Perilipin 5 Deletion Unmasks an Endoplasmic Reticulum Stress–Fibroblast Growth Factor 21 Axis in Skeletal Muscle

Lipid droplets (LDs) are intracellular organelles comprising a core of neutral lipids, including triglycerides and sterols, surrounded by a phospholipid monolayer and proteins that concentrate at the LD surface. LDs are highly conserved across species and in humans are stored in virtually every cell type, where they play critical roles in a variety of cellular processes, including the regulation of lipid metabolism (1).

LDs reside at the intersection of lipid catabolism and lipid storage, and proteins surrounding the LD coordinate the release of triglyceride-derived fatty acids that are used for mitochondrial β-oxidation and energy production, for generating cellular membranes and signaling lipids, and as activators of nuclear receptors that impact transcriptional regulation of metabolic pathways (2,3). Other LD-associated proteins are involved in sequestering fatty acids into LDs during times of energy surplus to protect the cell from accumulating excessive lipid intermediates (e.g., ceramides and diacylglycerol) and the associated cellular dysfunctions that are broadly referred to as “lipotoxic” stress (4). The importance of LD proteins is highlighted in a variety of human diseases that result from deletion or loss-of-function mutations of LD proteins, including neutral lipid storage disease with myopathy (adipose triglyceride lipase [ATGL]) (5), Chanarin-Dorfman syndrome (comparative gene identification 58 [CGI-58]) (6), and partial lipodystrophy, severe dyslipidemia, and insulin-resistant diabetes (perilipin 1 [PLIN1], seipin, and hormone-sensitive lipase [HSL]) (7).

Since the discovery of PLIN1, which is an abundant LD protein and is critical for the regulation of adipocyte metabolism, intense research efforts have attempted to identify the LD proteome and the proteins’ functions in many cell types. Prominent among this list is PLIN5, which is abundantly expressed in tissues with high oxidative capacity, such as the heart, liver, and skeletal muscle (8), and plays an important role in LD dynamics by modulating the flux of fatty acids from LDs to match the energy requirements of the cell (9). PLIN5 achieves this by interacting with and coordinating the actions of the key lipolytic proteins ATGL, CGI-58, and HSL (10,11). Moreover, PLIN5 appears to be important in the adaptation to nutrient and cellular stress, with PLIN5 expression increasing after high-fat feeding and endurance exercise training (9). Few studies have focused on understanding how PLIN5 impacts cell signaling pathways to modulate this adaptation to nutrient stress. A recent study showed that skeletal muscle PLIN5 has the capacity to stimulate fibroblast growth factor 21 (FGF21) expression in muscle and subsequently increase circulating FGF21, suggesting that muscle PLIN5 contributes to systemic metabolic control (12). The increase in muscle FGF21 was accompanied by endoplasmic reticulum (ER) stress, particularly Atf4 within the PERK/eIF2α pathway, which is a potent transcriptional activator of FGF21 (13). This is of special interest as LDs originate from and reside in close proximity to the ER and have been implicated in ER stress and activation of the unfolded protein response (UPR) (14).

Skeletal muscle plays a critical role in maintaining whole-body energy homeostasis because it is an important site for lipid disposal and oxidation and is the major tissue responsible for insulin-stimulated glucose uptake. Muscle insulin resistance is a pathological feature of type 2 diabetes, and LDs figure prominently in pathologies of insulin resistance, including obesity and type 2 diabetes (15). In this study, we have generated mice with muscle-specific deletion of Plin5 to determine the role of PLIN5 in regulating skeletal muscle metabolism, cellular stress, and FGF21 expression/secretion during dietary lipid overload and, in turn, to understand the impact of skeletal muscle PLIN5 on glycemic control.

Experimental procedures were approved by the Monash University Animal Ethics Committee (MARP/2013/050) and conformed to the National Health and Medical Research Council of Australia guidelines regarding the care and use of experimental animals. Generation of muscle-specific Plin5 knockout mice (Plin5^MKO) on a C57BL/6 background has been previously described (16). Male mice were housed at 22°C on a 12:12-h light-dark cycle and were fed either a rodent chow (5% energy from fat) (Specialty Feeds, Glen Forrest, Australia) or a high-fat, high-sucrose diet (HFD; 43% energy from fat) (High Fat Rodent Diet SF04-001; Specialty Feeds) starting at 8 weeks of age for a total of 8 weeks. Body weight was measured weekly. Mice were fasted from 0700 to 1100 h before all experiments unless otherwise stated.

