Diabetes Volume 65, December 2016 3561

Mainak Ghosh,1 Sougata Niyogi,1 Madhumita Bhattacharyya,2 Moumita Adak,1 Dipak K. Nayak,3 Saikat Chakrabarti,2 and Partha Chakrabarti1

Ubiquitin Ligase COP1 Controls Hepatic Fat Metabolism by Targeting ATGL for Degradation

Diabetes 2016;65:3561–3572 | DOI: 10.2337/db16-0506

Optimal control of hepatic lipid metabolism is critical for accumulation of triacylglycerol (TAG) in the liver (4), organismal metabolic fitness. In liver, adipose triglycer- which is caused by defects in lipid accumulation and mo- ide lipase (ATGL) serves as a major triacylglycerol (TAG) bilization (5,6). lipase and controls the bulk of intracellular lipid turnover. Adipose triglyceride lipase (ATGL) is the first and However, regulation of ATGL expression and its functional rate-limiting enzyme for the breakdown of cellular TAG implications in hepatic lipid metabolism, particularly in the (7–10). Mutation in the ATGL gene causes neutral lipid context of fatty liver disease, is unclear. We show that E3 storage disease and myopathy in humans (11). ATGL ex- ubiquitin ligase COP1 (also known as RFWD2) binds to the pression in adipose tissue is transcriptionally regulated by METABOLISM consensus VP motif of ATGL and targets it for proteasomal insulin through FoxO1 and EGR1, directly by peroxisome degradation by K-48 linked polyubiquitination, predom- proliferator–activated receptor-g (12–14), and posttran- inantly at the lysine 100 residue. COP1 thus serves as a scriptionally by G0S2 and CGI-58 (15). The importance critical regulator of hepatocyte TAG content, fatty acid of ATGL was evidenced by ectopic lipid accumulation in mobilization, and oxidation. Moreover, COP1-mediated regulation of hepatic lipid metabolism requires opti- many tissues of ATGL-null mice, including cardiac muscle, mum ATGL expression for its metabolic outcome. In vivo, skeletal muscle, and the liver (7,10). ATGL serves as the fi adenovirus-mediated depletion of COP1 ameliorates high- major TAG lipase in the liver (16), and liver-speci c ATGL fat diet–induced steatosis in mouse liver and improves liver knockdown or deletion in mice reveals progressive hepatic fl function. Our study thus provides new insights into the steatosis and in ammation (17) as well as changes in the regulation of hepatic lipid metabolism by the ubiquitin- lipid droplet (LD) lipidome (18) with uncoupling of glu- proteasome system and suggests COP1 as a potential cose tolerance from liver TAG accumulation (19). Lowered therapeutic target for nonalcoholic fatty liver disease. ATGL expression has also been found in patients with NAFLD (6). Molecular regulation of hepatic ATGL remained elusive, however. Nonalcoholic fatty liver disease (NAFLD) is the most Regulated cellular protein turnover via the ubiquitin- common chronic liver disease and is strongly associated proteasome system underlies a wide variety of signaling with obesity and type 2 diabetes (1). NAFLD describes a pathways, from cell-cycle control and metabolic homeo- spectrum of conditions characterized mainly by the his- stasis to development (20). Stepwise ubiquitination of a tological finding of macrovesicular hepatic steatosis (2) target protein is achieved through three enzyme classes: and now considered to be the hepatic component of the E1-ubiquitin–activating enzymes, E2-ubiquitin–conjugating metabolic syndrome (3). The hallmark feature of NAFLD enzymes, and E3 ubiquitin ligases (21,22). COP1 is an evo- pathogenesis, both histologically and metabolically, is the lutionarily conserved E3 ubiquitin ligase (23) essential for

