CGI-58/ABHD5 is phosphorylated on Ser239 by protein kinase A: control of subcellular localization

CGI-58/ABHD5 coactivates adipose triglyceride lipase (ATGL). In adipocytes, CGI-58 binds to perilipin 1A on lipid droplets under basal conditions, preventing interaction with ATGL. Upon activation of protein kinase A (PKA), perilipin 1A is phosphorylated and CGI-58 rapidly disperses into the cytoplasm, enabling lipase coactivation. Because the amino acid sequence of murine CGI-58 has a predicted PKA consensus sequence of RKYS239S240, we hypothesized that phosphorylation of CGI-58 is involved in this process. We show that Ser239 of murine CGI-58 is a substrate for PKA using phosphoamino acid analysis, MS, and immunoblotting approaches to study phosphorylation of recombinant CGI-58 and endogenous CGI-58 of adipose tissue. Phosphorylation of CGI-58 neither increased nor impaired coactivation of ATGL in vitro. Moreover, Ser239 was not required for CGI-58 function to increase triacylglycerol turnover in human neutral lipid storage disorder fibroblasts that lack endogenous CGI-58. Both CGI-58 and S239A/S240A-mutated CGI-58 localized to perilipin 1A-coated lipid droplets in cells. When PKA was activated, WT CGI-58 dispersed into the cytoplasm, whereas substantial S239A/S240A-mutated CGI-58 remained on lipid droplets. Perilipin phosphorylation also contributed to CGI-58 dispersion. PKA-mediated phosphorylation of CGI-58 is required for dispersion of CGI-58 from perilipin 1A-coated lipid droplets, thereby increasing CGI-58 availability for ATGL coactivation.

Perilipin 1A at the surfaces of lipid droplets is a master regulator of adipose lipolysis. Under basal (fed) conditions, perilipin 1A provides a protective barrier against lipolysis of stored triacylglycerols ( 7 ) and binds CGI-58 (also called ABHD5) ( 8,9 ), a coactivator of ATGL ( 10 ). When CGI-58 is sequestered on the perilipin scaffold, it cannot interact with or activate ATGL ( 11 ). When catecholamines stimulate adipose lipolysis, perilipin 1A is multiply phosphorylated by PKA. The phosphorylation of three N-terminal serine residues of perilipin 1A facilitates the docking of PKA-phosphorylated hormone-sensitive lipase through a protein-protein interaction with the perilipin scaffold ( 12 ), thus bringing the lipase to its substrate lipids. The PKA-mediated phosphorylation of two carboxyl-terminal serine residues of perilipin 1A facilitates the release of CGI-58 from perilipin 1A, enabling interaction of CGI-58 with ATGL ( 11 ), in turn, activating the lipase ( 10 ).
CGI-58 was identifi ed as an important factor in cellular triacylglycerol homeostasis when mutations in CGI-58 were established as the cause of Chanarin Dorfman syndrome ( 13 ), a neutral lipid storage disorder (NLSD) characterized by excessive storage of triacylglycerols in many cells and tissues (13)(14)(15)(16)(17)(18). CGI-58 was later identifi ed as a coactivator of ATGL ( 10 ), thus explaining its role in triacylglycerol turnover; however, the mechanism by which CGI-58 activates ATGL has not yet been elucidated. In this study, we asked whether CGI-58 is a substrate for PKA. We demonstrate that CGI-58 is indeed a substrate for PKA, and show that its phosphorylation is important for the subcellular traffi cking of CGI-58 in adipocytes.

Materials
All chemicals were reagent grade. Growth media for cultured cells were obtained from Mediatech, Inc. (Herndon, VA) or Sigma. FBS, isobutylmethylxanthine (IBMX), forskolin, and protease and phosphatase inhibitor cocktails were purchased from Sigma. The DNA purifi cation kit and nickel-nitrilotriacetic acid agarose matrix were purchased from Qiagen. Coomassie Plus protein assay reagent was purchased from Thermo Scientifi c/ Pierce; Bradford and DC protein assays were purchased from Bio-Rad. PfuUltra high-fi delity DNA polymerase was purchased from Stratagene, Inc. (La Jolla, CA). BODIPY 439/503, Alexa Fluor 546 goat-anti rabbit IgG, Alexa Fluor 546 donkey-anti goat IgG, and Alexa Fluor 488 donkey-anti rabbit IgG were purchased from Molecular Probes. Radioactive compounds were purchased from Perkin Elmer Biosciences. Oligonucleotide primers were purchased from Operon BioTechnologies, Inc. (Huntsville, AL). Phospholipids were purchased from Avanti Polar Lipids, Inc. (Alabaster, AL) and Sigma. Bovine heart PKA catalytic subunit and modifi ed trypsin were purchased from Promega. Peptides were synthesized and purifi ed by Bio-Synthesis, Inc. (Lewisville, TX). P81 phosphocellulose paper was from Whatman International Ltd. Hybond-P polyvinylidene difl uoride (PVDF) membranes to acid hydrolysis with 6 N HCl ( 22 ). Hydrolysates were dried under vacuum and applied to cellulose TLC plates (EM Science) with standard phosphoamino acids, phosphoserine, phosphothreonine, and phosphotyrosine. Amino acids were separated by 2D electrophoresis using formic acid:acetic acid:water (50:156:1,794, v/v) in the fi rst dimension and acetic acid: pyridine:water (100:10:1,890, v/v) in the second dimension ( 23 ). Following electrophoretic separation, the TLC plates were dried and subjected to phosphorimaging analysis with a Molecular Dynamics Storm phosphorimager. Standard phosphoamino acids were visualized by spraying the plate with 0.25% ninhydrin in acetone.