Fat and lean body mass were measured using DEXA (Lunar Pixi; PIXImus, Fitchburg, WI) as previously described (17). Oxygen consumption, carbon dioxide production, physical activity, and food intake were measured using a 12-chamber Oxymax indirect calorimetry system (Columbus Instruments, Columbus, OH). Studies were commenced after 8 h of acclimation to the metabolic chamber. Expired gases were assessed at 30-min intervals for 48 h at 22°C.

Mice were gavaged orally with glucose (2 g/kg body weight) or injected in the intraperitoneal (i.p.) cavity with insulin (0.75 units/kg body weight). Blood glucose levels were monitored (Accu-Chek II glucometer; Roche Diagnostics, Castle Hill, Australia), and plasma insulin and C-peptide levels were determined by ELISA (Ultra-Sensitive Mouse Insulin ELISA/Mouse C-peptide ELISA Kit; Crystal Chem, Elk Grove Village, IL).

Mice were fasted for 4 h and gavaged orally with glucose (2 g/kg body weight) and simultaneously injected i.p. with 2-[1-¹⁴C]-deoxy-d-glucose (2DG; 10 μCi/mouse) (PerkinElmer, Melbourne, Australia). Blood glucose concentration and tracer levels were determined as indicated (Fig. 4E). Blood was deproteinized and tissue-specific 2DG uptake determined as described previously (18). It is noted that the tracer and tracee were administered by different routes, which is not consistent with the principles of glucose kinetics analysis. While acknowledging this limitation, this approach allows determination of 2DG accumulation in tissues under conditions of hyperglycemia and hyperinsulinemia. In a separate set of experiments, 2DG uptake was measured in inguinal adipose tissue ex vivo, as described previously (19).

Lipids were extracted in chloroform:methanol (2:1 volume for volume), and the phases were separated with 4 mmol/L MgCl₂ as previously described (20). Triacylglycerol content was determined by colorimetric assay (Triglycerides GPO-PAP; Roche Diagnostics, Indianapolis, IN). Ceramide and diacylglycerol were analyzed by mass spectrometry as previously described (21). Plasma nonesterified fatty acids (NEFAs) and total plasma cholesterol were measured by colorimetric assay (Wako Diagnostics, Osaka, Japan). Thiobarbituric acid reactive substance (TBARS) and lipid hydroperoxides were determined as previously described (22,23).

Plasma β-hydroxybutyrate levels were measured by colorimetric assay (Sapphire Biosciences, Redfern, Australia). Plasma and medium FGF21 (Mouse/Rat FGF-21 Quantikine ELISA; R&D Systems, Minneapolis, MN) and tumor necrosis factor-α (TNFα) were measured by ELISA (Mouse TNFα ELISA; elisakit.com, Scoresby, Australia). Aspartate aminotransferase (AST) and alanine aminotransferase (ALT) activities were determined by enzymatic colorimetric assays (24,25).

Fatty acid metabolism in isolated soleus muscle was measured as previously described (17).

Lysates were prepared in radioimmunoprecipitation assay buffer, lanes were loaded relative to whole-tissue protein content (20 µg protein per lane), proteins were resolved by SDS-PAGE electrophoresis, and immunoblot analysis was conducted as described previously (16). Primary antibodies are listed in Supplementary Table 1 and have been validated (16,26–29).

RNA extraction, cDNA synthesis, and real-time PCR were performed as previously described in detail (29). The primer sequences are listed in Supplementary Table 2.

Mice were anesthetized with 5% isoflurane, the chest opened, the aorta clamped, and the heart removed into ice-cold calcium-free Hanks’ balanced salt solution. A 22-g stainless steel tube was introduced into the aorta, with the opening just above the coronary sinus, and secured in place. The heart was mounted on a Langendorff apparatus (ML870B2; AdInstruments, Bella Vista, Australia) and perfused through the coronary network with bubbled (95% O₂/5% CO₂) EX-CELL 325 Protein-Free CHO Serum-Free Medium (SAFC Biosciences) at 37°C for 2 h (recirculation of 6 mL of medium). The heart was enclosed by a heating jacket to ensure stable temperature and humidity. Flow was gradually increased (displayed on LabChart v7 software) until coronary network pressure was 80 mmHg (Gould pressure transducer).

