Perilipin 5 Deletion in Hepatocytes Remodels Lipid Metabolism and Causes Hepatic Insulin Resistance in Mice

The liver maintains normal lipid homeostasis by regulating a number of processes, including de novo lipogenesis; fatty uptake, oxidation, and esterification; and packaging of triglycerides within lipoproteins for secretion (1). Dysregulation of one or more of these processes can lead to hepatic steatosis, which is excessive accumulation of intracellular lipids and is the defining component of nonalcoholic fatty liver disease (NAFLD). NAFLD is of clinical importance because it is closely linked to metabolic abnormalities such as insulin resistance and dyslipidemia (2) and is a risk factor for life-threatening complications, including steatohepatitis, cirrhosis, and hepatocellular carcinoma (3).

The excess lipids stored in hepatic steatosis are predominately triglycerides, which are compartmentalized within intracellular lipid droplets. The storage of triglycerides within lipid droplets is postulated to protect tissues from free fatty acid (FFA)–induced “lipotoxicity.” Dysregulated lipid droplet metabolism is associated with the development of a range of cellular defects, including activation of stress signaling pathways, insulin resistance, lipoapoptosis, and organ failure (4). Lipids are mobilized from lipid droplets via the process of lipolysis, which is primarily regulated by adipose triglyceride lipase (ATGL) and hormone-sensitive lipase (HSL) (5). Proteomic approaches have identified many other lipid droplet–associated proteins (6) and prominent among these are the perilipin (PLIN) family of proteins, which induce tissue-specific metabolic effects, not least through their interactions with ATGL and HSL (7–9).

PLIN5 is actively involved in regulating lipid metabolism in tissues with high oxidative capacity, such as heart, skeletal muscle, and brown adipose tissue (7,10–13). Studies using cell lines and transgenic mice with Plin5 deficiency or overexpression show that PLIN5 interacts with ATGL and comparative gene identification (CGI)-58 to suppress lipolysis (14–16). This limits the production of lipid intermediates and cellular stress signals and, in turn, prevents disruption to tissue morphology and functions, including heart contraction, skeletal muscle insulin action, and β-cell insulin secretion (17). PLIN5 also promotes transcription of genes that mediate mitochondrial biogenesis and oxidative function in brown adipose tissue, thereby coupling protein kinase A–mediated lipolysis to transcriptional regulation of mitochondrial fatty acid metabolism (13).

Studies of PLIN5 functions in the liver have yielded equivocal results. PLIN5 levels are increased in murine models of obesity and hepatosteatosis (16,18), and adenovirus-mediated overexpression of PLIN5 worsened hepatosteatosis in high-fat diet–fed mice, but this did not cause liver injury or adversely affect systemic glucose tolerance or insulin sensitivity (16,18). Two studies in whole-body Plin5^−/− mice report divergent results. One group reported that Plin5 deletion reduces hepatosteatosis but causes hepatic inflammation, endoplasmic liver damage, and lipid peroxidation, without impacting glucose metabolism (16), whereas another reported no evidence of steatosis or liver damage and improved hepatic insulin sensitivity in Plin5^−/− mice (7).

In light of this uncertainty, we aimed to determine the impact of Plin5 deletion on hepatic lipid metabolism and insulin action and whole-body glucose homeostasis. Herein, we show that mice with liver-specific Plin5 deficiency (Plin5^LKO) have alterations in hepatic lipid metabolism that result in mild increases in hepatic triglycerides without significant liver damage. These changes were associated with activation of c-Jun N-terminal kinase (JNK), suppression of insulin signal transduction in the liver, and the development of hepatic insulin resistance and whole-body glucose intolerance in the context of diet-induced obesity.