MS
Purifi ed recombinant CGI-58 was phosphorylated in vitro by incubation with PKA and nonradioactive ATP. Phosphorylated residues of CGI-58 were identifi ed by LC-MS/MS after reduction, alkylation, and enzymatic digestion with AspN, modifi ed trypsin (Promega), or chymotrypsin (Roche). Digests of 0.5 g CGI-58 were acidifi ed with 0.5% trifl uoroacetic acid and separated by nano-HPLC (Dionex Ultimate 3000) equipped with aprecolumn (C18, 5 m, 100 Å, 5 × 0.3 mm) and an Acclaim Pep-Map RSLC nanocolumn (C18, 2 m, 100 Å, 150 × 0.075 mm) (all Thermo Fisher Scientifi c, Vienna, Austria). Samples were concentrated on the enrichment column for 2 min at a fl ow rate of 20 l/min with 0.5% trifl uoroacetic acid as an isocratic solvent. Separation was carried out on the nanocolumn at a fl ow rate of 300 nl/min using the following gradients, where solvent A is 0.3% formic acid in water and solvent B is a mixture of 80% acetonitrile in water containing 0.3% formic acid: 0-2 min 4% B, The LC-MS/MS data were analyzed by searching the mammalian SwissProt public database with Proteome Discoverer 1.4 (Thermo Fisher Scientifi c) and Mascot 2.4 (MatrixScience, London, UK). Carbamidomethylation on Cys was entered as a fi xed modifi cation. Oxidation on methionine and phosphorylation on serine or threonine was entered as a variable modifi cation. A precursor mass error tolerance of 10 ppm and a product mass error tolerance of 0.7 Da were used. A maximum false discovery rate of 5% using decoy database search and a Mascot ion score cut off of 30 were chosen as identifi cation criteria.

Measurement of CGI-58 coactivation of ATGL in vitro
Purifi ed recombinant CGI-58 was phosphorylated in vitro by incubation with PKA and nonradioactive ATP. Both phosphorylated and nonphosphorylated proteins were incubated with an Sf9 insect cell lysate expressing recombinant mouse ATGL ( 24 ) and a radioactive triacylglycerol substrate emulsifi ed with phospholipids, as described previously ( 10 ). Fatty acids were extracted from the reaction mixture with solvents and quantifi ed by scintillation counting ( 10 ).
In other experiments, recombinant CGI-58, S239D CGI-58, and S239E CGI-58 were expressed in E. coli , followed by disruption of the cells by sonication and removal of cell debris by centrifugation Expression of proteins was induced with 0.5 mM isopropyl ␤ -D-1thiogalactopyranoside when cells reached an optical density of 0.6 at 600 nm. Induced His-pSumo-CGI-58 cells were grown for 9-12 h, while His-pSumo-ATGL 1-288 and His-pSumo cells were grown for 4 h at 30°C. Expression of the proteins was confi rmed with SDS-PAGE.

Phosphorylation of recombinant mouse CGI-58 and synthetic peptides with PKA
Peptides of 10 amino acids surrounding and including the RKYSS sequence, with peptides containing RKYSS, RKYSA, RKYAS, and RKYAA, were tested for phosphorylation by PKA. Additionally, the PKA-mediated phosphorylation of comparable variants of purifi ed recombinant CGI-58 was examined. Both synthetic peptides and variants of purifi ed recombinant CGI-58 ( ‫ف‬ 0.5-0.8 g) were incubated with the indicated concentrations of PKA and 50 M [ ␥ 32 P]ATP ( ‫ف‬ 230,000-470,000 cpm/nmol) in 10 mM MgCl 2 , 60 mM DTT, 50 mM Tris-HCl (pH 8.0); reactions were incubated for 10 min at 30°C. Reactions were stopped by the addition of 4× Laemmli's sample buffer ( 21 ), followed by SDS-PAGE, immunoblot analysis, and autoradiography, or by spotting the reaction mixtures onto Whatman P81 phosphocellulose fi lters, followed by repeated treatments of the fi lters with 75 mM phosphoric acid and drying. All phosphorylation reactions were performed in triplicate. A unit of PKA was defi ned as the amount of enzyme that catalyzed the formation of 1 nmol of product/min. For some experiments, the PKA phosphorylation reaction was performed with nonradioactive ATP.

