A novel role for fatty acid transport protein 1 in the regulation of tricarboxylic acid cycle and mitochondrial function in 3T3-L1 adipocytes.

Fatty acid transport proteins (FATPs) are integral membrane acyl-CoA synthetases implicated in adipocyte fatty acid influx and esterification. Whereas some FATP1 translocates to the plasma membrane in response to insulin, the majority of FATP1 remains within intracellular structures and bioinformatic and immunofluorescence analysis of FATP1 suggests the protein primarily resides in the mitochondrion. To evaluate potential roles for FATP1 in mitochondrial metabolism, we used a proteomic approach following immunoprecipitation of endogenous FATP1 from 3T3-L1 adipocytes and identified mitochondrial 2-oxoglutarate dehydrogenase. To assess the functional consequence of the interaction, purified FATP1 was reconstituted into phospholipid-containing vesicles and its effect on 2-oxoglutarate dehydrogenase activity evaluated. FATP1 enhanced the activity of 2-oxoglutarate dehydrogenase independently of its acyl-CoA synthetase activity whereas silencing of FATP1 in 3T3-L1 adipocytes resulted in decreased activity of 2-oxoglutarate dehydrogenase. FATP1 silenced 3T3-L1 adipocytes exhibited decreased tricarboxylic acid cycle activity, increased cellular NAD+/NADH, increased fatty acid oxidation, and increased lactate production indicative of altered mitochondrial energy metabolism. These results reveal a novel role for FATP1 as a regulator of tricarboxylic acid cycle activity and mitochondrial function.

Interestingly, the majority of FATP1 remains on intracellular structures even under insulin-stimulated conditions ( 3,5 ), leading to the possibility of additional roles for FATP1. In this report, we used a proteomic approach to Abstract Fatty acid transport proteins (FATPs) are integral membrane acyl-CoA synthetases implicated in adipocyte fatty acid infl ux and esterifi cation. Whereas some FATP1 translocates to the plasma membrane in response to insulin, the majority of FATP1 remains within intracellular structures and bioinformatic and immunofl uorescence analysis of FATP1 suggests the protein primarily resides in the mitochondrion. To evaluate potential roles for FATP1 in mitochondrial metabolism, we used a proteomic approach following immunoprecipitation of endogenous FATP1 from 3T3-L1 adipocytes and identifi ed mitochondrial 2-oxoglutarate dehydrogenase. To assess the functional consequence of the interaction, purifi ed FATP1 was reconstituted into phospholipid-containing vesicles and its effect on 2-oxoglutarate dehydrogenase activity evaluated. FATP1 enhanced the activity of 2-oxoglutarate dehydrogenase independently of its acyl-CoA synthetase activity whereas silencing of FATP1 in 3T3-L1 adipocytes resulted in decreased activity of 2-oxoglutarate dehydrogenase. FATP1 silenced 3T3-L1 adipocytes exhibited decreased tricarboxylic acid cycle activity, increased cellular NAD + /NADH, increased fatty acid oxidation, and increased lactate production indicative of altered mitochondrial energy metabolism. These results reveal a novel role for FATP1 as a regulator of tricarboxylic acid cycle activity and mitochondrial function. from the gel and applied to a C18 Zip-Tip (Millipore), washed and eluted in 80% acetonitrile/0.1% trifl uoroacetic acid . Recovered peptides were subjected to lC-MS/MS analysis on an LTQ-Orbitrap mass spectrometer. MS/MS spectra were searched using SEQUEST (version 27, rev. 12, ThermoFinnigan, San Jose, CA) against a nonredundant mouse proteome sequence database assuming a maximum of two missed tryptic cleavages per peptide and the results were validated and organized using Scaffold (version Scaffold_2_02_00, Proteome Software Inc., Portland, OR). Reported peptide sequence matches met the following criteria: 1 ) all peptides were multiply charged, 2 ) the parent mass tolerance was 7 ppm, 3 ) two tryptic termini per peptide, and 4 ) a minimum of two unique peptides per match with a minimum peptide probability of 70% as specifi ed by the Peptide Prophet algorithm ( 25 ). Peptide sequence matches were considered positive interactions if enriched a minimum of 60-fold over the negative control based on normalized spectrum counts.

Immunofl uorescence and confocal microscopy
Immunofl uorescence analysis of 3T3-L1 adipocytes was performed as described ( 3 ) with minor modifi cations. 3T3-L1 adipocytes grown on glass cover slips were incubated with 100 nM Mitotracker Red CMXRos (Invitrogen) for 20 min at 37°C and fi xed in prewarmed 3.7% formaldehyde for 15 min at room temperature. Cells were permeabilized with ice-cold 75% methanol in water for 10 min at Ϫ 20°C, washed in PBS, and treated with Image IT-FX enhancer (Invitrogen) for 30 min at room temperature. The cover slips were washed and incubated in blocking buffer containing 0.3% Triton X-100, then stained with the rabbit anti-FATP1 antibody and Alexa Fluor 488-conjugated goat antirabbit IgG sequentially, each diluted in PBS and 0.3% Triton X-100, and incubated for 1 h at room temperature. Nuclei were stained with 0.2 g/ml DAPI in PBS for 1 min and viewed using an Olympus FluoView FV1000 inverted confocal microscope. The images were acquired and analyzed using FluoView software.

