Opposing roles of cell death-inducing DFF45-like effector B and perilipin 2 in controlling hepatic VLDL lipidation.

Regulation of hepatic very low density lipoprotein (VLDL) assembly and maturation is crucial in controlling lipid homeostasis and in the development of metabolic disorders, including obesity, hepatic steatosis, and insulin resistance. Cideb, a member of cell death-inducing DFF45-like effector (CIDE) protein family, has been previously shown to promote VLDL lipidation and maturation. However, the precise subcellular location of Cideb-mediated VLDL lipidation and the factors modulating its activity remain elusive. In addition to its localization to endoplasmic reticulum (ER) and lipid droplets (LD), we observed that Cideb was also localized to the Golgi apparatus. Mature and lipid-rich VLDL particles did not accumulate in the Golgi apparatus in Cideb(-/-) livers. Interestingly, we observed that hepatic perilipin 2/adipose differentiation-related protein (ADRP) levels were markedly increased in Cideb(-/-) mice. Liver-specific knockdown of perilipin 2 in Cideb(-/-) mice resulted in the reduced accumulation of hepatic triglycerides (TAG), increased VLDL-TAG secretion, and the accumulation of mature TAG-rich VLDL in the Golgi apparatus. These data reveal that Cideb and perilipin 2 play opposing roles in controlling VLDL lipidation and hepatic lipid homeostasis.

Hepatic lipid homeostasis, including lipid storage and very low density lipoprotein (VLDL) assembly and secretion, is important for controlling the plasma levels of of mature VLDL in this organelle. Interestingly, we found that perilipin 2 expression was markedly increased in the livers of Cideb Ϫ / Ϫ mice. Knocking down perilipin 2 in the livers of Cideb Ϫ / Ϫ mice resulted in increased lipid storage and enhanced VLDL-TAG secretion and accumulation of lipid-rich mature VLDL particles in the Golgi, rescuing the phenotypes caused by Cideb defi ciency.

Materials
All chemicals were purchased from Sigma-Aldrich (USA) unless otherwise stated. The rabbit anti-mouse Cideb antibody was used as previously described ( 45 ). The following antibodies were purchased from the indicated companies: guinea pig anti-perilipin 2 antibody (Fitzgerald Industries, USA), goat anti-apoB antibody (AB742, Chemicon, USA), mouse anti-MTP and mouse anti-GM130 antibodies (BD Biosciences, USA), mouse anti-Golgi 58K protein (p58K), mouse anti-␤ -actin and rabbit anti-calnexin (CNX) antibodies (Sigma-Aldrich, USA) and rat anti-GRP94 antibody (Santa Cruz Biotechnology, USA). The mouse anti-perilipin 5 antibody was a gift from Jing Ye's lab (Department of Pathology, Fourth Military Medical University). The protease inhibitor mixture tablets were purchased from Roche Diagnostics (Switzerland). The

Cideb
Ϫ / Ϫ mice were generated and maintained as previously described ( 45 ). Cideb Ϫ / Ϫ mice and wild-type C57BL/6J mice were housed in an environmentally controlled animal facility of the Center of Biomedical Analysis, Tsinghua University, Beijing, China. Mice had access to food and autoclaved water ad libitum. Twelve-week-old male mice were used in the studies. The mice were handled according to the Responsible Care and Use of Laboratory Animals guideline set by Tsinghua University, and the experimental protocols were approved by the animal ethics committee of Tsinghua University.

Isolation and culture of primary hepatocytes
Hepatocytes were isolated from livers of 12-week-old mice as previously described ( 48 ). The mice were anesthetized with 1% Pelltobarbitalum Natricum (AMRESCO, USA) and dissected. The livers were perfused through the portal veins to remove the blood, and the perfusion was continued until the livers became soft. The livers were immediately removed, cut into pieces, and mixed with a collagenase solution until the hepatocytes were dispersed. The hepatocytes were fi ltered to remove tissue debris and washed three times with cold Dulbecco's modifi ed Eagle's medium (DMEM, Invitrogen, USA). The isolated hepatocytes were suspended in DMEM containing 10% fetal bovine serum (FBS, Invitrogen, USA) and seeded at 1 × 10 6 cells/dish in 6 cm dishes. Cells were maintained in 5% CO 2 at 37°C.

