Overexpression of Perilipin A and B Blocks the Ability of Tumor Necrosis Factor a to Increase Lipolysis in 3T3-L1 Adipocytes*

Perilipins, a family of phosphoproteins, are specifically located at the surface of intracellular lipid (tria-cylglycerol) droplets, the site of lipolysis. Stimulation of lipolysis in 3T3-L1 adipocytes by tumor necrosis factor a (TNF- a ) is associated with a decrease in total cellular expression of perilipin A and B, consistent with the hypothesis that a decrease in perilipin protein expression is required for TNF- a -induced lipolysis. Adenovirus-me-diated overexpression of perilipin A or B maintains perilipin protein levels on the lipid droplet and blocks TNF- a -induced lipolysis. In contrast, overexpression of perilipin A or perilipin B does not inhibit isoproterenol-stimulated lipolysis and does not alter the isoprotere-nol-induced migration of perilipins from the lipid droplet. These results provide the first evidence of how perilipin functions and suggest that TNF- a regulates lipolysis, in part, by decreasing perilipin protein levels at the lipid droplet surface. were cotrans- fected into HEK 293 cells (Microbix, Toronto, ON, Canada), which provide in trans the missing adenoviral early region 1 functions of pJM17. Clonal viral stocks were isolated by single plaque purification, and their DNA was analyzed by polymerase chain reaction, amplified in HEK 293 cells, and purified and concentrated to 10 12 plaque-forming units/ml by CsCl ultracentrifugation. Cell Culture and Adenovirus Treatment— 3T3-L1 fibroblasts placed Cy-5 these studies were initiated, who encouraged and supported our studies on the mechanism of cytokine-activated lipolysis in adipocytes.

The major form of stored energy in the body is triacylglycerol contained in white adipose tissue (1). In times of energy need, such as fasting and exercise, catecholamines rapidly activate cAMP-dependent protein kinase (PKA), 1 resulting in hydrolysis of triacylglycerol to glycerol and free fatty acids (FFA). FFA are then used as substrates for ATP generation in diverse tissues. However, in certain pathophysiologic states such as obesity and diabetes, circulating FFA are chronically elevated (1,2). It has been suggested that chronic elevations of plasma FFA promote insulin resistance (2). The cytokine tumor necrosis factor ␣ (TNF-␣) has been associated with insulin resistance in obese animals (3) and humans (4,5). In contrast to catecholamines, which stimulate lipolysis within minutes (6), TNF-␣ increases lipolysis only after hours (Ͼ6 h) of incubation by an undetermined mechanism(s) (7,8). Infusion of TNF-␣ in mice results in an increase in plasma FFA and systemic insulin resistance (9). In addition, TNF-␣ knockout mice exhibit lower circulating FFA and are protected from the insulin resistance of obesity (10). Thus, TNF-␣-induced lipolysis may be one mechanism by which this cytokine is involved in the pathogenesis of the obese/diabetic state.
Whereas hormonal regulation of lipolysis is well described, little is known about the cellular and molecular mechanisms underlying this process. Lipolysis occurs at the surface of the intracellular lipid droplet where the perilipins, a family of phosphoproteins, are specifically located (11,12). The predominant perilipins in adipocytes, perilipin (Peri) A and B, arise by alternative RNA splicing from a single gene, generating predicted protein products of 57 and 46 kDa, respectively (13). Peri A is the most abundant of the perilipin proteins and is also the major phosphorylation target for PKA in adipocytes (12)(13)(14). Despite their abundance and unique location at the surface of the lipid droplet, the role of the perilipins in lipolysis is unknown.
In the present study we demonstrate that TNF-␣ decreases the protein levels of perilipins at the surface of the lipid droplet concurrent with an increase in adipocyte lipolysis. Adenovirusmediated overexpression of Peri A and Peri B blocks the ability of TNF-␣ to increase lipolysis. These results are consistent with the model in which TNF-␣ increases lipolysis by decreasing the expression of perilipin proteins.

EXPERIMENTAL PROCEDURES
Materials-A polyclonal antibody that recognizes both Peri A and Peri B (Ab 46) was generated as described previously (8). A specific polyclonal anti-Peri A antibody was generated using peptide: PREK-PARRVSDSFFRPSVC (Ab PREK). Antibodies were subsequently affinity-purified and used for Western blotting (1:1500) and for immunofluorescence (1:500). Recombinant murine TNF-␣ was purchased from Genzyme (Cambridge, MA), and 3T3-L1 cells were from the American Type Culture Collection (Manassas, VA). All other chemicals were purchased from Sigma.