Purifi cation and reconstitution of purifi ed FATP into SUVs
Purifi cation of murine FATP1-Myc/His was performed as previously described ( 8,10 ). Purifi ed FATP1 was buffer exchanged into buffer A [100 mM Tris, pH 7.5, 150 mM NaCl, 20% (v/v) glycerol] using a desalting column (Zeba Desalt Spin Columns, Thermo Scientifi c). For reconstitution, lipids dissolved in chloroform were mixed, dried under nitrogen, and buffer A was added for 1 h at room temperature. The lipid solution was vortexed briefl y and sonicated until the solution changed from milky to clear. The resulting small unilamellar vesicles (SUVs) were stored at 4°C until used. Purifi ed FATP1 was reconstituted into SUVs at a protein-lipid molar ratio of 1:1000. SUVs were detergent destabilized at room temperature via the addition of dodecylmaltoside to a fi nal concentration of 3.8 mM (0.195% w/v) that led to the onset of vesicle solubilization ( 26 ). The dodecylmaltoside concentration of purifi ed FATP was adjusted to 0.195%, added to the lipid-detergent solution, and allowed to incubate with gentle stirring for 10 min at 4°C. Excess dodecylmaltoside was removed by adding CALBIOSORB adsorbent resin (Calbiochem) equilibrated in buffer A. The resin was incubated with the proteinlipid-detergent solution at 4°C with gentle stirring and after 90 min, additional resin was added to remove residual detergent. The FATP1 proteoliposomes were kept on ice and used immediately.
To analyze the orientation of reconstituted FATP1, detergentpurifi ed or reconstituted FATP samples were incubated in the presence of trypsin (Promega) (1:50 trypsin:protein) for various lengths of time and the reaction stopped by boiling in the presence of reducing and denaturing SDS-PAGE loading buffer con-identify novel FATP1-interacting proteins in 3T3-L1 adipocytes and identifi ed mitochondrial 2-oxoglutarate dehydrogenase (OGDH), a kinetically rate-limiting step in the tricarboxylic acid cycle, as a binding partner. Using a reconstituted FATP1 proteoliposome system, we fi nd that FATP1 enhances the activity of OGDH in vitro that is independent of FATP1 acyl-CoA synthetase activity. Consistent with this, OGDH activity is decreased in 3T3-L1 adipocytes stably expressing short hairpin (sh)RNA directed at FATP1. These observations are coincident with altered mitochondrial tricarboxylic acid cycle metabolism in the FATP1 knockdown adipocytes, indicating that FATP1 is a novel regulator of mitochondrial function.

Reagents and cell culture
Cell culture reagents were obtained from Invitrogen. Cell culture-grade porcine insulin, puromycin, methylisobutylxanthine, and dexamethasone were obtained from Sigma-Aldrich. n -dodecyl-␤ -D-maltoside was obtained from MBL International Corp. Dioleoylphosphatidylcholine (DOPC), dioleoylphosphatidylethanolamine (DOPE), dioleoylphosphatidyl-L-serine (DOPS), and cholesterol were obtained from Avanti Polar Lipids, Inc. Alexa Fluor 488-conjugated goat anti-rabbit IgG was obtained from Invitrogen and IRDye 700-conjugated goat anti-rabbit IgG was obtained from Li-Cor Biosciences. All other reagents were of analytical grade and obtained from Sigma-Aldrich. Differentiation and maintenance of 3T3-L1 cell lines expressing either a shRNA targeting FATP1 or a scrambled sequence was previously described ( 3 ).

3T3-L1 adipocyte cross-linking, immunoprecipitation, and proteomic analysis
Formaldehyde cross-linking was performed as described ( 24 ), with modifi cations. Briefl y, day 8 3T3-L1 adipocytes were crosslinked with 0.5% formaldehyde for 10 min at room temperature and quenched with 1.5-fold molar excess Tris-HCL pH 7.5. Monolayers were lysed into RIPA buffer (50 mM Tris, pH 7.5, 100 mM NaCl, 50 mM sodium fl uoride, 5 mM sodium pyrophosphate, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS) containing protease and phosphatase inhibitors, sonicated and centrifuged at 13,000 g for 10 min at 4°C to prepare a detergent soluble extract. FATP1 was immunoprecipitated overnight at 4°C using rabbit anti-FATP1 with rabbit preimmune IgG used as a negative control at the same concentration; 20 g/ml. Immune complex was incubated with protein A-agarose beads for 1 h and washed in RIPA buffer. The immune complex beads were transferred to new tubes and boiled in SDS-loading buffer for 20 min to reverse the cross-linking. The supernatant was subjected to SDS-PAGE and stained for total protein using Sypro Ruby gel stain (Millipore).