1Division of Cell Biology and Physiology, Council of Scientific and Industrial This article contains Supplementary Data online at http://diabetes Research–Indian Institute of Chemical Biology, Kolkata, India .diabetesjournals.org/lookup/suppl/doi:10.2337/db16-0506/-/DC1. 2 fi Division of Structural Biology and Bioinformatics, Council of Scienti cand M.G., S.N., and M.B. contributed equally to this work. Industrial Research–Indian Institute of Chemical Biology, Kolkata, India © 2016 by the American Diabetes Association. Readers may use this article as 3Nuclear Medicine Division, Council of Scientific and Industrial Research–Indian long as the work is properly cited, the use is educational and not for profit, and the Institute of Chemical Biology, Kolkata, India work is not altered. More information is available at http://www.diabetesjournals Corresponding author: Partha Chakrabarti, [email protected] or partha.iicb@ .org/content/license. gmail.com. Received 25 April 2016 and accepted 14 September 2016. 3562 COP1 Ubiquitinates ATGL Diabetes Volume 65, December 2016 mouse development, because COP1-knockout mice were acquisition was done over 30 min using 180 time frames embryonically lethal (24). COP1 regulates the stabilities of 10 s each. of p53 (25), c-Jun (26), and acetyl-CoA carboxylase 1 (27), Cell Culture and Transfection each through different mechanisms. Recent discoveries Human hepatoma cell line HepG2, HUH7, and human have documented important roles of COP1 in the regula- embryonic kidney cell line HEK293A were cultured in high- tion of intermediary metabolism, including glucose (23,28) glucose DMEM supplemented with 10% FBS containing and lipid metabolism (27). The importance of COP1 in 1% penicillin/streptomycin. Transient transfections with mediating insulin secretion from pancreatic b-cells appre- plasmids and small interfering (si)RNA were done using ciates the role of COP1 as a master regulator of whole-body Lipofectamine 2000 reagent (Life Technologies) according glucose homeostasis (29). to the manufacturer’s instruction. For stable expression fi Inthisstudywehaveidenti edhepaticATGLasanovel and knockdown of ATGL, cells were transfected with target of COP1, and their interaction controls hepatic TAG pcDNA3.1-b myc-his vector and infected with lentivirus turnover. Moreover, depletion of COP1 in the liver abrogated generated from pooled shATGL plasmids (Santa Cruz Bio- hepatic steatosis, thereby suggesting that COP1 could be a technology), respectively. siRNA for human RFWD2 against novel target for ameliorating lipid accumulation in NAFLD. the GCUGUGGUCUACCAAUCUA sequence was from Euro- RESEARCH DESIGN AND METHODS gentec, Liège, Belgium. Antibodies Adenovirus Production Antibodies to ATGL, HA-tag, myc-tag, and GAPDH were Recombinant adenoviruses containing shCOP1 construct from Cell Signaling Technology (Boston, MA); the COP1 were generated by the BLOCK-iT Adenoviral Expression antibody was from Bethyl Laboratories Inc.; anti-Ub anti- System (Invitrogen) using the sequence CACCGAGTCCA body was from Santa Cruz Biotechnology; and anti-FLAG ATGTGCTGATTGC (28). Recombinant adenoviruses were M2 and b-actin antibody were from Sigma-Aldrich. purified and concentrated using the FAST-TRAP adenovi- rus purification and concentration kit (Millipore). Plasmids and Vectors Histology Human ubiquitin-HA and COP1-myc-FLAG clones were For histological analysis, tissues were fixed in 10% formal- from Addgene and OriGene, respectively. ATGL was cloned dehyde in PBS, embedded in paraffin, sectioned at 10 mm, in pcDNA 3.1(-) b myc-his vector using the forward primer and stained with hematoxylin and eosin (H&E) following 59 TCACCTCGAGATGTTCCCGAGGGAGACCAAGTGG 39 and standard staining protocol. reverse primer 59 TAGAAGCTTGGGCAAGGCGGGAGGC CAGGTGGATC 39 containing Xho1 and HindIII restric- Immunoprecipitation and Ubiquitination Assay tion sites, respectively. Mutagenesis of the ATGL clone Cells were harvested in radioimmunoprecipitation assay was done using the QuickChange II Site-Directed Muta- buffer (10 mmol/L Tris-HCl [pH 8.0], 1 mmol/L EDTA, genesis Kit (Agilent). Details of primers for mutagenesis 0.5 mmol/L EGTA, 1% Triton X-100, 0.1% SDS, 140 mmol/L are in Supplementary Table 1. NaCl with protease, and phosphatase-inhibitor cocktail; Roche). Then, 200 mg of protein was incubated overnight Animal Experiments with 2 mg primary antibody and 30 mLofProteinAmag- – C57BL/6 male mice 8 10 weeks of age were divided in two netic beads at 4°C. Immune complexes were separated by groups for normal chow and high-fat diet (HFD) containing Magnetic GrIP RAC (Millipore) and washed thrice. For af- 45% fat and 5.81 kcal/g diet energy content (# 960192; finity purification of COOH-terminal hexahistidine contain- MP Biomedicals). Animals were fed the HFD for 4 weeks, ing ATGL, cells were harvested in radioimmunoprecipitation 3 9 and 8 10 plaque-forming units short hairpin (sh)COP1 assay buffer and incubated overnight with 30 mLofNi adenovirus was injected retro-orbitally. Animals were sac- Sepharose beads (Roche). Beads were then warmed at fi ri ced after 1 week. All data are representative of two 60°C for 10 min in 23 sample buffer and centrifuged. independent experiments with four to seven animals per b-Marcaptoethanol was added to the supernatant to make group. The experimental protocols were approved by the In- the final concentration 5%. For immunoprecipitation, cells stitutional Animal Ethics Committee (approved by CPCSEA, were harvested in cell lysis buffer containing 20 mmol/L Ministry of Environment & Forest, Government of India). Tris-HCl (pH 7.4), 100 mmol/L NaCl, 2.5 mmol/L MgCl2, Technetium-99m–Mebrofenin Liver Activity Imaging and 0.05% Nonidet P-40 for 5 h on ice. Immunoprecipita- 99m 2 tion was performed as described previously and analyzed by Technetium-99m ( Tc)O4 was obtained by 2-butanone 99 2 Western blotting. extraction of a 5N NaOH solution of MoO4 , procured from Bhabha Atomic Research Centre (Mumbai, India). Western Blot Analysis 99mTc-labeled mebrofenin (15–18 MBq/animal) was admin- Proteins were separated by SDS-PAGE and transferred istered to each anesthetized animal through a tail vein. onto polyvinylidene fluoride membrane (Millipore). Mem- Scintigraphic imaging of animals was done in a GE Infinia branes were incubated in 5% nonfat dry milk, followed gamma camera equipped with a Xeleris Work Station. by incubation with primary and secondary antibody. Dynamic single-photon emission computed tomography Membranes were developed using Clarity Western ECL diabetes.diabetesjournals.org Ghosh and Associates 3563

