Consequences of Lipid Droplet Coat Protein Downregulation in Liver Cells: Abnormal Lipid Droplet Metabolism and Induction of Insulin Resistance

AML12 cells (Dr. Steven Farmer, Boston University, Boston, MA) were grown with the standard protocol (American Type Culture Collection, Manassas, VA). For siRNA experiments, cells were plated in 24 multiwell dishes (Costar; Thermo Fisher Scientific, Pittsburgh, PA) at a density of 1 × 10⁴ cells per well, transfected the next day with Hyperfect (Qiagen, Valencia, CA) according to the manufacturer's instructions. For immunocytochemistry, cells were plated in four-well chamber slides (Labtek; Thermo Fisher Scientific). For insulin-signaling assays, cells were deprived of insulin for 24 h and for the last 12 h were incubated in Dulbecco's modified Eagle's medium (DMEM)/F-12 media (1:1) containing 1% defatted BSA (Sigma-Aldrich, St. Louis, MO) and supplemented with 400 μmol/l oleic acid complexed to 0.4% BSA to promote TAG deposition. The next day, cells were incubated in DMEM/F-12 media (1:1) containing 1% defatted BSA (Sigma-Aldrich) for 6 h before a 10-min incubation with insulin at the concentration indicated. Cells were harvested in cold lysis buffer (150 mmol/l NaCl; 50 mmol/l Tris-HCl, pH 7.5; 1% Triton X-100; 0.5% NP-40; 0.25% sodium deoxycholate; 1 mmol/l EDTA; 1 mmol/l EGTA; 0.2 mmol/l ortho-vanadate; 1 mmol/l NaF; and protease inhibitor cocktail) (Roche Applied Science, Indianapolis, IN).

siRNA was purchased from Qiagen (HP-guaranteed siRNA). Positive siRNA for map kinase 1 and fluorescent oligos (Qiagen) were used to establish efficient conditions for transfection to obtain 80% inhibition by Western blot (12) versus All Star negative control. For off-target toxicity, we used siRNA negative for insulin signaling: a negative siRNA targeting ADFP with a two-base mismatch (Ctl1), sense sequence 5′AACGTCTGCTTGAGCCGAATA-3′ (online appendix Fig. S-9 [available at http://dx.doi.org/10.2337/db07-1383]); and a negative −siRNA targeting Tip47 (Ctl2), sense sequence 5′-GCGUGUCCAAUCAGUCAU-3′ (12). Sense sequences for Tip47 (5′-AACAGCACAGAGAAUGAGGAG-3′) and ADFP (5′-AACGTCTGTCTGGACCGAATA-3′) were selected by potency. Total siRNA per well was 10 nmol/l for all transfections; for double transfection using one-third Tip47 and two-thirds ADFP siRNA; for triple using one-fourth Tip47, one-half ADFP, and one-fourth adipose triglyceride lipase (ATGL) siRNA; and downregulation confirmed (online appendix Figs. S-4 and -7). siRNA for ATGL was reported (35) and confirmed by imunoblot (Fig. S-3). siRNA efficiencies were quantified by Western blot (Un-scan; Silk Scientific, Orem, UT) and results expressed as percentage of control.

C57BL/6 (8 weeks old) mice were ad libitum fed either a high-fat (60% kcal fat; Research Diets, New Brunswick, NJ) diet containing primarily lard or a low-fat (10% kcal fat, Research Diets) diet for 12 weeks. Mice on the high-fat diet exhibited diet-induced obesity (1.52-fold increase in body weight over low-fat–fed control mice) and impaired glucose tolerance (data not shown). On the day of the experiments, animals were killed by cervical dislocation and tissues were immediately frozen in liquid nitrogen. Tissue lipid was extracted (40) and expressed as percent lipid weight divided by total sample weight (mg/mg).