Subcellular fractionation and analysis of microsomal lumenal TAG contents
Subcellular fractionation was performed essentially as previously described ( 11 ). Livers were removed and rapidly homogenized with a loose-fi tted Dounce homogenizer in 5 ml of ice-cold solution containing 10 mM Hepes, pH 7.4, 150 mM sucrose, 0.5 mM DTT, and 1× EDTA-free protease inhibitors. The homogenates were centrifuged at 1,900 g for 10 min at 4°C, and the supernatants were centrifuged at 100,000 g for 90 min at 4°C in a Beckman SW60 rotor. The LD fraction fl oated on the top was is the main site of neutral lipid storage in all cell types ( 14,15 ). Recently, the LD has been recognized as a dynamic and functionally active organelle that regulates protein traffi cking, viral assembly, and insulin sensitivity (16)(17)(18)(19)(20). Two types of LDs, cytosolic LDs and ER-associated lumenal LDs, have been identifi ed ( 21,22 ). LDs have been demonstrated to be in close proximity to the ER-derived membrane in hepatocytes with downregulated perilipin 2/adipose differentiation-related protein (ADRP) ( 23 ) or stuck in the ER membrane by tight binding to lipidated apoB-100 ( 24 ). It is generally believed that cytosolic LDs represent a major source of lipids for VLDL maturation ( 25,26 ). However, the molecular mechanism by which TAG is transferred from cytosolic LDs to pre-VLDL particles remains a subject of intense discussion and debate. It has been proposed that TAG can be transferred from LDs in the ER lumen to pre-VLDL particles through their direct fusion ( 4,13 ). Alternatively, TAG from cytosolic LDs can be hydrolyzed to yield free fatty acids (FFA) that are then reesterifi ed on the lumenal side of secretory pathway organelles, such as the ER and the Golgi apparatus, to form a special form of lumenal LDs that can be incorporated into nascent VLDL particles (27)(28)(29)(30).
It remains to be determined whether TAG from cytosolic LDs can be directly transferred to pre-VLDL particles to form lipid-rich VLDL particles. One potential factor in this process is perilipin 2, a member of the PAT protein family. Perilipin 2 is associated with LDs and has been shown to regulate VLDL assembly and LD morphology in hepatocytes (31)(32)(33). Perilipin 2 defi ciency protects mice from the development of hepatic steatosis by reducing TAG storage and promoting VLDL assembly and secretion (33)(34)(35). Perilipin 2 and ApoE double-defi cient mice exhibit impaired foam cell formation and resistance to the development of atherosclerosis ( 36 ). The downregulation of perilipin 2 and perilipin 3/tail interacting protein of 47 kDa (Tip47) in AML12 cells increases lipolysis by enhancing the recruitment of adipose triglyceride lipase (ATGL) to the LD surface ( 18 ). In contrast, the overexpression of perilipin 2 inhibits the association of ATGL with LDs, reduces lipolysis ( 37 ), stimulates lipid biosynthesis, and inhibits fatty acid ␤ -oxidation ( 38 ). However, the exact mechanism by which perilipin 2 controls VLDL lipidation and lipid storage remains unknown.
The cell death-inducing DFF45-like effector (CIDE) proteins, including Cidea, Cideb, and Fsp27 (Cidec in human), are crucial regulators of lipid metabolism and play important roles in the development of metabolic disorders, including obesity, insulin resistance, and hepatic steatosis (39)(40)(41)(42)(43)(44). Cideb is expressed at high levels in the liver, and its defi ciency results in higher energy expenditure, reduced plasma TAG levels ( 45 ) and altered expression of genes in various metabolic and signaling networks ( 46 ). Furthermore, Cideb is localized to the LD and ER and facilitates VLDL lipidation and maturation, possibly through its interaction with apoB ( 47 ). Here, we demonstrate that, in addition to its subcellular localization on the LDs and ER, Cideb was associated with the Golgi apparatus, and Cideb defi ciency led to the decreased accumulation

Liver lipid extraction and TLC assay
The total lipids were extracted from livers as previously described ( 49 ). The livers were homogenized in PBS buffer containing protease inhibitors. The homogenates were rapidly mixed with a chloroform / methanol (2:1) solution by vortexing and centrifuged at 250 g for 10 min to separate the phases. The lower lipid-containing phase was carefully aspirated and dried in a 70°C metal bath with nitrogen steam. The dried lipids were reconstituted in methylbenzene and loaded onto a TLC plate. The lipids were separated in a hexane / diethyl ether / acetic acid (70:30:1, v/v) solution. The TLC plates were sprayed with 10% CuSO 4 in 10% phosphoric acid and developed by drying in a 120°C oven.

Analysis of plasma TAG contents
After the mice fasted for 12 h, blood was collected from the orbital plexus, and serum was isolated after centrifugation. The serum TAG levels were determined using the Serum Triglyceride Determination Kit (Sigma TR0100).

