10ϫ molar excess of trypsin inhibitor and an equal volume of 2ϫ Laemmli sample buffer (41). The samples were incubated at 95°C for 5 min and subjected to SDS-PAGE (12%, w/v) for immunoblot analysis.
Immunofluorescence Microscopy-Cells were fixed with methanol, permeabilized by treatment with 0.1% Triton X-100 (v/v), blocked with 5% goat serum (v/v), and stained with appropriate primary and secondary antibodies. Fluorescence micros-copy was done using an Olympus BX61 microscope, and Pearson's coefficient was calculated using Volocity 5.3 software.
Reconstitution of Purified Human CYP2C8 for Paclitaxel Hydroxylation-Human CYP2C8 expressed in Escherichia coli cells was purified as described before (42,43), and the CYP content was measured by the Fe 2ϩ -CO difference spectra described later in this section. Enzyme reactions with purified CYP2C8 (full-length form) were performed in 0.3-ml final volumes of 50 mM potassium phosphate (pH 7.4) buffer containing 0.2 nmol of purified CYP protein reconstituted with or without 0.5 nmol of cytochrome P450 reductase or 0.4 nmol of purified Adx and 0.04 nmol of purified AdxR and 10 M paclitaxel. Reconstitutions with cytochrome P450 reductase also contained phospholipid vesicles (42,43). Inhibition studies were carried out by preincubating the enzyme for 20 min on ice with 10 mg/ml inhibitory antibody (BD Gentest, BD Biosciences) or similar amounts of control mouse ascites fluid or 5 M montelukast at 37°C for 20 min. Reactions were initiated by addition of 1 mM NADPH and carried out for 20 min at 37°C in a shaking water bath. Reactions were terminated by addition of 150 l of 100% CH 3 CN, mixed well using a vortex device, and centrifuged at 9,000 ϫ g for 20 min at 4°C. Approximately 250 -300 l of clear supernatant was transferred to a fresh tube and frozen at Ϫ80°C until analysis.
Paclitaxel Oxidation by Mitochondrial and Microsomal Enzymes from Stable Cell Lines-Reactions were run in 0.3-ml final volumes of 50 mM potassium phosphate (pH 7.4) buffer using 300 -500 g of freeze-thawed mitochondrial protein or microsomal protein. Mitochondrial samples were reconstituted with 0.4 nmol of purified Adx and 0.04 nmol of purified AdxR for supplementing the loss of Adx and AdxR during mitochondrial isolation (44). Microsomal samples were not supplemented with cytochrome P450 reductase because they contain optimal levels of NADPH-cytochrome P450 reductase for reconstitution of enzyme activity. Either 10 M paclitaxel or 70 M arachidonic acid was used as a substrate. Inhibition studies were performed by preincubating enzymes for 20 min with 1.5 mg of inhibitory antibody for 20 min on ice or 5 M montelukast at 37°C for 20 min. Reactions were initiated by addition of 1 mM NADPH and continued for 20 min at 37°C for paclitaxel and 5 min at 37°C for arachidonic acid in a shaking water bath. Reactions with paclitaxel were processed as described above, and those with arachidonic acid were snap frozen at Ϫ80°C until analysis. To this, 4.0 ml of a mixture of CHCl 3 :CH 3 OH (2:1, v/v) and 116 l of HCl were added. After mixing with a Vortex device, the organic layer was transferred to a new tube, and the aqueous layer was extracted again with 5 ml of CHCl 3 . The combined organic layers were pooled, 100 l of 5 M KOH was added, and the samples were evaporated to dryness under a stream of N 2 at 45°C. Each sample was dissolved in 1.0 ml of 0.4 M KOH in CH 3 OH:H 2 O (4:1, v/v), and the tubes were purged with argon gas and sealed (Teflon liners). The samples were heated at 50°C for 45 min, and then 3.0 ml of 0.15 M KCl and 90 l of glacial CH 3 CO 2 H were added to each. Each sample was extracted twice with diethyl ether, and the solvent was evaporated under a stream of N 2 at 45°C. Each sample was dissolved in 2.0 ml of 0.5% CH 3 CO 2 H in hexane and loaded onto a silica cartridge pre-equilibrated with the same solvent followed by a 10-ml wash with the same solvent. The columns were eluted with 10 ml of a mixture of hexane:diethyl ether:CH 3 CO 2 H (50: 50:0.5, v/v/v). The samples were dried under a stream of N 2 at 45°C, 150 l of 9 M aqueous CH 3 CO 2 H was added to each, and the samples were purged with argon gas and left at 23°C overnight to convert the EETs to dihydrodiols. The samples were dried under a stream of N 2 and dissolved in 35 l of a mixture of 15 mM aqueous NH 4 CH 3 CO 2 (pH 8.5): Aliquots were analyzed by LC-MS using a Waters Acquity octadecylsilane LC column (1.7 m, 1.0 ϫ 100 mm) operating at 60°C. Solvent A was 15 mM aqueous NH 4 CH 3 CO 2 (pH 8.5), and Solvent B was CH 3  LC-MS Analysis of Paclitaxel Products-LC-MS analysis of the oxidation products was performed on a Waters Acquity UPLC system connected to a Thermo LTQ mass spectrometer (Thermo Fisher, Waltham, MA) using an Acquity UPLC BEH octadecylsilane (C 18 ) column (1.7 m, 2.1 ϫ 100 mm). LC conditions were as follows: Solvent A contained 0.1% HCO 2 H acid in H 2 O (v/v), and solvent B contained 0.1% HCO 2 H in CH 3 CN (v/v). The following gradient program was used with a flow rate of 0.2 ml/min: 0 -2 min, hold at 10% B (v/v); 2-7 min, linear gradient from 10% B to 90% B (v/v); 7-7.5 min, hold at 90% B (v/v); 7.5-8 min, linear gradient from 90% B to 10% B (v/v); 8 -10 min, hold at 10% B (v/v). The temperature of the column was maintained at 40°C. Samples (20 l) were infused using an autosampler. MS analyses were performed in the electrospray ionization positive ion mode. The mass spectrometer was tuned using 6-hydroxypaclitaxel. Product ion spectra were acquired over the range m/z 500 -1000 and quantified using the target parent m/z 870, calibrating against external standards.