Studies were approved by the Monash University School of Biomedical Science Animal Ethics Committee (MARP-2013-050) and conformed to the National Health and Medical Research Council of Australia guidelines regarding care and use of experimental animals. Plin5^lox/+ mice were generated as described previously (19). Mice with liver-specific deletion of Plin5 were generated by crossing these mice with mice expressing Cre recombinase under the control of the liver-specific murine promoter (Alb-Cre; provided by T. Tiganis, Monash University) to create mice with liver-specific ablation of Plin5 (Plin5^lox.lox.Cre⁺, denoted herein as Plin5^LKO mice) and control wild-type (Wt) littermates (lox/lox, i.e., Plin5^lox.lox.Cre^– mice). In separate experiments, an adeno-associated virus serotype 8 (AAV) driven by the albumin promotor and containing Plin5 cDNA or GFP (1 × 10¹² genome copies/mouse; Vector Biolabs, Malvern, PA) was injected via the tail vein in 20-week-old high-fat–diet fed Plin5^LKO. Experiments were conducted in mice 8 weeks later. Male mice were housed at 22°C on a 12:12-h light-dark cycle. Wt and Plin5^LKO mice aged 8 weeks were fed either a regular low-fat chow (4.6% energy from fat) or high-fat diet (43% energy from fat, SF04-001; Specialty Feeds, Western Australia, Australia) for 12 weeks prior to metabolic assessments. Mice were fasted from 0700 h to 1100 h before all experiments, unless stated otherwise.

Body composition was measured using DEXA (Lunar Pixi; PIXImus, Madison, WI) (20). Oxygen consumption, carbon dioxide production, physical activity, and food intake were measured using an Oxymax indirect calorimetery system (Columbus Instruments, Columbus, OH).

Mice fasted for 4 hours received an oral gavage of d-glucose (2 g/kg) or intraperitoneal injection of insulin (1 unit/kg, Actrapid) for glucose and insulin tolerance tests, respectively. Blood obtained from a tail nick was assessed for blood glucose (Accu-Chek) as indicated.

Clamps were conducted in conscious and restrained mice (4 mU/kg/min insulin) as described previously (7,21).

Hepatocytes were isolated from Wt and Plin5^LKO livers by collagenase digestion (22). Metabolic assessment of hepatocytes was previously described (23). Hepatocytes were incubated for 2 h with low-glucose DMEM GlutaMAX (Life Technologies) containing 2% fatty acid–free BSA, 2 μCi [1-¹⁴C]palmitate (CFA23; GE Healthcare), and 0.5 mmol/L palmitate (Sigma-Aldrich). Fatty acid oxidation was calculated as the sum of complete oxidation to ¹⁴CO₂ and “incomplete” oxidation (i.e., acid-soluble metabolites). Fatty acid uptake was calculated as the sum of fatty acid oxidation and fatty acids stored in complex lipids (i.e., ¹⁴C in the organic fraction of the lysed cells). Lipogenesis was measured for 2 h in DMEM containing 2 μCi/mL d-[2-³H]glucose (NET238C001MC; PerkinElmer). Triglyceride secretion was assessed in medium (DMEM) after 8 h incubation and was determined using a colorimetric assay kit (Triglycerides GPO-PAP; Roche Diagnostics, Indianapolis, IN). All data are expressed per milligram protein.

Hepatocytes grown in M199 medium were incubated with MitoTracker (Molecular Probes) diluted (2 μmol) in prewarmed media for 15 min. Cells were washed with PBS. BODIPY (2 μg/mL) (Thermo Fisher Scientific, San Jose, CA) was applied to cells and incubated for 10 min. Cells were washed with PBS and viewed using time-lapse confocal microscopy using a TCS SP8 confocal microscope (Leica). To quantify the distance between lipid droplets and mitochondria over several time points, a ×40 lens was calibrated at 1 pixel = 0.01679 µm, i.e., 100 pixels = 1.1679 µm. A Euclidean pixel-distance map was derived with any MitoTracker signal above threshold set as distance zero. The mass centers of lipid droplets were localized by their position on the Euclidean distance map, across three time points, to determine proximity time away from the mitochondria. For example, if the lipid droplet touches a mitochondrion, the lipid droplet is zero pixels away from a mitochondrion; thus, for that time point, that lipid droplet will have a measurement of zero. Additionally, if the lipid droplet sits 10 pixels away from the nearest mitochondria, then that droplet gets assigned a value of 10 for that time point. Therefore, if a droplet spends three time points touching a mitochondria, two time points being two pixels away, and one time point being five pixels away, the lipid droplet will receive a value of (3 × 0) + (2 × 2) + (1 × 5) = 9 units of proximity time.