Rabbit anti-Tip47, LSDP-5, and goat anti-ADFP were used (12,13). Neutral lipid dye bodipy 558/568, alexa fluor 488, hoescht, and alexa fluor 598 species-specific secondary antibodies were from Molecular Probes (Invitrogen, Carlsbad, CA). Rabbit antibodies against human ATGL (Rockland Immunochemicals for Research, Gilbertsville, PA) were tested in human preadipocytes and 3T3-L1 adipocytes and in human and mice adipose tissues (M.B., C.S., and Mee Jeong Lee, unpublished data) or purchased, as were Akt, phosph-Akt, PDK-1, GS3-Kα, phosphor–GS3-K α, phospho-specific antibodies to protein kinase C (PKC), extracellular signal–regulated kinase (ERK), ERK-phosp antibodies (Cell Signaling technology, Beverly, MA), phosphor–insulin receptor substrate (IRS)-1 (Ser307) (Upstate Biotechnology, Lake Placid, NY), IRS-1 and Foxo-1 (Santa Cruz Biotechnology, Santa Cruz, CA), and β-actin (Abcam, Cambridge, U.K.). CGI-58 was a gift (Dr. Brasaemle, Rutgers University, Newark NJ) (31). Phospho-specific antibodies to PDK-1 were a gift from Dr. Quon. Immunoprecipitation for IRS-1 was performed as described (36).

Fixation and staining were performed as described (12). Cells were viewed with a confocal laser microscope using a ×63 or ×40 oil objective lens. LD size was determined (LSM510; Carl Zeiss MicroImaging). Cellular extracts for immunoblot were obtained by scraping cells in Laemmli sample buffer. Each SDS-PAGE lane was loaded with protein from a single well of a 24-multiwell dish. Four wells were pooled for each data point for insulin signaling immunoblots. Quantitative analysis was performed using Un-scan (Silk Scientific Corporation). ³²PI3P (phosphatidylinositol 3 phosphate) bands were revealed by PhosphorImager (Storm 860; GE Healthcare) and band intensities quantified (ImageQuant 5.0; GE Healthcare). Live cell imaging was performed using an axiovert 200 microscope equipped with a camera (Carl Zeiss).

Triglyceride synthesis and lipolysis were performed (21,39). Briefly, cells were incubated for 12 h with growth medium supplemented with 400 μmol/l oleic acid complexed to 0.4% BSA to promote triacylglycerol deposition. [³H]oleic acid, at 1 × 10⁶ dpm/well, was included as a tracer. In lipolysis experiments, reesterification of fatty acids in AML12 cells was prevented by inclusion of 10 μmol/l triacsin C (Biomol, Plymouth Meeting, PA), an inhibitor of acylcoenzyme A synthetase, in the medium (21,38,39). Quadruplicate wells were tested for each condition. Lipolysis was determined by measuring radioactivity release (12,21,39).

Cell monolayer was washed with ice-cold PBS and scraped into 1 ml PBS. For total TAG, lipids were extracted by the Dole method (40) from five wells of a 24-mutiwell dish. The total upper phase was dried down, resuspended in isopropanol, and assayed with a triglyceride kit (Wako Chemicals, Richmond, VA). Protein was measured by a kit (Pierce Biotechnology, Rockford, IL). For triglyceride synthesis, lipids were extracted by the Bligh-Dyer method (41) and 10% of the total lipids were analyzed with thin-layer chromatography extractions performed as reported (37,38) (see online appendix). Intracellular diacylglycerol, NEFAs, and ceramide were measured by analytical service (Avanti Polar Lipids, Albaster, AL) (43).

Measurements were performed according to Berusi et al. (48). Briefly, 2 days following siRNA transfection, cells were incubated in DMEM/F12 medium containing 1% BSA, 1 μmol/l dexamethasone (Sigma-Aldrich), and 1 μmol/l cAMP (EMD Chemicals, Gibbstown, NJ) for 24 h. Cells were incubated in 0.35 ml (per well) of phenol red–free, glucose-free DMEM containing 2 mmol/l pyruvate and 20 mmol/l lactate containing dexamethasone and cAMP. Some wells also contained insulin (10⁻⁸ mol/l concentration). Media was collected 5 h later for glucose measurement (48) with a fluorimeter (Molecular Devices, Union City, CA) in triplicate. Glucose output rate was normalized by cellular protein concentration and expressed as nanomols of glucose per milligram of protein every 5 h. Two separate experiments were performed.