Intact soleus and extensor digitorum longus (EDL) muscle (each 10–12 mg) were excised from HFD-fed mice and placed in 0.5 mL pregassed (95% O₂/5% CO₂) EX-CELL medium at 37°C for 6 h. Liver slices (∼400 µm thickness, 20 mg) were incubated under the same conditions. FGF21 was measured in the “conditioned medium” (CM; i.e., medium containing secreted factors). In addition, the CM from muscles was applied to primary murine hepatocytes (26) for 24 h after 4-h adherence of hepatocytes to the culture plates (50% CM, 50% M199 hepatocyte medium). These experiments were replicated using heart CM. In separate experiments, 10% serum was added to the hepatocyte medium for 24 h. FGF21 secretion from hepatocytes was measured and corrected for FGF21 present in the muscle CM (contributed 1–2% of total medium FGF21) and in the plasma (3–11% of total FGF21). Interassay variation was minimized by combining hepatocytes from three mice for each experiment.

Plin5^MKO mice were injected i.p. with 5 ng recombinant murine FGF21 (Sino Biologicals, Beijing, China) 1 h before an oral glucose tolerance test (OGTT). Blood glucose, plasma FGF21, insulin, and C-peptide were assessed as indicated (Fig. 6).

Results are presented as mean ± SEM. Data were analyzed with an unpaired two-tailed Student t test or a two-way ANOVA with Bonferroni post hoc analysis, where appropriate. Statistical significance was accepted at P < 0.05.

Similar to our recent publication (16), Plin5 mRNA was reduced by 73% and 97% in quadriceps muscle and heart of Plin5^MKO mice, respectively (Fig. 1A), without changes in other tissues or other LD-associated proteins (PLIN2 and CGI-58) (Supplementary Fig. 1A–C). Whereas body weight was unchanged in Plin5^MKO compared with wild-type (WT) mice (Fig. 1B), there was an increase in fat mass in Plin5^MKO mice (Fig. 1C) and no change in lean mass between genotypes (Fig. 1D). Energy expenditure was not different between genotypes (Fig. 1E), whereas respiratory exchange ratio (RER) was decreased in Plin5^MKO mice (Fig. 1F), demonstrating increased whole-body fatty acid oxidation. Food intake was not different between genotypes (WT 3.33 ± 0.07 [n = 9], Plin5^MKO 3.03 ± 0.13 [n = 8] g/mouse/day). Glucose clearance during an OGTT was indistinguishable between WT and Plin5^MKO mice (Fig. 1G), as was plasma insulin measured in fasted mice and during the OGTT (Fig. 1H). There was no difference in insulin action (Fig. 1I).

We next determined whether increasing dietary fat content would reveal metabolic alterations in Plin5^MKO mice because PLIN5 is upregulated by high-fat feeding (30). Whereas total body weight (Fig. 2A) and individual tissue weights (Supplementary Table 3) were similar between genotypes, lean mass was elevated by 15% (Fig. 2B) and whole-body adiposity tended to be decreased (P = 0.099) (Fig. 2C) in Plin5^MKO mice. Food intake was similar between genotypes (WT 2.40 ± 0.05 [n = 5], Plin5^MKO 2.35 ± 0.16 [n = 5] g/mouse/day).

Energy expenditure was similar (Fig. 2D) and whole-body fatty acid oxidation was increased in Plin5^MKO compared with WT mice (Fig. 2E). Skeletal muscle fatty acid oxidation was significantly increased in Plin5^MKO mice (Fig. 2F), whereas triglyceride esterification tended to be reduced (Fig. 2G), pointing toward fatty acids being preferentially shuttled into oxidative pathways in skeletal muscle of Plin5^MKO mice. Fatty acid uptake was not different between genotypes (Fig. 2H).

The abundance of electron transport chain complexes was similar in skeletal muscle of WT and Plin5^MKO mice (Fig. 2I and Supplementary Fig. 1D–H). Interestingly, mRNA and protein contents of acetyl-CoA carboxylase 2 (ACC2), an enzyme that regulates the entry of fatty acids into the mitochondria, was decreased in Plin5^MKO mice (Fig. 2J–L). Whereas phosphorylation at the deactivating Ser212 residue of ACC2 was markedly decreased in Plin5^MKO mice (Fig. 2K), there was no difference between genotypes when expressed per unit of ACC2 protein (Fig. 2M). There was no difference in phosphorylated (Thr172) or total AMPK (Supplementary Fig. 1I and J), the upstream regulator of ACC. Hence, the profound decrease in ACC2 abundance in skeletal muscle is likely to contribute to the observed increase in fatty acid oxidation.