Lipids were extracted as previously described (24). Triglycerides were determined by colorimetric assay (Triglycerides GPO-PAP; Roche Diagnostics). Liver cytosolic and membrane fractions were isolated using a well-accepted ultracentrifugation method (25), and lipids were extracted using a modified Folch extraction (26). Samples were spiked with internal standards and were analyzed using an Orbitrap Fusion Lumos mass spectrometer coupled to a Vanquish UHPLC (Thermo Fisher Scientific). For both analyses at positive and negative ionization mode, each sample was injected into an Accucore C30 column (2.1 × 250 mm, 2.6 µm) (Thermo Fisher Scientific). For analysis at both polarities, top-speed data-dependent acquisition was performed with a cycle time of 1 s. Diglyceride and ceramide species content was determined by comparing ratios of unknowns with internal standards and referencing to a standard curve. Plasma FFA and β-hydroxybutyrate were measured using a colorimetric kit (Wako Diagnostics, Osaka, Japan, and Cayman Chemical, Ann Arbor, MI, respectively). Thiobarbituric acid–reactive substance (TBARS) and lipid hydroperoxides were determined as previously described (27,28). AST and alanine aminotransferase (ALT) activities were determined by enzymatic colorimetric assays (29,30). Triglyceride secretion rates were assessed by injecting 300 mg/kg Triton WR 1339 (Tyloxapol, T8761; Sigma-Aldrich) into the tail vein of conscious mice, and plasma triglyceride was determined over 2 h.

Livers were fixed in 10% neutral buffered formalin and paraffin embedded, and immunohistochemistry was performed as described previously (7). Immunofluorescence images were obtained using a TCS SP8 confocal microscope (Leica).

RNA was extracted from tissues with QIAzol lysis reagent (Qiagen, Doncaster, Victoria, Australia) and reverse transcribed with iScript Reverse Transcriptase (Bio-Rad Laboratories, Invitrogen, Mt. Waverley, Victoria, Australia), and the gene products were determined by real-time quantitative RT-PCR (ep realplex4 Mastercycler; Eppendorf, Hamburg, Germany) using SYBR Green PCR Master Mix (Brilliant II SYBR Green QPCR Master Mix; Agilent Technologies, Santa Clara, CA) for Plin5 and TaqMan Universal PCR Master Mix (AmpliTaq Gold DNA Polymerase; Applied Biosystems, Scoresby, Victoria, Australia) for other genes. 18S was used as a reference gene and did not vary between groups. The mRNA levels were determined by the ΔΔCT method. Primer sequences are listed in Supplementary Table 1.

Liver lysates were prepared in 100 mmol/L TRIS-HCl (pH 7.4; for PLIN5) or RIPA buffer, proteins were resolved by SDS-PAGE electrophoresis, and immunoblot analysis was conducted as described previously (19). Antibodies are listed in Supplementary Table 2. Stain-free images were collected after transfer for loading control (ChemiDoc MP and ImageLab software version 4.1; Bio-Rad Laboratories, New South Wales, Australia).

Statistical analysis was performed using unpaired two-way Student t tests or two-way ANOVA with Bonferroni post hoc analysis where required. Statistical significance was established at P < 0.05. Data are reported as means ± SEM.