Proteomic analysis was applied to either isolated gel bands or the entire sample lane. To analyze the sample lane, the gel segments containing the immunoglobulin heavy chain and light chain bands were excised and discarded. The remaining lane fragments were divided into 10 fractions, reduced, and cysteine residues alkylated with iodoacetic acid. Samples were subjected to trypsin digestion for 24 h at 37°C and peptides were extracted containing 50 M L-carnitine overnight and isolated mitochondria were added to 1 ml pyruvate oxidation buffer (1 mM [2-14 C] pyruvate (0.5 Ci/ mole pyruvate), 20 mM Tris, pH 7.4, 100 mM sucrose, 10 mM potassium phosphate, 100 mM KCl, 1 mM MgCl 2 , 2 mM ADP, 0.1 mM CoA, and 1 mM DTT) and incubated for 30 min at 37°C. Reactions were terminated with injection of 200 l 9 M sulfuric acid and 14 CO 2 was assessed as described previously.

Metabolite quantitation
To measure lactate production, cells were incubated overnight in low-serum DMEM (high-glucose DMEM, 0.5% FBS, and 50 M L-carnitine) and the medium was collected for lactate quantitation. To measure levels of cellular NAD + and NADH, cells were grown as above under normal serum conditions and washed twice in PBS before nucleotide extraction. Lactate and NAD + / NADH levels were assessed using their respective kits (BioVision) according to the manufacturer's instructions.

Statistical analysis
The data are represented as the mean ± SD or as the mean ± SEM as indicated. Statistical signifi cance was determined using the two-tail Student's t -test and one-way ANOVA where appropriate. P < 0.05 was considered to be statistically signifi cant.

RESULTS
Previous studies on FATP1 have focused on its role in mediating long-and very long-chain fatty acid infl ux into adipocytes and muscle cells. As shown originally by Stahl et al. ( 5 ) and subsequently by Lobo et al. ( 3 ), FATP1 translocates from intracellular sites to the plasma membrane in response to insulin and such translocation mediates, in part, insulin-stimulated fatty acid infl ux. However, immunolocalization analysis by Lobo et al. indicated that, as opposed to GLUT4 translocation, only a small percentage of FATP1 migrated to the plasma membrane; the vast majority of all FATP1 remained intracellular. Moreover, the appearance of punctate regions of immunofl uorescence suggested that FATP1 might be organellar. These observations suggested that FATP1 plays a role independent of mediating plasma membrane fatty acid infl ux.
Inspection of the FATP1 primary sequence and bioinformatic analysis predicted that the protein is associated with the mitochondrion. To assess this, each of the FATP family members was evaluated using MitoProt II ( 32 ) to identify putative mitochondrial targeting sequences and the probability of mitochondrial import ( Table 1 ). The analysis predicts with a probability greater than 95% that FATPs 1, 2, and 4 may be imported into the mitochondrion. Work by Stahl et al. ( 5 ) has previously shown that FATP1 fractionates with mitochondria and we were able to confi rm that observation through a combination of differential centrifugation and immunoblotting (results not shown). As an independent method to demonstrate FATP1 localization with the mitochondrion, immunofl uorescence microscopy was used in conjunction with the mitochondrion-specifi c dye, Mitotracker Red. Figure 1 shows that FATP1 colocalized with Mitotracker Red and taken together, confi rms that FATP1 is in part localized to the adipocyte mitochondrion. taining 4% SDS for 10 min. The samples were separated via SDS-PAGE using a 5-15% SDS-polyacrylamide gel and analyzed via immunoblot analysis as previously described ( 3 ). Purifi ed or reconstituted FATP were assayed for acyl-CoA synthetase activity by the conversion of [ 3 H]palmitic acid or [ 3 H]lignoceric acid to their CoA derivatives by a modifi ed method from Nagamatsu et al. ( 27 ) as previously described ( 3,10 ).

In situ and in vitro oxoglutarate dehydrogenase activity assay
In situ oxoglutarate dehydrogenase activity was measured as described ( 28 ) with modifi cations. Briefl y, cells were washed in HBSS and the assay was initiated upon addition of OGDH reaction buffer containing 50 mM Tris, pH 7.6, 5 mM MgCl 2 , 0.3 mM thiamine pyrophosphate, 3 mM ␣ -ketoglutarate, 3 mM NAD, 0.2 mM CoA, 0.1 mM CaCl 2 , 0.05 mM EDTA, 0.5 g/ml rotenone, 0.2% Triton X-100, 3.5% polyvinyl alcohol (MW 7,000-10,000), 0.75 mM nitrotetrazolium blue, and 0.05 mM phenazine methosulfate. After 10 min at room temperature, cells were aspirated and washed in HBSS. Cells were incubated in 5% NP-40 in HBSS overnight at 37°C and sonicated to dissolve the formazan product. Samples were centrifuged at 8,000 g for 10 min to remove lipid and absorbance was measured at 555 nm.