Generation of Stable Cell Lines-HepG2 cells were stably transduced with lentiviral vectors carrying Var_1 and Var_3 cDNAs. The lentiviral plasmids (12 g each) and the packaging plasmid DNA (6 g of Gag-Pol and MD.2G) were co-transduced into 293T cells using FuGENE HD transfection reagent to obtain fully functional lentiviral particles. After 48 h of transfection, the virus particles secreted in the cell culture medium were harvested by centrifugation at 600 ϫ g for 5 min, filtered through a 0.2-m syringe filter, and used for transducing HepG2 cells. The colonies were screened using puromycin (4 g/ml) as the selection marker. Cell colonies were selected and amplified. All the experiments were conducted in cells that were cultured without puromycin for at least three passages to rule out the adverse effects of puromycin on mitochondrial function.
Alkaline Extraction of Membrane Proteins-The membrane topology of the mitochondrial and microsomal CYP2C8 was studied by extraction with alkaline Na 2 CO 3 buffer essentially as described by Clark and Waterman (45) and modified by Anandatheerthavarada and co-workers (26,28). Freshly isolated mitochondria and microsomes suspended in sucrose-mannitol buffer (10 mg of protein/ml) were diluted to 20-fold with 0.1 M Na 2 CO 3 (pH 11) to a final concentration of 0.5 mg of protein/ ml. This mixture was mixed with a Vortex device three times (30 s each) to obtain an even suspension. The suspension was incubated on ice for 30 min and then centrifuged at 210,000 ϫ g for 90 min at 4°C. The pellet was suspended in 10 mM Tris-HCl buffer (pH 7.0) containing 2% SDS (w/v), and dissociated proteins were subjected to immunoblot analysis. The supernatant was mixed with an equal volume of ice-cold 20% trichloroacetic acid (w/v) and held at 4°C for 2 h. The precipitated proteins were collected by centrifugation at 100,000 ϫ g for 20 min and washed twice with ice-cold acetone. The resulting protein pellet was dissolved in SDS-containing buffer. The pellet and soluble protein fractions (100 g of protein) were subjected to immunoblot analysis.
Analysis of CO Difference Spectra-The CYP content of stable cell mitochondria and microsomes was measured by the Fe 2ϩ -CO versus Fe 2ϩ difference spectra as described by Omura and Sato (46) and modified by Guengerich (47)  Quantitative Real Time PCR-For mRNA quantitation by PCR analysis, RNA was digested with Turbo DNase I (Ambion, Austin, TX). Total RNA (5 g) was reverse transcribed using a High Capacity cDNA Archive kit, and 25 ng of the resulting cDNA was used in a standard Power SYBR Green real time PCR on an ABI 7300 real time PCR machine and analyzed using Primer Express 3.0 (Applied Biosystems). The levels of the integrated puromycin resistance gene (in relation to the cellular ␤-actin gene used as a control) were determined by real time PCR using total cell DNA as templates.
Measurement of ROS Production Using 2Ј,7Ј-Dichlorodihydrofluorescein Diacetate-Mitochondrial ROS was assayed using 2Ј,7Ј-dichlorodihydrofluorescein diacetate oxidation (Molecular Probes, Eugene, OR) (48) as modified by Prabu et al. (49). The addition of 5 g of brain cytosolic fraction was used as the source of esterase for assaying mitochondrial fractions (49). Stable non-fluorescent 2Ј,7Ј-dichlorodihydrofluorescein diacetate (2.5 M in CH 3 OH) was used in 200 l of assay mixture containing 135 l of PBS, 20 g of mitochondrial protein, 5 g of brain cytosol, and 0.1 mM NADPH. Appropriate controls, such as membrane-permeable superoxide dismutase (from bovine erythrocytes; Sigma) and catalase (from bovine liver; Sigma), were added (at 100 and 500 units/ml, respectively) for ascertaining whether the fluorescence signal was due to H 2 O 2 . The fluorescence was recorded with an excitation wavelength of 485 nm and emission wavelength of 535 nm (for 20 min) in a MicroWin Chameleon multilabel detection platform.
Measurement of Extracellular H 2 O 2 by Amplex Red-H 2 O 2 in cells (hereafter referred to as ROS) was measured using an Amplex Red hydrogen peroxide/peroxidase assay kit from Invitrogen following the manufacturer's suggested protocol. The fluorescence was recorded with an excitation wavelength of 530 nm and emission wavelength of 590 nm in a MicroWin Chameleon multilabel detection platform.
Analysis of Cellular O 2 Consumption Using a Seahorse Bioscience Extracellular Flux Analyzer-Oxygen consumption rates (OCRs) were measured using an XF24 high sensitivity respirometer (Seahorse Bioscience) as described by Wu et al. (50). Cells were cultured in Dulbecco's modified Eagle's medium (DMEM) without or with arachidonic acid (70 M). Cells (30,000 each) were cultured in DMEM overnight, and the medium was changed to XF assay medium, low buffered bicarbonate-free DMEM (pH 7.4), for 1 h before the measurement. The final concentrations of inhibitors used were 2 g/ml oligomycin, 90 M 2,4-dinitrophenol (used as uncoupler), and 1 M rotenone (Complex I inhibitor). Each plate (along with the cartridge) was loaded into the XF analyzer. OCR was measured under basal conditions and after the sequential addition of oligomycin, 2,4-dinitrophenol, and rotenone. Respiration rates were measurements in triplicate wells.
Statistical Analysis-Means Ϯ S.E. were calculated from three to five independent experiments. Statistical significance (p values) between control and experimental or paired experiments was calculated using Student's t test. A p value of 0.05 was considered significant.