Liver Plin5 mRNA expression was reduced by 85% in Plin5^LKO compared with Wt mice (Fig. 1A). The absence of PLIN5 protein in liver was confirmed by immunoblot analysis (Fig. 1B) and immunohistochemistry (Fig. 1C). Specificity of liver Plin5 knockdown was confirmed by qPCR, showing expression similar to that in Wt mice in skeletal muscle, heart, kidney, and white adipose tissue (Fig. 1D). There were no changes in Plin2, Plin3, or Plin4 expression in the liver of Plin5^LKO mice or changes in the mRNA expression of prominent lipases including Pnpla2 and Lipe compared with Wt mice (Fig. 1E). PLIN4 was not detectable by immunoblot, whereas PLIN2 and PLIN3 protein content was not different between Plin5^LKO and Wt mice (PLIN2 1.00 ± 0.06 vs. 1.13 ± 0.12 and PLIN3 1.00 ± 0.13 vs. 1.17 ± 0.18 arbitrary units for Wt and Plin5^LKO, respectively; n = 4 per group) (Fig. 1F).

We first investigated the metabolic effects of Plin5 deletion in primary hepatocytes (Fig. 2A). Although triglyceride content was not different between genotypes (Fig. 2B), detailed analysis revealed significant remodeling of fatty acid metabolism in Plin5^LKO hepatocytes. Fatty acid uptake was reduced (Fig. 2B) and was accompanied by a reduction in fatty acid oxidation (Fig. 2C) and fatty acid esterification into triglycerides (Fig. 2D) and diglycerides (Fig. 2E). These changes were not accompanied by significant changes in genes encoding key proteins of lipid metabolism or mitochondrial capacity (Fig. 2F). PLIN5 promotes a close interaction between lipid droplets and mitochondria in cardiomyocytes, which may enhance fatty acid oxidation (31). Live cell imaging revealed a significant reduction in the interaction and proximity distance between lipid droplets and mitochondria in Plin5^LKO hepatocytes, indicating reduced contact between these organelles (Fig. 2G and H). De novo lipogenesis was not significantly impacted in Plin5^LKO hepatocytes (Fig. 2I), whereas triglyceride secretion from Plin5^LKO hepatocytes was markedly reduced compared with Wt hepatocytes (Fig. 2J).

We next investigated the metabolic phenotype of Plin5^LKO and Wt mice fed a standard chow diet. Liver histology showed no evidence of increased lipid deposition or fibrosis (Fig. 3A). Consistent with this notion, liver triglyceride (Fig. 3B), diglyceride (Fig. 3C), and ceramide (Fig. 3D) levels were not different between Plin5^LKO and Wt mice. Fasting plasma FFAs (Fig. 3E), β-hydroxybutyrate (Fig. 3F), and triglyceride levels (Fig. 3G), triglyceride secretion (Fig. 3H), and microsomal triglyceride transfer protein (MTTP) content (data not shown) were not different between genotypes, which contrasted with the reduction in triglyceride secretion measured from cultured hepatocytes of Plin5^LKO mice (Fig. 2K). Plasma ALT is a clinical measure of liver damage and was not different between genotypes (Fig. 3I), nor were the liver contents of dinitrophenylhydrazine, a marker of protein carbonylation and oxidative damage (Fig. 3J). The malondialdehyde content was decreased in the livers of Plin5^LKO mice (Fig. 3K), indicating a reduction in lipid peroxidation and oxidative stress.

Wt and Plin5^LKO mice had similar body weights from 5 to 20 weeks of age (Fig. 4A). In addition, there was no difference in lean or fat mass (Fig. 4B) or liver mass (Fig. 4C) between genotypes. Whereas food intake was similar (Fig. 4D), energy expenditure was modestly decreased in Plin5^LKO mice (Fig. 4E) and was reflective of decreased activity (Fig. 4F). Despite the reduction in hepatocyte fatty acid oxidation (Fig. 2B), there were no differences in the respiratory exchange ratio between genotypes (Fig. 4G), indicating similar whole-body fat oxidation. Whole-body glycemic control (Fig. 4H) and insulin sensitivity (Fig. 4I) were impaired in Plin5^LKO compared with Wt mice. Fasting blood glucose and plasma insulin levels were increased in the Plin5^LKO mice (Fig. 4J and K), which is consistent with the presence of insulin resistance.