Generation of Recombinant Adenovirus-Plasmids pJM17 (15) and pAC.CMVpLpA (16) have been described. Plasmids pACCMV-␤-Gal, pACCMV-GFP, pACCMV-Peri A, and pACCMV-Peri B were obtained by subcloning the cDNA for ␤-galactosidase (␤-Gal) with an SV40 nuclear localization sequence, the Aequoria victoria green fluorescent protein, mouse Peri A (gift of C. Londos, A. R. Kimmel, and J. Gruia-Gray, National Institutes of Health) and mouse Peri B (cloned by screening a murine adipocyte cDNA library) into the multiple cloning site of the vector pACCMVpLpA. Recombinant, replication-defective adenoviruses (Ad) were generated by homologous recombination as described previously (15,17,18). Briefly, pACCMV constructs and the pJM17 vector, which contains a modified Ad5 genome, were cotransfected into HEK 293 cells (Microbix, Toronto, ON, Canada), which provide in trans the missing adenoviral early region 1 functions of pJM17. Clonal viral stocks were isolated by single plaque purification, and their DNA was analyzed by polymerase chain reaction, amplified in HEK 293 cells, and purified and concentrated to 10 12 plaque-forming units/ml by CsCl ultracentrifugation.
Cell Culture and Adenovirus Treatment-3T3-L1 fibroblasts placed in 12-well plates were cultured in Dulbecco's modified Eagle's medium containing 10% bovine calf serum and differentiated using standard protocols (19). Adipocytes were infected on day 5 of differentiation, when small lipid droplets were visible, with a multiplicity of infection of ϳ100 plaque-forming units/cell, for 18 h at 37°C. Four days later, infected adipocytes were serum-deprived overnight in Dulbecco's modified Eagle's medium with 2% bovine serum albumin. For most of the assays, recombinant adenovirus expressing ␤-galactosidase was used as a control (Ad ␤-Gal). Adenoviruses expressing either green fluorescent protein or no exogenous proteins were also used as controls, and the results were identical to those observed with Ad ␤-Gal. The following morning, cells were treated as described under "Results." Lipolysis-Glycerol content of the incubation medium was determined using a colorimetric assay (GPO-Trinder, Sigma). Protein content was determined using the BCA protein assay (Pierce).
Western Analysis-Adipocytes were rinsed briefly with 1 ml of phosphate-buffered saline. Proteins were extracted as described previously (8), separated in 10% SDS-polyacrylamide gel electrophoresis, and electrophoretically transferred to nitrocellulose membranes. Equivalent amounts of protein were loaded onto the gel for each treatment as described in figure legends. Proteins were detected with the ECL system (Amersham Pharmacia Biotech).
Immunofluorescence-For determination of perilipin fluorescence, cells were cultured in 35-mm coverslip bottomed dishes (MatTek Corp., Ashland, MA) and infected as described above. After treatment, adipocytes were fixed in 2% paraformaldehyde for 10 min at 25°C, washed, and treated with either polyclonal antiserum that recognizes both Peri A and Peri B or a specific anti-Peri A polyclonal antibody (Ab. PREK, 1:500 dilution) and a donkey-anti-rabbit Cy-5 labeled antibody (Jackson Immunoresearch). Cy-5 fluorescence imaging was assessed by confocal microscopy as described (20).
Statistical Analysis-Results are expressed as the means Ϯ S.E. Effects were assessed by using single-factor analysis of variance. Tukey's honestly significant differences were used to make pairwise comparisons. All calculations were performed using SYSTAT version 7 for Windows (SPSS, Inc., Chicago, IL).
Overexpression of Peri A and Peri B Blocks TNF-␣-induced Lipolysis-To ascertain the role of perilipin in TNF-␣-induced lipolysis, we generated two recombinant type 5 adenoviruses expressing either Peri A or Peri B (Ad Peri A or Ad Peri B). 3T3-L1 adipocytes were infected with adenovirus control (Ad ␤-Gal) and Ad Peri A and Ad Peri B (see "Experimental Procedures") and incubated in the presence or absence of 10 ng/ml TNF-␣ for 24 h. Western blotting of protein lysates revealed that Peri A and Peri B expression were increased by 5-and 49-fold, respectively, as compared with Ad ␤-Gal-transduced cells (Fig. 2A, lane 1 versus lanes 3 and 5). Importantly, in adipocytes transduced with Ad Peri A or Ad Peri B and subsequently incubated with TNF-␣, the expression of Peri A or B was not decreased as compared with untreated cells (Fig. 2A,  lane 3 versus lane 4 and lane 5 versus lane 6). In fact, perilipin levels were higher with TNF-␣-treated cells than in controls. This is presumably because of up-regulation of the CMV promoter by TNF-␣-induced activation of stress-activated mito-gen-activated protein kinases (21)(22)(23).