Blotting substrate (Bio-Rad), and images were captured in Molecular Modeling and Docking the Bio-Rad Gel Documentation system using ImageLab Homologs of ATGL and COP1 were searched in the Software. structural databases Protein Data Bank (PDB) (32) and Structural Classification of Proteins (SCOP) (33). The Quantitative PCR highest identity and coverage was selected, and sequences Total RNA from cell homogenates was isolated using were subjected to secondary structures prediction using Trizol (Invitrogen), and cDNA was synthesized using the online servers Psi-Blast Based Secondary Structure iScript Reverse Transcription Supermix (Bio-Rad). The Prediction (PSIPRED) (34) and Phyre2 (35). Homology gene expression was measured by SYBR green chemistry modeling by Modeler V9 (36) and ab initio modeling by (Fast Start Universal SYBR Green Master; Roche) in the online server Iterative Threading Assembly Refine- LightCycler 96 real-time PCR (Roche) and normalized by ment (I-TASSER) (37) were both used to build full-length 18S RNA expression by the DDCt method. The primer structures. Loop modeling and ligand modeling was per- sequences are in Supplementary Table 2. formed, and minimal energy full-length models were Fatty Acid Oxidation, TAG, and Cholesterol Assay selected and optimized by the Normal Mode Analysis, Fatty acid (FA) oxidation (FAO) was measured as de- Deformation and Refinement (NOMAD-REF) server (38) 14 fi scribed earlier (30). C-labeled released CO2 was esti- using GROMACS force elds. Models were subjected to mated using a b-counter (PerkinElmer) and normalized validation tools, such as Verify3D (39), which determines to protein concentrations. Lipids were extracted from cells the compatibility of an atomic model (three dimensional) or from the liver using hexane/isopropyl alcohol (3:2, with its own amino acid sequence (one dimensional) and vol/vol). TAG was assayed using the Serum Triglyceride Rampage (40), which checks the stereo chemical quality of Determination Kit (TR0100; Sigma-Aldrich). Cholesterol protein structure by analyzing a Ramachandran plot. Sol- levels were assayed in extracted lipid or in animal serum vation free energy of folding for each model was calcu- using ferric chloride/sulfuric acid reagent (31). TAG and lated by the Protein Data Bank in Europe Proteins, cholesterol content were normalized with cellular protein Interfaces, Structures and Assemblies (PDBePISA) web or liver weight, and data were expressed as means 6 SD server (41) and compared with the corresponding tem- for duplicate samples repeated thrice. plates. Stand-alone version of PatchDock 1.1 package (42) was used for protein-protein docking in a generalized Oil Red O Staining protocol. Directed docking was done to generate 100 solu- Oil Red O staining was performed as described before tions, followed by refinement and rescoring by the FireDock (30). Photographs were taken with a Canon SX10 Power- refinement package (43) and clustered based on their overall Shot camera under optimal illumination. orientation using ensemble clustering of Chimera1.8.1 (44). FA Mobilization Assay, Confocal Microscopy, and The top three largest clusters were inspected manually and Image Analysis statistically analyzed. Top 10 solutions identified by geo- At 30 h after transfection, cells were incubated with metrical shape complementarity score (by PatchDock), 1 mmol/L boron-dipyrromethene (BODIPY) 558/568 C12 solvation free energy of binding (by PDBePISA), and fi (Red C12; Thermo Fisher Scienti c) for 6 h. Cells were size of interface area (by PDBePISA). Finally, the best so- washed and kept in complete media. Subsequently, lution was identified by manual inspection, which satis- 100 mmol/L oleic acid-BSA (both Sigma-Aldrich) conjugate fied favorable interaction between two proteins. was treated for 16 h. BODIPY 493/503 (Life Technologies) For further validation of the selected docking interface, was used to label LDs at a concentration of 200 ng/mL, eight protein complexes were identified from PDB having 20 min before imaging. Hoechst 33342 (5 mg/mL; similar interface area (6 100 Å2) by the PDBePISA web fi Thermo Fisher Scienti c) was added for 5 min to stain server (41) and refined by distance calculation. Residues 3 the nucleus. Cells were washed twice with 1 PBS and of two different proteins, located at the distance #10 Å, viewed under a Leica TCS (True Confocal System) SP8 were considered at interface, and residues located #5Å microscope with HC PLAN APO objective (original mag- apart were considered to be interacting. Hydrogen bonds fi 3 3 ni cation: 63*6 /1.40, 63*1.83 /1.40). Images were and salt bridges were identified by the PDBePISA web analyzed using LAS X (Leica) and ImageJ (National In- server. Non-Polar contacts were identified by an in-house stitutes of Health) software. Brightness and contrast were perl script. Approximate ΔG of binding (solvation free en- adjusted in Adobe Photoshop Elements 10. LD number ergy gained upon complex formation) was calculated. The was calculated manually for at least 30 cells from a dif- protein complexes used for comparison of binding proper- fi fi 3 ferent eld (original magni cation: 63*1.83 /1.40) for ties with respect to the ATGL-COP1 complex are reported each experimental condition. For the FA mobilization in Supplementary Table 3. assay, the area ratio of the red channel (C12 BODIPY 558/568) and green channel (BODIPY 493/503) was cal- Statistical Analysis culated from the thresholded image using the Analyze Par- The Student unpaired two-tailed t test was used to eval- ticles tool of ImageJ software. This ratio was called the FA uate the statistical significance between two groups. Sta- mobilization index. tistical significance was predefined as P # 0.05. 3564 COP1 Ubiquitinates ATGL Diabetes Volume 65, December 2016

RESULTS HepG2 cell lysate and found that endogenous ATGL and ATGL Is Ubiquitinated and Proteasomally Degraded endogenous ubiquitin were reciprocally coprecipitated in Hepatocytes (Fig. 1C). To assess whether ATGL is ubiquitinated in hepatocytes, ATGL Ubiquitination Occurs Predominantly on myc-ATGL and hemagglutinin (HA)-ubiquitin constructs Lysine100 Residue in the Patatin-Like Domain Via were cotransfected in HepG2 and HUH7 cells, and ATGL Lysine 48 Residue of Ubiquitin was immunoprecipitated using myc-tag antibody. Im- We next sought to identify the lysine residue(s) of ATGL mune complexes were analyzed by immunoblotting using that are ubiquitinated. To this end, we first examined HA-tag antibody. In both cell lines, cotransfection of the polyubiquitination of different ATGL domains. The ATGL and ubiquitin resulted in the development of pro- polyubiquitin signal of ATGL was obtained in full-length tein smears that indicate an accumulation of polyubiqui- (486 amino acids) or patatin-like (11–178 amino acids) tinated ATGL protein (Fig. 1A and Supplementary Fig. or cytosolic phospholipase A (cPLA) domain (299–388 1A). Furthermore, treatment with proteasomal inhibitor amino acids) containing vectors (Fig. 1D)indicatingthat MG-132 and not with lysosomal inhibitor chloroquine ATGL ubiquitination occurs between 1 and 178 amino acid. significantly increased the intensity of protein smear, sug- Sequence analysis of the patatin-like domain revealed gesting polyubiquitinated ATGL proteins are degraded that lysine residues of murine amino acid positions 74, predominantly in the proteasome (Fig. 1A and B and Sup- 78, 92, and 100 are conserved in mammals. As shown in plementary Fig. 1A and B). To confirm that endogenous Fig. 1E,lys→argmutationonlyat100thpositionsig- ubiquitin binds with ATGL protein, we performed re- nificantly reduced ATGL polyubiquitination. We next ciprocal immunoprecipitation by ATGL and ubiquitin in sought to determine the role of lysine 48 and 63 residues