Increasing fatty acid–derived mitochondrial electron flow can lead to oxidative stress. Oxidative stress was increased in the skeletal muscle of Plin5^MKO mice compared with WT mice as indicated by increased lipid hydroperoxide (Fig. 2N) and TBARS (Fig. 2O). The increased muscle fatty acid oxidation in Plin5^MKO mice also coincided with a decrease in intramyocellular triglyceride content (Fig. 2P), indicative of increased lipolysis and oxidation of triglyceride-derived fatty acids. No differences were observed in liver triglyceride content (WT 17.4 ± 1.5 [n = 6], Plin5^MKO 20.6 ± 1.2 [n = 7] µmol/g tissue) or in circulating triglycerides, NEFAs, and total plasma cholesterol content between WT and Plin5^MKO mice (Table 1). There were no differences in muscle ceramide and diacylglycerol contents (Supplementary Fig. 2A–D) between genotypes.

Changes in lipid catabolism and LD formation are likely to impact ER stress (14), and deletion of Plin5 impacts these processes in skeletal muscle (Fig. 2). Accordingly, we analyzed the three major ER stress–induced signaling pathways in skeletal muscle of WT and Plin5^MKO mice, which include the IRE1/XBP1, PERK/eIF2α, and ATF6 pathways. Under ER stress, IRE1α is autophosphorylated at Ser724 and removes a 26–base pair intron to produce spliced XBP1, which in turn regulates expression of genes involved in the UPR. Plin5 deletion in skeletal muscle decreased IRE1α Ser724 compared with WT muscle (Fig. 3A), the mRNA expression (data not shown) and protein contents of the spliced isoform of XBP (XBP1s) (Fig. 3B), and mRNA expression of various transcription targets of XBP1s, including Edem1, Dnajb11 (HEDJ), Dnajc3 (P58IPK), and Hspa5 (GRP78/Bip) (Fig. 3C). In accordance with reduced IRE1α activation, phosphorylation of the downstream stress kinase JNK was substantially reduced in the muscle of Plin5^MKO compared with WT mice (Fig. 3D). Together, these data demonstrate a marked reduction of the IRE1 pathway in Plin5^MKO skeletal muscle.

There was also evidence for mild inhibition of the PERK/eIF2α pathway; whereas eIF2α phosphorylation (Ser51) was unchanged, ATF4 mRNA content was markedly reduced (Fig. 3E). We were unable to reliably detect PERK in skeletal muscle using several antibodies. Nuclear factor Nrf2 regulates ATF4 transcription (31) and was reduced in Plin5^MKO muscle (Fig. 3E). XBP1s and ATF4 regulate the transcription factor C/EBP homologous protein (CHOP). CHOP mRNA was decreased by 77% and CHOP protein content was reduced by 30% (not significant) (Fig. 3F). In contrast, the ATF6 pathway was unaffected by Plin5 deletion (Fig. 3G).

In addition to reduced ER stress signaling, the mRNA content of the proinflammatory markers TNFα, IL6, and MCP1 was decreased in the skeletal muscle of Plin5^MKO compared with WT mice (Fig. 3H). Plasma TNFα was not different between genotypes, indicating localized rather than systemic effects (Table 1). Finally, PLIN2 has been associated with p62-mediated lipophagy in cultured myotubes (32), indicating a potential role for PLIN5 in autophagy-regulated clearance of lipids. There was no difference between genotypes for the autophagy markers p62 and the LC3 (II/I) ratio (Fig. 3I).

MCK-Cre induced a substantial reduction in Plin5 in the heart (Fig. 1A), and, similar to skeletal muscle, Plin5^MKO mice exhibited a 55% reduction in cardiac triglyceride content (Supplementary Fig. 3A). In contrast to skeletal muscle, ACC2 phosphorylation and protein content and markers of ER stress and inflammation were not different between genotypes (Supplementary Fig. 3B–N), and lipid hydroperoxide levels (Supplementary Fig. 3O) and TBARS (Supplementary Fig. 3P) were significantly reduced in the hearts of Plin5^MKO mice. These results demonstrate cell-autonomous roles for PLIN5 in skeletal and cardiac muscle.

