Fatty Acid-binding Proteins Interact with Comparative Gene Identification-58 Linking Lipolysis with Lipid Ligand Shuttling

Background: A multiprotein complex designated as lipolysome degrades intracellular triglycerides and contains proteins such as adipose triglyceride lipase (Atgl) and its co-activator Cgi-58. Results: Cgi-58 interacts with fatty acid-binding proteins (Fabps), which impact Atgl-mediated lipolysis and lipid signaling. Conclusion: Fabps modulate Atgl-mediated TG hydrolysis and link lipolysis with intracellular lipid ligand shuttling. Significance: Novel mechanistic insights into the regulation of lipid catabolism and energy homeostasis are presented.

Intracellular lipolysis includes the stepwise hydrolytic breakdown of triglycerides (TGs), 3 stored in lipid droplets (LDs), resulting in the formation of glycerol and fatty acids (FAs). TG catabolism involves the consecutive action of adipose triglyceride lipase (Atgl), which hydrolyzes TGs to diglycerides, hormone-sensitive lipase (Hsl) converting diglycerides to monoglycerides, and monoglyceride lipase, hydrolyzing monoglycerides to FA and glycerol (1). These enzymes constitute the core of a multiprotein network (1, 2) encompassing a variety of regulatory, accessory, and scaffolding proteins (3)(4)(5). Lipolysis basically occurs in all cell types and tissues to meet the cellular requirements for FAs as energy substrates or building blocks for membrane lipids. It is most prominent in white adipose tissue (WAT) that supplies oxidative tissues with FAs via the bloodstream during times of starvation and physical exercise. Lipolysis is a tightly regulated process, and it critically depends on specific protein/protein interactions. Such transient complexes are common in the regulation of metabolic pathways and signaling cascades enabling instant cellular responses to diverse stimuli (6,7). Additionally, metabolic lipases hydrolyzing TGs often require specific co-activators for maximal enzyme activity, presumably by mediating enzyme/ substrate interaction at the water-lipid interphase. Examples include the activation of pancreatic lipase by colipase or lipoprotein lipase by apolipoprotein CII (8,9).
Cloning of the cgi-58_S239E Variant with Increased Solubility-Mouse cgi-58 (mcgi-58, coding for Lys-2-Asp-351 (UniProt ID Q9DBL9)) (13), preceded by a TEV protease cleavage site, was amplified by PCR and the primers forward 5Ј-CGAAGCAGAGAGCTCGAAAACCTGTATTTTCAGG-3Ј and reverse 5Ј-GGAACCCTCGAGTCATCAGTCTACTG-TGTGGC-3Ј. The TEV-mcgi-58-containing PCR product and a His-pSumo (kindly provided by C.D. Lima, Sloan-Kettering Institute) vector lacking a functional BamHI endonuclease cleavage site were digested with the appropriate endonucleases and ligated with T4 ligase (New England Biolabs, Ipswich, MA) to yield His-pSumo-TEV-mcgi-58. The mutation S239E was introduced using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA) with the forward primer 5Ј-CCT-GATTTCAAGCGGAAGTACGAGTCTATGTTTGAAGAT-GACACG-3Ј and the reverse primer 5Ј-CGTGTCATCTTCA-AACATAGACTCGTACTTCCGCTTGAAATCAGG-3Ј. The fusion protein encoded by the resulting construct His-pSumo-TEV-cgi-58_S239E was subsequently cleaved by the TEV protease to obtain the mutant variant mCgi-58 S239E , which showed enhanced solubility, and was used for NMR experiments only. For reasons of simplification, this variant is called m sol Cgi-58 throughout the text.
Expression and Purification of Gst-Cgi-58 in E. coli-Gsttagged Cgi-58 and truncation variants thereof used in solid phase assays were expressed and purified as described previously (13).
Expression of Recombinant Proteins in Eukaryotic Cells and Preparation of Cell Extracts-Monkey embryonic kidney cells (COS-7, ATCC CRL-1651) were cultivated in Dulbecco's modified Eagle's medium (DMEM) (Invitrogen) supplemented with 10% FCS (Sigma), penicillin (100 IU/ml), and streptomycin (100 g/ml) in a standard humidified 5% CO 2 atmosphere at 37°C. For transfection, 6 g of plasmid DNA were complexed to 30 l of Metafectene (Biontex GmbH, Munich, Germany), unless otherwise stated, and the mixture was added to serum-free cell culture medium. After 4 h, medium was replaced with standard medium containing 10% FCS. Cells were harvested after 24 -48 h, suspended in solution A, disrupted by sonication (Missonix sonicator, QSonica LLC, Newtown, CT), and centrifuged at 13,000 ϫ g for 15 min. Protein concentrations were determined by protein assay (Bio-Rad, Munich, Germany) according to the manufacturer's protocol using BSA as standard.