Four 24-mutiwell dishes for each condition were treated with siRNA negative or ADFP and Tip47. LD isolation was as reported (25). Total homogenate protein was determined and adjusted to 1 μg/μl. The lipid fat cake was isolated and suspended in 200 μl PBS containing 5% SDS.

Statistical significance was by one-way ANOVA or two-tailed Student's t test. (GraphPad software, San Diego, CA).

After incubation of AML12 liver cells with 400 μmol/l oleic acid, numerous LDs appear, and most are coated with ADFP and fewer with Tip47 (Fig. 1A). Relatively few exhibit colocalized ADFP and Tip47 at the surface. ADFP was found exclusively at LD surfaces, while Tip47 was found in both cytosolic and LD compartments. Other PAT proteins were not detected by imunocytochemistry or immunoblot (results not shown). Hence, AML12 cells express two PAT proteins, ADFP and Tip47.

The effect from downregulation of Tip47, ADFP, or both with siRNA are shown in Fig. 1B and C. siRNA transfection produced minimal cellular toxicity from unchanged levels of β-actin. Efficiency of Tip47 inhibition was >95%, and efficiency of ADFP inhibition was 80%. siRNA treatments effectively produced droplets coated with only one of two PAT proteins or none (Fig. 1B). ADFP downregulation led to a dramatic increase in Tip47 protein at the LD surface, although total Tip47 protein in homogenates was not affected by increased exogenous oleic acid (online appendix Fig. S-1). Thus, increased Tip47 at the LD surface is recruited from preexisting cytosolic Tip47. Behavior of PAT proteins in control cell homogenates in response to increased exogenous NEFAs is also different (online appendix Fig. S-1); while ADFP increased proportionally to the exogenous lipid, Tip47 was barely affected. Importantly, knockdown of one or more PAT proteins resulted in substantial changes in LD size and number, as observed by bodipy staining (Fig. 2) (online appendix Figs. S-5A–D and S-7), phase contrast microscopy (online appendix Fig. S-2), or quantify (Fig. 3). Absence of PAT proteins induced a marked increase in LD size and decrease in LD number (28 ± 2.6 vs. 48.09 ± 6.06 for control; P < 0.05). When Tip47 was predominant, the LD morphology appeared similar to control treatment cells. When ADFP was predominant, LD number increased (251.7 ± 22.6 vs. 48.09 ± 6.06 for control; P < 0.01) but LD size was smaller. Despite changes in LD morphology, TAG content in cellular extracts of oleic acid–loaded cells was unchanged (Fig. 4B). However, total TAG decreased in PAT protein–deprived cells when grown without oleic loading. (Fig. 4A). This difference prompted examination of PAT protein effects on lipid metabolism.

In cells lacking either ADFP or Tip47, no significant alteration in lipolysis was observed (Fig. 5A). However, lipolysis significantly increased up to twofold (Fig. 5A) (one-way ANOVA; P < 0.02) in AML12 cells lacking both proteins. A recent phenotype reported for ATGL-null mice suggests that ATGL is operative in liver (14). Presence of ATGL in AML12 cells was revealed by ATGL-targeted siRNA treatment (online appendix Fig. S-3), which increased LD, as expected from decreased lipolytic activity from loss of ATGL. In cells lacking PAT proteins, ATGL increased at the LD surface (Fig. 5B) (online appendix Fig. S-4). Immunoblot analysis revealed an increase in both ATGL and CGI-58 in fat cakes from these cells (Fig. 5C). CGI-58 is known to increase ATGL activity and to interact with PAT proteins (30). Increase of ATGL at LDs in cells lacking ADFP and Tip47 supports the hypothesis that these proteins limit access of endogenous lipase to the LDs. In contrast, lack of PAT proteins did not affect uptake of ³H exogenous oleic acid and had little influence on the ability of cells to utilize exogenous NEFAs for TAG or phospholipid synthesis (online appendix Fig. S-4). High-performance liquid chromatography analysis did not reveal significant differences in the intracellular content of diacylglycerol (1.45 ± 0.25% [wt/wt] for Ctl vs. 1.4 ± 0.2% [wt/wt]) for cells lacking PAT proteins or ceramide (1.1% in both conditions), but 1.25% (wt/wt) intracellular NEFAs were found in cells lacking PAT proteins, while NEFAs remain undetected in control cells (n = 2 experiments).