Given that oxidative and ER stress and inflammation cause glucose intolerance and insulin resistance (33), we next investigated glucose metabolism in mice. Glucose clearance was impaired in Plin5^MKO compared with WT mice after an oral glucose load (Fig. 4A), and this was accompanied by a mild reduction in plasma insulin levels (Fig. 4B) and plasma C-peptide levels (Fig. 4C) in Plin5^MKO mice. Plin5^MKO mice also exhibited a mild impairment in whole-body insulin action (Fig. 4D). To ascertain tissue-specific differences in glucose disposal, mice were administered 2DG during an OGTT. Blood glucose was higher in Plin5^MKO mice 30 min after glucose administration (WT 19.1 ± 1.0 vs. Plin5^MKO 22.3 ± 1.2 mmol/L, P = 0.05), and the radiotracer concentration in the blood was similar between genotypes (Fig. 4E). Accumulation of 2DG uptake in skeletal muscle (Fig. 4F) and heart (Fig. 4G) was similar between genotypes, whereas 2DG content was significantly reduced in the white adipose tissue of Plin5^MKO compared with WT mice (Fig. 4H). The finding was confirmed in ex vivo 2DG uptake experiments in adipose tissue (Fig. 4I).

Given that the data demonstrated no role of muscle-specific Plin5 deletion on muscle glucose uptake, but rather in adipose tissue glucose uptake and reduced plasma insulin and C-peptide levels, we hypothesized that an endocrine factor might be responsible for actions in tissues distant to skeletal muscle. XBP1s, ATF4, and CHOP are powerful transcriptional activators of FGF21 (13,34), and these proteins were reduced in the skeletal muscle of Plin5^MKO mice (Fig. 3B, E, and F). Consistent with this molecular regulation, FGF21 mRNA content was markedly reduced in the skeletal muscle of Plin5^MKO compared with WT mice fed an HFD (Fig. 5A), and this was accompanied by a 55% reduction in plasma FGF21 (Fig. 5B). We also assessed liver FGF21 because this is thought to be the primary source of circulating FGF21 (35). Liver FGF21 mRNA was significantly reduced in Plin5^MKO compared with WT mice (Fig. 5C), suggesting that factors secreted from skeletal muscle may impact FGF21 expression in the liver. FGF21 mRNA was not different between genotypes in quadriceps muscle or liver of chow-fed mice, or in the heart and inguinal adipose tissue of HFD mice (Supplementary Fig. 4A–D).

To test the possibility of skeletal muscle–liver tissue cross talk, secreted products were collected from glycolytic (EDL) and oxidative (soleus) skeletal muscle ex vivo and the CM was added to primary murine hepatocytes. Intriguingly, muscle CM from Plin5^MKO mice, but not WT mice, reduced FGF21 secretion from hepatocytes (Fig. 5D). Secreted factors from the heart of Plin5^MKO mice did not impact liver FGF21 secretion (Fig. 5D). We also assessed FGF21 secretion from muscle and liver to ascertain the potential contribution of each tissue toward circulating FGF21 levels in WT and Plin5^MKO mice. Interestingly, FGF21 secretion was per unit mass higher in muscle compared with liver (Fig. 5E). Moreover, FGF21 secretion was reduced in the soleus muscle and liver of Plin5^MKO mice, whereas no differences were observed between genotypes in EDL muscle or heart (Fig. 5F). Interestingly, addition of 10% serum from Plin5^MKO mice to the hepatocyte culture medium resulted in a 30% reduction in hepatocyte FGF21 secretion, compared with hepatocytes incubated with WT serum (Fig. 5G). Thus, FGF21 appears to be decreased in the circulation of high fat–fed Plin5^MKO mice as a result of reduced expression and secretion from skeletal muscle and liver, the latter effect being regulated by an unknown muscle-secreted factor.

To test if the reduced plasma FGF21 levels in Plin5^MKO mice are likely to impact adipose tissue, we examined the expression of known FGF21 target genes in adipose tissue. Ucp1, Cidea, and Cpt1a expression was significantly reduced in Plin5^MKO mice (Fig. 6A–C). Expression of FGF21 nontargets (36) Kctd20 and Spata13 was unchanged (data not shown).

We next tested whether correction of the reduced serum FGF21 could “rescue” the metabolic defects observed in Plin5^MKO mice. Injection of recombinant murine FGF21 in Plin5^MKO mice was sufficient to increase plasma FGF21 to levels observed in WT mice (Fig. 6D), to rescue glucose intolerance (Fig. 6F), and to restore insulin secretion (Fig. 6G).