NMR Resonance Assignment-Backbone resonance assignment and interaction mapping by NMR were performed with 500 M 15 N-and 13 C-labeled A-Fabp samples in buffer 4 (100 mM NaPP, 150 mM NaCl, pH 6.5) and NMR titration buffer, respectively. 10% D 2 O was added to all samples. Spectra were recorded on a Varian/Agilent Inova 600 MHz spectrometer equipped with a cryogenically cooled probe and a 500 MHz Bruker spectrometer equipped with a TXI room temperature probe at 298 K, respectively. Backbone assignment was obtained by analysis of three-dimensional NMR experiments (HNCA/HNCOCA, HNCO/HNCACO, HNCACB, and CCONH experiments (26)). Backbone experiments were recorded nonuniformly with 15% Poisson gap weighted sampling of the grid in indirect dimensions (27) and reconstructed with the iterative soft thresholding approach (28). The CCONH experiment and 1 H-15 N HSQC spectra were recorded linearly and were processed with NMRpipe (29). All spectra were analyzed with CcpNmr (30).
Heteronuclear Single Quantum Coherence Spectroscopy ( 1 H-15 N HSQC) Titration-Titration of 50 M 15 N labeled A-Fabp was carried out in NMR titration buffer with increasing concentrations of m sol Cgi-58 (0, 10, 25, 50, 100, and 200 M). 1 H-15 N HSQC spectra were recorded after each titration step and compared in CcpNmr (30). Titration of A-Fabp with m sol -Cgi-58 resulted in the loss of peak intensities and disappearance of peaks at early points of the titration. To analyze the interaction of A-Fabp with Cgi-58, peak intensities in the 1 H-15 N HSQC spectrum of A-Fabp recorded in the presence of 100 M m sol Cgi-58 were compared with peak intensities in the 1 H-15 N HSQC spectrum of A-Fabp acquired in absence of m sol Cgi-58. Those residues, for which the corresponding peak intensities in the 1 H-15 N HSQC spectrum experienced a reduction larger than twice the standard deviation compared with the reference spectrum, were mapped on the surface of an A-Fabp crystal structure (Protein Data Bank code 3Q6L).
Solid Phase Assay-96-Well polystyrene plates (MaxiSorp, Nalgen Nunc Int., Rochester, NY) were coated overnight with 3 g of purified Fabp or 1 g of purified Atgl (as described in Ref. 10) per well. Unspecific binding sites were blocked with 5% (w/v) delipidated BSA (Sigma) dissolved in TBS (50 mM Tris, pH 8.0, 150 mM NaCl) for 2 h. Cell extracts (10 -60 g of total protein) containing equal amounts of His-tagged proteins supplemented with 2% BSA and 0.05% Tween 20 were added and incubated overnight. The plate was rinsed three times with TBS containing 0.05% Tween 20 and incubated with mouse anti-His antibody (GE Healthcare) in the same buffer containing 0.5% delipidated BSA for 1 h. The plate was again washed three times and incubated with the secondary sheep horseradish peroxidase-conjugated (Hrp) anti-mouse antibody (GE Healthcare) in the same buffer as the primary antibody for 1 h. After three final washing steps, antibody binding was visualized with tetramethylbenzidine as chromogenic substrate, and absorbance was measured at 450 nm with 620 nm as reference wavelength. The correspondingWesternblotswereperformedaccordingtostandard procedures using mouse anti-His antibody (1:5000, GE Healthcare) and anti-mouse Cy3-conjugated secondary antibody (1:1000, GE Healthcare). Fluorescence was detected in a Typhoon instrument (GE Healthcare), and signal intensities were quantified by densitometry within a linear range.
Microscale Thermophoresis-Microscale thermophoresis analyses were carried out with a Monolith NT.115 instrument (Nano Temper, Munich, Germany). Purified Gst-tagged Cgi-58 ( Preparation of Cell and Tissue Extracts-COS-7 cells, transfected or co-transfected with expression vectors encoding Atgl, Cgi-58, and/or A-Fabp (as described above), were disrupted by sonication (Virsonic 475, Virtis, Gardiner, NJ) in solution A. Murine gonadal WAT was homogenized in the same buffer using an Ultra Turrax homogenizer (Ika GmbH, Staufen, Germany), and the lysates were centrifuged at 20,000 ϫ g for 30 min. The infranatants were collected and used for the subsequent assays.
The inhibitor BMS309403 was synthesized as described in Sulsky et al. (34).