PLIN5 is upregulated in hepatocytes with dietary lipid overload (16,32,33), presumably to facilitate the transfer of excess fatty acids into lipid droplets to limit lipotoxic stress (7,13). Hence, we surmised that the metabolic consequences of Plin5 ablation would be more prominent in mice fed a high-fat diet. Plin5^LKO mice were marginally heavier than Wt mice (Fig. 5A). There was no significant difference in lean or fat mass (Fig. 5B) or food intake (Fig. 5C) between genotypes, whereas energy expenditure was lower in Plin5^LKO mice (Fig. 5D). There was no difference between genotypes for fasting plasma triglycerides (Wt 1.01 ± 0.05 vs. Plin5^LKO 1.04 ± 0.08 mmol/L), FFA (Wt 0.90 ± 0.06 vs. Plin5^LKO 0.91 ± 0.07 mmol/L), β-hydroxybutyrate (Wt 415 ± 80 vs. Plin5^LKO 468 ± 54 μmol/L), or cholesterol (Wt 67.4 ± 5.4 vs. Plin5^LKO 72.9 ± 7.8 mg/dL).

Liver mass was increased in Plin5^LKO compared with Wt mice (Fig. 5E) and coincided with a 70% increase in triglyceride content (Fig. 5F). Analysis of hematoxylin-eosin (H-E)–stained liver sections (Fig. 5G) confirmed the presence of fatty liver, with higher liver steatosis scores in the livers of Plin5^LKO compared with Wt mice (data not shown). Lobular inflammation and hepatocellular ballooning were not different between genotypes (data not shown). Fibrosis was assessed by Masson’s trichrome stain and was unaffected by Plin5 deletion (Fig. 5G). Plasma ALT was increased in Plin5^LKO compared with Wt mice, suggesting mild damage in the livers of these mice (Fig. 5I). In contrast, liver oxidative stress, assessed as TBARS (Fig. 5I) and protein carbonylation (Fig. 5J), was not different between genotypes.

Hepatic lipid accumulation is commonly associated with the development of insulin resistance. Consistent with this notion, fasting plasma insulin, but not blood glucose, levels were increased in Plin5^LKO compared with Wt mice (Fig. 6A and B). Plin5^LKO mice had impaired glucose tolerance (Fig. 6C), whereas plasma insulin levels during the oral glucose tolerance test (GTT) were not different between genotypes (Fig. 6D). In addition, Plin5^LKO mice exhibited impaired insulin sensitivity as assessed by an insulin tolerance test (Fig. 6E). Hyperinsulinemic-euglycemic clamps were performed to assess whole-body and tissue-specific insulin action. Glucose was clamped at euglycemia (Fig. 6F), and a steady-state glucose infusion rate was achieved in both genotypes (Fig. 6G). Plasma insulin levels were higher during the clamp (main treatment effect) but were not different between Plin5^LKO and Wt mice (Fig. 6H). Whole-body insulin action was impaired in Plin5^LKO mice, as reflected by a 50% reduction in the glucose infusion rate (Fig. 6I). The insulin-stimulated glucose disposal rate was not different between genotypes (Fig. 6J), indicating no differences in peripheral glucose uptake. Rather, Plin5^LKO mice displayed marked hepatic insulin resistance compared with Wt mice, as evidenced by a failure to suppress endogenous glucose production (Fig. 6K and L) and reduced Akt Ser⁴⁷³ phosphorylation after insulin administration (Fig. 6M and N).