The results reveal an unanticipated link between lipid metabolism, ER homeostasis, and the regulation of myokine production. Our data show that PLIN5 is required to maintain the appropriate balance between fatty acid storage and fatty acid oxidation during periods of lipid overload and that failure to do so causes oxidative stress in skeletal muscle. When comparing the responses of high fat–fed WT and Plin5^MKO mice, the data suggest that intracellular lipid accumulation activates the UPR in skeletal muscle and that this drives the production and secretion of the myokine FGF21, which is known to improve adipose tissue insulin sensitivity and pancreatic insulin secretion. This is of interest as an activation of the UPR is commonly associated with a deterioration of insulin sensitivity (37). Moreover, this study has identified novel interorgan communication between muscle and liver. PLIN5 deletion alters the factors secreted by skeletal muscle, which in turn reduce the production and secretion of FGF21 by the liver, suggesting a potential mechanism by which altered intramyocellular lipid metabolism can regulate systemic metabolic homeostasis.

The role of LD proteins in the regulation of lipid and glucose metabolism has garnered much attention in recent years, particularly in skeletal muscle, which is a critical regulator of whole-body fatty acid oxidation and insulin-stimulated glucose uptake. There is particular interest in PLIN5’s role in these processes because of its high expression in skeletal muscle and its regulation in response to physiological perturbations that modulate metabolism, including high-fat feeding, fasting, and exercise (9). Whole-body fat oxidation was increased in Plin5^MKO mice and was accompanied by a partitioning of fatty acids away from storage and toward oxidation, resulting in reduced intramyocellular triglyceride accumulation in both skeletal and heart muscle. This finding is consistent with previous reports (12,38,39) and the premise that LDs are a central hub for fatty acid trafficking in muscle (3,40) and that PLIN5’s major function is to reduce lipolysis by binding ATGL and CGI-58 and preventing their interaction (11,41). The increase in fatty acid oxidation was not related to molecular changes reflecting altered mitochondrial content or oxidative capacity, which contradicts a recent report describing PLIN5 as a nuclear protein that promotes transcriptional regulation of genes that mediate oxidative function (42), but rather was associated with a substantial reduction in ACC2 mRNA and protein content in muscle. ACC2 produces malonyl CoA, which allosterically inhibits the mitochondrial fatty acid transporter CPT1 and thereby suppresses fatty acid oxidation. ACC2 transcription is regulated by several transcription factors, including XBP1s (i.e., spliced isoform of XBP1), which is activated as part of the UPR during ER stress (43). XBP1 was substantially decreased in the muscle of Plin5^MKO mice, concomitant with a drastic reduction in ACC2 mRNA and protein, accelerated fatty acid oxidation, and oxidative stress. Changes in ACC2 were independent of changes in the upstream regulator AMPK. Collectively, these data suggest that PLIN5 is required to coordinate shuttling of intracellular triglyceride lipolysis toward fatty acid oxidation and that deleting this function disrupts the normal sensing of ER stress and results in excessive fatty acid oxidation. Although speculative, this lipid-sensing mechanism may be required to protect cells from further cellular stress (e.g., oxidative stress) in the face of lipid accumulation and ER stress.

ER stress is induced by several metabolic perturbations, and, although it remains unclear how lipids cause ER stress and activate the UPR, accumulating evidence shows that this conserved response plays an important role in maintaining metabolic and lipid homeostasis. LDs originate from and reside in close proximity to the ER and have been implicated in the UPR (14). In this regard, LD accumulation is closely associated with ER function in yeast (44) and in the livers of mammals, where the accumulation of LDs is associated with ER stress (45) and resolution of ER stress prevents lipid accumulation (46). Skeletal muscle contains an extensive network of ER, called the sarcoplasmic reticulum, and high-fat feeding induces the UPR in skeletal muscle of mice (37,47). We show that deletion of Plin5 in skeletal muscle increases fatty acid oxidation, reduces intracellular lipid accumulation, and blunts activation of the UPR, most notably the IRE1-XBP1s and PERK-eIF2α pathways. Such regulation is consistent with mice overexpressing Plin5 in skeletal muscle that shows excessive muscle triglyceride accumulation (12,39) and an induction of ER stress pathways (12), and it is consistent with the notion that PLIN5 is required to regulate healthy intramyocellular lipid flux to buffer lipid loading in myocytes (38,48). We did not observe any differences in ER stress in the heart despite a complete deletion of Plin5, pointing toward skeletal muscle–specific regulation.