Firefly Luciferase Gene Transactivation Assay-COS-7 cells were seeded in 24-well plates and co-transfected with expression plasmids using TurboFect (Thermo Scientific). Control cells were co-transfected with pPPRE-Luc, a Ppar-responsive expression vector of Photinus pyralis firefly luciferase, and pRLL-Luc encoding Renilla reniformis luciferase under the control of a constitutive promoter (both plasmids were kind gifts from B. Steals, University of Lille, France). To determine the impact of Atgl, Cgi-58, and/or Fabps on Ppar activation, cells were additionally transfected with expression vectors encoding these proteins. The total DNA amount was kept constant for all transfections by addition of empty vector-DNA. Luciferase activities were measured after passive cell lysis using the Dual-Luciferase Reporter assay system (Promega, Madison, WI) in a GloMax-Multiϩ Microplate Multimode Reader (Promega). Firefly luminescence units were divided by Renilla lumincescence units to normalize to transfection efficiency. Signals of control transfections (cells transfected with pPPRE-Luc and PLL-Luc) were set to 100%, and all other transfections were expressed relative to it.

A-Fabp and Some Fabp Isoforms Bind to
Cgi-58 -To identify potential binding partners of Cgi-58, we performed a preliminary label-transfer screen using adipose tissue lysates and labeltransfer reagent-modified Cgi-58 as bait. The initial exploratory approach revealed a large number of potential interaction partners summarized in Table 1. The most frequently detected Cgi-58-binding protein in MS analysis was A-Fabp as well as another Fabp family member, namely E-Fabp. To confirm and substantiate this initial observation of an interaction between Cgi-58 and A-Fabp, we applied four independent methods as follows: solid phase assays, microscale thermophoresis, co-immunoprecipitation, and NMR titration. For the solid phase assays, purified A-Fabp was immobilized on microtiter plates and incubated with COS-7 cell lysates containing recombinant murine or human Cgi-58, murine Atgl, Hsl, or LacZ as negative control. Specific binding of His-tagged Cgi-58 to A-Fabp was detected in a sandwich assay using His tag-specific antibodies and Hrp-coupled secondary antibodies against mouse IgG. Both mouse and human Cgi-58 bound to A-Fabp in a dose-dependent manner (Fig. 1a). Hsl also dose-dependently bound to A-Fabp confirming previous observations (35)(36)(37). Atgl or LacZ protein did not bind to A-Fabp. All solid phase assays contained similar amounts of recombinant protein as estimated by Western blotting signal intensities (Fig. 1b). Solid phase assays with other members of the Fabp family revealed specific binding of mouse and human Cgi-58 to H-Fabp, I-Fabp, L-Fabp, and E-Fabp (Fig. 1c).
Next, microscale thermophoresis, a novel method for immobilization-free analysis of protein/protein interactions, was used to assess the Cgi-58/A-Fabp interaction in vitro. Thermo-

TABLE 1 List of potential Cgi-58 interaction partners identified by label transfer experiments with subsequent nano-LC/MS-MS analysis
Purified Gst-tagged Cgi-58 was labeled with the heterobifunctional cross-linker Sulfo-SBED and incubated with murine WAT lysates to label interacting proteins with a biotin tag. After affinity purification, the interacting proteins were separated by SDS-PAGE, tryptically digested in gel, and analyzed with nano-LC-MS/MS. Proteins found in the respective samples including the NCBI accession numbers, their molecular masses, the number of identified distinct peptides, the Spectrum Mill protein score, and the percentage of amino acid coverage are given in the table. The database used for re-analysis of the raw data was the "mouse" subset of the "SwissProt" database, downloaded on 7.1.2014 from "ncbi.nih.gov" (456,996 total entries). Database searching was performed with the "MS/MS Search" feature of the "Agilent-Spectrum Mill" software (revision 2.7). Acceptance parameter: individual peptide score: min 10; protein score: min 20, and minimum number of distinct peptides: 2. Keratin was excluded from the list. phoresis is the directed movement of proteins in response to a temperature gradient and depends on particle charge, size, conformation, and hydration state. Thus, under constant buffer conditions thermophoresis of unbound proteins typically differs from the thermophoresis of proteins bound to interaction partners. Microscopic temperature gradients are generated by an infrared laser focused on a capillary. Thermophoretic movement of a fluorescently labeled protein is then measured by monitoring the fluorescence distribution. When fluorescently labeled Cgi-58 was titrated against increasing amounts of purified A-Fabp, its thermophoretic movement markedly changed indicating protein/protein interaction (Fig. 1d). Calculation of an equilibrium dissociation constant revealed a K d of 192 Ϯ 22 M. According to Ozbabacan et al., (7) a K d value in this micromolar range is commonly found for weak and transient interactions.