Figure 1—ATGL is polyubiquitinated at lysine 100 in patatin-like domain by the lysine K48 linked chain of ubiquitin and degraded in proteasome. A: HepG2 cells were transfected with myc-tagged ATGL and HA-tagged ubiquitin constructs. After 48 h of transfection, cells were treated with 50 mmol/L MG-132 for 4 h. Cell lysates were then immunoprecipitated (IP) with myc-tag antibody and analyzed by immunoblot (IB) using anti-HA antibody (top). The bottom panels show the respective protein levels of ATGL and actin. B:Cellswere transfected with indicated plasmids and after 48 h treated with respective concentrations of chloroquine (Chlor) for 4 h. Ubiquitination assay was performed as described in A. C: Immunoprecipitation experiments were performed in HepG2 cells. Endogenous ATGL and ubiquitin (Ub) were immunoprecipitated and probed with Ub (upper panel) and ATGL (lower panel) antibody, respectively. D: Different fragments of ATGL containing different combination of domains (top); myc-tagged clones containing patatin-like and cPLA domain were transfected in HepG2 cells together with HA-ubiquitin, and polyubiquitination of ATGL was assayed by immunoblotting. EV, empty vector. E: Mutant clones with arginine in place of lysine at four positions of ATGL were transfected and checked for their ability to be polyubiquitinylated. F: Ubiquitination of ATGL with indicated lysine mutant variants of ubiquitin. The immunoblots are representative of one of at least three independent experiments. diabetes.diabetesjournals.org Ghosh and Associates 3565 on ubiquitin, the two most well-characterized residues FLAG-COP1 in HEK293A cells and immunoprecipitated implicated in polyubiquitination. ATGL is polyubiquiti- with anti-FLAG antibody. ATGL was detected in the im- nated in K63R, but ubiquitinated ATGL was undetectable mune complex (Fig. 2B). Similarly, in the HepG2 cell line, in the K48R-transfected cells (Fig. 1F). stably expressing myc-tagged ATGL (Supplementary Fig. 2A), COP1, and myc-ATGL were, respectively, coimmuno- COP1 Is the E3 Ligase for ATGL in Hepatocytes precipitated (Fig. 2C). Moreover, endogenous ATGL and To identify the putative E3 ligase responsible for ATGL COP1 in rat liver lysates were also coimmunoprecipitated ubiquitination, we searched for the consensus E3 ligase– with each other (Supplementary Fig. 2B). To further de- binding amino acid sequences on the ATGL protein. We termine whether COP1 binds to ATGL via its VP motif, we found one putative COP1 binding motif E-W-L-P-D-VP- find that mutation in the VP motif (V361A/P362A) led to E-D (consensus binding motif being D/E-x-x-x-VP-D/E), loss of COP1 binding (Fig. 2D) and significant attenuation the so called VP motif in the cPLA domain of the murine in ATGL ubiquitination (Fig. 2E). ATGL protein (Fig. 2A). To examine whether COP1 in- To obtain the structural insight into the ATGL-COP1 teracts with ATGL, we cotransfected myc-ATGL and binding, we generated in silico molecular models for ATGL

Figure 2—COP1 directly interacts with ATGL through VP binding motif in the cPLA domain of ATGL. A: Conserved lysine residues in the patatin-like domain in different mammalian taxa. ATGL and COP1 interaction in coimmunoprecipitation (IP) assays in HEK293A cells (B) and in HepG2 cells stably expressing ATGL (C). D: Coimmunoprecipitation experiment of COP1 with WT and VP mutant (V361A/V362A) of ATGL showing the importance of the VP motif in COP1 and ATGL interaction. E: Ubiquitination status of WT and VP mutant of ATGL. The immunoblots (IB) are representative of one of at least three independent experiments. F, left panel: Three-dimensional structure of ATGL protein predicted by ab initio and comparative modeling. The residues, which interact with COP1, are highlighted in orange and presented in sphere orientation. The active site (red color) and ubiquitination site (blue color) are also presented in sphere orientation. Right panel: Three-dimensional structure of COP1 protein predicted by ab initio and comparative modeling. The residues, which interact with ATGL, are highlighted in green and presented in sphere orientation. G, left panel: Docked complex of ATGL (blue) and COP1 (pink). Interface residues are shown as spheres in orange (ATGL) and green (COP1) colors. ATGL and COP1 interact through two interfaces marked as IF1 and IF2. Right panel: IF1 and IF2 are enlarged and the residues are represented in stick orientation. H: Interface area (top panel) and gain of solvation free energy (bottom panel) of the docked complex are compared with the same obtained from eight different protein complexes having interface area of similar sizes. 3566 COP1 Ubiquitinates ATGL Diabetes Volume 65, December 2016 and COP1. The full-length ATGL model generated by WD40 propeller region interacts with the VP motif joining the N-terminal homology model and the COOH- (E354–D361: EWLPDVPED) exposed at the surface of terminal ab initio model (Fig. 2F and Supplementary Fig. ATGL COOH-terminal, whereas IF2 is formed by the 3A) was subjected to the GROMACS energy minimization ring finger of COP1 and N-terminal domain of ATGL function using the NOMAD-REF web server (38) and val- (Fig. 2G,rightpanel).Thetotalinterfaceareawas708Å2, idated by a Ramachandran plot (Supplementary Fig. 3C). and the ΔG of interaction (solvation free energy) was Templates for COP1 were available for ring finger (region: 27.49 kcal/mol (Fig. 2H). To validate the ATGL-COP1 114–203) and WD40 repeat region (region: 323–731). docking, the solvation free energy of interaction of the Full-length COP1 model structure was also constructed docked complex was compared with other complexes by combining template-based homology and ab initio with an interface area of a similar size (Fig. 2H). Collec- modeling. Optimized COP1 structure was selected based tively, using in silico modeling and docking, we provide on energy and overall structural stability (Fig. 2F and support for direct binding of COP1 with ATGL. Supplementary Fig. 3B) and was validated by a Ramachan- To test whether COP1 functions as an E3 ligase for dran plot (Supplementary Fig. 3D). We then performed ATGL, we overexpressed and knocked down COP1 in directed docking for COP1 and ATGL using the stand- HepG2 cells (Fig. 3A). Overexpression of COP1 signifi- alone version of PatchDock 1.1 package (42) (Fig. 2G, cantly decreased, but siRNA-mediated depletion of COP1 left panel). The complex was stabilized by interprotein increased the ATGL protein level without any change in interactions at interface (IF) 1 and IF2. In IF1, the the ATGL mRNA expression (data not shown). Moreover,