Localization of Full-length and Truncated CYP2C8 in Mitochondria and Microsomes of Human Liver
Samples-We analyzed mitochondria and microsomes from 26 human liver samples for CYP2C8 levels by immunoblot analysis. Representative blots for 12 samples in Fig. 1A show that in addition to fulllength CYP2C8 (apparent molecular mass of 54 kDa) there is a faster migrating (ϳ44-kDa) species in many samples. Some samples also showed additional antibody reactive species of low abundance that may be additional molecular forms or degradation products. The relative levels of both 54-and 44-kDa species in the mitochondrial and microsomal fractions also markedly vary, suggesting interindividual variations. For example, samples HL97 and HL3718 showed very low mitochondrial 54-kDa protein but high microsomal content. Samples HL9016, HL3904, HL3987, and HL3053 showed relatively high mitochondrial 54-kDa protein and nearly equal microsomal levels of protein (Fig. 1, A and quantification in B). The blots were co-developed with NADPH-cytochrome P450 reductase and TOM20 antibodies to evaluate the relative extent of cross-contamination of subcellular fractions used for analysis. The mitochondrial isolates from almost all human liver samples contain CYP2C8, ranging from 20 to 50% of the total tissue pool.
Immunoblots in Fig. 1A also show that the level of 44-kDa protein in the mitochondrial fraction was significant in many liver samples, whereas the level in the microsomal fraction was negligible in almost all samples. The ratios of mitochondrial 44-kDa species:54-kDa species ranged from 0.1 to 0.8 in different samples, indicating marked variability.
Alternatively Spliced mRNAs as a Source of Different CYP2C8 Molecular Species-At least three differentially spliced mRNAs for human CYP2C8 with open reading frames of 490, 420, and 388 amino acids have been reported (Ensembl Transcript CYP2C8-201 ENST00000535898 and GenBank accession number NM_001198853.1). We therefore reasoned that some of the faster migrating proteins on SDS gels (Fig. 1A) could be translated from alternatively spliced mRNAs. The two predicted mRNAs (Var_2 and Var_3) lack different stretches of nucleotide sequences from exons I and II and are therefore predicted to yield three different sized amplicons on PCR amplification using both forward and reverse primers from conserved regions as indicated under "Materials and Methods." The splice junctions of the full-length and Var_3 mRNAs are shown in Fig. 2B. Both Var_2 and Var_3 mRNAs encode N-terminal truncated proteins compared with the full-length 54-kDa (Var_1) protein (Fig. 2, A and B). Var_2 and Var_3 showed predicted reading frames of 420 and 388 amino acids, respectively, compared with the 490-amino acid reading frame for full-length Var_1 protein. It should be noted that 11 transcripts, including additional splice variants not chosen for analysis, are derived by differential splicing of the CYPC8 gene-encoded primary transcript; eight of these are likely non-functional, three of them undergo nonsense-mediated decay, and the remaining five contain part of the introns and encode no protein products.
The agarose gel patterns in Fig. 2C show the relative levels of full-length Var_1, Var_2, and Var_3 mRNAs. Two important points are noteworthy regarding the Var_2 amplicon. First, it is about 100 nucleotides larger than the full-length mRNA amplicon because within the mRNA region being amplified fulllength mRNA lacks a stretch of sequence that is present in Var_2 mRNA. Second, the relative levels of Var_2 mRNA were generally very low, consistent with the negligible to very low steady-state levels of corresponding protein. For these reasons, we did not pursue further characterization of this protein. The Var_3 mRNA amplicon was seen in most liver samples and represented a significant part of the total CYP2C8 mRNA pool (Fig. 2C, panels i and ii). The relative distribution of Var_3 and Var_1 amplicons showed marked variations in steady-state levels as indicated by the ratios of band intensities (values in parentheses below the gel patterns). Samples 9016, B019, HL97, HL3718, and 9438 showed equal or marginally higher ratios of Var_3:Var_1. Samples 9304, 500A, and 7017D showed higher levels of Var_1 and lower levels of Var_3. Other samples, such as 3987 and HL111 showed low levels of both Var_1 and Var_3 amplicons. The low levels of amplification with H111 RNA were surprising in view of the high levels of both 54-and 44-kDa proteins detected in this sample (Fig. 1A), although degraded RNA in the tissue may be the reason. These results show marked variations in the levels of differentially spliced Var_3 mRNA.
The possible molecular identity of Var_3 with the 44-kDa protein detected in the human liver samples was verified by immunoblot analysis of liver mitochondrial samples run alongside stable HepG2 cells expressing Var_3 protein. Two liver mitochondrial samples from each of the three panels in Fig. 1A were used for electrophoresis. The immunoblot in Fig. 2D shows that the 44-kDa protein co-migrates with the Var_3 protein expressed in HepG2 cells. Notably, both the liver and HepG2 cell mitochondria also show slower migrating nonspecific protein bands.
Three-dimensional Molecular Modeling and in Silico Analysis of Heme Binding Residues of Var_1 and Var_3 CYP2C8 Proteins-The next question we wanted to address was whether Var_3 protein has an intact heme binding pocket. Using the Blots were developed with polyclonal antibodies to CYP2C8 (1:500 dilution, v/v) and TOM20 (1:2,000 dilution, v/v) and monoclonal antibody to cytochrome P450 reductase (CPR) (1:1,500 dilution, v/v). In addition to the full-length CYP2C8, a smaller form of 44 kDa (Var_3 (V3)) was seen predominantly in the mitochondrial fraction. The numbers in parentheses below the CYP2C8 immunoblot represent the ratios of Var_3 and full-length (Var_1) proteins in terms of band intensities. B, relative distribution of full-length CYP2C8 in mitochondria and microsomes of the liver samples analyzed in A. The percent distribution was calculated based on the densitometry of band intensities in A. Results represent averages from two blots.