To determine whether maintenance of elevated perilipin protein levels blocked TNF-␣-stimulated adipocyte lipolysis, media from the treated adipocytes described in Fig. 2A were assayed for glycerol release, an index of lipolysis. In adipocytes overexpressing Peri A or Peri B, TNF-␣-induced lipolysis was reduced by 80 and 72%, respectively (Fig. 2B). This decrease was observed even when the rate of lipolysis was maximal, i.e. during the last 3 h of a 24-h incubation (Fig. 2C).
Overexpression of Perilipins Does Not Alter Isoproterenolinduced Lipolysis-To determine whether overexpression of perilipins affected catecholamine-stimulated lipolysis, adipocytes transduced with Ad ␤-Gal, Ad Peri A, or Ad Peri B were treated with 10 M isoproterenol (Iso) for 3 h (Iso-induced lipolysis is maximal). 2 Overexpression of Peri A or Peri B had no significant effect on catecholamine-stimulated lipolysis (Fig.  3A), suggesting that TNF-␣ and Iso stimulate lipolysis through different mechanisms. Moreover, Western blotting analyses of adipocytes treated as described in Fig. 3A demonstrated that overexpression of either Peri A or Peri B did not affect the ability of Iso treatment to hyperphosphorylate Peri A (Fig. 3B). Iso treatment is known to alter the electrophoretic migration of Peri A but not of Peri B. The alterations in Peri A migration are attributed to hyperphosphorylation by PKA (12,14).
Different Mechanisms Underlie Stimulation of Lipolysis by TNF-␣ or Iso-To determine the mechanisms by which overexpression of the perilipins selectively blocks TNF-␣-induced lipolysis but not Iso-stimulated lipolysis, we examined the subcellular localization of the perilipin proteins in 3T3-L1 adipocytes incubated with either 10 ng/ml of TNF-␣ for 24 h or with 10 M Iso for 3 h. Simultaneous perilipin immunofluorescent (IF) and differential interference contrast (transmission) (DIC) images allowed us to correlate immunoreactivity with intracellular structure. IF studies were performed using perilipin antiserum that recognizes a common epitope in Peri A and Peri B. These studies confirmed that perilipins are located at the surface of the lipid droplet (Fig. 4). IF and DIC images demonstrated that TNF-␣ reduced perilipin expression at the surface of lipid droplets as compared with untreated cells (Fig.  4, A, B, D, and E). In contrast, adipocytes treated with Iso exhibit a reduction of droplet-associated perilipins, with some 2 Y. Kang and A. S. Greenberg, unpublished data. cells showing a punctate pattern and others a diffuse cytoplasmic immunostaining (Fig. 4, C and F). Although reducing perilipin levels at the surface of intracellular lipid droplets appears to be required for both TNF-␣ and Iso to stimulate lipolysis, this reduction is achieved by different mechanisms. TNF-␣ decreases the total amount of perilipin protein, whereas Iso does not.
To determine the subcellular localization of the perilipin proteins in adipocytes overexpressing Peri A, IF analyses were performed using antiserum specific for Peri A. Adipocytes overexpressing Peri A generated stronger fluorescence (Fig. 5, B, C, and D), relative to Ad ␤-Gal-transduced adipocytes (Fig. 5A). Overexpressed perilipins localized to the lipid droplet similar to the endogenous perilipin (compare Figs. 4A with 5B). In addition, perilipin immunostaining at the surface of the lipid droplet did not decrease in cells transduced with Ad Peri A that were treated with 10 ng/ml of TNF-␣ for 24 h (Fig. 5, B versus C). In contrast, perilipin immunostaining in Iso-treated cells was similar to that observed with endogenous perilipins (compare Figs. 5D with 4C). The pattern of immunofluorescence in adipocytes overexpressing Peri B was similar to that observed with overexpression of Peri A (data not shown). Thus, maintaining high levels of perilipin at the surface of the lipid droplet impairs the ability of TNF-␣ to increase lipolysis, whereas Iso-stimulated lipolysis and its apparent redistribution of perilipins to the cytoplasm are not altered. DISCUSSION The perilipin phosphoproteins are highly expressed in adipocytes and are located at the surface of the lipid droplet, the site of lipolysis in adipocytes. Despite the location of the perilipins, their role in lipid hydrolysis is unknown. Using TNF-␣-induced lipolysis as a model of reduced perilipin expression in adipocytes, the experiments described in this paper demonstrate for the first time that Peri A and B can regulate the hydrolysis of triacylglycerol. We observed that a decrease in perilipin protein levels at the lipid droplet surface is associated with TNF-␣-induced lipolysis. Adenovirus-mediated overexpression of Peri A and B blocked TNF-␣-induced lipolysis in 3T3-L1 adipocytes, thus delineating one mechanism whereby this cytokine could increase the breakdown of triacylglycerol. Moreover, this effect is specific for TNF-␣, because Iso-stimulated lipolysis was not altered when Peri A or B was overexpressed.