Phosphoamino acid analysis
For phosphoamino acid analysis, a portion of a PVDF membrane containing 32 P-labeled recombinant CGI-58 was subjected PBS for 20 min, followed by a PBS wash step, and prepared for microscopy, as described previously ( 31 ). Cells expressing perilipin 1A and CGI-58 were probed with goat polyclonal antisera raised against an N-terminal peptide of perilipin 1A (kindly donated by Dr. Constantine Londos, formerly of National Institutes of Health, Bethesda, MD, deceased) and rabbit polyclonal antisera raised against recombinant mouse CGI-58 ( 8 ), followed by anti-goat Alexa Fluor 546 and anti-rabbit Alexa Fluor 633. Nuclei were stained with Hoechst 33422 and lipid droplets were stained with BODIPY 493/503 (Life Technologies) ( 32 ), each at 0.1 g/ml in PBS for 10 min. Cells were viewed with either a Nikon Eclipse E800 fl uorescence microscope equipped with a Photometrics CoolSNAP EZ digital camera or a LSM510 Meta confocal laser scanning microscope (Zeiss, Oberkochen, Germany) using a 63× oil immersion lens. Images were processed using Zen 2008 and ImageJ software. Cells were manually scored for protein localization patterns by at least two observers blinded to sample identity and automatically scored using the ImageJ JACoP plugin ( 33 ) to determine Manders coeffi cients; more than 50 cells were counted for each determination.

Measurement of cellular triacylglycerol
Human NLSD fi broblasts were transduced with adenovirus to drive the expression of WT CGI-58 or S239A/S240A-mutated CGI-58. At various times after transduction, cells were harvested and lysed in hypotonic lysis solution (10 mM Tris-HCl, 10 mM NaF, 1 mM EDTA, 10 g/ml leupeptin, 100 M 4-(2-aminoethyl)benzenesulfonylfl uoride hydrochloride, and 500 M benzamidine) and homogenized by probe sonication (Branson Sonifi er) for 10 s. Lipids were extracted from a portion of samples using isopropanol:hexane:water (80:20:2), and the lipid phase was collected and evaporated. Lipid extracts of corn oil were used as triacylglycerol standards. The cellular triacylglycerol content was determined using an enzymatic assay from Thermo Electron Trace adapted for use with cultured cells ( 34 ). Absorbance of samples was determined at 540 nm using a VersaMax microplate reader (Molecular Devices). The triacylglycerol measurements were expressed relative to cellular protein content measured by the Bio-Rad DC Protein Assay.
In one experiment, 6 h after transduction of NLSD fi broblasts with intact or S239A/S240A-mutated CGI-58, 10 M forskolin and 0.5 mM IBMX were added to increase adenylyl cyclase activity, in turn, activating PKA. Cells were collected at 0, 0.5, 1, and 2 h for the determination of triacylglycerol levels.

Adipose tissue collection
Animal care and handling were performed in accordance with the standards established by the Austrian Federal Ministry of Science and Research, Division of Genetic Engineering and Animal Experiments (Vienna, Austria) and protocols were approved by an institutional review board. Mice had ad libitum access to food and water under a 12 h light/12 h dark cycle in a temperaturecontrolled environment. Male C57Bl/6 mice (age 8-12 weeks, body weight 20-30 g) were fed a chow diet (Ssniff ® , Soest, Germany) and fasted during the daytime for 6 h prior to euthanization. Gonadal white adipose tissues were collected and cut into small pieces. Tissue samples were incubated in low glucose DMEM supplemented with 2% fatty acid-free BSA, with half of the tissue fragments additionally incubated with 10 M isoproterenol and 0.5 mM IBMX for 20 min at 37°C with gentle shaking. Tissues were then lysed by sonication in ice-cold PBS (pH 7.4), with protease and phosphatase inhibitors. Lysates were centrifuged at 10,000 g for 10 min at 4°C. Protein content of supernatants was determined by Bradford Assay. for 10 min at 2,700 g and determination of the protein concentration of the extracts by Bradford assay. Twenty-six micrograms of the supernatants with variants of recombinant CGI-58 were used in triacylglycerol hydrolase assays with 8 g of partially purifi ed recombinant ATGL 1-288, as described ( 25 ).

Generation of recombinant adenovirus
Adenoviral expression vectors driving the expression of mouse CGI-58 and ␤ -galactosidase were described previously ( 19 ). Adenoviral expression vectors driving expression of the S239A/ S240A-mutated variant of CGI-58 and perilipin 1A were prepared by ligating the cDNA for S239A/S240A-mutated CGI-58, mouse perilipin 1A, or mutated perilipin 1A lacking serine residues in six consensus sites for PKA phosphorylation into the shuttle vector for the AdEasy XL adenoviral vector system (Stratagene), and then following the manufacturer's protocols for recombination of the shuttle vector to make adenoviral expression vectors and for assembly of virions in cultured AD293 cells. The mutated perilipin 1A cDNA encoded alanine substitutions for serine residues at positions 81, 222, 276, 433, and 492 to prevent phosphorylation by PKA, and a glutamate substitution for serine 517 to permit targeting of the mutated perilipin 1A to lipid droplets ( 26 ). Adenoviral preparations were purifi ed over cesium chloride gradients.