Spectra
To further corroborate the Cgi-58/A-Fabp interaction, we performed co-immunoprecipitation experiments with murine WAT lysates. A monoclonal A-Fabp antibody bound to protein A/G-Sepharose co-precipitated Cgi-58. Conversely, a protein A/G-Sepharose-bound Cgi-58 antibody co-precipitated A-Fabp (Fig. 1e). Neither A-Fabp nor Cgi-58 were detected when the lysates were incubated with a protein A/G-Sepharose-bound Gfp antibody as negative control. These data show that the interaction between Cgi-58 and A-Fabp can be detected with endogenous proteins under close-to-physiological conditions. Finally, we confirmed the Cgi-58/A-Fabp interaction by NMR titration experiments using a more soluble but functional variant of Cgi-58 termed m sol Cgi-58 (38). A two-dimensional heteronuclear single quantum coherence spectrum ( 1 H-15 N HSQC) was recorded with a sample of 15 N-labeled A-Fabp. The observed peaks (depicted as contour plots) in the spectrum correspond to bonded 15 NH pairs, mainly present in the protein backbone. The peak dispersion of the 1 H-15 N HSQC spectrum of 15 N-labeled A-Fabp indicated a folded protein with a high amount of ␤-sheet content as expected from the three-dimensional structure. The chemical shifts are very sensitive to the chemical environment of each 15 NH pair. Therefore, addition of an interacting protein leads to marked changes of the chemical shift and/or the intensities of the respective peaks. As seen in Fig. 2a, addition of unlabeled m sol Cgi-58 to 15 N-labeled A-Fabp led to significant changes in the two-dimensional spectrum indicating interaction of these two proteins (Fig. 2a).
A  (Fig. 2b). Mapping these residues onto the known three-dimensional structure of A-Fabp (Protein Data Bank code 3Q6L) revealed that a region encompassing helix ␣1 and ␣2 and two spatially adjacent loops (including Phe-58, Lys-59, Asp-77, and Asp-78) very likely represent the binding interface between Cgi-58 and A-Fabp (Fig.  2c). To investigate whether these amino acids are crucial for A-Fabp interaction with Cgi-58, we generated mutant A-Fabp variants and tested their binding to Cgi-58 in solid phase assays. A-Fabp F58S , A-Fabp K59E , A-Fabp D77G , and A-Fabp D78G showed 20 -50% decreased binding to Cgi-58 as compared with wild type A-Fabp (Fig. 3a), indicating that residues Phe-58, Lys-59, Asp-77, and Asp-78 are involved in the binding of A-Fabp to Cgi-58. Decreased but not abolished binding suggests that none of the residues are indispensable per se. We also tested A-Fabp mutant variants with amino acid exchanges at Asp-18, Lys-22, FIGURE 1. A-Fabp interacts with Cgi-58. a, in solid phase assays, polystyrene plates were coated with purified A-Fabp and incubated with lysates of COS-7 cells transfected with plasmids encoding His-tagged murine Atgl, murine or human Cgi-58, murine Hsl, and as control ␤-galactosidase (LacZ). Bound proteins were detected using anti-His primary and Hrp-conjugated secondary antibody. Plates were developed using tetramethylbenzidine as substrate, and the absorbance was measured at 450/620 nm. Absorbance values were normalized to the relative expression levels of prey proteins determined by densitometry. b, Western blot analysis of respective cell lysates using anti-His primary and Cy3-conjugated secondary antibody. Fluorescence was quantified in a typhoon instrument. c, Cgi-58 interacts with various Fabp isoforms. Polystyrene plates were coated with purified A-, H-, L-, I-, or E-Fabp and incubated with COS-7 cell lysates (10, 30, and 60 g total protein) containing His-tagged murine Cgi-58, human Cgi-58, and as control LacZ, expressed at comparable levels. Binding of proteins was detected using anti-His primary and Hrp-conjugated secondary antibody. d, A-Fabp/Cgi-58 interaction demonstrated by microscale thermophoresis. Ten mol of purified and fluorescently labeled Cgi-58 was titrated against increasing amounts of unlabeled A-Fabp and fluorescence distribution inside the capillary determined using the Monolith NT.115. F norm , normalized fluorescence. e, co-immunoprecipitation of endogenous A-Fabp and Cgi-58. White adipose tissue from overnight fasted mice was lysed and incubated with antibodies directed against A-Fabp, Cgi-58, or Gfp as negative control. Antibody/antigen complexes were precipitated using protein A/G-Sepharose and analyzed together with the lysate (ϭ input) for the presence of antigens by Western blotting. Data are shown as mean Ϯ S.D. (n ϭ 4) and are representative of three independent experiments. Statistical difference was determined as compared with LacZ control (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001) and of 60 and 30 g versus 10 g of each lysate ( §, p Ͻ 0.05; § §, p Ͻ 0.01; § § §, p Ͻ 0.001).
and Arg-31. These residues, together with Asp-19, constitute a charged amino acid quartet that has been shown to be essential for the binding of A-Fabp to Hsl (37). However, the mutant variants A-Fabp D18K , A-Fabp K22E , A-Fabp R31E , and A-Fabp K22E,R31E exhibited similar binding to Cgi-58 as WT A-Fabp (Fig. 3b) indicating that these amino acid residues may not be required for the A-Fabp/Cgi-58 interaction.