TAG synthesis and fatty acid ␤ -oxidation
Fifty thousand viable hepatocytes per well were cultured in 6-well plates for 12 h in DMEM containing 10% FBS. The hepatocyte monolayers were washed twice with PBS buffer before being tested for TAG synthesis and fatty acid ␤ -oxidation. The reactions were carried out in triplicate (2 ml per well), each containing [ 3 H]oleic acid (1 Ci/well) with 22 µM unlabeled oleic acid and 0.5 mg/ml fatty acid-free BSA in PBS. After incubation at 37°C for 2 h, the monolayers were scraped into centrifuge tubes in 2 ml PBS, and 8 ml hexane/isopropanol (3:2) was added. After vigorously mixing, the tubes were centrifuged at 3,000 g for 5 min. The upper phase, containing the lipids, was transferred to vials and dried under nitrogen steam. The TAG was separated on a TLC plate and collected. Five milliliters of scintillation solution were added, and the radioactivity was measured. The protein pellets were dried and dissolved in 1 ml 0.2 M KOH solution overnight. The protein content was determined using the Bradford method.
For the measurement of fatty acid ␤ -oxidation, the reaction medium was transferred to glass tubes after incubation at 37°C for 2 h or 4 h, and 8 ml of methanol/chloroform (2:1) was added. After vigorously mixing, the mixtures were centrifuged at 3,000 g for 5 min, the aqueous phase containing 3 H 2 O was transferred to a new tube, and the radioactivity was measured.

Analysis of in vivo VLDL-TAG and apoB secretion
The rate of VLDL secretion was determined by blocking the catabolism and clearance of VLDL with Triton WR-1339 ( 50 ). Mice were injected with 500 mg/kg body weight of Triton WR-1339 (in PBS) by tail-vein injection after 12 h of fasting. One hundred microliters of blood was drawn from the retro-orbital plexus at 1 h intervals after the injection, and the serum TAG concentrations were measured using the Serum Triglyceride Determination Kit (Sigma TR0100). Levels of apoB-100 and apoB-48 were visualized by Western blot analysis, and the intensities of the bands were semiquantitatively determined using Bio-Rad Quantity One software.

Preparation of the recombinant adenovirus and tail-vein injection
The recombinant adenoviruses carrying perilipin 2 shRNA or control shRNA were constructed using the AdEasy-1 System (Stratagene). The perilipin 2-targeting shRNA sequence was as follows: 5 ′ -GAATATGCACAGTGCCAAC -3 ′ . The viruses were packaged and amplifi ed in AD293 cells and purifi ed by CsCl density gradient ultracentrifugation, dialyzed against PBS supplemented collected. The pellets were resuspended in 2.3 ml of an 8.58% sucrose solution and loaded on top of a sucrose density gradient containing the following layers (from the bottom): 56% (0.46 ml), 50% (0.92 ml), 45% (1.38 ml), 40% (2.3 ml), 35% (2.3 ml), 30% (1.38 ml), and 20% (0.46 ml) sucrose. After ultracentrifugation at 39,000 rpm for 18 h at 4°C in a Beckman SW41 rotor, 12 fractions were unloaded from top to bottom of each centrifuge tube. The distribution patterns of the subcellular compartment markers were determined by Western blotting (ER marker: CNX, GRP94, MTP; Golgi marker: GM130, p58), and the intensities of the bands were determined semiquantitatively using Bio-Rad Quantity One software.
The Golgi or ER fractions were combined and centrifuged at 100,000 g for 90 min at 4°C. The microsome pellets were resuspended in lysis buffer (20 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton-X100, and protease inhibitors, at pH 7.4), and the TAG levels were measured using the Serum Triglyceride Determination Kit (Sigma TR0100).

Sucrose gradient separation of apoB-containing lipoproteins from the isolated ER and Golgi lumenal contents
Separation of the lumenal apoB-containing lipoproteins was performed as described ( 11 ). The lumenal contents were released from microsomes by treatment with 0.1 M sodium carbonate (pH 11) and deoxycholic acid (0.025%) for 25 min at room temperature. BSA was added to a fi nal concentration of 5 mg/ml, and the samples were centrifuged at 50,000 rpm in a Beckman SW60 rotor for 1 h at 4°C. The lumenal contents (supernatant) were adjusted to pH 7.4 by the addition of 10% acetic acid and adjusted to a sucrose concentration of 12.5% (w/v). The adjusted supernatants were placed on the top of a step gradient consisting of 1.9 ml of 49% sucrose and 1.9 ml of 20% sucrose. Next, 2.8 ml of phosphate-buffered saline (PBS) was layered on the top of the supernatants. All solutions contained protease inhibitors. After centrifugation at 35,000 rpm for 65 h at 10°C in a Beckman SW41 rotor, 12 fractions (0.95 ml each) with densities ranging from 1.0 to 1.125 g/ml were collected from the top of the tube. ApoB was immunoprecipitated from each fraction. The density of each fraction was determined as the ratio of the weight of a 1 ml aliquot to the weight of a 1 ml aliquot of water. The apoB distribution patterns in the Golgi and ER lumens were determined by immunoblotting, and the intensities of the bands were determined semiquantitatively using Bio-Rad Quantity One software.

Western blots
Livers were homogenized in lysis buffer (20 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton-X100, and protease inhibitors, at pH 7.4) and centrifuged to discard the cell debris. The protein concentrations were determined using a Bio-Rad kit (USA). Equivalent amounts of protein homogenate were resolved by SDS-PAGE, transferred to a polyvinylidene difl uoride (PVDF) membrane, and probed with specifi c antibodies for visualization.