Cell culture and adenoviral transduction
Normal human skin fi broblasts (WS1) were obtained from American Type Culture Collection (Manassas, VA); human skin fi broblasts from an individual with NLSD (27)(28)(29) were generously provided by Dr. Rosalind A. Coleman (University of North Carolina, Chapel Hill, NC). Normal and NLSD fi broblasts were cultured as described previously (27)(28)(29). NLSD fi broblasts were transduced with purifi ed adenoviral preparations for expression of intact (WT) mouse CGI-58 (WT) or S239A/S240A-mutated CGI-58; cells were collected at various times following transduction for the determination of triacylglycerol levels.
For other experiments, purifi ed adenoviruses were used to transduce Cos-7 cells or NIH3T3CAR ⌬ fi broblasts. Cos-7 cells were cultured in DMEM supplemented with 10% FBS, 100 units/ ml penicillin, and 100 g/ml streptomycin. NIH3T3CAR ⌬ fi broblasts are NIH3T3 fi broblasts stably expressing a truncated version of the Coxsackie and adenovirus receptor (CAR ⌬ ) lacking the cytoplasmic signaling domain ( 30 ) to increase the uptake of adenoviral vectors. The cells were cultured in DMEM supplemented with 10% calf serum, 100 units/ml penicillin, and 100 g/ml streptomycin, with 800 g/ml G418 to maintain selection of cells expressing CAR ⌬ . Cells were transduced with adenoviruses for the expression of either WT mouse perilipin 1A or the mutated variant of mouse perilipin 1A that lacks six serine residues in PKA consensus sequences. At the same time, the cells were also transduced with either WT CGI-58 or S239/S240A CGI-58 and incubated for 48 h before processing the cells for immunofl uorescence microscopy.

Immunofl uorescence microscopy
Twenty-four hours after transduction, Cos-7 and NIH3T3CAR ⌬ cells expressing perilipin 1A or mutated perilipin 1A and CGI-58 or S239A/240A-mutated CGI-58 were transferred to glass coverslips. Thirty-six hours after transduction, 200 M oleic acid complexed to fatty acid-free BSA at a 4:1 molar ratio ( 7 ) was added to increase the synthesis and storage of triacylglycerol in lipid droplets. Forty-eight hours after transduction, cells were incubated with 10 M forskolin and 0.5 mM IBMX in culture medium supplemented with 1% fatty acid-free BSA for 30 min for stimulated conditions, or in culture medium with 1% fatty acid-free BSA for 30 min for basal conditions. Cells were fi xed with 4% paraformaldehyde in of CGI-58 from vertebrate species, but not in the sequences of CGI-58 orthologs from Caenorhabditis elegans or Drosophila melanogaster ( Fig. 1A ). To test the hypothesis that CGI-58 is a substrate for PKA, we studied the phosphorylation of recombinant mouse CGI-58 that was purifi ed from E. coli lysates ( 19 ). Recombinant CGI-58 was incubated with mammalian PKA and [ ␥ - 32 P]ATP in vitro; the incorporation of radioactive phosphate into CGI-58 was monitored by phosphorimaging analysis of blots and revealed that recombinant CGI-58 is a substrate for PKA ( Fig. 1B ). Phosphoamino acid analysis of the 32 P-labeled protein showed that PKA phosphorylates CGI-58 at a serine residue ( Fig. 1C ).
The kinetics of the in vitro phosphorylation reaction were studied in detail. Phosphorylation of CGI-58 by PKA displayed saturation kinetics and was dependent on time ( Fig. 2A ), concentration of PKA ( Fig. 2B ), concentration of ATP ( Fig. 2C ), and concentration of recombinant CGI-58 ( Fig. 2D ). Analysis of the data using the Michaelis-Menten equation yielded apparent K m values for ATP and recombinant CGI-58 of 14.0 ± 2.8 M and 71.5 ± 13.9 g/ml, respectively.

Data and statistical analyses
Kinetics data were analyzed according to the Michaelis-Menten equation using the GraphPad Prism kinetic model-fi tting program. Data are reported as mean ± standard deviation or standard error and were analyzed using two-way ANOVA and Tukey's post hoc test with SPSS software. Differences between samples were considered signifi cant at P < 0.05.