Figure 3—COP1 regulates ATGL stability and ubiquitination in hepatocytes. A: ATGL protein levels in COP1-overexpressed and -depleted conditions in hepatocytes. Densitometric quantification of ATGL band intensity (right) normalized with actin (n = 3). B: Ubiquitination status of ATGL in COP1-depleted hepatocytes. IB, immunoblot; IP, immunoprecipitation. C: Effect of COP1 in ubiquitination of WT and K100R mutant of ATGL assayed by Ni pull-down. The immunoblots are representative of one of at least three independent experiments. D: Cycloheximide (Chx) chase experiment in HepG2 cells transfected with WT or K100R mutant ATGL for indicated time durations (top). Densitometric analysis of ATGL levels normalized to actin for three independent experiments (bottom). EV, empty vector; sc, scrambled. Data are expressed as mean 6 SD. *P < 0.05; **P < 0.01. diabetes.diabetesjournals.org Ghosh and Associates 3567 an increase in the ATGL protein after COP1 knockdown RedC12 fluorescence remains mostly localized to LDs, as resulted in a significant reduction in the level of ATGL revealed by the yellow color in the merge panel. In con- ubiquitination (Fig. 3B). Further, we find that COP1- trast, siRNA-mediated depletion of COP1 showed diffused dependent ATGL ubiquitination was considerably re- Red C12 resulting a significant drop in the yellow signal duced in the K100R ATGL mutant (Fig. 3C), suggesting (Fig. 4H). Similar to ATGL overexpressing cells, we also that COP1 ubiquitinates ATGL predominantly at the found a significant increase in the FA mobilization index 100th lysine residue. Albeit with appreciably decreased in the siCOP1 transfected cells (Fig. 4I). COP1-dependent intensity, the presence of background smear could be a FA mobilization would also correspondingly modulate the result of low abundant ubiquitination on other lysine res- FAO rate; thus, FAO rates were, respectively, increased or idues or of nonspecific ubiquination by overexpressed diminished upon COP1 knockdown or overexpression COP1. To further test this possibility, we asked whether (Fig. 4J). Collectively, our findings clearly indicate an im- K100R mutation increases the ATGL protein’s half life. portant role of COP1 in hepatocyte lipid metabolism. Thus HepG2 cells were transfected with wild-type (WT) ATGL Is Required for COP1-Dependent Hepatic Lipid or K100R mutant ATGL and treated with protein trans- Metabolism lation inhibitor cycloheximide for different time periods, The metabolic output of COP1 could be mediated via one D fi as shown in Fig. 3 . K100R mutant had signi cantly or more of its targets. We thereby asked whether ATGL is $ longer half life ( 24 h) than WT ATGL (;16 h), suggest- necessary for COP1-dependent hepatocyte lipid metabo- ing COP1 predominantly ubiquitinates at lysine the 100th lism. Knockdown of COP1 in ATGL-depleted cells would D lysine residue (Fig. 3 , bottom panel). Taken together, we not significantly alter COP1-dependent lipid metabolism. uncover a hitherto unknown posttranscriptional regula- In line with this argument, we transiently depleted COP1 tion of ATGL in hepatocytes where COP1 directly binds to (siCOP1) in shATGL cells. Although shATGL cells accu- fi ATGL via the VP motif and serves as the bona de E3 mulate higher LDs and TAG, siRNA-mediated knockdown ubiquitin ligase for ATGL. of COP1 (siCOP1+shATGL) did not show any difference A– COP1 Controls Lipid Content and Mobilization even after oleate induction (Fig. 5 C). Consistently, in in Hepatocytes the FA mobilization assay, siCOP1 in shATGL cells did If COP1 targets ATGL for proteasomal degradation, not show any difference over shATGL cells in merged D depletion of COP1 should increase the rate of lipolysis yellow signal (Fig. 5 ) or in FA mobilization index (Fig. E and, hence, decrease cellular lipid accumulation and 5 ), implying no further change in FA mobilization from increase in lipid mobilization. To examine this possibility, cellular lipid stores. The FAO rate also remains similar cells were treated with oleate-BSA conjugate for 16 h after siRNA-mediated COP1 knockdown in shATGL cells F and stained with Oil Red O, which stains neutral lipids, (Fig. 5 ). or BODIPY 493/503, a hydrophobic fluorescent dye that On the basis of these results, we argue that ectopic stains LDs. As shown in Fig. 4A and C knockdown or expression of COP1 in ATGL-overexpressing cells cellular overexpression of COP1, respectively, attenuated and lipid turnover would be unaffected. To this end, we enhanced oleate-dependent TAG accumulation. Likewise, expressed COP1 with increasing levels in HepG2 cells the number of cellular LDs was also significantly de- stably expressing ATGL and found that COP1 dose- creased when COP1 was knocked down (Fig. 4B)and dependently reduced ATGL proteins even in these cells C increased when COP1 was overexpressed (Fig. 4D). Accu- (Supplementary Fig. 4 ). To address our hypothesis, we mulation of LDs also corroborates with the cellular TAG thus chose an optimum level of COP1 expression where content. We thus found a significant drop in oleate- ATGL proteins are present at a physiologically relevant induced TAG accumulation when COP1 was depleted (Fig. level. In contrast to shATGL cells, ATGL-overexpressing G–I 4E) and a significant increase in TAG when COP1 was over- cells had a lower LD number and TAG content (Fig. 5 ). expressed (Fig. 4F). Coexpression of COP1 partially reversed the effects of fi ATGL serves as a major hepatic TAG hydrolase (16), ATGL expression; however, they remained signi cantly and modulation of its expression concomitantly alters lower compared with the cells transfected with empty G–I cellular FA mobilization. As COP1 degrades ATGL, its dif- vector (Fig. 5 ). In agreement with that, COP1 could J K ferential expression would also correspondingly modulate not reduce FA mobilization (Fig. 5 and ) and the rate of L FA mobilization. To test how COP1 controls FA mobiliza- FAO (Fig. 5 ) in the ATGL-overexpressing cells. Taken tion in hepatocytes, we performed a pulse-chase assay, as together, these rescue experiments indicate that ATGL described by Rambold et al. (45) (Fig. 4G). The ratio of red is required for COP1-mediated regulation of hepatic lipid and green channel area that represents the distribution of storage and its mobilization. FAs outside the LDs was calculated as the FA mobilization COP1 Knockdown Ameliorates Hepatic Steatosis index. Thus, stable knockdown (shATGL) or overexpres- in HFD-Fed Mice sion of ATGL in HepG2 cells respectively decreased or Given the important role of COP1 in TAG accumulation in increased FA mobilization (Supplementary Fig. 4A and cultured hepatocytes, we next sought to examine the B). Both in control and in COP1-overexpressing cells, effect of COP1 inhibition in an HFD-induced fatty liver 3568 COP1 Ubiquitinates ATGL Diabetes Volume 65, December 2016