For the in vitro activity assay, purifi ed OGDH (Sigma) (or purifi ed pyruvate dehydrogenase complex (Sigma)) was buffer exchanged into 20 mM potassium phosphate, pH 7.3, 20% glycerol. Ten mU of dehydrogenase was assayed for 15 min at 37°C in 250 l of a modifi ed reaction buffer containing 50 M fatty acid-free BSA, 50 mM Tris, pH 7.6, 5 mM MgCl 2 , 0.3 mM thiamine pyrophosphate, 3 mM ␣ -ketoglutarate or 3 mM pyruvate, 3 mM NAD, 0.2 mM CoA, 0.1 mM CaCl 2 , 0.05 mM EDTA, 5 mM potassium phoshphate, 0.75 mM nitrotetrazolium blue, and 0.05 mM phenazine methosulfate. The reaction was placed on ice for 5 min and centrifuged at 16,000 g for 30 min at 4°C. The formazan pellet was dissolved by sonication and the absorbance at 555 nm measured.

Analysis of cellular and mitochondrial fatty acid oxidation
Cellular fatty acid oxidation was performed as described ( 29 ). Briefl y, 3T3-L1 adipocytes were incubated in growth medium containing 50 M L-carnitine overnight and serum-starved for 1 h in Krebs-Ringer's HEPES containing 5.4 mM glucose, 1 mM L-carnitine, and 0.1% fatty acid-free BSA. Palmitate oxidation was initiated upon addition of 400 M [1-14 C]palmitate (2 Ci/ mol palmitate) buffered with fatty acid-free BSA (4:1 fatty acid-BSA) and incubated for 1 h at 37°C and 5% CO 2 . Media and cells were transferred to glass vials and acidifi ed with 70% perchloric acid. Volatilized 14 CO 2 was absorbed in 1 M NaOH and transferred to liquid scintillation vials for counting. The remaining acidifi ed sample was centrifuged at 2,000 g at 4°C and the radioactivity in the supernatant (acid soluble metabolites) determined by liquid scintillation counting.
For mitochondrial fatty acid oxidation, mitochondria were isolated as described ( 30 ). A total of 100 l isolated mitochondria were added to 1 ml fatty acid oxidation buffer (150 M [1- 14

Analysis of mitochondrial pyruvate oxidation
Pyruvate oxidation was performed as described ( 31 ) with modifi cations. 3T3-L1 adipocytes were incubated in growth medium FATP1 immunoprecipitation was estimated to be at least 60-fold greater than in the control immunoprecipitates based on normalized spectral counts.
To address the association between FATP1 and OGDH directly, we evaluated the effects of FATP1 on the activity of purifi ed OGDH. We previously reported that the activity of purifi ed FATP1 in DDM-micelles is unstable at 37°C and is rapidly lost within minutes although being stable for several hours at 4°C ( 10 ). Because many integral membrane proteins are unstable unless reconstituted into lipid bilayers, we produced FATP1-containing proteoliposomes and evaluated the effects of reconstitution on FATP1 activity. FATP1 was reconstituted using glycerophospholipids that contained oleoyl side chains at the sn -1 and sn -2 positions because these phospholipids exhibit gel-to-liquid crystal transition temperatures below 4°C, keeping the lipid bilayer fl uid during the reconstitution. Preformed SUVs were detergent-destabilized prior to the addition of purifi ed protein and were successfully reconstituted into mixed DOPC-DOPE (3:1) vesicles based on the migration of FATP1 in a 15-60% sucrose gradient (results not shown).