Because dysregulation of cellular lipid metabolism has been linked to insulin resistance, we measured the effect of PAT siRNA treatment on insulin activation of protein kinase B (Akt) in AML12 cells by immunoblot analysis of cellular protein extracts using phosphospecific antibodies. In both control- and siRNA-treated cells, insulin stimulated phosphorylation Akt phosphorylation at Ser473, an effect previously shown to correlate with the extent of Akt activation in hepatocytes (16). However, absence of PAT proteins resulted in a significant decrease in insulin responsiveness and sensitivity compared with the control (Fig. 6A and B). Akt phosphorylation was not affected when only one PAT protein was downregulated (results not shown). Effect on Akt phosphorylation was also observed with decreasing concentration of oleic acid in media (online appendix Fig. S-8). To test if increased lipolysis contributed to the observed decrease in insulin responsiveness, cells lacking all PAT proteins were additionally treated with ATGL siRNA. Figure 6C shows that the triple inhibition normalized insulin stimulation of Akt to control levels. The total amount of Akt did not differ among the various treatments. Upstream regulation of Akt was also affected by lack of PAT proteins: phosphorylations of PDK-1 (Fig. 7A), IRS-1/phosphoinositide 3-kinase activity were decreased in cells lacking PAT proteins (Fig. 7B). IRS-1–Ser307 phosphorylation, which promotes general inhibition of IRS-1 signaling (17), was increased in cells lacking PAT proteins (Fig. 7C). However, this did not result in a detectable decrease in tyrosine phosphorylation of IRS-1 (results not shown). Downstream targets of Akt, Foxo-1, and GS3-Kα/β were affected (Fig. 7E and F). Activation of PKC-è was previously shown to be involved in insulin resistance induced by fatty acids (18). Upon insulin stimulation, cells lacking PAT proteins also increased PKC-è protein and phosphorylation in the total cellular membrane fraction (Fig. 6D). ERK1/2 phosphorylation was increased in both basal and stimulated conditions in cells lacking PAT proteins (Fig. 7G). Overall glucose output was very low in AML112 cells; however, 10⁻⁸ mol/l insulin was able to suppress glucose output in cells transfected with control siRNA to a greater extent than in cells lacking ADFP and Tip47, confirming a defect in these latter cells in insulin signaling (online appendix Table 1)

Lack of information on PAT protein and lipase content of liver fat cake in established models of insulin resistance prompted an examination of DIO mice. Liver fat cakes isolated from DIO mice have fourfold-increased lipid content (11.7 ± 2.2% [wt/wt] vs. 3.4 ± 1% [wt/wt]) for ad libitum–fed mice (P < 0.05). Importantly, ATGL content also increased (Fig. 8). Despite the large increase in lipids, a matching increase was observed only for LSPD-5, while a significant change in Tip47 was not found and the increase in ADFP appears modest.

Using siRNA technology, we were able to develop evidence supporting the role of the PAT proteins in the regulation of ectopic fat deposition and demonstrate the importance of defected LD utilization in the development of cellular insulin resistance. First, the composition of PAT proteins at the LD surface dictates their size and number. Second, the affinity of ADFP may be greater than Tip47 for binding to LDs. Third, the PAT proteins help maintain intracellular NEFA homeostasis by protecting the LDs against lipolysis by decreasing the recruitment of ATGL at the LD surface. Fourth, lack of PAT proteins induces cellular insulin resistance and affects multiple steps in the insulin signaling pathway. Fifth, DIO liver fat cakes differ in their content of PAT proteins and ATGL.