Figure 4—COP1 regulates hepatocyte lipid metabolism. A and C: Differential intracellular LD staining using Oil Red O or BODIPY (493/503) for indicated experimental conditions. EV, empty vector; OA, oleic acid; sc, scrambled. Scale bars, 10 mm. B and D: Intracellular LD number for mentioned experimental conditions. Data are presented as mean LD number per cell 6 SD. E and F: Cellular TAG concentrations for different conditions normalized to protein concentrations. G: Schematic presentation of fluorescent FA pulse-chase assay. Cells were labeled with Red C12, a saturated FA analog, followed by treatment with hydrophobic BODIPY 493/503 (green) for LD staining and visualizing fluorescence with confocal microscopy. H: Confocal microscopy showing LDs and Red C12 following the assay as described in G for indicated experimental conditions. BF, bright field. Scale bars, 10 mm. I: FA mobilization index for different experimental conditions. J: FAO rate for the indicated experimental conditions. Data are expressed as mean 6 SD. NS, not significant. *P < 0.05; **P < 0.01.

mouse model. Because of hepatic tropism of adenovirus, liver (Fig. 6A–C). Compared with livers from chow-fed we used an adenovirus-based gene-targeting approach to mice, TAG levels were expectedly higher in HFD-fed ani- specifically knockdown COP1 (shAd) in livers of male mals, whereas shAd lowered its levels (Fig. 6D). Knock- C57Bl/6 mice fed the HFD for 4 weeks. We found a down of hepatic COP1 also consistently decreased serum significant reduction of COP1 expressions with a con- TAG levels but did not reach statistical significance (Fig. comitant increase in ATGL protein levels in the mouse 6E). H&E staining of shAd liver sections revealed a diabetes.diabetesjournals.org Ghosh and Associates 3569

Figure 5—COP1-mediated hepatic fat metabolism is ATGL dependent. A and G: BODIPY (493/503) staining of intracellular LDs in indicated experimental conditions. EV, empty vector; OA, oleic acid. Scale bars, 10 mm. B and H: Intracellular LD number for each condition. Data are expressed as mean LD number per cell 6 SD. C and I: Measurement of intracellular TAG for indicated conditions. sc, scrambled. D and J: Confocal microscopy showing LDs and Red C12 following the assay as described in Fig. 4G for different conditions. BF, bright field. Scale bars, 10 mm. E and K: FA mobilization index for mentioned experimental conditions. F and L: FAO rate for indicated conditions. Protein normalized data are presented. Data are expressed as mean 6 SD. NS, not significant. *P < 0.05; **P < 0.01.

reduction in hepatic lipid accumulation (Fig. 6F). Con- metabolically active tissues, including the liver, is essen- versely, no corresponding alterations in serum and liver tially unknown. Although ubiquitination of ATGL has cholesterol levels were observed (Fig. 6G and H). Collec- been reported in adipose tissue (47), where ATGL expres- tively, our results demonstrate that knockdown of COP1 sion is predominantly controlled transcriptionally, the in mouse liver increases ATGL levels and prevents TAG molecular details of such ubiquitination are lacking. Using accumulation in diet-induced hepatic steatosis. various biochemical and in silico tools, we describe the We next examined the effects of hepatic COP1 de- ubiquitin-proteasome system as a critical regulator of he- pletion on the functional restoration of HFD-induced patic ATGL content, which in turn is linked to hepatic liver damage. To assay liver function in vivo, we injected TAG turnover, particularly in the context of HFD-induced 99mTc-mebrofenin through a tail vein and monitored its NAFLD. Molecular pathogenesis of NAFLD and its down- clearance from the liver into the gall bladder over 30 min stream sequelae involve a multitude of metabolic path- using a gamma camera, as described earlier (46). No dif- ways, including hepatic insulin resistance, augmented de ference was detected in hepatic 99mTc-mebrofenin uptake novo lipogenesis and TAG synthesis, impaired FAO, al- in animals fed chow or the HFD. However, HFD-fed mice tered TAG secretion, and lipid sequestration by lipolysis, showed dampened 99mTc-mebrofenin clearance from the (48) of which the latter has been the focus of the current liver into the gall bladder (Fig. 6I and J and Supplemen- study. tary Fig. 5). Conversely, shAd animals fed the HFD In summary, we identify COP1 as a novel player in revealed significant recovery of 99mTc-mebrofenin clear- controlling hepatic TAG metabolism that is mediated by ance comparable to chow-fed mice (Fig. 6I and J and its direct interaction through VP-motif and K48-linked Supplementary Fig. 5). We thus find that COP1 plays an polyubiquitination of ATGL predominantly at lysine 100. important role in liver function and hepatobiliary trans- However, we cannot exclude the existence of other E3 port in diet-induced steatosis. ubiquitin ligases that can also target ATGL. In adipose, COP1 degrades acetyl CoA carboxylase, the rate-limiting DISCUSSION enzyme for lipogenesis, and stimulates lipolysis (27), Control of ATGL expression has extensively been studied whereas in the liver it ubiquitinates FA synthase (49). in adipocytes (12,13,30); however, its regulation in other Interestingly, COP1 was also shown to control hepatic 3570 COP1 Ubiquitinates ATGL Diabetes Volume 65, December 2016

Figure 6—Knockdown of hepatic COP1 reduces hepatic lipid accumulation and improves liver function in HFD-fed mice. HFD-fed mice were injected with adenovirus carrying shRNA against murine COP1 (shAd, green) or with vehicle (mock, red). Relative mRNA expression (A) and protein levels (B) of COP1 and ATGL in livers of indicated mice. C: Relative quantification of the band intensity normalized with actin (n = 6). Data are expressed as mean 6 SD relative to mock. Liver (D)andserumTAG(E) levels of chow-fed control (black), HFD (red), and shAd (green) mice. F: The H&E staining of liver sections from indicated mice. Scale bars, 100 mm. Cholesterol levels were assayed from serum (G) and liver tissue (H). Liver activity measured by analyzing 99mTc-mebrofenin signal 30 min after injection (n =4)(I)andquantification of 99mTc-mebrofenin distribution in liver and gall bladder by calculating the area under the curve (AUC) at different time points for 30 min through 180 time frames, with data expressed as mean 6 SD relative to chow (J). *P < 0.05, **P < 0.01 compared with chow or mock.