CYP2C8 and CYP2D6 as the template (33,36,51), we generated the three-dimensional model for CYP2C8 (WT2C8; full length) and Var_3 (V3-2C8) and superimposed the two images (Fig. 3A,  third panel). As seen from the superimposed image, the binding pocket and active site appear to be conserved in Var_3. However, because the F-G helixes are missing, the substrate entrance tunnel is altered, which might alter entry and subsequently specificity toward different substrates. Kelley 2 and 3, respectively). Conserved residues in all three sequences are marked with asterisks. The reported Var_2 and Var_3 lack the N-terminal stretch of amino acids from the full-length protein. B, schematic representation of differential splicing of pre-mRNA for the generation of full-length CYP2C8 (Var_1) and Var_3 mRNAs. Ex, exon; aa, amino acids. C, DNA amplicons generated by RT-PCR of total RNAs using the common 3Ј-and 5Ј-primers were resolved on a 2% agarose gel (w/v) and stained with ethidium bromide. WT CYP2C8, Var_3 (V3), and a slow migrating minor component, Var_2 (V2), are shown for the liver samples analyzed in Fig. 1A. M, DNA marker. Relative band intensities of Var_3 and Var_1 amplicons are presented as ratios in parentheses below the gel patterns. C, panels i and ii, immunoblot analysis of mitochondrial proteins (50 g each) from individual liver samples from Fig. 1A. Mitochondrial protein (Mito) from HepG2 cells expressing Var_3 cDNA was run alongside.
sequence. The authors conclude that family-specific signatures of the heme-binding region are more predictive as compared with the overall PROSITE pattern. For example, in CYP2D6, Arg-441 appears to be essential for its enzymatic function and heme binding. In consideration of this, because CYP2C also belongs to CYP2 family, we did a cluster sequence alignment of full-length CYP2D6, full-length (Var_1) CYP2C8, and CYP2C8 splice Var_3 to find the amino acids interacting with heme, and the results are summarized in Fig. 3C. Hydrogen-bonding interactions between the carboxyl groups of the two heme propionate groups and the side chains of Arg-97, Trp-120, Arg-124, His-368, Ser-429, and Arg-443 in Var_1 CYP2C8 (full length) and Trp-18, Arg-22, His-266, Ser-327, and Arg-341 of Var_3 CYP2C8 anchor the heme in the heme-binding site (Fig. 3B). The homology modeling of these proteins shows that the hemebinding Arg is well conserved in CYP2C8 with Arg at position 443 in wild type and position 341 in Var_3. Furthermore, all of the CYP proteins in CYTOCHROME_P450, PS00086 have a consensus pattern: (F/W)(S/G/N/H)X(G/D)(F)(R/K/H/P/ T)(P)C(L/I/V/M/F/A/P)(G/A/D) where C is the heme iron ligand. This signature pattern is conserved in both WT CYP2C8 (428 -437; FSaGKRICAG) and splice Var_3 (326 -335; FSaGKRICAG). These results suggested that despite the absence of the N-terminal Arg-97 residue the Var_3 protein has the capability to bind the heme group. However, it is apparent that the lack of the F-G helical region creates a very wide substrate binding pocket, raising questions about its ability to bind large substrates.

Mitochondrial Targeting of Full-length and Variant CYP2C8 Proteins Under in Vitro and in Cell
Conditions-The mitochondrial targeting efficiency of the WT, Var_2, and Var_3 proteins was studied with in vitro import by isolated rat liver mitochondria as described previously (24,40). The fluorogram presented in Fig. 4A, panel i, shows that all three molecular forms were imported into a trypsin-resistant compartment with Var_2 imported at the lowest efficiency (ϳ7% of input) and Var_3 at . Heme is colored in red, and felodipine is colored blue. B, amino acid residues interacting with the heme (in red) through hydrogen-bonding interactions in WT CYP2C8 (at left) and Var_3 (at right). C, list of amino acid residues (one-letter notation) involved in anchoring the heme in the two molecular forms.

JOURNAL OF BIOLOGICAL CHEMISTRY 29621
the highest efficiency (ϳ34% of input). In all three cases, the trypsin-resistant, putative imported molecular species co-migrated with the input protein on SDS gels, suggesting that the imported protein is not processed by mitochondrial matrix proteases. A control experiment showed that dihydrofolate reductase, a cytoplasmic protein, is not imported significantly, but SU9-dihydrofolate reductase protein (with an N-terminally fused mitochondrial targeting signal from ATPase subunit 9 (52)) was imported efficiently (Fig. 4A, panel ii). As expected, the N-terminal cleavable SU9 targeting signal was clipped as part of the import process.
Subcellular targeting of the three proteins was studied by three different approaches. In the first approach, we used the MitoProt II-v1.101 program to predict signal efficiencies. The Var_3 protein showed a mitochondrial export probability of 0.86, whereas the full-length Var_1 protein yielded a score of 0.39, and the variant 2 gave a score of 0.013 (Table 1). In the second approach, mitochondrial targeting was studied by transient transfection of COS-7 cells with the WT and variant cDNAs cloned in pCMV6-AC-GFP vector followed by subcellular fractionation. The immunoblot in Fig. 4B, panel i, shows the relative levels of expression of three cDNAs in COS-7 cells. Var_3 was expressed at a significantly higher level than the WT and Var_2 cDNAs. The mitochondrial (Mt) and microsomal (Mc) distributions of Var_1, Var_2, and Var_3 proteins in transfected cells are shown in Fig. 4B, panel ii. Both Var_1 and Var_2 proteins were targeted at higher levels to the microsomes than to mitochondria, whereas Var_3 protein was targeted at a substantially higher level to mitochondria. All three mitochondrion-associated proteins were relatively resistant to trypsin but became sensitive to trypsin following disruption of the mitochondrial membrane by treatment with Triton X-100 (Fig. 4B,   panel iii, TT), suggesting that they are localized in the mitochondrial matrix compartment. Nearly 60 -80% of mitochondrion-associated Var_1 and Var_3 proteins were resistant to trypsin, whereas only 5-10% of Var_2 protein was resistant to trypsin. A noteworthy point is that Var_3 protein appears as a doublet (Fig. 4B, panels i, ii, and iii) on gels. Although the precise reason remains unclear, it is likely due either to translation starting from an alternate position or to posttranslational modifications.