To test whether A-Fabp binding to Cgi-58 depends on its loading status with FAs, we generated the mutant variant A-Fabp R127Q , which is unable to bind FAs (24). Solid phase assays  JULY 24, 2015 • VOLUME 290 • NUMBER 30 revealed similar binding efficiencies of the mutant and WT A-Fabp to murine Cgi-58 indicating that the interaction occurs independently of the FA loading status of A-Fabp (Fig. 3b). In contrast, the interaction of A-Fabp with Hsl occurs only when A-Fabp is loaded with a FA (36).

Fabps Interact with Cgi-58
A-Fabp Binds to the C-terminal Region of Cgi-58 and Does Not Bind to Atgl-Next, we asked whether binding of A-Fabp to Cgi-58 affects the interaction of Cgi-58 with Atgl. In competitive solid phase assays, partially purified Atgl was adsorbed to polystyrene plates and incubated with bacterial lysates containing Strep-tagged Cgi-58 together with increasing amounts of purified A-Fabp. After excessive washing, the amount of retained Cgi-58 was determined using a Strep-tag-specific antibody and an Hrp-coupled secondary antibody against mouse IgG. As shown in Fig. 4, the addition of A-Fabp did not interfere with the binding of Cgi-58 to Atgl at any concentration tested. As assumed, LacZ as a negative control failed to bind to Atgl. This result implicates that Atgl and A-Fabp binding to Cgi-58 occurs at different noncompetitive sites. Consistent with our data above (Fig. 1a), we could not detect a direct interaction between Atgl and A-Fabp.
To identify potential binding regions in Cgi-58 that interact with A-Fabp, we tested various structural variants of Cgi-58 in solid phase assays. These binding studies revealed that fulllength Cgi-58 and N-terminally truncated Cgi-58 variants lacking amino acids between 1 and 10 to 104 of the N terminus (Cgi-58-N-trunc10, Cgi-58-N-trunc30, Cgi-58-N-trunc79, and Cgi-58-N-trunc104) bound equally well to immobilized A-Fabp (Fig. 5, a and b, shows a control SDS-PAGE of purified Cgi-58 and the truncated variants), indicating that the first 104 amino acids of Cgi-58 are not involved in A-Fabp binding. Apparently, the mutant Cgi-58 variants are able to bind to A-Fabp despite the fact that they at least partially lose their abilities to stimulate Atgl and to localize to LDs (13).
We also investigated the A-Fabp interaction of Cgi-58 with mutations causative for neutral lipid storage disease also named Chanarin-Dorfman syndrome (10,39). Consistent with the results of the deletion mutants analyzed above, a C-terminally truncated variant of Cgi-58 carrying a premature stop codon at residue 190 (Cgi-58 190TER ) expressed in COS-7 cells almost entirely lost its ability to bind to immobilized A-Fabp. Cgi-58 Q130P and Cgi-58 E260K , two variants with single amino acid exchanges, showed a 46 and 27% decrease in binding capability to A-Fabp, respectively. The amino acid exchange E7K (Cgi-  58 E7K ) did not affect A-Fabp binding (Fig. 5, c and d, shows a control Western blot of the respective COS-7 cell lysates). Overall, these data demonstrate that Cgi-58 binds to A-Fabp via residues in the C-terminal half of the protein and that binding of Cgi-58 occurs independently of its ability to activate Atgl.
Binding of A-Fabp to Cgi-58 Stimulates Atgl-mediated Lipolysis-The impact of A-Fabp on the enzymatic function of Atgl was determined by analyzing neutral lipid hydrolase activities of COS-7 cell lysates overexpressing Atgl, Cgi-58, and/or A-Fabp. Overexpression of Atgl alone led to a 38% increase in TG hydrolase activity compared with samples overexpressing LacZ (negative control). The additional presence of Cgi-58 increased the enzyme activity 3.2-fold, although co-expression of A-Fabp did not increase the TG hydrolase activity. Co-expression of all three proteins, A-Fabp, Atgl, and Cgi-58, led to highest TG hydrolase activities demonstrating that A-Fabp stimulates Atgl-mediated TG hydrolysis in a Cgi-58-dependent manner (Fig. 6, a and b, shows a control Western blot analysis of the respective lysates).