RNA extraction and semiquantitative real-time PCR
Total RNA was isolated from mice livers with TRIzol (Invitrogen, USA). First-strand cDNA synthesis was performed using the Superscript III RT kit and oligo-dT primers (Invitrogen, USA). Semiquantitative real-time PCR (RT-PCR) reactions were performed using the SYBR Green PCR system (Applied Biosystems, USA) in an ABI 7500 thermal cycler (Applied Biosystems, USA). ␤ -actin was used as a reference gene. The primers for real-time PCR are described in supplementary Table I. The sections were observed using a Zeiss 200M inverted microscope, and images were collected with an AxioCam MRm camera and AxioVision software.

Histology
The livers were excised and fi xed in 10% formalin buffer. The fi xed specimens were sectioned and stained with Oil Red O. Isolated hepatocytes were fi xed in 10% buffered formalin and stained with Oil Red O.
with 15% glycerol, and stored at Ϫ 80°C. The adenoviruses were injected via the tail vein at a dosage of 1 × 10 12 pfu/mouse. The plasma and liver TAG were examined 6 days after viral injection.

Immunofl uorescent staining
Hepatocytes were cultured for 24 h and fi xed with 4% paraformaldehyde for 1 h at room temperature. Bodipy 493/503 (Molecular Probes, USA) was used for neutral lipid staining. The nuclei were stained with Hoechst (Molecular Probes, USA). mice (57.30 ± 3.25 versus 115.60 ± 5.46 mg/g protein, P < 0.001, Fig. 2A ). In contrast, the amount of TAG in the ER fraction was similar between the wild-type and Cideb Ϫ / Ϫ livers (50.61 ± 4.58 versus 40.32 ± 5.16 mg/g protein, Fig. 2A ). Therefore, signifi cantly fewer mature and lipidrich VLDL particles accumulated in the Golgi apparatus in Cideb Ϫ / Ϫ livers. Next, we isolated the Golgi and ER microsomes from the livers of wild-type and Cideb Ϫ / Ϫ mice and treated them with sodium carbonate and deoxycholate to release the apoB-containing lipoproteins ( 11 ). The released lipoproteins from the ER and Golgi were separated by sucrose density gradient centrifugation, and the apoB in individual density fractions was immunoprecipitated. The presence of apoB in each fraction was evaluated by Western blot analysis. As shown in Fig. 2B , the Golgi microsomes from the livers of wild-type mice contained lipoproteins with densities ranging from 1.0 to 1.053, representing VLDL and IDL/LDL particles. However, the lipoproteins isolated from the Golgi of Cideb Ϫ / Ϫ livers exhibited densities

Statistical analysis
The data are presented as means ± SEM. Student t -tests were used to determine signifi cant differences between treatment groups. P < 0.05 was considered statistically signifi cant.

Cideb is localized to the Golgi apparatus
To further evaluate the role of Cideb in regulating VLDL lipidation and maturation, we assessed the presence of Cideb in the Golgi apparatus, which plays an important role in VLDL lipidation and maturation. We fi rst isolated ER and Golgi membrane fractions from mice livers using a sucrose density gradient centrifugation approach ( 11 ). The fractions were loaded according to their density, and the Golgi membranes were enriched in fractions 3-5 based on the detection of two Golgi-specifi c markers, 130 kDa cis -Golgi matrix protein (GM130) and Golgi 58K protein (p58), in these fractions. Higher levels of glucose-regulated protein 94 kDa (GRP94), CNX, and MTP, three ER-specifi c markers, were detected in fractions 6-12, suggesting ER enrichment in these fractions. Interestingly, Cideb exhibited a broad distribution pattern, and this protein was detected in fractions 3-12, spanning both the Golgi and ER fractions ( Fig. 1A ). As a LD-associated protein, perilipin 2 was only detected in the LD but not in the ER or Golgienriched fractions. Cideb was also detected in the LD fraction. Consistent with previous observations that VLDL lipidation and maturation occurs in the Golgi apparatus ( 11,12 ), apoB-100 and apoB-48 were mainly detected in the low-density fractions (fractions 1-4 ), with the highest levels present in the Golgi-enriched fractions (fractions 3-4 ). These data indicate that, in addition to its ER and LD association, Cideb is associated with the Golgi apparatus. Subcellular fractionation of liver tissues from Cideb Ϫ / Ϫ mice showed no difference in the distribution of various markers in the appropriate fractions ( Fig. 1B ). Interestingly, we observed a slight but consistent shift of the apoB peak from fraction 3 to fraction 4, which may correlate with the accumulation of less lipidated VLDL particles in the livers of Cideb Ϫ / Ϫ mice. When the relative levels of Cideb in the Golgi, ER, and LD fractions were normalized and compared, we observed that the amounts of Cideb in the ER and LDs were approximately 3-and 3.5-fold higher, respectively, than those in Golgi apparatus ( Fig. 1C ).