We performed biochemical and molecular analyses in the livers of high-fat diet–fed Plin5^LKO and Wt mice to determine the link between Plin5 ablation and hepatic insulin resistance. We assessed diglyceride and ceramide species in membrane or cytosol fractions of liver. Ceramides were not different in either fraction between genotypes (data not shown). No diglyceride species were different in the membrane fractions of Plin5^LKO and Wt livers. In contrast, 16 diglyceride species were increased in the cytosolic fraction of Plin5^LKO compared with Wt livers (Fig. 7A). Although cytosolic diglycerides have been associated with hepatic PKCε activation, as reflected by PKCε translocation to the plasma membrane, we observed no differences in membrane PKCε between genotypes, suggesting that this mechanism is unlikely to explain the impaired insulin signaling in Plin5^LKO livers (Fig. 7B). There was no evidence for increased endoplasmic reticulum (ER) stress in Plin5^LKO mice, as indicated by immunoblot analysis of phosphorylation in IRE1 Ser⁷²⁴ and eIF2α Ser⁵¹, as well as changes in total protein content of XBP1, CHOP, and ATF6 (Fig. 7C). Similarly, there was no difference in mRNA expression of the ER stress genes Atf4, Chop, Gadd34, Grp94, and Xbp1 (Fig.7D). There were no differences in the autophagy markers LC3B-II and p62 (Fig. 7E) or markers of inflammation, including IκBα content (Fig. 7F) and Il-6, Tnfa, and Adgre1 mRNA expression (Fig. 7G). We have previously shown that Plin5 deletion in muscle reduces FGF-21 (34); however, neither FGF-21 mRNA expression nor circulating levels were altered in Plin5^LKO compared with Wt mice (Fig. 7H and I). Of interest, livers of Plin5^LKO mice showed substantial phosphorylation (activation) of the serine/threonine kinase JNK (Fig. 7J and K). JNK activation was supported by the finding of increased gene expression and protein content of c-Jun, which is regulated by JNK (Fig. 7J–L). A JNK target protein is insulin receptor substrate 1 (IRS-1), which once phosphorylated at Ser³⁰⁷, inhibits its activity, and contributes to hepatic insulin resistance (35). JNK phosphorylation was increased by 50% in the livers of Plin5^LKO compared with Wt mice, and this coincided with an increase in IRS-1 Ser³⁰⁷ phosphorylation (Fig. 7M and N). JNK phosphorylation was also increased in the livers of chow-fed Plin5^LKO compared with Wt mice (Wt 1.00 ± 0.10 vs. Plin5^LKO 1.52 ± 0.17 arbitrary units, respectively; n = 8 per genotype, P < 0.05), demonstrating conservation of this response.

Administration of an AAV containing Plin5 restored PLIN5 protein content to Wt levels (Fig. 8A). This was associated with no change in body mass (Fig. 8B) and restoration of liver triglyceride content to levels of Wt littermate control mice (Fig. 8C). Whole-body glucose tolerance (Fig. 8D) and insulin action (Fig. 8E), which were impaired in Plin5^LKO mice, were not different between Wt mice and Plin5^LKO mice with PLIN5 re-expression. These changes in glycemic control and insulin action were associated with a dampening of JNK phosphorylation in Plin5^LKO mice to levels measured in Wt mice (Fig. 8F). As expected, re-expression of GFP alone did not rescue the impaired glycemic control, insulin action, or defective JNK signaling in Plin5^LKO mice.

Hepatic insulin resistance is often associated with dysregulated lipid metabolism (36), which is partly driven by the expression levels, activity, and interactions of proteins located at the surface of lipid droplets. PLIN5 appears to be required for the adaptation to lipid overload, as its expression is increased upon exposure to fatty acids in cultured cells (33,37), with high-fat feeding (18) and prolonged fasting (37) in mice, and in humans with NAFLD (16). Here, we have shown that Plin5^LKO mice exhibit systemic glucose intolerance and insulin resistance when fed a standard chow diet, and that these effects were exacerbated upon high-fat feeding and were associated with substantial triglyceride accumulation, activation of JNK, inhibition of hepatic insulin signaling, and impaired insulin-mediated suppression of hepatic glucose output. Altogether, these studies define important roles for PLIN5 in hepatic lipid metabolism and insulin action without the pleiotropic effects that occur in global knockout mice (7,11,16).