FATP1 is predicted to have a single transmembrane domain separating a 4-10 amino acid N-terminal domain from the catalytic region ( 34 ). This suggested that limited tryptic proteolysis could be used to determine the fraction of FATP1 orientated toward the outside of the vesicle versus that orientated inside of the vesicle. Using trypsin sensitivity and an antibody directed toward the catalytic region (amino acids 192-215), FATP1 in DDM-micelles was rapidly proteolyzed producing an immunoreactive stable core peptide of 25 kDa ( Fig. 3A ). Reconstituted FATP1 was also rapidly and completely proteolyzed, generating an immunoreactive stable core peptide of 15 kDa. These results suggested that FATP1 was orientated essentially unidirectionally in an inside-out manner with the catalytic domain located on the exofacial side of the SUV. This characteristic allowed subsequent assessment of acyl-CoA synthetase activity as previously performed in DDM-micelles. Reconstituted FATP1 exhibited a slightly altered structural conformation resulting in the increased accessibility of at least one trypsin cleavage site as observed by the appearance of a core 15 kDa cleavage product rather than the 25 kDa product found for trypsinized FATP1 in DDM-micelles. To compare the thermostability of reconstituted FATP1 in SUVs to that purifi ed in DDM, the acyl-CoA synthetase activity of FATP1 at 37°C was monitored as a function of time ( Fig. 3B ). The acyl-CoA synthetase activity of FATP1 in DDM-micelles was lost rapidly with a half-life ( t 1/2 ) of ‫ف‬ 1 min. In contrast, reconstituted FATP1 exhibited increased thermostability at 37°C with a half-life of 30-60 min. The If FATP1 resides in the mitochondrion, it is likely that it functions in conjunction with other proteins. To test this hypothesis, we undertook a proteomics approach to identify such FATP1-interacting partners. To that end , endogenous FATP1 was immunoprecipitated from detergent-solubilized 3T3-L1 adipocytes and subjected the immune complex and associated polypeptides to SDS-PAGE ( Fig. 2 ) and Sypro Ruby staining. A protein at ‫ف‬ 100 kDa that was routinely observed in multiple experiments but not present in control immunoprecipitates was excised from the gel, trypsin digested, and the resultant peptides analyzed via LC-MS/MS . The peptides identifi ed matched the E1 subunit of mitochondrial OGDH, a rate-limiting step in the tricarboxylic acid cycle ( Fig. 2 , Table 2 ). A 90 kDa band was also observed and processed in parallel; however, its identity was not determined.
Because the association between FATP1 and OGDH could have formed following detergent extraction, we used formaldehyde cross-linking as an additional method to evaluate FATP1-OGDH interaction. Mild formaldehyde cross-linking of cells has been used previously to stabilize protein-protein interactions while minimizing nonspecifi c interactions ( 33 ). In this analysis, proteins that directly associate with FATP1 plus secondary interactions between other associated proteins are revealed through the crosslinking. Experimentally, 3T3-L1 adipocyte proteins were cross-linked with formaldehyde, immunoprecipitated with anti-FATP1 antibodies, heated to reverse the cross-links, and the resultant proteins resolved by SDS-PAGE. The regions corresponding to the heavy and light chains were discarded and the remaining segment of the gel divided into 10 sections. Protein in each segment was digested with trypsin and the peptides recovered and analyzed via LC-MS/MS sequencing. The resulting matches included the target protein FATP1 and OGDH (i.e., the E1, E2, and E3 subunits) ( Table 3 ) as well as proteins involved in fatty acid synthesis and oxidation, glucose metabolism, tricarboxylic acid cycle, and branched-chain amino acid metabolism. The level of each protein determined from the Peptide sequences of FATP family members were analyzed for mitochondrial targeting sequences and the probability of mitochondrial import was calculated using MitoProt II. Surprisingly, FATP1 enhanced OGDH activity over twofold and was independent of ATP and fatty acids ( Fig. 4A ). Additionally, the increase in OGDH activity was dependent on the concentration of FATP1 ( Fig. 4B ). Taken together, these results indicate that FATP1 is an OGDH activator independently of acyl-CoA production.
To determine if the ability of FATP1 to enhance OGDH activity occurred in a cellular context, we took advantage of a 3T3-L1 adipocyte cell line stably expressing shRNA directed against FATP1 that we had previously established ( Fig. 5A) ( 3 ) and accessed OGDH activity. Corroborating the in vitro data, OGDH activity was decreased 20% in the FATP1 knockdown adipocytes compared with the scramble shRNA-expressing adipocytes ( Fig. 5B ). This suggests that FATP1 is a physiological regulator of OGDH.
Because OGDH is a rate-limiting step in the tricarboxylic acid cycle, the decrease in OGDH activity in the FATP1 knockdown adipocytes could also result in decreased tricarboxylic acid cycle activity. To assess this, the rate of [2-14 C]pyruvate oxidation was evaluated as a measure of tricarboxylic acid cycle activity ( 31 ). As shown in Fig. 6A , mitochondria isolated from FATP1 knockdown adipocytes exhibited a 60% reduction in the rate of [2-14 C]pyruvate oxidation compared with that in the control mitochondria, indicating a decrease in tricarboxylic acid cycle activity that is consistent with the decrease in OGDH activity.
Because tricarboxylic acid cycle function plays an important role in the cellular redox status, we measured the NAD + and NADH levels in the FATP1 knockdown cells. NAD + levels were unchanged in the FATP1 silenced adipocytes compared with scramble adipocytes whereas NADH levels were decreased 33% ( Fig. 6B, C ). There was no sta-activity of reconstituted FATP1 was essentially unaffected by changes in the phospholipid composition or addition of cholesterol (results not shown).
OGDH activity is inhibited in vitro by fatty acyl-CoAs at micromolar levels (35)(36)(37)(38)(39)(40). Because of this, we hypothesized that FATP1 would potentially act as an OGDH inhibitor via its acyl-CoA synthetase activity. Having established functional and stable FATP1 proteoliposomes, we assessed the activity of OGDH in the absence or presence of FATP1.  The 100 kDa band in Fig. 2 was excised and subjected to MS/MS analysis as described in Materials and Methods using the X! Tandem search engine (v2007.01.01.1) in addition. Xcorr and DCn scores generated from SEQUEST and -log(e) scores generated from X! Tandem.