Recombinant CGI-58 is phosphorylated by PKA in vitro
Analysis of the predicted amino acid sequence of mouse CGI-58 using PROSITE ( 36 ) revealed a PKA consensus sequence of RKYSS containing two potential targets for serine phosphorylation, Ser239 and Ser240. This consensus sequence is highly conserved in the amino acid sequences CGI-58 with the intact RKYSS sequence, and S240A-mutated CGI-58 are good substrates for PKA ( Fig. 3 ). These data suggest that the major site of phosphorylation is Ser239.
MS was used to further study the in vitro phosphorylation of CGI-58. Recombinant mouse CGI-58 was purifi ed from E. coli lysates, incubated with PKA and ATP, reduced, alkylated and digested with trypsin, chymotrypsin, or AspN. Digests were then analyzed by LC-MS/MS using different fragmentation and scanning techniques. A search substrates for PKA, 10-amino acid peptides containing the intact PKA consensus sequence of RKYSS or comparable peptides with alanine substitutions for these serine residues both individually and in combination (S239A, S240A, S239A/S240A) were synthesized. We examined whether these peptides serve as substrates for PKA in vitro. Both the peptide with the intact RKYSS consensus sequence and the S240A peptide were highly phosphorylated ( Table 1 ). In contrast, the S239A peptide was a poor substrate for PKA, showing only 6% of the activity of the peptide with intact RKYSS. The S239A/S240A peptide was not phosphorylated by PKA in vitro. These data indicate that Ser239, in the context of the local amino acid environment, serves as a substrate for PKA in vitro.
We then extended these studies to full-length recombinant CGI-58. Recombinant variants of CGI-58 with the intact RKYSS sequence, or comparable mutations (S239A, S240A, S239A/S240A), were expressed and purifi ed from E. coli lysates, and then incubated with PKA and [ ␥ -32 P]ATP in vitro. Autoradiography of SDS-PAGE gels containing the recombinant proteins revealed that the S239A-mutated variant of recombinant CGI-58 is a poor substrate for PKA, whereas  Ten-amino acid peptides (0.5 mM) were incubated with PKA (0.32 unit/ml) and 50 M [ ␥ -32 P]ATP for 10 min and then spotted onto Whatman P81 phosphocellulose fi lters, followed by rinses with 75 mM phosphoric acid and drying. Incorporation of phosphate into peptides was determined by scintillation counting. Data are mean ± standard deviation of triplicate reactions. observed in the PKA-stimulated tissue when compared with control tissue ( Fig. 4A ). Additionally, CGI-58 was immunoprecipitated from the stimulated and control tissues and immunoblotted with an anti-phospho-S/T PKA substrate antibody or CGI-58 antiserum ( Fig. 4B ). Only the stimulated tissue showed a positive signal for phosphorylation, while similar amounts of CGI-58 were detected in immunoprecipitates from both the stimulated and control samples. Finally, to confi rm that Ser239 is the major site for phosphorylation of CGI-58 by PKA in intact cells, CGI-58 was immunoprecipitated from stimulated and control NIH3T3CAR ⌬ cells expressing intact CGI-58 (WT) or S239A/S240A-mutated CGI-58 ( Fig. 4C ). As expected, only stimulated cells expressing intact CGI-58 showed detectable phosphorylation.

PKA-mediated phosphorylation of Ser239 does not alter CGI-58 function in either coactivation of ATGL in vitro or turnover of triacylglycerol in NLSD fi broblasts
To gain understanding of the function of CGI-58 phosphorylation, CGI-58-mediated coactivation of the triacylglycerol hydrolase activity of ATGL was assessed in vitro. Purifi ed recombinant CGI-58 was incubated with PKA and nonradioactive ATP prior to incubation of phosphorylated and nonphosphorylated CGI-58 with Sf9 insect cell lysates containing recombinant mouse ATGL and an emulsifi ed radioactive triacylglycerol substrate. Phosphorylated and nonphosphorylated CGI-58 showed equivalent capacity to activate the triacylglycerol hydrolase activity of ATGL over hydrolysis catalyzed by ATGL alone ( Fig. 5A ). Similar results were obtained when CGI-58 with S239D or S239E phosphomimetic mutations were used to activate a partially purifi ed truncated variant (amino acids 1-288) of recombinant ATGL ( Fig. 5B ). The data indicate that the phosphorylation of CGI-58 neither facilitates nor impedes the coactivation of ATGL.
To test CGI-58 function in intact cells, adenoviral expression vectors were employed to drive the expression of either intact CGI-58 (WT) or S239A/S240A-mutated CGI-58 in cultured human NLSD fi broblasts that lack functional CGI-58 ( 10 ), and have perilipin 2-containing lipid droplets (not shown). Immunoprecipitations revealed that NLSD cells express ATGL (data not shown), yet are unable to turn over triacylglycerol normally in the absence of CGI-58 ( 27 ). Without ectopic expression of CGI-58, NLSD cells expressing ␤ -galactosidase as a control protein show approximately 15-fold higher triacylglycerol levels relative to fi broblasts from control humans cultured in the same media (supplementary Fig. 1). By 6 h after transduction with adenoviruses for either the intact (WT) or mutated variant of CGI-58, the triacylglycerol content of NLSD cells decreased signifi cantly (relative to untransduced cells; t = 0 h), and continued to decrease over 48 h ( Fig. 5C ). The rate of triacylglycerol turnover was similar for cells expressing WT and mutated CGI-58. Immunoblotting of cell lysates for CGI-58 revealed that 6 h is suffi cient to detect ectopic CGI-58 (data not shown); levels of CGI-58 increased with longer time of protein expression. Moreover, equivalent levels of intact and mutated CGI-58 were of the mammalian SwissProt database unambiguously identifi ed Ser239 as the major phosphorylation site in CGI-58 peptides derived from three different digests ( Table 2 ). The majority of peptides containing Ser239 were phosphorylated.
To determine whether CGI-58 is a substrate for PKA in intact cells or tissues, two types of experiments were conducted. Murine white adipose tissue was stimulated ex vivo with isoproterenol and IBMX to activate adenylyl cyclase, sustain elevated levels of cAMP, and, in turn, activate PKA. Immunoblots of 2D gels of the tissue lysates revealed three to four distinct spots detected by the CGI-58 antiserum that likely represent CGI-58 isoforms. Because phosphorylation adds negative charge, the isoelectric point of a phosphorylated protein is shifted to an acidic pH, as was 1A lacking serine residues in six consensus PKA sites. Adenoviral titers were adjusted to drive equivalent levels of protein expression for both variants of perilipin 1A and both variants of CGI-58 (supplementary Fig. 2). For 12 h prior to starting the experiment, the cells were incubated with oleic acid to increase triacylglycerol synthesis and lipid droplet formation, and consequently, stabilize perilipin 1A ( 38 ). Cells were then incubated for 30 min with forskolin and IBMX, or under basal conditions, before fi xation and immunostaining of the cells for perilipin and CGI-58. The data show that both intact (WT) and S239A/ S240A-mutated CGI-58 localized to perilipin-coated lipid droplets under basal conditions, whether WT or mutated variants of perilipin 1A were expressed ( Fig. 6A, B ; Table  3 ). In contrast, the mutation of PKA-site serine residues in CGI-58 and perilipin 1A decreased the dispersion of CGI-58 into the cytoplasm following incubation of the cells with forskolin and IBMX, whether only one protein was mutated or both were mutated in combination. Thus, the PKA-mediated phosphorylation of both CGI-58 and perilipin 1A contribute to the dispersion of CGI-58 from perilipin 1A-coated lipid droplets.
We conducted a similar experiment using cells expressing S239D CGI-58 and perilipin 1A. Under basal conditions, S239D CGI-58 localized effi ciently to perilipin 1A-coated lipid droplets (supplementary Fig. 3), suggesting that the addition of negative charge to amino acid 239 of CGI-58 is insuffi cient to cause dispersion of CGI-58 from perilipin 1A-coated lipid droplets without the activation of PKA. When cells were stimulated with forskolin and IBMX, S239D CGI-58 effi ciently dispersed into the cytoplasm (supplementary Fig. 3). These data suggest that negative charge at position 239 is important for CGI-58 dispersion when PKA has been activated.