Reduced VLDL accumulation in the Golgi of Cideb Ϫ / Ϫ liver tissue
To further evaluate the role of Cideb in VLDL lipidation and maturation in the Golgi, we isolated the Golgi and ER microsomal fractions from the livers of wild-type and Cideb Ϫ / Ϫ mice and measured the amount of TAG in the Golgi and ER, respectively. In wild-type liver tissue, the amount of TAG in the Golgi was 2-fold higher than that in the ER, consistent with the role of the Golgi in VLDL lipidation. Interestingly, we observed an approximately 50% reduction in the amount of TAG in the Golgi fraction from Cideb Ϫ / Ϫ livers compared with livers from wild-type  effectively reduced hepatic perilipin 2 expression compared with adenoviruses containing scrambled shRNA (Ad-ctrl shRNA) in both wild-type and Cideb Ϫ / Ϫ mice. The knockdown of perilipin 2 did not change the levels of perilipin 3 but slightly reduced the total levels of perilipin 5 in the livers of both wild-type and Cideb Ϫ / Ϫ mice (supplementary Fig. I-B). However, the amounts of LD-associated perilipin 3 and perilipin 5 were signifi cantly increased in Cideb Ϫ / Ϫ livers with a liver-specifi c knockdown of perilipin 2 (supplementary Fig. I-C). The knockdown of perilipin 2 in the livers of wild-type mice moderately reduced hepatic TAG accumulation (41.98 ± 2.52 versus 73.43 ± 11.96 mg/g protein for perilipin 2 knockdown and control, respectively, P < 0.05). Intriguingly, perilipin 2 knockdown markedly reduced hepatic TAG levels in the livers of Cideb Ϫ / Ϫ mice (117.50 ± 10.66 versus 356.00 ± 62.13 mg/g protein for the perilipin 2 knockdown and control, respectively, P < 0.01, Fig. 3E ). The decreased TAG accumulation was also evident in isolated Cideb Ϫ / Ϫ hepatocytes that were stained with Oil Red O ( Fig. 3G ). The decreased hepatic TAG levels in the perilipin 2 knockdown and Cideb Ϫ / Ϫ hepatocytes were not due to the downregulation of TAG synthesis, because the hepatic TAG synthesis rate was not changed ( Fig. 3H ) and the mRNA levels of lipogenic genes involved in fatty acid synthesis [acetyl-CoA carboxylase 1 (ACC1), fatty acid synthase (FAS), and stearol-CoA desaturase 1 (SCD1)] and TAG synthesis [diacylglycerol O-acyltransferase 1 (DGAT1) and diacylglycerol O-acyltransferase 2 (DGAT2)] were similar between the perilipin 2-knockdown and control mice (data not shown). The reduced hepatic TAG levels were not due to the increased fatty acid ␤ -oxidation, as the rate of fatty acid ␤ -oxidation in Cideb Ϫ / Ϫ hepatocytes with perilipin 2 knockdown was signifi cantly reduced ( Fig. 3I ). Consistent with a decreased rate of fatty acid oxidation, the expression levels of genes involved in fatty acid ␤ -oxidation, including peroxisome proliferator-activated receptor ␣ (PPAR ␣ ), acetyl-CoA carboxylase 2 (ACC2), carnitinepalmitoyl transferase 1 (CPT1), carnitine-palmitoyl transferase 2 (CPT2), and cytochrome c oxidase subunit IV (COX4), were reduced ( Fig. 3J ). In contrast to the reduced hepatic TAG levels, the plasma levels of TAG were signifi cantly increased in the Cideb Ϫ / Ϫ mice with liverspecifi c knockdown of perilipin 2 (0.386 ± 0.032 versus 0.164 ± 0.014 mg/ml for the perilipin 2 knockdown and control, respectively, P < 0.001, Fig. 3F ). These data indicate that perilipin 2 knockdown in the livers of Cideb Ϫ / Ϫ mice leads to reduced hepatic TAG accumulation and increased plasma TAG levels.
ranging from 1.013 to 1.053 (IDL/LDL lipoproteins) and consisted of signifi cantly fewer VLDL particles. In contrast, the ER fractions from the livers of both wild-type and Cideb Ϫ / Ϫ mice contained apoB-containing lipoproteins with densities of 1.041-1.053 (i.e., LDL range) but no VLDL or IDL density particles. In addition, the ER fractions from the livers of Cideb Ϫ / Ϫ mice contained a unique high-density lipoprotein species. These data not only confi rm our previous observation that Cideb promotes VLDL lipidation and maturation but also further suggest that Cideb may mediate VLDL lipidation in the Golgi apparatus.
Increased perilipin 2 expression in the livers of Cideb Ϫ / Ϫ mice As Cideb is a LD-associated protein, we hypothesized that other LD-associated proteins may infl uence Cidebmediated VLDL lipidation and maturation. To identify potential factors in the Cideb-mediated VLDL lipidation pathway, we assessed the expression levels of several LDassociated proteins, including PAT family proteins (perilipin 1, perilipin 2, perilipin 3, and perilipin 5) and the other two CIDE family proteins (Cidea and Fsp27) in the livers of Cideb Ϫ / Ϫ mice. The perilipin 1, Cidea, and Fsp27 proteins were not detected in the livers of wild-type and Cideb Ϫ / Ϫ mice (data not shown). Interestingly, we observed a marked increase in the expression of perilipin 2 at both the mRNA and protein levels in the livers of Cideb Ϫ / Ϫ mice under fasting conditions ( P < 0.001, Fig. 3A , B ). Hepatic levels of perilipin 5 were slightly increased in Cideb -deficient mice ( Fig. 3B ). However, the hepatic levels of perilipin 3 were similar between wild-type and Cideb Ϫ / Ϫ mice under both fed and fasting conditions ( Fig. 3B ). Markedly increased perilipin 2 levels and slightly higher levels of perilipin 3 and perilipin 5 were also observed in the LD fractions of Cideb -/mice compared with that of wild-type mice (supplementary Fig. I-A). In addition, the protein levels of perilipin 2 were signifi cantly higher in isolated Cideb Ϫ / Ϫ hepatocytes treated with oleic acid (OA) ( Fig. 3C ). These data indicate that the levels of perilipin 2 were signifi cantly upregulated in the livers of Cideb Ϫ / Ϫ mice.