In this study, a loss-of-function approach was developed using siRNA-targeting PAT proteins in a liver cell line because such a system provides a model for one of the most important tissues where ectopic fat develops. The value of a loss-of-function approach for PAT proteins has been proven with the perilipin-null mice, which exhibit a lean phenotype and enable identification of its regulatory role of lipolysis (19,20). However, the ADFP-null mice exhibit few phenotypic alterations (22), attributed tentatively to a compensatory replacement by Tip47 (12). Further analyses of ADFP-null fibroblast cells treated with siRNA against Tip47 supported the importance of the PAT proteins’ role in lipid metabolism (12). AML12 cells are attractive for studies of LDs and insulin signaling (15,23).

Only two PAT proteins, ADFP and Tip47, were found in AML12 cells, predominately ADFP at the liver LD surface (25), whereas Tip47 was observed on some droplets but mainly in the cytosol. This corroborates previous findings that Tip47 and ADFP behave differently in most cells in that Tip47 can be present both in the cytosol and at the LD while ADFP is seen primarily at the LD upon lipid-loaded conditions (25). Inhibition of ADFP expression resulted in increased Tip47 at the LD surface in AML12 cells, similar to our earlier report with cells from ADFP-null mice, suggesting that ADFP has a greater affinity for LDs than Tip47. Thus, when ADFP expression is inhibited, Tip47 moves to the LD surface without major change in LD size and number. However, if Tip47 is repressed, ADFP is then the only PAT protein present and the LD number increases while the sizes decrease. This morphological change may reflect ADFP's greater ability to bind to LDs and greater surfactant function. Finally, absence of both PAT proteins generated fewer but larger LDs. Since PAT proteins have surfactant properties, in their absence LDs probably fuse to minimize surface area contact with the surrounding aqueous cytosol. These studies demonstrate that the PAT protein composition in liver cells is a critical determinant of LD size and number.

Interestingly, varying the LD coat protein did not affect the ability of cells to take up exogenous NEFAs and to esterify it into TAG. Thus, ADFP and Tip47 are not important factors in determining the amount of TAG produced, but they are important factors in determining how it is packaged. However, without PAT proteins, TAG lipolysis increased, demonstrating a function of these two PAT proteins to inhibit LD hydrolysis by endogenous lipase(s). This property is a feature shared for most PAT proteins. Perilipin is known to regulate lipid storage in adipose cells (21,42), and published studies support that ADFP (34), Tip47 (12), and, recently, LSDP-5 (13) protect LD TAG against lipolysis.

Endogenous lipases in liver cells responsible for LD turnover have not yet been catalogued to the same extent as in adipose tissue (26). The phenotype of ATGL-null mice indicates that it may be an important lipase not only in adipose tissue but also in nonadipose tissue. Notably, hepatic fat content was doubled in ATGL-null mice compared with wild-type (14). Interestingly, by imunocytochemistry and immunoblot analysis, we found an ATGL increase at the LD surface when all PAT proteins were absent, confirming a recent report (44) that the lack of ADFP increases ATGL presence in fibroblasts. Furthermore, by downregulating ATGL expression, we could show an active role of ATGL in LD turnover, as judged by increased LDs in cells lacking this lipase. These results led us to hypothesize that ATGL is one of the enzymes participating in LD hydrolysis in liver cells and, importantly, that nonadipose tissue PAT proteins also have a role in inhibiting LD hydrolysis.