In the third approach, the level of mitochondrial targeting in COS-7 cells was further ascertained by confocal immunofluorescence microscopy of transfected cells. Transfected cells were stained with CYP2C8 antibody and co-stained with either an antibody to a mitochondrial marker (Fig. 4C, panel i, cytochrome-c oxidase subunit I (CcOI)) or a microsomal marker (Fig. 4C, panel ii, calreticulin (CRT)), and the CYP staining patterns were superimposed on the cytochrome-c oxidase subunit I-or calreticulin-stained patterns. WT CYP2C8 co-localized with the mitochondrial marker with a marginal Pearson coefficient of 0.61. Var_2 protein co-localized insignificantly with the mitochondrial marker (Pearson coefficient of 0.4), whereas Var_3 protein co-localized with the highest Pearson coefficient of 0.9. Co-localization with the microsomal marker (calreticulin) was in the reverse order with WT and variant 2 exhibiting high efficiencies (Pearson coefficients of 0.82 and 0.9, respectively) and Var_3 protein exhibiting a very low efficiency (Pearson coefficient of 0.55). These results together suggest that although all three proteins localize to mitochondria and that there is a marked difference in relative efficiency with Var_3 showing the highest level, WT protein showing an intermediate level, and Var_2 protein showing the lowest level. This is consistent with our previous predictions on bimodal targeting of CYP1A1, -2E1, and -2D6 proteins to mitochondria that proteins with more hydrophilic N-terminal targeting domains with lower affinity of binding to the signal recognition particle are targeted to mitochondria at higher levels (28 -32).
Reconstitution of Purified Full-length CYP2C8 with Microsomal and Mitochondrial Electron Transfer Proteins-The enzyme activity of purified CYP2C8 (Var_1) could be reconstituted with the mitochondrial electron transfer proteins Adx and AdxR using a large substrate, paclitaxel (854 Da), and a smaller substrate, dibenzylfluorescein (189 Da). Paclitaxel  35 S-labeled translation products of wild type (WT2C8), Var_2 (V2-2C8), and Var_3 (V3-2C8). Radiometric imaging of gels was performed to determine the level of import of input protein for each construct in trypsin-treated samples (T). The input protein level was considered to be 100% in each case. Panel ii, import of dihydrofolate reductase (DHFR) and SU9-dihydrofolate reductase (Su9-DHFR) proteins as negative and positive controls, respectively. The lanes marked "In" or "I" (for input) were loaded with 20% of the counts used for the import reactions. "C" represents control experiments in which total protein bound and imported into mitochondria is present, "T" represents trypsin-treated mitochondria in which only the protein imported into mitochondria is present. B, translocation of WT and variant 2 and 3 proteins in transiently transfected COS-7 cells. Panel i, total cell lysates (50 g each) from transiently transfected COS-7 cells were resolved by 12% SDS (w/v)-PAGE and probed with antibodies to CYP2C8 and ␤-actin for assessing loading levels. Panel ii, mitochondria and microsomes were isolated from transfected COS-7 cells, and 50 g of protein each was resolved by 12% SDS (w/v)-PAGE and probed with antibodies to CYP2C8, cytochrome P450 reductase (CPR), and cytochrome-c oxidase subunit I. Panel iii, relative resistance of mitochondrion-associated proteins to trypsin treatment (T). In some cases, mitochondria were lysed by treatment with 1% Triton X-100 (v/v) before trypsin treatment (TT). Proteins (50 g each) were resolved by SDS-PAGE and probed with antibodies to CYP2C8 and TOM20 for immunoblot analysis. C, immunofluorescence microcopy of COS-7 cells transfected with WT (Var_1), Var_2, or Var_3 cDNA. Panel i, a-c, co-localization of CYP2C8 with a mitochondrial marker, cytochrome-c oxidase subunit I (CcOI). Cells were stained with a 1:1,000 dilution (v/v) of primary anti-goat antibody to CYP2C8 (green) (Abcam, Cambridge, MA) and co-stained with a 1:500 dilution (v/v) of cytochrome-c oxidase subunit I (red) (anti-mouse) antibody (Abcam). Panel ii, a-c, co-localization of CYP2C8 with microsomal membrane marker calreticulin. Cells were stained with CYP2C8 (green) as above and co-stained with a 1:500 dilution (v/v) of calreticulin (CRT) (anti-rabbit) antibody (red) (Santa Cruz Biotechnology, Santa Cruz, CA). The cells were subsequently incubated with secondary Alexa Fluor 546-conjugated anti-mouse and then anti-rabbit IgG and Alexa Fluor 488-conjugated anti-goat IgG and imaged through a confocal microscope. Numbers in the bottom panels indicate Pearson coefficients for coincidence calculated using Volocity 5.3 software.

TABLE 1 The mitochondrial targeting efficiency of N-terminal signals of the full-length (Var_1; WT), Var_2 (V*2), and Var_3 (V*3) proteins
The mitochondrial targeting efficiency of the three proteins was analyzed using the MitoProt II-v1.101 program. aa, amino acids.

Protein sequence length
Probability of export to mitochondria aa CYP2C8 WT 490 0.39 CYP2C8 V*2 420 0.014 CYP2C8 V*3 388 0.87 6-hydroxylation activity was relatively higher with Adx plus AdxR electron transfer proteins, whereas dibenzylfluorescein oxidation was lower than the activity reconstituted with cytochrome P450 reductase (Fig. 5, A and B). Montelukast, a specific inhibitor of CYP2C8, and an inhibitory antibody raised against CYP2C8 inhibited the activities with both substrates. These results clearly show that, similar to results reported for other Family 2 CYPs (CYP2B1, CYP2D6, and CYP2E1) (26, 28 -30, 32, 34), human CYP2C8 can accept electrons from the soluble Adx and AdxR electron transfer system.