Reduced TAG accumulation in the livers of perilipin 2 knockdown and Cideb ؊ / ؊ mice
To evaluate the role of perilipin 2 in modulating Cidebmediated VLDL lipidation, we knocked down perilipin 2 expression in the livers of wild-type and Cideb Ϫ / Ϫ mice using a targeted adenoviral system. As shown in Fig. 3D , adenoviruses containing shRNA against perilipin 2 (Ad-shPerilipin 2)

VLDL-TAG secretion rate was signifi cantly lower in Cideb
Ϫ / Ϫ mice compared with wild-type mice. When perilipin 2 was knocked down in the Cideb Ϫ / Ϫ mice, the VLDL-TAG secretion rate was dramatically increased and reached the levels observed in wild-type mice, indicating that knocking down perilipin 2 rescued the VLDL-TAG secretion defect in Cideb Ϫ / Ϫ mice. The VLDL-TAG secretion was also markedly increased in wild-type mice with hepatic perilipin 2 knockdown. In addition, the plasma collected from Cideb Ϫ / Ϫ Increased VLDL-TAG secretion in Cideb ؊ / ؊ mice with liver-specifi c knockdown of perilipin 2 The reduced hepatic TAG accumulation and increased plasma TAG levels in Cideb Ϫ / Ϫ mice with liver-specifi c knockdown of perilipin 2 suggest that hepatic TAG secretion might be increased. To test this in vivo, we injected the animals with Triton-WR1339 to block VLDL catabolism and clearance and measured the plasma TAG levels at various time points after injection. As shown in Fig. 4A , the
Consistent with the role of CIDE proteins in controlling LD size ( 51 ), Cideb Ϫ / Ϫ hepatocytes contained much smaller and more clustered LDs compared with wild-type cells ( Fig. 6A ).
The accumulation of a large number of small LDs in Cideb Ϫ / Ϫ hepatocytes was further confi rmed by electron microscopy ( Fig. 6B ). When perilipin 2 was knocked down in wild-type hepatocytes, the sizes of the LDs were increased ( Fig. 6C ). Statistical analysis revealed that the majority of the LDs in wild-type hepatocytes manifested sizes ranging 1-2 m in diameter ( Fig. 6D ), with a small population (less than 2%) of LDs larger than 3 m in diameter. However, more than 30% of the LDs in wild-type hepatocytes with perilipin 2 knockdown were larger than 3 m in diameter ( Fig. 6D ). Knocking down perilipin 2 in Cideb Ϫ / Ϫ hepatocytes also signifi cantly increased the LD size; more than 50% of LDs were 1-2 m in diameter, whereas 65% of the LDs in Cideb Ϫ / Ϫ hepatocytes were less than 1 m in diameter ( Fig. 6E, F ). These data indicate that Cideb and perilipin 2 also play opposing roles in controlling LD size, with Cideb as a positive and perilipin 2 as a negative regulator of LD size in hepatocytes.
mice with liver-specifi c perilipin 2 knockdown appeared creamy and turbid, whereas the control plasma was clear and bright ( Fig. 4B ). These data indicate that perilipin 2 is a negative regulator of VLDL-TAG secretion, and its knockdown completely rescues the VLDL-TAG secretion defect caused by Cideb defi ciency.
To confi rm that the increased TAG secretion is indeed due to increased VLDL lipidation and not to enhanced apoB synthesis and secretion, we measured the levels of hepatic apoB-100/apoB-48 and MTP, two proteins crucial for VLDL assembly, and observed no difference in their expression levels in Cideb Ϫ / Ϫ mice with liver-specifi c perilipin 2 knockdown compared with controls ( Fig. 4C ). In addition, the plasma levels of apoB were similar in Cideb Ϫ / Ϫ mice with or without perilipin 2 knockdown ( Fig. 4D ). Furthermore, the secretion rates of apoB-100 and apoB-48 were similar in Cideb Ϫ / Ϫ mice with or without perilipin 2 knockdown ( Fig. 4E ). Consistent with previous analysis in McA-RH7777 cells ( 32 ), the overexpression of perilipin 2 in the livers of wild-type mice increased hepatic TAG accumulation, reduced plasma TAG levels, and signifi cantly decreased VLDL-TAG secretion (data not shown). Therefore, perilipin 2 is a negative regulator of VLDL lipidation, and its defi ciency enhances VLDL lipidation and maturation in the livers of Cideb Ϫ / Ϫ mice.