The inability to derive energy from mitochondrial metabolism is frequently accompanied by increased metabolism of glucose and fatty acids to compensate for the energy defi ciency. This suggests that the FATP1 silenced tistical change in total NAD levels, however (data not shown). This resulted in a 50% increase in the cellular NAD + /NADH ratio ( Fig. 6D ) and is consistent with decreased tricarboxylic acid cycle activity. Differentiated 3T3-L1 adipocytes were cross-linked with 0.5% formaldehyde and FATP1 immunoprecipitated. The resultant proteins were separated by SDS-PAGE and subjected to MS/MS analysis as described in Methods and Materials. All identifi ed proteins were enriched at least 60-fold compared with control IgG immunoprecipitate. difference may arise from the absence of glucose metabolism in the isolated mitochondria. Consistent with this, the presence of pyruvate attenuates CO 2 production from fatty acids without effecting total fatty acid oxidation (data not shown).

DISCUSSION
FATP1 has been previously demonstrated to play a vital role in insulin-stimulated LCFA infl ux both in adipocytes ( 3,5 ) and skeletal muscle ( 5 ). Consistent with the role of FATP1 in LCFA uptake, FATP1 translocates from intracellular structures to the plasma membrane upon insulinstimulation; however, a large fraction of FATP1 remains on intracellular structures ( 3,5 ). Several reports have noted the wide subcellular distribution of FATP1 in adipocytes and muscle ( 3,5,41,42 ), implying that FATP1 has additional roles in these tissues. Using 3T3-L1 adipocytes as a model system, we took a multidimensional approach to elucidate additional functions of FATP1. Using a combination of immunoprecipitation of endogenous FATP1 and mild formaldehyde cross-linking to stabilize protein-adipocytes may exhibit increased glucose metabolism and lactate production. Consistent with this view, lactate production was increased 2.4-fold in the FATP1 knockdown adipocytes ( Fig. 7A ). Similar metabolic changes were seen in cellular fatty acid oxidation where total fatty acid oxidation increased 13% ( Fig. 7B ). This was due to a 24% increase in the acid soluble metabolites (ASM) produced, whereas complete oxidation of fatty acids (i.e., fatty acids oxidized to CO 2 ) decreased 30%, consistent with a defect in tricarboxylic acid cycle activity. To eliminate the possibility that the change in fatty acid oxidation is due to other peripheral changes in lipid and glucose metabolism, fatty acid oxidation was measured in isolated mitochondria from the FATP1 silenced cells ( Fig. 7C ). Similarly, total fatty acid oxidation in isolated mitochondria was increased 17% compared with scramble adipocytes and predominantly due to a 17% increase in ASM. However, unlike what was measured for cellular fatty acid oxidation, complete fatty acid oxidation in isolated mitochondria was increased 11% compared with the scramble adipocytes. This  protein interactions, we identifi ed proteins of the OGDH as a likely binding partner for FATP1. In the absence of cross-linking, the E1 subunit of OGDH was identifi ed as the major FATP1 associated protein ( Fig. 2 , Table 2 ). Using cross-linking, a large number of mitochondrial and mitochondrially associated proteins were identifi ed consistent with a large supramolecular complex of multiple proteins and functional activities. Among those identifi ed in the cross-linking experiment was long-chain acyl-CoA synthetase 1 (ACSL1). Richards et al. ( 15 ) previously coimmunoprecipitated epitope-tagged FATP1 and ACSL1 using lentiviral overexpression in 3T3-L1 adipocytes. Recent work has shown that ACSL1 is not involved in LCFA infl ux in adipocytes but may be linked to reesterifi cation of fatty acids following lipolysis ( 43 ). Therefore, the true function of the FATP1-ACSL1 interaction is still unknown. These results, in sum, conclude that FATP1 is a mitochondrially-associated protein. Consistent with this, both bioinformatic analysis and immunofl uorescence microscopy indicated that FATP1 is localized to mitochondria in 3T3-L1 adipocytes ( Table 1 , Fig. 1 ). Moreover, during the preparation of this manuscript, Guitart et al. ( 44 ) published work demonstrating that FATP1 localizes to mitochondria in both cultured myotubes and human skeletal muscle. As such, fi ndings in multiple systems conclude that FATP1 has additional functions beyond facilitating fatty acid infl ux.