Characterization of HepG2 Cell Lines Stably Expressing Human Full-length and Var_3
CYP2C8 Proteins-HepG2 cell lines stably expressing full-length (Var_1) and Var_3 cDNAs were generated using a lentiviral vector. Immunoblots (Fig. 6A) show the mitochondrial and microsomal protein levels in mock-transfected, WT CYP2C8, and CYP2C8 Var_3 cells. The microsomal protein content in Var_1 (WT) CYP2C8 cells was significantly higher than the mitochondrial content. In cells expressing CYP2C8 Var_3, however, the mitochondrial content was more than twice the microsomal level. Notably, in contrast to the overall level of expression of the two cDNA constructs in transiently transfected cells (Fig. 3), the steadystate level of Var_3 protein in the stable cell line was significantly lower. The reason for this difference remains unclear. The membrane topology of proteins associated with the mitochondrial fraction was evaluated using the alkaline Na 2 CO 3 extraction method (26,49). Both full-length and CYP2C8 Var_3 in the mitochondrial fraction were nearly quantitatively extracted with the alkaline aqueous phase, whereas the microsomal WT CYP2C8 protein partitioned more into the insoluble fraction, suggesting a transmembrane topology. A small part of the microsomal Var_3 protein partitioned in the insoluble phase, suggesting that the bulk of the ER-associated Var_3 protein was associated mainly through membrane-extrinsic interaction. The PCR results in Fig. 6C show that all three cell lines contained nearly the same levels of integrated puromycin acetyltransferase gene, suggesting comparable levels of viral DNA integration.
The mitochondrial and microsomal heme contents of the stable cell lines were measured by ferrous CO difference spectra (Fig. 6, D and E). The WT CYP2C8 cell line showed a higher CYP content (ϳ900 pmol/mg of protein) than the mitochondrial fraction (ϳ700 pmol/mg of protein). The Var_3-expressing cells showed about 500 pmol/mg of protein, whereas the microsomal fraction from these cells showed Ͻ300 pmol/mg of protein. The mock-transfected cells showed very low CYP contents. The overall CYP contents of cell fractions from these cells are consistent with the relative CYP2C8 antibody-reactive proteins detected in these fractions (Fig. 6A). Both the mitochondrial and microsomal fractions from WT CYP2C8 cells showed comparable activity for paclitaxel 6-hydroxylation, which was inhibited by the CYP2C8-specific inhibitor montelukast and an inhibitory antibody (Fig. 6F). However, despite containing measurable CYP heme, mitochondria from CYP2C8 Var_3-expressing cells did not show any detectable paclitaxel 6-hydroxylase activity. We therefore decided to test the effects of the CYP2C8 inhibitors proadifen and montelukast on mitochondrial ROS production using a 2Ј,7Ј-dichlorodihydrofluorescein diacetate method. Both inhibitors and N-acetylcysteine, an antioxidant, marginally inhibited ROS production in isolated mitochondria from mock-transfected cells but significantly inhibited ROS production in mitochondria from WT CYP2C8and CYP2C8 Var_3-expressing cells. The overall ROS production was substantially higher in the CYP2C8 Var_3-expressing cells. These results suggested that mitochondrial CYP2C8  (5 M) and inhibitory antibody to CYP2C8 (2C8Ab) (10 mg/ml) were used. Control ascites fluid (CAF; 10 mg/ml) was used as a negative control. B, reconstitution of dibenzylfluorescein oxidation was carried out essentially as described above in A. The activities in all cases represent the means Ϯ S.E. (error bars) of three to five separate assays. Purified CYP2C8 was preincubated with inhibitors and control ascites fluid as described under "Materials and Methods." ࡗ in A indicates no detectable activity.
Var_3 may be catalytically active and may accept electrons from Adx/AdxR. We next tested the metabolic activity of WT CYP2C8 and CYP2C8 Var_3 with a smaller substrate, arachidonic acid, which belongs to a group of polyunsaturated fatty acids that are present in the phospholipids of membranes and are abundant in brain, muscle, and liver cells (14 -16). Once released from the phospholipids, arachidonic acid is converted to endogenous bioactive eicosanoids by cyclooxygenase, lipoxygenases, and microsomal CYP2C8, CYP2C9, CYPJ2, and CYP4F2 (14,(17)(18)(19)(20). We identified three different EETs (8,11,and 14, and 20-OH (20-HETE) in reactions catalyzed by mitochondrial WT CYP2C8 and CYP2C8 Var_3 (Fig. 7, A-D). The mitochondrial WT CYP2C8 in the presence of Adx and AdxR electron transfer proteins showed higher levels of 11,12and 14,15-EETs than the corresponding microsomal activity driven by cytochrome P450 reductase. The mitochondrial CYP2C8 Var_3 protein, however, produced significantly lower levels of both products (Fig. 7, A and B). Notably, mitochondrial CYP2C8 Var_3 yielded a higher level of 8,9-EET and nearly a 2-fold higher level of 20-HETE than mitochondrial WT CYP2C8. The level of 20-HETE produced by mitochondrial CYP2C8 Var_3 was also more than 2-fold higher than that produced by the microsomal WT CYP2C8 reconstituted with NADPH-cytochrome P450 reductase. These results not only suggest that mitochondrial Var_3 CYP2C8 is metabolically active but also that it may be responsible for producing large amounts of 20-HETE, which is implicated to have several physiological effects (53). In all cases, montelukast inhibited the activity by 50 -80%.
The catalytic activity of mitochondrially targeted Var_3 was further ascertained by reconstituting the arachidonic acid metabolism in the presence and absence of added Adx ϩ AdxR proteins and using the CYP2C8-specific inhibitor montelukast. Fig. 7E shows that the total EET activity of Var_3-expressing mitochondria was dependent on the addition of Adx and AdxR and was inhibited by montelukast. Bacterially expressed purified Var_1 CYP2C8 also showed a high level of activity for arachidonic acid metabolism when reconstituted with Adx ϩ AdxR (Fig. 7E).