Knockdown of perilipin 2 restores mature VLDL accumulation in the Golgi of Cideb ؊ / ؊ livers under fasting conditions
To provide further evidence that perilipin 2 knockdown in the livers of Cideb Ϫ / Ϫ mice leads to increased VLDL lipidation, we isolated the Golgi and ER microsomal fractions from the livers of Cideb Ϫ / Ϫ mice with or without perilipin 2 knockdown, and then separated the lipoproteins from the fractions by sucrose density gradient centrifugation. Under fasting conditions, lower levels of mature VLDL particles (densities ranging from 1.0 to 1.005) were detected in the Golgi microsomal fractions from the livers of Cideb Ϫ / Ϫ mice with control adenovirus ( Fig. 5A ). However, high levels of mature VLDL were observed in the Golgi fractions from the livers of Cideb Ϫ / Ϫ mice with liver-specifi c knockdown of perilipin 2 ( Fig. 5A ). In contrast, the ER fractions from the livers of Cideb Ϫ / Ϫ mice with or without perilipin 2 knockdown exhibited similar distributions of lipoprotein particles ( Fig. 5B ). Increased mature VLDL accumulation in the Golgi fraction (fraction 1) was also observed in wildtype mice with liver-specifi c knockdown of perilipin (supplementary Fig. II). These data indicate that, under fasting conditions, perilipin 2 is a negative regulator of VLDL lipidation. Under fed conditions, the amount of mature VLDL particles in the liver Golgi fractions and the levels of hepatic TAG were similar between wild-type and Cideb Ϫ / Ϫ mice with or without liver-specifi c knockdown of perilipin 2 (supplementary Fig. III). These data indicate that Cideb and perilipin 2 play specifi c roles in controlling hepatic TAG accumulation and VLDL lipidation under fasting conditions.

Perilipin 2 defi ciency enhances LD size in
Because both Cideb and perilipin 2 are LD-associated proteins, we further assessed the morphology of the LDs in using similar approaches demonstrated that perilipin 2 is specifi cally localized to LDs but not to the ER or Golgi. Furthermore, other ER-associated proteins (MTP and GRP94) are not associated with the Golgi or LDs. The broad distribution of Cideb in the ER and Golgi apparatus overlaps with apoB, providing a physical basis for its role in interacting with apoB and controlling VLDL lipidation. Based on our previous and current analyses, Cideb appears to be an important positive regulator of VLDL lipidation and maturation in the livers as Cideb Ϫ / Ϫ mice. In addition, isolated Cideb Ϫ / Ϫ hepatocytes exhibit reduced VLDL lipidation and TAG secretion. In contrast, the overexpression of Cideb in the livers of Cideb Ϫ / Ϫ mice increases VLDL lipidation and TAG secretion ( 47 ). Furthermore, Cideb has been shown to mediate PGC-1 ␣ -stimulated VLDL-TAG secretion in HepG2 cells ( 54 ). Finally, we demonstrate here that the Golgi fractions in the livers of Cideb Ϫ / Ϫ mice contain signifi cantly reduced levels of mature VLDL. What is the molecular mechanism of Cidebmediated LD growth and VLDL lipidation? Cidea and Fsp27, two proteins highly related to Cideb, are clustered and enriched at LD contact sites (LDCS) and promote lipid transfer and LD growth ( 51 ). By analogy, Cideb may promote VLDL lipidation through its interaction with apoB and direct the transfer of TAG from cytosolic LDs to pre-VLDL particles that are anchored to the membrane of the ER and/or Golgi apparatus. Alternatively, Cidea and Fsp27 have been reported to control LD size and lipid storage by regulating lipolysis in adipocytes, as Fsp27 and Cidea defi ciencies result in increased lipolysis ( 41,43,55 ),