DISCUSSION
In this study, we employed a variety of approaches to demonstrate that CGI-58 is phosphorylated on Ser239 by PKA. The fi nding is supported by data from the phosphorylation of recombinant CGI-58 in vitro, immunoblotting of immunoprecipitated proteins from mammalian tissue and cells in which phosphorylation occurred in intact cells, and MS. All of our studies were conducted using mouse CGI-58; however, human CGI-58 is likely phosphorylated, given the complete conservation of the PKA consensus sequence in chordates. detected ( Fig. 5D ). In summary, S239A/S240A-mutated CGI-58 was as effective as WT CGI-58 in facilitating the turnover of triacylglycerols in NLSD cells.
Initial experiments to test CGI-58 function in NLSD fibroblasts were conducted in the absence of compounds to activate PKA. We next investigated whether the activation of PKA would alter the rate of triacylglycerol turnover in NLSD cells expressing ectopic CGI-58. Adenoviruses driving the expression of either intact (WT) or S239A/S240Amutated CGI-58 were added to NLSD cells 6 h prior to the addition of forskolin and IBMX. Measurement of the triacylglycerol content of cell lysates over the next 2 h revealed no increase in the rate of lipolysis due to the activation of PKA, and no difference in the rate of triacylglycerol turnover when comparing data from cells expressing WT CGI-58 to those from cells expressing S239A/S240A-mutated CGI-58 ( Fig. 5E ). Thus, PKA is not a major modulator of lipolysis in NLSD fibroblasts, and phosphorylation of Ser239 (or Ser240) is not required for the function of CGI-58 in facilitating triacylglycerol turnover in human skin fibroblasts. It is important to note that fi broblasts lack both perilipin 1 and hormone-sensitive lipase; hence, the major players in PKA-mediated lipolysis are absent in these cells.