Roles of Var_1 (WT) CYP2C8 and CYP2C8 Var_3 in Arachidonic Acid-induced Oxidative Stress-We used a Seahorse Bioscience metabolic flux analyzer to study the mitochondrial respiratory parameters in mock-transfected and Var_1 CYP2C8-and Var_3 CYP2C8-expressing cells. Stable cells expressing both Var_1 CYP2C8 and Var_3 proteins showed significantly lower basal respiration, maximum uncoupled respiration, and ATP-dependent respiration (Fig. 8A, panels i-iii). Treatment with 70 M arachidonic acid had a marginal effect on all three respiratory parameters in WT CYP2C8 cells. However, arachidonic acid induced a profound 80 -90% inhibition of basal, maximum uncoupled, and ATP-coupled respiration in Var_3-expressing cells. It is possible that the metabolic activity of mitochondrial Var_3 protein contributes to oxidative stress. In support of this, measurement of H 2 O 2 production by the Amplex Red method (Fig. 8B) showed that Var_3-expressing cells produced the highest level of H 2 O 2 in the presence of arachidonic acid. These results together support a possible role of CYP2C8 Var_3 in inducing oxidative stress.

DISCUSSION
CYP2C8, which is expressed in hepatic and many extrahepatic tissues, is involved in the metabolism of an array of drugs used in human medicine in addition to a proposed contribution to oxidative stress during cardiac ischemia (14 -19). A number of different point mutations have been reported that affect either the heme binding ability or substrate binding (7)(8)(9)(10)(11)(12)(13).
Here we report the molecular and functional characterization of some of the previously reported but not biochemically characterized splice variants of CYP2C8. Splice Var_3 (1n⌬1a⌬2a) was shown to be predominately targeted to mitochondria and shows markedly altered catalytic activity for different substrates. This appears to be the first study showing different subcellular distributions of the WT form and the N-terminal truncated Var_3 of CYP2C8 and distinctive metabolic activities of the two molecular forms of the enzyme. Analysis of human liver samples revealed that although the WT (full-length) CYP2C8 is the most abundant form expressed in the liver a faster migrating, putative 44-kDa form, characterized as Var_3, is expressed at markedly variable levels in almost all liver samples. The Var_2 form with a reading frame of 420 amino acids was detected only in some livers and at very low levels. Because of its low abundance, we did not characterize this form in terms of its catalytic activity.
The evidence for mitochondrial localization of WT and Var_3 CYP2C8 comes from multiple set of experiments. First, the in vitro import of translation products into isolated mitochondria (Fig. 4) in which resistance to limited digestion with trypsin but sensitivity to Triton X-100 plus trypsin is regarded as evidence for the matrix side localization of the proteins. With this approach, we observed that ϳ10% of input WT CYP2C8 was imported, whereas nearly 34% of Var_3protein was imported. The second approach involved transient transfection of COS-7 cells with cDNA constructs and quantification of mitochondrial and microsomal targeting of proteins by subcellular fractionation under minimal cross-contaminating condi- The blots were also probed with antibodies to TOM20 (bottom panel) as a mitochondrion-specific marker and NADPH-cytochrome P450 reductase (top panel) as a microsome-specific marker. Std, standard. B, membrane-extrinsic and -intrinsic nature of wild-type CYP2C8 (WT 2C8) and Var_3 (V3 2C8) in the mitochondrial (Mito) and microsomal (Micro) fractions was analyzed by the alkaline Na 2 CO 3 extraction method described under "Materials and Methods." The soluble (E) and insoluble (P) protein fractions were subjected to SDS-PAGE separation and probed with CYP2C8 antibody (2C8Ab) by immunoblot analysis. C, the relative levels of viral vector DNA (puromycin acetyltransferase gene) were determined by real time PCR using total cell DNA as template and the actin gene as an internal reference. D, spectrophotometric scans of CYP heme in the mitochondrial and microsomal fractions obtained from HepG2 stable cells expressing WT and Var_3 proteins. Fe 2ϩ -CO versus Fe 2ϩ spectra were recorded as described under "Materials and Methods." Abs, absorbance. E, relative CYP contents of mitochondria and microsomes from mock-, wild-type CYP2C8-, and CYP2C8 Var_3expressing HepG2 cells. CYP content was measured by CO difference spectra as described under "Materials and Methods." F, paclitaxel 6-hydroyxlation activity reconstituted with mitochondria and microsomes from stable HepG2 cells expressing WT CYP2C8 and Var_3 CYP2C8 and mock-transfected cells. Assays were carried out as described under "Materials and Methods." G, ROS production in isolated mitochondria from stable HepG2 cell lines with or without treatment with the antioxidant N-acetylcysteine (NAC) or the inhibitors proadifen and montelukast. Mitochondria (50 g each) were seeded in 96-well plates for ROS measurements using the 2Ј,7Ј-dichlorodihydrofluorescein (DCF) diacetate method as described under "Materials and Methods." Results represent means Ϯ S.E. (error bars) of three to four separate assays. * indicates a p value Ͻ0.05, and ** represents a p value Ͻ0.001. ࡗ in F indicates no detectable activity. CPR, cytochrome P450 reductase.