Our IF studies confirm that Peri A and B are located at the surface of lipid droplets in untreated cells and indicate that TNF-␣-and Iso-stimulated lipolysis are associated with reduced perilipin protein levels at the surface of the lipid droplet. TNF-␣ reduces total cellular perilipins and thus lipid dropletassociated perilipin protein levels while increasing adipocyte lipolysis. Overexpression of the perilipins prevents both the TNF-␣-mediated reduction in perilipin protein accumulation at the surface of the lipid droplet and the ability of the cytokine to increase lipolysis. However, in Iso-treated adipocytes, the pattern of perilipin immunostaining is consistent with a reduction of perilipins at the lipid droplet surface and redistribution in the cell. 3 Consistent with the confocal experiments, in cell fractionation studies, Iso induces a redistribution of the perilipins from the fat (fat cake) to the infranatant. 2 PKA-mediated phosphorylation, secondary to Iso stimulation, is presumably required for the redistribution of the perilipin proteins. Based on these data we propose a model of lipolysis in which perilipin accumulation at the lipid droplet surface may limit lipid hydrolysis. 4 TNF-␣ reduces perilipin protein levels at the lipid droplet surface, resulting in an increase in lipolysis. In contrast, Iso-induced phosphorylation of perilipins would cause relocation of the perilipins, potentially increasing the accessibility of stored triacylglycerol to lipases. Further studies are necessary to extend our understanding of the role of perilipins in catecholamine-stimulated lipolysis and to delineate their mechanisms in lipid hydrolysis.
The thiazolidinediones (TZDs) are a class of antidiabetic agents that, by unknown mechanisms, ameliorate insulin resistance and type II diabetes mellitus. One suggested mechanism is that the TZDs antagonize the actions of TNF-␣, resulting in decreased plasma FFA levels (9,25). Increased plasma FFA have been suggested to induce systemic insulin resistance (2). We have previously demonstrated that the TZDs partially blocked both the TNF-␣-mediated reduction of Peri A and TNF-␣-induced lipolysis (8). The results of this prior study are consistent with the data presented in this paper and suggest a possible role for the perilipins and TNF-␣-induced lipolysis in the pathogenesis of the obese-diabetic state.
Acknowledgments-We thank Dr. I. Rosenberg, R. Coleman, and S. Fried for scientific support and encouragement. We thank Dr. M. Obin for his editorial review of the manuscript. We thank Dr. Londos and colleagues (National Institutes of Health) for discussions on their observations prior to publication, which suggest that perilipins protect against lipid hydrolysis in unstimulated cells. We also thank D. M. Levin, T. Barber, J. Blanchette-Mackie, and C. Londos for performing the initial confocal study and sharing the observations that TNF-␣ causes a disruption of the perilipin shell at the lipid droplet surface in 3T3-L1 adipocytes. Finally, we are grateful to Dr. Londos, in whose laboratory these studies were initiated, who encouraged and supported our studies on the mechanism of cytokine-activated lipolysis in adipocytes.

FIG. 5.
Immunofluorescence of transduced cells demonstrates the localization of overexpressed Peri A to the lipid droplet and the differential effects of TNF-␣ and Iso. Confocal IF images of 3T3-L1 adipocytes transduced with Ad ␤-Gal (A) and Ad Peri A (B, C, and D) using a specific Peri A antibody. Adipocytes were incubated in the presence or absence of TNF-␣ (10 ng/ml) for 24 h or Iso (10 M) for 3 h. Data shown are representative of four experiments. Laser bean intensity was 1 ⁄10 of the one used in Fig. 4.