PKA-mediated phosphorylation of Ser239 alters the subcellular localization of CGI-58 in forskolin-treated cells expressing perilipin 1A
We next asked whether the PKA-mediated phosphorylation of CGI-58 affects subcellular localization of the protein. PKA controls lipolysis in adipocytes in part by modulating the subcellular localization of CGI-58 ( 8,11,37 ). Under basal conditions, CGI-58 binds to perilipin 1A on adipocyte lipid droplets, but when PKA is activated, CGI-58 disperses into the cytoplasm ( 8,11 ). The PKAmediated phosphorylation of carboxyl terminal serine residues of perilipin 1A has been shown to facilitate the dispersion of CGI-58 from lipid droplets ( 11 ); this event is necessary for CGI-58 to gain access to ATGL. We hypothesized that the PKA-mediated phosphorylation of CGI-58 might also contribute to the translocation of CGI-58 into the cytoplasm in lipolytically stimulated cells. Experiments were conducted in both cultured Cos-7 cells and NIH3T3CAR ⌬ fi broblasts; the latter cells lack detectable endogenous CGI-58, as determined by immunoprecipitation followed by immunoblotting (Fig. 4C). Using adenoviral expression vectors, either intact CGI-58 (WT) or S239A/S240A-mutated CGI-58 was expressed with either intact (WT) perilipin 1A or a mutated variant of perilipin PKA-mediated phosphorylation of CGI-58 also plays a role in the dispersion of CGI-58 from the perilipin scaffold, to enable CGI-58 interaction with and coactivation of ATGL. A previous study has demonstrated that perilipin 1A sequestration of CGI-58 prevents the interaction of CGI-58 and ATGL, reducing basal lipolysis; human truncation mutations in perilipin 1A that prevent CGI-58 sequestration lead to elevated rates of basal lipolysis ( 45 ). Thus, precise control of the subcellular localization of CGI-58 is an important component of the regulation of lipolysis in adipocytes.
Although the PKA-mediated phosphorylation of CGI-58 is an important step in the activation of lipolysis for cells in which perilipin 1A is the major lipid dropletassociated protein, it is likely less important for cells in In adipocytes, the PKA-mediated phosphorylation of key mediators of lipolysis is a crucial event in the initiation of lipolysis. Phosphorylation of hormone-sensitive lipase is required to trigger movement of the lipase from a diffuse cytoplasmic localization to the surfaces of lipid droplets (39)(40)(41)(42), and to activate lipase activity in an as yet poorly understood mechanism ( 12,43,44 ). The phosphorylation of perilipin 1A is also required to enable hormone-sensitive lipase docking on lipid droplets through protein-protein interactions between the lipase and perilipin ( 12 ). Additionally, phosphorylation of perilipin 1A is required to release CGI-58 from its binding site on a carboxyl terminal sequence of perilipin 1A ( 8,11 ); this redistribution of CGI-58 into the cytoplasm increases the interaction of CGI-58 with ATGL ( 11 ). We now demonstrate that skin fi broblasts, have small lipid droplets coated with perilipin 2 ( 35 ), which does not appear to be phosphorylated by PKA. In these cells, it is unlikely that phosphorylation of CGI-58 plays a role in facilitating lipolysis. Consistent with this concept, the addition of forskolin and IBMX to cultured fi broblasts from an individual with NLSD did not accelerate lipolysis in the presence of ectopic CGI-58. Moreover, CGI-58 does not localize strongly to lipid droplets when perilipin 2 is the predominant lipid droplet-associated protein, but instead displays a diffuse localization pattern throughout the cytoplasm ( 8,49 ). It is possible that PKA-mediated phosphorylation of CGI-58 plays a role in lipolysis in myocytes from cardiac and skeletal which lipid droplets are coated with other members of the perilipin family. Perilipin 1A is an abundant lipid droplet protein in adipocytes from white and brown adipose tissue ( 46 ), is less abundant in steroidogenic cells of adrenal cortex, testes, and ovaries ( 47 ), and is largely absent from other cells and tissues under normal physiological conditions. In steroidogenic cells, hormonal signaling activates adenylyl cyclase, in turn, activating PKA, and initiating the lipolysis of stored cholesterol esters to provide substrate for steroid hormone synthesis ( 48 ). Although CGI-58 is expressed in steroidogenic cells, it has not been shown to play a role in lipolysis of cholesterol esters. Most other cells of the body, including human PKA-mediated phosphorylation of Ser239 does not alter the role of CGI-58 in coactivation of ATGL or turnover of triacylglycerol in NLSD fi broblasts. A: Purifi ed recombinant CGI-58 was phosphorylated in vitro prior to incubation with Sf9 cell lysates containing recombinant ATGL and radioactive triolein emulsifi ed with 3:1 phosphatidylcholine:phosphatidylinositol. Released fatty acids were solventextracted and quantifi ed by scintillation counting. Phosphorylated and nonphosphorylated CGI-58 were equally effective in increasing the triacylglycerol hydrolytic activity of ATGL. Experiment is a representative experiment out of three, each with triplicate samples; data are mean ± standard deviation of triplicate samples. B: Recombinant WT or S239D-or S239E-mutated CGI-58, or extracts from E. coli expressing the control vector driving the production of a 6-His-Sumo protein (smt), were mixed with partially purifi ed recombinant truncated ATGL (amino acids 1-288) and radiolabeled triacylglycerol