DISCUSSION
One long-standing question about VLDL assembly and maturation is how cytosolic TAG is transferred from the lipid storage organelles, LDs, to pre-VLDL particles. We have previously shown that Cideb is localized to the LDs and ER, and hepatocytes with Cideb defi ciency exhibited increased TAG accumulation under fasting conditions and reduced VLDL-TAG secretion ( 47 ). Here, we demonstrate that Cideb is also detected in the Golgi apparatus, an important site for VLDL lipidation and maturation. Cideb Ϫ / Ϫ mice exhibited signifi cantly reduced accumulation of mature VLDL particles in the Golgi fractions from the liver. Interestingly, we observed that the expression of perilipin 2, a LD-associated protein, is markedly upregulated in the livers of Cideb Ϫ / Ϫ mice. Knocking down perilipin 2 in the livers of Cideb Ϫ / Ϫ mice restored VLDL lipidation, increased VLDL-TAG secretion, and decreased hepatic lipid storage in Cideb Ϫ / Ϫ mice. Consistent with the role of CIDE proteins in promoting LD growth, Cideb Ϫ / Ϫ hepatocytes accumulate small LDs. Interestingly, perilipin 2 defi ciency in both wild-type and Cideb Ϫ / Ϫ hepatocytes resulted in the accumulation of larger LDs. Cideb has been previously shown to localize to the LDs and ER by fl uorescent staining and biochemical fractionation, similar to Cidea and Fsp27 ( 43,44,47,52,53 ). Here, using a sucrose density gradient centrifugation approach, we observed that Cideb is also localized to the Golgi apparatus. The localization of Cideb to three subcellular compartments (LDs, ER, and Golgi) is a unique feature, as curs in the ER or Golgi apparatus? Furthermore, how do we explain the increased VLDL lipidation in the absence of both positive (Cideb) and negative (perilipin 2) regulators? In addition to associating with LDs, perilipin 2 is found clustered at the cytoplasmic leafl ets of the ER membrane next to the LDs ( 61 ), and reduced perilipin 2 expression causes the apposition of LDs and the ER membranes ( 23 ). In addition, apoB processing and LD formation have been shown to be closely connected, as LDs can be trapped in the ER membrane by binding to lipidated apoB ( 24 ). Cideb may act as a positive factor to bring cytosolic LDs into close proximity with pre-VLDL particles associated with ER and Golgi membranes to promote VLDL lipidation. The presence of perilipin 2 on the LD surface may prevent this close contact between the cytosolic LDs and pre-VLDL particles ( Fig. 7 ). Therefore, the loss of perilipin 2 results in the close apposition of LDs with the ER or Golgi membranes and enhanced VLDL lipidation in the hepatocytes of wild-type and Cideb -/mice. Alternatively, Cideb and perilipin 2 may control VLDL lipidation through their cooperative regulation of the lipolysis and reesterifi cation pathways. Interestingly, we observed drastically increased perilipin 5 levels in LD fractions of Cideb -/mice that had a perilipin 2 knockdown, suggesting that perilipin 5 may provide a compensatory effect for the loss of perilipin 2 in controlling LD formation and VLDL lipidation. However, LDs lacking Cideb and perilipin 2 (but increased perilipin 5) in hepatocytes may not be compatible with cytoplasmic storage, and their TAG would need to be shuttled into the secretory pathway (ER and Golgi) in the form of lumenal LDs for use in the fi nal steps of VLDL lipidation in the Golgi. Overall, our data demonstrate that Cideb and perilipin 2 exhibit opposing effects in controlling LD growth and VLDL lipidation in hepatocytes.
Interestingly, we observed dramatically increased perilipin 2 expression in the livers of Cideb Ϫ / Ϫ mice. Perilipin 2 has been shown to be a negative regulator of hepatic microsomal TAG accumulation and VLDL secretion ( 32,34,35,60 ). We have demonstrated that the depletion of perilipin 2 in the livers of Cideb Ϫ / Ϫ mice results in signifi cantly decreased hepatic TAG accumulation and increased levels of VLDL-TAG secretion, reversing the phenotype caused by Cideb defi ciency. Reducing the expression of perilipin 2 in the livers of Cideb Ϫ / Ϫ mice also restores VLDL lipidation, as the amount of mature VLDL particles accumulated in the Golgi apparatus of Cideb Ϫ / Ϫ hepatocytes is markedly increased. Unlike Cideb, which is localized to the LD, ER, and Golgi apparatus, perilipin 2 is specifi cally localized to the LDs but not to the ER or Golgi. How does a LD-associated protein like perilipin 2 control VLDL lipidation, which oc- Fig. 7. Model of the possible functions of Cideb and perilipin 2 in regulating LD growth and VLDL lipidation. Cideb is localized to the Golgi, ER, and LDs. Perilipin 2 only localizes to LDs. Cideb may mediate VLDL lipidation in the ER or Golgi through the direct transportation of TAG from cytosolic LDs to pre-VLDL particles that are attached to the membrane of the ER and Golgi. Perilipin 2 may act as a physical barrier on LDs to block the Cideb-mediated close apposition of cytosolic LDs and pre-VLDL particles and inhibit the direct transportation of TAG from the LDs to pre-VLDL particles. Cideb and perilipin 2 may work cooperatively to regulate lipolysis and reesterifi cation and control VLDL lipidation.