To effectively test a functional signifi cance to the proposed FATP1-OGDH interaction, we developed detergentfree FATP1 proteoliposomes for in vitro analysis. FATP1 purifi ed in DDM-micelles is unstable at 37°C ( Fig. 3B ). However, reconstitution of FATP1 in SUVs resulted in the surprising fi nding that the enzyme activity was markedly stabilized when maintained at 37°C. Characterization of FATP1 proteoliposomes revealed an approximately 30fold increase in thermostability compared with FATP1 in Fig. 5. Analysis of OGDH activity in FATP1 silenced adipocytes. Expression of FATP1 in 3T3-L1 adipocytes stably expressing a scrambled (Scr) or FATP1 shRNA (FATP1 kd) (A) were assayed for OGDH activity for 10 min (B) as described in Materials and Methods; n = 6, mean ± SEM. ** P < 0.01 relative to Scr. DDM-micelles ( Fig. 3B ). These changes in protein function may be explained by an altered structural conformation in the N-terminal catalytic domain. Consistent with this view, tryptic cleavage of reconstituted FATP1 resulted in a stable 15 kDa protein fragment containing amino acids 192-215 of the N-terminal catalytic domain in contrast to the stable 25 kDa protein fragment generated from FATP1 in DDM-micelles ( Fig. 3A ). The location of the conformationally sensitive trypsin cleavage site is unknown; however, the RHRR 73 motif resides adjacent to the predicted amino terminal transmembrane region, suggesting that reconstitution of the protein into SUVs may affect this site, leading to changes in trypsin accessibility. Interestingly, an additional trypsin cleavage site at Lys 238 resides just C-terminal to the antibody recognition site. This suggests that if the RHRR 73 motif were inaccessible to trypsin in the DDM-micelles, a 238 amino acid fragment would be produced whereas, in SUVs, a 166 amino acid product would be generated. Because the lipid vesicles represent a more physiological environment and result in increased thermostability of FATP1, the altered structure of FATP1 is believed to represent the natural conformation found in vivo. This suggests that FATP1 proteoliposomes could be used in subsequent in vitro studies. strated that tricarboxylic acid cycle activity is decreased substantially (60%) in the FATP1 knockdown adipocyte mitochondria as measured by [2-14 C]pyruvate oxidation ( Fig. 6A ). This is consistent with the decrease in the NADH levels and, hence, increases in the NAD + /NADH ratio ( Fig. 6C, D ), indicating altered mitochondrial energy metabolism. Attenuated tricarboxylic acid cycle activity has also been linked to increases in compensatory glycolysis in the cytoplasm ( 52-55 ), in part due to the decrease fl ux of pyruvate through the tricarboxylic acid cycle as well as increases in fatty acid oxidation ( 56 ). Accordingly, lactate production was increased over two-fold ( Fig. 7A ) and total mitochondrial fatty acid oxidation increased 17% in the FATP1-silenced cells ( Fig. 7B, C ). Previous work from Lobo et al. ( 3 ) has shown that FATP1 knockdown adipocytes have a 10% increase in basal glucose uptake. Although this would in part contribute to an increase in lactate production ( 57,58 ), the magnitude of this increase correlates with the magnitude of the decrease in tricarboxylic acid cycle activity. Further work is needed to explore the mechanistic links between FATP1, tricarboxylic acid activity, and alterations in mitochondrial function.
The work of Guitart et al. ( 44 ) also linked FATP1 to glucose oxidation in skeletal myotubes. In their studies, FATP1 overexpression in myotubes resulted in enhanced glucose oxidation whereas fatty acid oxidation was decreased. These results mirror our own observations in the FATP1 knockdown adipocytes. Guitart et al. also found that the increase in glucose oxidation was consistent with an increase in the activity of pyruvate dehydrogenase complex (PDH) although the mechanism through which FATP1 increases the activity of PDH remained unknown. PDH belongs to the 2-oxo acid dehydrogenase superfamily and includes OGDH and the branched-chain keto acid dehydrogenase complex. All three of these dehydrogenase complexes are structurally similar and are comprised of Using FATP1 proteoliposomes and purifi ed OGDH, we found that FATP1 enhanced OGDH activity in a concentration-dependent manner that was independent of ATP and fatty acids and, therefore, independent of the acyl-CoA synthetase activity of FATP1 ( Fig. 4 ). This observation was corroborated in the FATP1 knockdown adipocytes where OGDH activity was decreased 20% ( Fig. 5 ). Because we were unable to measure OGDH protein levels due to lack of a functional commercial antibody, we cannot rule out the possibility that OGDH levels are altered in the FATP1 silenced adipocytes. However, because FATP1 was able to enhance purifi ed OGDH activity in vitro, it is strongly suggestive that the interaction is direct and that the loss of FATP1 in the silenced cells is at least in part directly responsible for the observed decrease in the OGDH activity. Fatty acyl-CoAs have been previously shown to attenuate OGDH activity in vitro at low micromolar levels (35)(36)(37)(38)(39)(40). Although we were unable to observe any OGDH inhibition in the presence of 6 M palmitoyl-CoA (data not shown), it is likely that the lipid vesicles in this study act as a sink for fatty acyl-CoAs due to their amphiphilic nature. Therefore, we cannot exclude the possibility that FATP1 acyl-CoA synthetase activity infl uences OGDH in vivo.