tions followed by immunoblot analysis and by immunofluorescence co-localization of CYP2C8 with mitochondrial and microsomal markers. By all these criteria, we show that WT CYP2C8 is preferentially targeted to the ER, and a small frac-tion (Ͻ25%) is targeted to mitochondria. Conversely, the Var_3 protein is predominately targeted to the mitochondria and minimally associated with the ER. The ER association of this N-terminal truncated form, which lacks the transmembrane FIGURE 7. Oxidation of arachidonic acid by mitochondria and microsomes isolated from stable HepG2 cells expressing Var_1 (WT) CYP2C8 and Var_3 proteins. A-D, the mitochondrial (Mt) and microsomal (MIC) proteins (300 -500 g) were assayed for arachidonic acid metabolism as described under "Materials and Methods" using 70 M arachidonic acid as substrate. Inhibition studies were performed by preincubating enzymes with 5 M montelukast at 37°C for 20 min. Reactions were initiated by the addition of 1 mM NADPH and continued for 5 min at 37°C in a shaking water bath, and the metabolites were extracted and analyzed as described under "Materials and Methods." Four major arachidonate products (11,14,8,EET, and 20-HETE) were quantified as described under "Materials and Methods." E, mitochondrial (Mito) proteins from Var_3 (V3)-expressing cells were reconstituted with or without added Adx/AdxR or added montelukast using arachidonic acid as substrate. In one case, purified Var_1 (V1) CYP2C8 was reconstituted with Adx/AdxR as described in Fig. 5. The total EET metabolites were quantified using the LC-MS method to ascertain the dependence of the enzyme on Adx/AdxR. The results represent means Ϯ S.E. (error bars) of three independent assays. * indicates p Ͻ 0.05, and ** indicates p Ͻ 0.001. CPR, cytochrome P450 reductase. domain, appears to be extrinsic, and other results (not presented) show that ER-associated CYP2C8 Var_3 is not catalytically active. As shown previously (27), the transmembrane organization of CYP with the N-terminal anchoring domain is important for interaction of the catalytic domain with the similarly anchored cytochrome P450 reductase.
A three-dimensional molecular model of the Var_3 protein superimposed on a CYP2C8 model (Fig. 3) shows that both the heme binding pocket and the substrate (felodipine) binding pocket are nearly intact in the truncated form. A molecular model of CYP2C8 based on x-ray crystal structure coordinates of CYP2D6, a member of the CYP2 family, and CYP2C8 (33,36,38,51) shows that six Arg, His, and Tyr residues conserved in both proteins are involved in heme binding. Of these, the N-terminal Arg-97 (homologous to Arg-101 in CYP2D6) is lacking in the Var_3 protein. Despite this, the energy minimization model (Fig. 3B) suggests that the N-terminal truncated Var_3 protein is capable of binding to the heme moiety. This was further confirmed by results showing that mitochondria from stable cells expressing CYP2C8 Var_3 exhibit the characteristic CYP heme spectrum as determined by CO absorbance spectroscopy. Conversely, mitochondria from mock vector-transduced cells showed no detectable heme content. These results combined with the catalytic assays provide evidence for the heme binding ability of CYP2C8 Var_3.
Similar to CYP3A4, the substrate pocket of CYP2C8 is large but has more of a long channel rather than an open space (33,51). This structural attribute enables both the enzymes to bind to large substrates and inhibitors, such as paclitaxel (834 Da), erythromycin (734 Da), montelukast (586 Da), etc. The N-terminal sequence stretch that forms the G and F helices is positioned over the large substrate pocket (33) and may thus facilitate the stabilization of large substrates bound on top of the enzyme. The Var_3 enzyme, which lacks the N-terminal portion of the molecule, exhibited no activity with the large substrate paclitaxel. However, the enzyme was highly active with smaller substrates, such as arachidonic acid and dibenzylfluorescein (data not shown for the latter). It is likely that the N-terminal portion of the protein forming the F and G helices is critical for the stable association of large substrates with the enzyme. Notably, the catalytic activity of Var_3 was inhibited by montelukast and proadifen, suggesting that its large substrate binding pocket is intact. These two inhibitors and a CYP2C8-specific inhibitory antibody inhibited ROS production by mitochondria from CYP2C8 Var_3-expressing cells, suggesting a possible role in the metabolism of some physiolog- ical substrate(s). These results together suggest a difference in substrate preference between the intact WT CYP2C8 and its N-terminal truncated Var_3.
Several studies have shown that CYP2C8 is expressed in the endothelial cells as well in the myocardium (16 -20) and suggest its role in endothelial cell and myocardial toxicity. Arachidonic acid is a polyunsaturated -6 fatty acid produced in mammalian tissues and absorbed through dietary sources. Once released from the phospholipids by phospholipases, arachidonic acid is converted to endogenous bioactive eicosanoids by three different classes of enzymes, including CYPs belonging to CYP2C, CYP2J, and CYP4F families (11)(12)(13)(14)(15)(16). The four EET regioisomers formed by epoxygenation reactions are 5,6-, 8,9-, 11,12-, and 14 -15-EETs. Hydroxylation activity of CYPs also produces HETEs, mainly 19-HETE and 20-HETE, the latter of which is biologically active. Several in vitro studies implicate a role for 20-HETE in cell proliferation (54,55), angiogenesis (56,57), and oxidative stress and inflammation (58), although the in vivo pathophysiological effects of this compound remain unclear. Prostaglandin E 2 , a metabolite of 20-HETE, has been shown to induce mature inflamed adipocyte hypertrophy in mesenchymal stem cells undergoing differentiation (51). In extension of these studies, we now show that mostly mitochondrially targeted Var_3 produced 2-fold more 20-HETE than WT CYP2C8 associated with the microsomal fraction of cells. Furthermore, arachidonic acid induced a high level of ROS production and mitochondrial respiratory dysfunction in stable cells expressing Var_3 protein (Figs. 7 and 8). These results further support the possibility that subcellular localization of various drug-metabolizing CYPs can have important pathophysiological significance.
Analysis of human liver samples showed widely varying levels of CYP2C8 Var_3 mRNA and proteins in different individuals. The observed variation in the level of CYP2C8 Var_3 is important because of its propensity to generate high levels of biologically active 20-HETE in different tissues. Currently, many aspects of the regulation of CYP2C8 gene expression and the regulation of differential splicing remain unclear. However, this study demonstrates the importance of mitochondrially targeted CYPs and their roles in drug metabolism and toxicity.