gluconeogenesis through FoxO1 (23) or TORC2 (29), both and Sounak Bhattacharya, Council of Scientific and Industrial Research–Indian of which are transcriptional regulators of gluconeogene- Institute of Chemical Biology, Kolkata, for help with confocal microscopy. sis. Together with previous studies, our results thus sup- Funding. M.G., M.B., and D.K.N. received research fellowships from the fi port the notion that COP1 integrates multiple metabolic Council of Scienti c and Industrial Research, India. S.N. and M.A. received research fellowships from the Department of Science and Technology–Innovation outputs to maintain hepatic metabolic homeostasis. fi in Science Pursuit for Inspired Research and University Grants Commission, India, Thus identi cation of COP1-mediated ATGL degrada- respectively. This work was supported by Council of Scientific and Industrial tion in the current study not only provides novel insight Research, India, grants BSC 0121 to S.C. and BSC 0206 to P.C. and Department into how hepatic TAG turnover is maintained but also of Science and Technology, Ministry of Science and Technology, India, grant offers new therapeutic strategies by targeting COP1 in SB/S0/BB/007/2013 to P.C. hepatic steatosis. Duality of Interest. No potential conflicts of interest relevant to this article were reported. Author Contributions. M.G. performed experiments for Figs. 1–3 and 6. Acknowledgments. The authors thank Mita Chatterjee Debnath, Council of M.G., S.N., S.C., and P.C. analyzed data. S.N. conducted experiments for Figs. 4–6. Scientific and Industrial Research–Indian Institute of Chemical Biology, Kolkata, and M.B. performed the in silico studies. M.A. generated cell lines and screened. Samarendu Sinha, Regional Radiation and Medicine Centre, Variable Energy Cyclo- D.K.N. conducted the 99mTc-mebrofenin experiments. S.C. and P.C. designed tron Centre, Kolkata, for their assistance with the 99mTc-mebrofenin experiments, the research. P.C. wrote the paper. P.C. is the guarantor of this work and, as such, diabetes.diabetesjournals.org Ghosh and Associates 3571