OGDH is one of the rate-limiting steps in the tricarboxylic acid cycle ( 45,46 ) and uses NAD + to convert ␣ -ketoglutarate to succinyl-CoA, producing NADH. The NADH produced at this step and subsequent steps in the tricarboxylic acid cycle provide the electrons necessary for maintaining the mitochondrial membrane potential. Because tricarboxylic acid cycle activity is intimately linked to the cellular redox state and energy homeostasis (47)(48)(49)(50)(51), we hypothesized that the decrease in OGDH activity in the FATP1 knockdown adipocytes would manifest in decreased tricarboxylic acid cycle activity and further lead to alterations in other energy-producing pathways. We demon- Fig. 7. Lactate production and fatty acid oxidation are increased in the FATP1 silenced adipocytes. A: Lactate in the culture media; n = 6, mean ± SEM. B, C: 3T3-L1 Scr (light gray) or FATP1 knockdown (dark gray) adipocytes were pretreated with 50 M L-carnitine overnight and fatty acid oxidation assessed in the adipocytes (B) or isolated mitochondria (C) using [1-14 C]palmitate. B: n = 6, mean ± SEM. C: n = 4, mean ± SEM. ASM, acid soluble metabolites; Total, ASM + CO 2 . ** P < 0.01; *** P < 0.001 relative to Scr. the FATP1 knockdown adipocytes have an increased NAD + /NADH ratio due to decreased NADH levels. These changes in the cellular redox status correlate with the observed increase in lactate production and fatty acid oxidation in the FATP1 knockdowns. This work, along with the work of Guitart et al., reveals a new and novel role for FATP1 in the regulation of tricarboxylic acid cycle function and energy homeostasis via its protein interactions with key proteins in the tricarboxylic acid cycle. three subunits (E1, E2, and E3) where the E1 and E2 subunits are unique to each of the dehydrogenases while sharing the same E3 ( 59 ). While assessing FATP1 activation of purifi ed OGDH using our FATP1 proteoliposome system, we found that purifi ed PDH was similarly activated by FATP1 in a concentration-dependent manner (data not shown). This suggests that the enhancing ability of FATP1 on OGDH activity also pertains to PDH activity. Furthermore, branched-chain keto acid dehydrogenase was also identifi ed as a candidate FATP1-protein interaction ( Table 3 ). Together, our data suggest the ability of FATP1 to enhance dehydrogenase activity may be conserved for all 2-oxo acid dehydrogenase superfamily members. It is unclear as to why none of the unique PDH subunits were identifi ed as candidate FATP1-protein interactions. One explanation is that the proteins were simply excluded when the IgG gel bands were excised and discarded. Another possibility is that endogenous FATP1 interacts with PDH at relatively low stoichiometry in adipocytes, making it diffi cult to detect the interaction. Although our current study correlates the decrease in OGDH activity with the decrease in tricarboxylic acid cycle activity in the FATP1 knockdown adipocytes, PDH activity within the knockdowns was not assessed. Therefore, it is possible that a decrease in PDH activity exists in the FATP1 knockdowns and may be contributing to the decreased tricarboxylic acid cycle activity. Further studies are needed to better understand the nature of the physical interactions between FATP1 and the dehydrogenases and the contribution of each of the dehydrogenases to tricarboxylic acid cycle activity in adipocytes.
Interestingly, cardiomyocyte-specifi c overexpression of FATP1 results in cardiac lipotoxicity ( 19 ) whereas Fatp1 Ϫ / Ϫ mice maintain insulin sensitivity even on a high-fat diet ( 7 ). Although the effects of FATP1 in these systems has been solely considered with regard to fatty acid infl ux, it is possible that the changes in mitochondrial FATP1 levels and the associated modulation of tricarboxylic acid cycle function also contribute to these phenotypes. Fatp1 Ϫ / Ϫ mice are also cold intolerant due to a defect in nonshivering thermogenesis in brown adipose tissue ( 60 ). Wu et al. ( 60 ) indicated that FATP1 plays an important role in the early phase of thermogenesis as indicated by a decrease in oxygen consumption and CO 2 production prior to the increase in FATP1 protein expression during the late phase of thermogenesis. These observations are also consistent with the decreases in tricarboxylic acid cycle activity and NADH levels we observed in the FATP1 knockdowns and may even result in attenuation of mitochondrial membrane potential in brown adipocytes ( 61 ). This would retard the rate of electron transport uncoupling and, in part, attenuate thermogenesis. Additional studies are needed to confi rm the role of mitochondrial FATP1 in these systems.
In summary, we demonstrated that FATP1 is localized to mitochondria and enhances OGDH activity in vitro. This is consistent with an attenuation of OGDH activity in cultured FATP1 knockdown adipocytes and correlates with a decrease in tricarboxylic acid cycle activity. Furthermore,