Despite several physiological and molecular studies, the role of PLIN5 in regulating hepatic lipid metabolism remained unresolved. To better understand the role of PLIN5 in hepatic lipid metabolism, we developed hepatocyte-specific Plin5-null mice and examined PLIN5’s cell-autonomous role in hepatocytes. Plin5 deletion impacted lipid homeostasis by reducing uptake, storage, and oxidation of extracellular-derived fatty acids and reduced triglyceride secretion. These findings contrast earlier work by Wang et al. (16) that reported increased fatty acid oxidation, decreased fatty acid storage, and increased triglyceride secretion from hepatocytes isolated from whole-body Plin5^−/− mice. Although the molecular events mediating the changes in lipid metabolism remain unresolved, the absence of marked changes in gene expression of proteins controlling these key metabolic processes indicates that posttranscriptional regulation may be important, and in particular, protein interactions. In this regard, PLIN5 regulates lipolysis in nonhepatic cells by physical interactions with ATGL, HSL, and CGI-58 (8,9) and appears to modulate lipid metabolism by facilitating the interaction between lipid droplets and mitochondria (31). Studies aimed at identifying PLIN5 protein-protein interactions are ongoing and may explain the unique reciprocal roles of PLIN5 in hepatic lipid metabolism described here. Alternatively, the decrease in fatty acid oxidation and storage could be explained most simply by reduced uptake of extracellular fatty acids, as fatty acid flux per se can impact distal processes and was previously reported in hearts of Plin5^−/− mice (38).

VLDL assembly of triglycerides occurs via a process involving lipolysis/re-esterification and is likely to involve triglyceride hydrolase (39) rather than the PLIN5-interacting lipases ATGL or HSL (23). Another novel observation observed here was the decreased triglyceride secretion from Plin5^LKO hepatocytes, which contrasts with a previous report showing that triglyceride secretion is increased in cultured Plin5-deficient hepatocytes (16). Although triglyceride secretion was markedly impaired in cultured hepatocytes, this was not associated with altered expression of key regulators of this process, including triglyceride hydrolase or MTTP (data not shown). Moreover, our in vivo analysis demonstrated no effect of Plin5 deletion on triglyceride secretion or plasma triglyceride levels, the latter aligning with previous studies reporting normal plasma triglyceride levels in three independent Plin5^−/− mouse colonies (7,11,16). Taken together, these data indicate that PLIN5 appears to regulate triglyceride secretion from hepatocytes but that humoral or neural signals overcome this cell autonomous regulation, resulting in normal triglyceride secretion in vivo.