ER stress and activation of the UPR is associated with the development of insulin resistance (37). We conclusively show that the IRE1-XBP1 and the PERK-eIF2α arms of the UPR are downregulated in the skeletal muscle of Plin5^MKO mice. This coincided with decreased insulin and C-peptide levels during an oral glucose load and a modest impairment in whole-body insulin action, which was associated with reduced adipose tissue insulin sensitivity and, surprisingly, no changes in muscle 2DG after an oral glucose challenge. XBP1s, ATF4, and CHOP are transcriptional regulators of FGF21 (13,34,49), and our analysis demonstrated an ∼50% decrease in muscle and liver FGF21 expression and circulating FGF21 levels in Plin5^MKO mice. FGF21 is a regulator of β-cell insulin secretion (50) and adipose tissue insulin sensitivity (51,52), and the marked downregulation of FGF21 provides a plausible explanation for this metabolic phenotype in Plin5^MKO mice. The substantial increase in muscle and plasma FGF21 in mice overexpressing Plin5 in skeletal muscle supports this notion (12). Our data also support the premise that upregulation of FGF21 by peripheral tissues might be a more general response to cellular stresses (53,54).

Whether muscle-derived FGF21 impacts whole-body physiology remains controversial, with some reports suggesting that FGF21 is not typically expressed in muscle and is only induced in situations of muscle stress (55,56), whereas others describe FGF21 as an important myokine (57,58). Our results linking ER stress with FGF21 production and our finding under ex vivo conditions that FGF21 secretion was greater from muscle than from liver supports both interpretations. Extending on these data, we have also demonstrated that muscle-secreted factors in Plin5^MKO mice significantly reduce hepatocyte FGF21 secretion, whereas heart-secreted factors had no impact, highlighting the likely importance of skeletal muscle–liver communication in regulating systemic FGF21 levels. This is further highlighted by plasma of Plin5^MKO mice having the capacity to reduce hepatocyte FGF21 secretion. Identifying the specific muscle-secreted factors driving the inhibition of FGF21 secretion in Plin5^MKO mice is beyond the scope of this study, but such information will provide insight in understanding myokine regulation of metabolism. Our results support the recent findings in mice overexpressing Plin5 in skeletal muscle, which reported increased skeletal muscle and serum FGF21 levels that were proposed to induce the “browning” of adipose tissue and improved systemic metabolism (12). In this respect, we show that correction of plasma FGF21 levels in Plin5^MKO mice by recombinant FGF21 injection can rescue their glucose intolerance and restore insulin secretion to rates of WT mice. In addition, the present studies extend on these previous observations by elucidating the likely mechanism mediating this effect; that is, muscle-specific Plin5 deletion leads to remodeling of LD dynamics and lipid accumulation, a decrease in UPR/ER stress, and downregulation of essential transcription factors for FGF21 expression and secretion in muscle.

In conclusion, these findings demonstrate that PLIN5 deficiency in skeletal muscle reduces intramyocellular triglyceride accumulation, increases fatty acid oxidation, and causes oxidative stress. Our studies comparing WT and Plin5^MKO mice have unmasked the likely importance of lipid sensing by the ER for regulating systemic glucose metabolism, showing that under conditions of dietary lipid oversupply, PLIN5 is required for the adaptive maintenance of UPR-driven transcription of FGF21 in skeletal muscle, which leads to secretion of FGF21 that can positively impact adipose tissue insulin sensitivity and systemic glycemic control. Moreover, skeletal muscle secretes as yet unidentified factors to promote FGF21 production and secretion by the liver, highlighting the growing appreciation of muscle-liver cross talk for systemic metabolic control.

Acknowledgments. The authors thank Maria Matzaris (Monash University) for technical assistance.

Funding. This work was funded by the Australian National Health and Medical Research Council (NHMRC) (1047138). M.K.M. is supported by a Peter Doherty Fellowship and M.J.W. is supported by a Senior Research Fellowship from the NHMRC (1077703). C.R.B. is supported by an Australian Research Council Future Fellowship (FT160100017).

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

Author Contributions. M.K.M. and M.J.W. planned and conducted the experiments, analyzed the data, and wrote the manuscript. R.M. planned and conducted the experiments and analyzed the data. J.B. and V.M.S. conducted the experiments. H.C.P. and C.R.B. planned and conducted the 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.