substrate emulsifi ed with phospholipids for determination of ATGL activity. WT CGI-58 and S239D-and S239E-mutated CGI-58 were equally effective in activating triacylglycerol hydrolytic activity of ATGL in vitro. Data are mean ± standard deviation of triplicate samples from one representative experiment out of two. Data were analyzed by ANOVA; data with the same letter subscript are not different from each other; data with different letters are different ( P < 0.05). C: Adenoviral vectors were used to express either WT CGI-58 or S239A/S240A-mutated CGI-58 in NLSD fi broblasts. At the indicated times, lipids were solvent-extracted and the triacylglycerol content determined by an enzymatic assay. Triacylglycerol levels decreased at 6 h and all subsequent time points. Data are the mean ± standard error of the mean for triplicate samples from a representative experiment out of two. Data were analyzed by ANOVA; no signifi cant differences were observed between WT CGI-58 and S239A/S240A CGI-58. D: Immunoblotting was used to determine the relative expression of intact (WT) CGI-58 and S239A/S240A-mutated CGI-58 at 24 h (D1) and 48 h (D2). E: Adenoviral vectors were used to express WT CGI-58 and S239A/S240A-mutated CGI-58 in NLSD fi broblasts. Six hours after the addition of adenoviruses, cells were incubated with forskolin and IBMX. Lipids were extracted and triacylglycerol levels determined. The addition of forskolin and IBMX to NLSD cells does not increase the rate of triacylglycerol turnover in the presence of either WT or mutated CGI-58. Data are mean ± standard error of the mean for triplicate samples. When error bars are not visible they are contained within the symbols.
Although the phosphorylation of CGI-58 increases the availability of CGI-58 to bind to ATGL in adipocytes, it does not affect the coactivation mechanism itself. In vitro lipolysis assays suggest that phosphorylated CGI-58 activates ATGL as effi ciently as nonphosphorylated protein.
The mechanism by which CGI-58 increases ATGL-mediated hydrolysis of triacylglycerols is as yet unknown. Control of ATGL activity is complex; triacylglycerol hydrolase activity is inhibited by a small protein named G0S2 ( 54,55 ), yet there does not appear to be competition between G0S2 and CGI-58 for access to ATGL. The mechanism of G0S2 inhibition of ATGL is as yet unknown, as is how a transition muscle; perilipin 5 is a major lipid droplet protein in these cells (50)(51)(52). CGI-58 localizes to lipid droplets by binding to carboxyl-terminal sequences of perilipin 5, but does not appear to interact with ATGL when bound to perilipin 5 ( 49 ). Recent experiments have suggested that PKA-mediated phosphorylation events increase lipolysis in cells expressing perilipin 5, and that perilipin 5 may be a substrate for PKA ( 53 ). Additional experimentation is required to determine whether the phosphorylation of CGI-58 facilitates lipolysis in cells expressing perilipin 5 through a dispersion mechanism similar to that observed in perilipin 1A expressing cells.  6. CGI-58 and S239/S240A-mutated CGI-58 localize equally well to lipid droplets coated with perilipin 1A, but mutation of serine residues in PKA consensus sequences of either CGI-58 or perilipin 1A reduces forskolin-induced dispersion of CGI-58 into the cytoplasm. A: Representative micrographs of Cos-7 cells expressing WT CGI-58 [CGI-58-WT (i, ii)] or S239A/S240A-mutated CGI-58 [CGI-S239A/ S240A (iii, iv)] and WT perilipin 1A [Plin1A-WT (i, iii)] or perilipin 1A with mutations in six serine residues within PKA consensus sequences [Plin1A-all6 (ii, iv)] under basal (left side panels) or forskolin + IBMX stimulated conditions (right side panels). Cells were fi xed and stained with antisera against perilipin and CGI-58, Hoechst 33422 for the detection of nuclei, and BODIPY 493/503 for the detection of lipid droplets. B: Averaged cell counts for two observers counting more than 50 cells per condition for both basal (left side) and forskolin + IBMX incubated (right side) cells. "Lipid droplets" designates signal for CGI-58 primarily on lipid droplets; "On/Off" designates signal for CGI-58 both on lipid droplets and diffuse throughout the cytoplasm; "Dispersed" designates signal for CGI-58 primarily diffuse throughout the cytoplasm. Experiment was repeated at least three times in each of two locations.
is made between G0S2 inhibition and CGI-58 activation of lipase activity.
We have shown that Ser239 of CGI-58 is phosphorylated by PKA in vitro, as well as in tissue samples and intact cultured cells. Interestingly, 2D electrophoresis revealed three to four isoforms of CGI-58, suggesting additional posttranslational modifi cations. Additional studies are needed to address this hypothesis. Cos-7 cells expressing WT CGI-58 ( Fig. 6Ai, ii ; CGI-WT) or S239A/S240A-mutated CGI-58 ( Fig. 6Aiii, iv ; CGI-S239A/S240A), and WT perilipin 1A (Plin1A-WT) or perilipin 1A with mutations in six serine residues within PKA consensus sequences (Plin1A-all6) were incubated under basal or stimulated conditions before fi xation and staining for CGI-58, perilipin 1, nuclei, and lipid droplets. Micrographs were analyzed automatically with the ImageJ JACoP plugin. Colocalization of CGI-58 with either perilipin 1A or lipid droplets was assessed by determination of Manders coeffi cients 2 (M2) (mean ± standard error of the mean). An M2 of 0.00 indicates no colocalization, while 1.00 indicates complete colocalization.