had full access to all the data in the study and takes responsibility for the integrity of 19. Ong KT, Mashek MT, Bu SY, Mashek DG. Hepatic ATGL knockdown un- the data and the accuracy of the data analysis. couples glucose intolerance from liver TAG accumulation. FASEB J 2013;27:313– 321 References 20. Nalepa G, Rolfe M, Harper JW. Drug discovery in the ubiquitin-proteasome 1. Ahmed A, Wong RJ, Harrison SA. Nonalcoholic fatty liver disease review: system. Nat Rev Drug Discov 2006;5:596–613 Diagnosis, treatment, and outcomes. Clin Gastroenterol Hepatol 2015;13:2062– 21. Hershko A, Heller H, Elias S, Ciechanover A. Components of ubiquitin- 2070 protein ligase system. Resolution, affinity purification, and role in protein 2. Ludwig J, Viggiano TR, McGill DB, Oh BJ. Nonalcoholic steatohepatitis: breakdown. J Biol Chem 1983;258:8206–8214 Mayo Clinic experiences with a hitherto unnamed disease. Mayo Clin Proc 1980; 22. Nakatsukasa K, Okumura F, Kamura T. Proteolytic regulation of metabolic 55:434–438 enzymes by E3 ubiquitin ligase complexes: lessons from yeast. Crit Rev Biochem 3. Marchesini G, Brizi M, Morselli-Labate AM, et al. Association of non- Mol Biol 2015;50:489–502 alcoholic fatty liver disease with insulin resistance. Am J Med 1999;107:450– 23. Kato S, Ding J, Pisck E, Jhala US, Du K. COP1 functions as a FoxO1 455 ubiquitin E3 ligase to regulate FoxO1-mediated gene expression. J Biol Chem 4. Parks EJ, Krauss RM, Christiansen MP, Neese RA, Hellerstein MK. Effects of 2008;283:35464–35473 a low-fat, high-carbohydrate diet on VLDL-triglyceride assembly, production, and 24. Migliorini D, Bogaerts S, Defever D, et al. Cop1 constitutively regulates clearance. J Clin Invest 1999;104:1087–1096 c-Jun protein stability and functions as a tumor suppressor in mice. J Clin Invest 5. Cohen JC, Horton JD, Hobbs HH. Human fatty liver disease: old questions 2011;121:1329–1343 and new insights. Science 2011;332:1519–1523 25. Dornan D, Wertz I, Shimizu H, et al. The ubiquitin ligase COP1 is a critical 6. Kato M, Higuchi N, Enjoji M. Reduced hepatic expression of adipose tissue negative regulator of p53. Nature 2004;429:86–92 triglyceride lipase and CGI-58 may contribute to the development of non-alcoholic 26. Wertz IE, O’Rourke KM, Zhang Z, et al. Human De-etiolated-1 regulates fatty liver disease in patients with insulin resistance. Scand J Gastroenterol 2008; c-Jun by assembling a CUL4A ubiquitin ligase. Science 2004;303:1371–1374 43:1018–1019 27. Qi L, Heredia JE, Altarejos JY, et al. TRB3 links the E3 ubiquitin ligase COP1 7. Zimmermann R, Strauss JG, Haemmerle G, et al. Fat mobilization in to lipid metabolism. Science 2006;312:1763–1766 adipose tissue is promoted by adipose triglyceride lipase. Science 2004;306: 28. Dentin R, Liu Y, Koo SH, et al. Insulin modulates gluconeogenesis by in- 1383–1386 hibition of the coactivator TORC2. Nature 2007;449:366–369 8. Villena JA, Roy S, Sarkadi-Nagy E, Kim KH, Sul HS. Desnutrin, an adipocyte 29. Suriben R, Kaihara KA, Paolino M, et al. b-Cell Insulin Secretion Requires gene encoding a novel patatin domain-containing protein, is induced by fasting the Ubiquitin Ligase COP1. Cell 2015;163:1457–1467 and glucocorticoids: ectopic expression of desnutrin increases triglyceride hy- 30. Chakrabarti P, English T, Shi J, Smas CM, Kandror KV. Mammalian target of drolysis. J Biol Chem 2004;279:47066–47075 rapamycin complex 1 suppresses lipolysis, stimulates lipogenesis, and promotes 9. Jenkins CM, Mancuso DJ, Yan W, Sims HF, Gibson B, Gross RW. Identi- fat storage. Diabetes 2010;59:775–781 fication, cloning, expression, and purification of three novel human calcium- 31. Zlatkis A, Zak B, Boyle AJ. A new method for the direct determination of independent phospholipase A2 family members possessing triacylglycerol serum cholesterol. J Lab Clin Med 1953;41:486–492 lipase and acylglycerol transacylase activities. J Biol Chem 2004;279:48968– 32. Berman HM, Westbrook J, Feng Z, et al. The Protein Data Bank. Nucleic 48975 Acids Res 2000;28:235–242 10. Haemmerle G, Lass A, Zimmermann R, et al. Defective lipolysis and altered 33. Lo Conte L, Ailey B, Hubbard TJ, Brenner SE, Murzin AG, Chothia C. SCOP: a energy metabolism in mice lacking adipose triglyceride lipase. Science 2006; structural classification of proteins database. Nucleic Acids Res 2000;28:257– 312:734–737 259 11. Fischer J, Lefèvre C, Morava E, et al. The gene encoding adipose tri- 34. Jones DT. Protein secondary structure prediction based on position-specific glyceride lipase (PNPLA2) is mutated in neutral lipid storage disease with my- scoring matrices. J Mol Biol 1999;292:195–202 opathy. Nat Genet 2007;39:28–30 35. Kelley LA, Mezulis S, Yates CM, Wass MN, Sternberg MJ. The Phyre2 web 12. Chakrabarti P, Kandror KV. FoxO1 controls insulin-dependent adipose tri- portal for protein modeling, prediction and analysis. Nat Protoc 2015;10:845– glyceride lipase (ATGL) expression and lipolysis in adipocytes. J Biol Chem 2009; 858 284:13296–13300 36. Eswar N, Eramian D, Webb B, Shen MY, Sali A. Protein structure modeling 13. Chakrabarti P, Kim JY, Singh M, et al. Insulin inhibits lipolysis in adipocytes with MODELLER. Methods Mol Biol 2008;426:145–159 via the evolutionarily conserved mTORC1-Egr1-ATGL-mediated pathway. Mol Cell 37. Roy A, Kucukural A, Zhang Y. I-TASSER: a unified platform for automated Biol 2013;33:3659–3666 protein structure and function prediction. Nat Protoc 2010;5:725–738 14. Nielsen R, Pedersen TA, Hagenbeek D, et al. Genome-wide profiling of 38. Lindahl E, Azuara C, Koehl P, Delarue M. NOMAD-Ref: visualization, de- PPARgamma:RXR and RNA polymerase II occupancy reveals temporal activation formation and refinement of macromolecular structures based on all-atom nor- of distinct metabolic pathways and changes in RXR dimer composition during mal mode analysis. Nucleic Acids Res 2006;34(Web Server issue):W52–W56. adipogenesis. Genes Dev 2008;22:2953–2967 39. Eisenberg D, Lüthy R, Bowie JU. VERIFY3D: assessment of protein models 15. Subramanian V, Rothenberg A, Gomez C, et al. Perilipin A mediates the with three-dimensional profiles. Methods Enzymol 1997;277:396–404 reversible binding of CGI-58 to lipid droplets in 3T3-L1 adipocytes. J Biol Chem 40. Davis IW, Leaver-Fay A, Chen VB, Block JN, et al. MolProbity: all-atom 2004;279:42062–42071 contacts and structure validation for proteins and nucleic acids. Nucleic Acids 16. Ong KT, Mashek MT, Bu S, Greenberg AS, Mashek DG. Adipose triglyceride Res 2007;35(Web Server issue):W375–W383 lipase is a major hepatic lipase that regulates triacylglycerol turnover and fatty 41. Krissinel E, Henrick K. Inference of macromolecular assemblies from acid signaling and partitioning. Hepatology 2011;53:116–126 crystalline state. J Mol Biol 2007;372:774–797 17. Wu JW, Wang SP, Alvarez F, et al. Deficiency of liver adipose triglyceride 42. Schneidman-Duhovny D, Inbar Y, Nussinov R, Wolfson HJ. PatchDock and lipase in mice causes progressive hepatic steatosis. Hepatology 2011;54:122– SymmDock: servers for rigid and symmetric docking. Nucleic Acids Res 2005; 132 33(Web Server issue):W363–W367. 18. Chitraju C, Trötzmüller M, Hartler J, et al. The impact of genetic stress by 43. Mashiach E, Schneidman-Duhovny D, Andrusier N. FireDock: a web server ATGL deficiency on the lipidome of lipid droplets from murine hepatocytes. J Lipid for fast interaction refinement in molecular docking. Nucleic Acids Res 2008; Res 2013;54:2185–2194 36(Web Server issue):W229–W232. 3572 COP1 Ubiquitinates ATGL Diabetes Volume 65, December 2016

44. Pettersen EF, Goddard TD, Huang CC, et al. UCSF Chimera–a visualization 47. Dai Z, Qi W, Li C, et al. Dual regulation of adipose triglyceride lipase by system for exploratory research and analysis. J Comput Chem 2004;25:1605–1612 pigment epithelium-derived factor: a novel mechanistic insight into progressive 45. Rambold AS, Cohen S, Lippincott-Schwartz J. Fatty acid trafficking in obesity. Mol Cell Endocrinol 2013;377:123–134 starved cells: regulation by lipid droplet lipolysis, autophagy, and mitochondrial 48. Sun Z, Lazar MA. Dissociating fatty liver and diabetes. Trends Endocrinol fusion dynamics. Dev Cell 2015;32:678–692 Metab 2013;24:4–12 46. Neyt S, Huisman MT, Vanhove C, et al. In vivo visualization and quantifi- 49. Yu J, Deng R, Zhu HH, et al. Modulation of fatty acid synthase degradation cation of (Disturbed) Oatp-mediated hepatic uptake and Mrp2-mediated biliary by concerted action of p38 MAP kinase, E3 ligase COP1, and SH2-tyrosine excretion of 99mTc-mebrofenin in mice. J Nucl Med 2013;54:624–630 phosphatase Shp2. J Biol Chem 2013;288:3823–3830