To examine whether A-Fabp inhibition affects TG hydrolase activities of lysates prepared from differentiated 3T3-L1 adipocytes or murine WAT, we used the A-Fabp inhibitor BMS309403 (Fi) in TG hydrolase activity assays. Fi is known to specifically inhibit FA binding of A-Fabp (32). Addition of Fi to 3T3-L1 cell lysates decreased TG hydrolase activity in a dosedependent manner. At an inhibitor concentration of 100 M, TG hydrolase activity was inhibited by 65% (Fig. 6c). We also performed similar experiments using murine adipose tissue lysates and the A-Fabp inhibitor (Fi), the Atgl inhibitor Atglistatin (Ai) (33), and/or the Hsl inhibitor 76-0079 (Hi) (10). In lysates of murine adipose tissue, the presence of Fi decreased total TG hydrolase activity by 85% (Fig. 6d) suggesting that functional A-Fabp is required for lipase function in adipocytes. The addition of Hi decreased TG hydrolase activity in mouse adipose tissue lysates by 56%. This decrease was further potentiated by the presence of Fi suggesting that Fi inhibits lipolysis independently of Hsl. In contrast, Ai did not diminish the remnant hydrolase activity observed in lysates treated with Fi. Thus, FIGURE 5. Identification of Cgi-58 regions involved in binding to A-Fabp. a, in solid phase assays, polystyrene plates were coated with purified A-Fabp and increasing amounts of purified Gst, Gst-Cgi-58, or N-terminally truncated Cgi-58 variants. Binding of proteins was detected using anti-Gst primary and Hrp-conjugated secondary antibody. Plates were developed using tetramethylbenzidine as substrate, and the absorbance was measured at 450/620 nm. b, SDS-PAGE of purified Gst, Gst-Cgi-58, and N-terminally truncated Cgi-58 variants after Coomassie Brilliant Blue staining. c, solid phase assay detecting the interaction between purified A-Fabp and His-tagged human Cgi-58, the naturally occurring mutant variants of human Cgi-58, and LacZ (control) contained in lysates of transfected COS-7 cells. d, Western blot analysis of respective cell lysates using anti-His primary and Cy3-conjugated secondary antibody. Statistical difference was determined as compared with Gst/LacZ control (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001) and of 0.6/60 and 0.3/30 g versus 0.1/10 g of each lysate ( §, p Ͻ 0.05; § §, p Ͻ 0.01; § § §, p Ͻ 0.001).

Fabps Interact with Cgi-58
Atgl-mediated lipolysis is not functional when A-Fabp is inhibited.
Fabps Promote Ppar Signaling-Haemmerle et al. (40) demonstrated that Atgl-mediated lipolysis is required for Ppar␣ function and the expression of Ppar␣ target genes in the heart. A conceivable mechanism linking lipolysis to Ppar activity involves the Fabp-mediated transport of FA from the site of production at LDs to the nucleus, where they act as potential ligands for Ppars (41). To elucidate whether the interaction of Fabps with Cgi-58 affects Ppar activation, we performed luciferase gene transactivation assays in transfected COS-7 cells. The assay was based on a reporter plasmid expressing firefly luciferase under transcriptional control of multiple Ppar-responsive elements (PPRE), which results in the production of luminescence directly proportional to Ppar activity.
Because L-Fabp has been shown to bind to Ppar␣ (42-44), we co-transfected COS-7 cells with the reporter plasmid and various combinations of expression plasmids for Ppar␣ and L-Fabp, Atgl, the enzymatically inactive mutant variant Atgl S47A , and Cgi-58. Western blotting analysis revealed similar expression efficiency for the same constructs (Fig. 7a). We found that the expression of L-Fabp alone increased Ppar␣driven luciferase production by 23%. The additional expression of Atgl and Atgl/Cgi-58 further increased luciferase activity by 54 and 140%, respectively, although the inactive variant of Atgl S47A failed to further induce luciferase activity (Fig. 7b). Additionally, we measured TG hydrolase activity and observed a strong correlation between luciferase activities with TG hydrolase activities (Fig. 7c).
Similar results were obtained when Ppar␥ was co-expressed with Atgl, Cgi-58, and A-Fabp in COS-7 cells (Fig. 8a). The expression of A-Fabp alone increased Ppar␥-driven luciferase production by 36%. The expression of Atgl and Atgl/Cgi-58 increased luciferase activity by 32 and 121%, respectively. Highest firefly luciferase activities were achieved by co-transfection of Ppar␥, Atgl, Cgi-58, and A-Fabp, indicating that the presence of A-Fabp further promotes Ppar␥ activation.