NAFLD and type 2 diabetes are common comorbidities (40), and hepatic steatosis often develops before insulin resistance, suggesting a causative role of hepatic lipid accumulation in the pathogenesis of hepatic insulin resistance (1). We found that liver-specific Plin5 ablation reduced hepatic insulin action, without affecting insulin sensitivity of other tissues, and was sufficient to induce systemic insulin resistance and impair glucose tolerance. Further, reintroducing PLIN5 into the livers of Plin5^LKO mice was sufficient to restore whole-body glycemic control and insulin action. We systematically evaluated the most common mediators of lipid-induced insulin resistance and observed no significant effect of hepatic Plin5 deletion on the accumulation of ceramides, oxidative stress, ER stress, autophagy, or inflammation. Instead, we report activation of JNK and suppression of insulin signal transduction, which is a plausible explanation for the impaired hepatic insulin sensitivity (41,42). By precedence, Plin5 deletion activates JNK in aortic valve tissues of ApoE^−/− mice (43). Others have shown that adenoviral overexpression of PLIN5 in the liver prevented the development of late-onset whole-body insulin resistance in high-fat diet–fed mice but that this was insufficient to impact whole-body glycemic control (18). In contrast, we have previously reported that whole-body Plin5^−/− mice have improved whole-body insulin sensitivity, which was mediated by enhanced liver insulin sensitivity but was accompanied by paradoxical reductions in muscle insulin sensitivity (7). These discordant results between mice with whole-body (7) and liver-specific Plin5 deletion could be explained by altered intertissue communication. Supporting this premise, mice with muscle-specific Plin5 deficiency have impaired activation of the unfolded protein response, which was associated with reduced fibroblast growth factor 21 (FGF-21) expression and secretion and FGF-21–dependent changes in systemic glycemic control (34). Previous work also showed that Plin5 overexpression in muscle increases FGF-21 production and improves glycemic control (44), but this was not the case with Plin5 deletion in the liver. Evidently, potential changes in intertissue cross-talk could extend to hormones other than FGF-21, and this will require further investigation. In addition, our results and others (45) support the notion that some PLIN5 functions are highly conserved across cell types/tissues (e.g., maintenance of intracellular triglyceride levels) but that other tissue-specific actions of PLIN5 impact local insulin action through distinct mechanisms (e.g., altered production of signaling lipids, including ceramides in skeletal muscle but not in liver [oxidative stress in the heart but not liver]).

Previous reports indicate that PLIN5 facilitates physical contact between lipid droplets and mitochondria within the heart (7,31), which is proposed to facilitate tighter coupling of triglyceride lipolysis to fatty acid oxidation. In keeping with this, live cell imaging studies demonstrated reduced lipid droplet–mitochondria contact in hepatocytes derived from Plin5^LKO mice compared with Wt mice, and this was associated with reduced fatty acid oxidation in Plin5^LKO hepatocytes, independent of changes in mitochondrial protein abundance. However, immunofluorescent and electron microscopy approaches have not confirmed this relationship in skeletal muscle (7), raising the distinct possibility of tissue specificity in the interacting/adaptor proteins that facilitate this interorganelle contact. Future studies are clearly required to elucidate the proteins that facilitate the interaction between lipid droplets and mitochondria under physiologically relevant conditions and determine whether the functional interplay between these organelles is an important determinant of metabolic efficiency, intracellular signaling, and insulin action.

In sum, this study affirms the importance of lipid droplet proteins in regulating cell metabolism and expands our knowledge of PLIN5 biology by demonstrating multiple effects on hepatic lipid metabolism and the requirement of PLIN5 for normal insulin action in the liver, especially in mice fed a high-fat diet. Since PLIN5 expression is increased in response to lipid oversupply in mice (18) and in steatotic livers of humans (16), and overexpression of PLIN5 improves insulin action (18), our data suggest that PLIN5 is part of an adaptive response to protect against insulin resistance and type 2 diabetes.

Acknowledgments. The authors thank Michelle Kett, Robert Lee, Camden Lo, and Maria Matzaris (Monash University) for technical assistance. The authors thank the Melbourne Mass Spectrometry and Proteomics Facility of the Bio21 Molecular Science & Biotechnology Institute at The University of Melbourne for the support of mass spectrometry analysis.

Funding. This work was funded by the National Health and Medical Research Council of Australia (NHMRC) (1047138). S.N.K. was supported by a Biomedicine Discovery Scholarship from Monash University, and M.J.W. (APP1077703) and M.K.M. (APP1143224) were supported by research fellowships from the NHMRC.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. S.N.K. and M.J.W. planned and conducted the experiments, analyzed the data, and wrote the manuscript. R.C.M., J.C.Y.L., and M.K.M. planned and conducted the experiments and analyzed the data. A.R. and S.N. planned and conducted experiments. All authors edited the manuscript. M.J.W. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.