A recent finding on ATGL regulation is that this enzyme appears to require CGI-58 for full activity in cytosolic extracts (27). CGI-58 is a protein that when mutated or truncated has been found to be responsible for LD accumulation in most tissues (Chanarin-Dorfman syndrome) (28,29). Most recently, CGI-58 was identified to bind to Perilipin and ADFP (30,31). In our studies, fat cake extract from cells lacking PAT proteins was enriched in CGI-58 protein. This led us to hypothesize that common mechanisms exist among PAT proteins to regulate ATGL access to the LD surface, both in adipose and nonadipose tissues.

We found that PAT protein downregulation is sufficient to induce insulin resistance in liver cells. These results led us to conclude that the mechanism underlying the defect in insulin resistance in the absence of PAT proteins is due at least in part to an increase in lipolytic rate and/or uncoupling to NEFA utilization. Absence of PAT proteins at the LD surface facilitates access of endogenous triglyceride lipases such as ATGL to the TAG substrate, releasing NEFAs that in turn affect the signaling cascade at multiple levels. These findings support our working hypothesis that ADFP and Tip47 play important roles in the regulation of ectopic fat, similar to the key role of perilipin in adipose tissue, and that their function can explain at least partially the connection between ectopic fat and development of insulin resistance. If true, then alteration in the composition and/or activity of PAT proteins will result in dysfunctional lipolysis triggering a cascade of deleterious events in cellular function. Hence, a strong connection should be visible in nonadipogenic tissues between PAT protein composition, surface protein content, lipolysis, and insulin signaling. However, an in vivo study (45) of ADFP antisense downregulation in liver reported improved insulin sensitivity; no change in VLDL secretion, contrary to cell culture study results (47); and no compensation by Tip47, at least in the whole cellular extract. However, no attempt was reported to assess LPSD-5 presence/activity, while its expression has been shown to correlate with increased β-oxidation and is present in liver (46). Knowing the redundancy with the PAT protein family, LSDP-5 may be responsible for the apparent discrepancy existing between these in vivo results and cell culture (47). The presence of LSDP-5 in liver LDs brings another level of complexity to the regulation of LD utilization. To date, little knowledge exists of LD PAT protein composition in insulin resistance models. We show here that liver LDs from the DIO disease model differ in the ratio of PAT proteins at the LD surface and, importantly, that ATGL is increased at the LD surface. Since ATGL-null mice have increased insulin sensitivity, we hypothesize that defects increasing ATGL access to the lipid surface and/or activity will have the opposite effect (i.e., promote insulin resistance). Thus, we hypothesize that DIO alters liver lipid turnover, contributing to insulin resistance.

It is widely accepted that excessive free fatty acid–induced insulin resistance involves intramyocellular and intrahepatocellular accumulation of TAG, activation of several serine/threonine kinases, reduction in tyrosine phosphorylation of IRS-1/2, and impairment of IRS/phosphoinositide 3-kinase pathway (32) and ERK1/2 pathway (32). Attention has been given in the literature thus far to a failure of mitochondrial oxidative function, but little thought has been given to the failure of the LD compartment to appropriately store the excess NEFAs (34). The results from these studies indicate that defects in LD lipolysis/utilization contribute to loss of ability of nonadipose tissue to maintain NEFA homeostasis and thus are responsible for deleterious consequences observed in the signaling pathway. These defects likely occur in the function/activity of the PAT family LD surface proteins regulating lipolysis. Future research focusing on LD utilization and its regulation will provide us with important clues for understanding regulation of cellular energy homeostasis.

This work was supported by a career development award (1-05-CD-17) from the American Diabetes Association (to C.S.); a grant from the National Institutes of Health (1RO1 DK 075017-01A2 to C.S.); the Geriatric Research, Education and Clinical Center, Baltimore Veterans Affairs Health Care Center; a grant from the Clinical Nutrition Research Unit of Maryland (DK072488 to S.K.F.); and Intramural Research Programs of the National Institute of Diabetes and Digestive and Kidney Diseases and the National Center for Complementary and Alternative Medicine, National Institutes of Health.

We thank Dr. M.C. Woodle for his constant help and support and Dr. A.K. Kimmel for his helpful criticisms.