To test whether functional FA binding to A-Fabp was required for the activation of Ppar␥, we performed luciferase gene transactivation assays with A-Fabp R127Q , a variant unable to bind FAs. This mutant Fabp variant did not further stimulate firefly luciferase activities when co-expressed with Ppar␥, Atgl, and Cgi-58 (Fig. 8b), indicating that FA binding is a prerequisite for Ppar␥-driven reporter gene activation. Similar effects were observed when applying the mutant variant A-Fabp F58A with a defect in ligand-induced nuclear translocation (47). Also, A-Fabp F58A failed to increase firefly luciferase activity upon coexpression with Atgl and Cgi-58, implying that functional nuclear import of ligand-bound A-Fabp is essential for A-Fabpmediated Ppar␥ activation (Fig. 8b). To further corroborate these findings, we investigated the subcellular distribution of endogenous A-Fabp in response to forskolin-stimulated lipolysis. Therefore, we analyzed localization of A-Fabp in differentiated, nonstimulated (basal), or forskolin-stimulated 3T3-L1 cells by Western blotting (Fig. 8c). A-Fabp was predominantly detected in the cytosol under both conditions. Notably, however, the amount of A-Fabp in the nuclear fraction markedly increased upon forskolin stimulation arguing for an increased A-Fabp translocation into the nucleus under lipolytic conditions. Thus, both FA binding and nuclear translocation of A-Fabp are required for Atgl/Cgi-58-mediated Ppar␥ stimulation.

Discussion
The identification and characterization of components that constitute a functional lipolysome are essential for a detailed understanding of fat catabolism and energy homeostasis. Several of the recently discovered enzymes and regulatory proteins not only affect lipolysis per se but are also linked to other cellular processes such as inflammation, cancer, apoptosis, or cell signaling. For example, Pedf stimulates lipolysis via Atgl (5) but also has anti-angiogenic, neuroprotective, and anti-tumor activities (45). Other examples include the noncompetitive Atgl inhibitor G0s2 (20) that also interacts with OXPHOS complex V leading to increased cellular ATP synthesis (46) or the Atglbinding partner Gbf-1, which is part of the Arf1/CopI machinery, involved in membrane trafficking pathways (23). Even the indispensable Atgl co-activator Cgi-58 has partially undefined non-Atgl-related functions in liver, skin, and macrophages (47)(48)(49). This plurality of factors regulating TG hydrolysis and their additional biochemical functions beyond lipid degradation highlights the close link of lipolysis with other cellular pathways.
In this study, we show that A-Fabp forms a physical complex with the Atgl co-activator Cgi-58, both in vitro and in living cells. A-Fabp belongs to a conserved nine-member multigene family of intracellular lipid-binding proteins with ubiquitous tissue distribution. Even though the individual Fabp isoforms exhibit marked differences in ligand affinity, selectivity, and binding mechanisms, all Fabps bind long-chain FAs with high affinity (50,51). Because this process is essential to prevent (lipo)toxic effects via FA membrane interaction or micelle formation, we propose that A-Fabp/Cgi-58 binding is required to absorb lipolysis-derived FAs. Interestingly, the plant ortholog of Cgi-58 interacts with the ABC transporter 1 (Pxa1), a protein that transports FAs or lipophilic hormone precursors into peroxisomes for metabolic processing (52). We also show that Cgi-58 interacts with five of nine members of the Fabp family suggesting that the Cgi-58/A-Fabp interaction may be relevant in multiple cell types and tissues, a conclusion that is also consistent with the broad tissue expression pattern of both proteins.
The interface responsible for the protein/protein interaction involves the helix ␣1-turn-helix ␣2 region of A-Fabp and two adjacent loops. Interestingly, single amino acid exchanges within the two adjacent loops led to significantly decreased binding of A-Fabp to Cgi-58. Because members of the Fabp family are highly related and share similar tertiary structures (53), we assume that the Cgi-58 binding interface is analogous in all Fabp isoforms. Furthermore, the mutant variant of A-Fabp that lacks more than 99% of the FA binding capacity of the wild type protein was still able to bind to Cgi-58 suggesting that FA loading of A-Fabp is not required for the interaction between the two proteins.
Notably, using pulldown, co-immunoprecipitation, isothermal titration calorimetry, and FRET experiments, Bernlohr and co-workers (37,54) demonstrated that A-Fabp interacts with Hsl. In contrast to the A-Fabp/Cgi-58 interaction, Hsl only interacts with A-and E-Fabp but not with L-or I-Fabp, and the A-Fabp/Hsl interaction only occurs when A-Fabp is loaded with a FA (36,55). The contact region between A-Fabp and Hsl involves a cluster of four residues (Asp-18, Asp-19, Lys-22, and Arg-31) (37) and the authors hypothesized that this region may be a general motif for protein/protein interactions of A-Fabp. However, solid phase assays with mutant A-Fabp variants revealed that Asp-18, Lys-22, and Arg-31 were not required for binding of A-Fabp to Cgi-58. These data suggest that Hsl and Cgi-58 do not share an identical contact region.
The interaction of A-Fabp and Cgi-58 has important functional consequences. First, A-Fabp promotes Atgl-mediated TG hydrolysis. Consistent with our findings that A-Fabp only binds to Cgi-58 but not to Atgl, the activation of Atgl's hydrolytic activity by A-Fabp depends on the presence of Cgi-58. Apparently, Cgi-58 acts as an adaptor protein enabling A-Fabp to bind the FA produced by Atgl.
Inhibition of A-Fabp by the specific inhibitor BMS309403 led to markedly reduced TG hydrolase activities in WAT and 3T3-L1 cell lysates. Inhibition of TG hydrolase activity by BMS309403 was observed in the presence of an Hsl inhibitor but not in the presence of an Atgl inhibitor providing compelling evidence that A-Fabp stimulates Atgl/Cgi-58-mediated lipolysis. In our view, direct binding of A-Fabp to the Atgl/ Cgi-58 complex or Hsl is required to sequester FAs from the first and the second step of TG hydrolysis. The immediate FA . Relative luminescence units of firefly and Renilla luciferase were determined in a 24-well plate luminometer and calculated relative to Renilla luciferase activities (control values were set to 100%). c, Western blot analysis of cytoplasmic and nuclear fractions obtained from differentiated 3T3-L1 adipocytes under nonstimulated ( ϭ basal, ϪF) and forskolin-stimulated (ϩF) conditions. A-Fabp was detected using an antibody specific for A-Fabp and Hrp-conjugated secondary antibody. Purity of the fractions was assessed by Western blotting using anti-histone H3 and anti-Gapdh antibodies, respectively. Data are shown as mean Ϯ S.D. (n ϭ 4) and are representative of three independent experiments. Statistical difference was determined as compared with control (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001).
capture in statu nascendi likely prevents product inhibition by increased FA concentrations as both Hsl (56) and Atgl (57) are known to be inhibited by high concentrations of oleic acid. In accordance with this conclusion, genetic abrogation of A-Fabp in mice leads to reduced lipolysis (58).
Besides their crucial role as energy substrates and membrane lipid precursors, FAs also act as important signaling molecules. Among many other ligands, they can bind to members of the Ppar family of nuclear receptors that dimerize with retinoid X receptor to regulate the expression of numerous genes involved in energy metabolism (59). The nuclear import of FAs has been shown to involve L-Fabps to activate Ppar␣ in hepatocytes, K-Fabp to activate Ppar␤/␦, or A-Fabp to activate Ppar␥ in adipocytes. Fabps bind to Ppars in the nucleus, and it is assumed that they "hand over" the FA to the Ppar ligand-binding site (40,43). Nuclear translocation of Fabps appears to depend on the stabilization of an "activated" state of Fabps by a subset of small molecule ligands (60 -63).
We show that the transfection of COS-7 cells with either Ppar␣ or Ppar␥ and a luciferase reporter construct under the control of multiple PPREs leads to increased luciferase reporter gene expression when lipolysis is increased by co-expression of Atgl and Cgi-58. Lipolysis-induced transcription increased further when COS-7 cells were additionally transfected with L-Fabp-or A-Fabp-expressing plasmids. Mutant A-Fabp variants that either cannot bind FAs or exhibit impaired nuclear translocation were unable to stimulate Ppar␥-activated reporter gene expression. These data support the concept that the provision of specific FAs by lipolysis, FA binding by Fabps, and the nuclear translocation of Fabps are all important for Ppar signaling. The interaction between Cgi-58 and A-Fabp may play a crucial role in this process. Interestingly, residue Phe-58, which together with the nuclear localization signal residues Lys-22, Arg-31, and Lys-32, which mediates the nuclear import of A-Fabp (62), is also part of the A-Fabp/ Cgi-58 binding interface. Because the interaction of both proteins does not prevent A-Fabp nuclear translocation, we speculate that the binding of A-Fabp to Cgi-58 occurs transiently on the surface of LDs, where FAs are generated by the Atgl reaction. It is also conceivable that the quick FA binding to Fabps prevents their activation by acyl-CoA synthetases and permits their translocation to the nucleus in an unactivated state.
Another important aspect of the A-Fabp/Cgi-58 interaction refers to a FA transport-independent function of A-Fabp. Similar to other regulatory factors involved in lipolysis (e.g. Cgi-58, Pedf, or G0s2), A-Fabp also has demonstrated bioactivities unrelated to its function in FA transport and lipolysis. For example, A-Fabp has been shown to act as an adipokine regulating hepatic glucose production (64) or induces the ubiquitination and degradation of Ppar␥ (65). Whether and how these activities of A-Fabp are affected by its interaction with Cgi-58 remains elusive and requires clarification.
In summary, we show that Fabps interact with the Atgl coactivator Cgi-58. This interaction represents the basis for a novel mechanism to modulate lipolysis and Ppar-mediated gene expression.