Muscle-Derived Angiopoietin-Like Protein 4 Is Induced by Fatty Acids via Peroxisome Proliferator–Activated Receptor (PPAR)-δ and Is of Metabolic Relevance in Humans

A detailed description of the methods is given in the online appendix (found at http://diabetes.diabetesjournals.org/cgi/content/full/db07-1438/DC1).

Primary human myotubes and murine C2C12 myocytes were used for cell experiments. Microarray analysis was performed with Affymetrix Human Genome U133 Plus 2.0 arrays. Real-time RT-PCR was performed with SYBR Green I dye on a LightCycler. The anti-ANGPTL4 antibody from BioVendor was used for immunoblotting. Intracellular and secreted ANGPTL4 was quantified by ELISA. For RNA interference (RNAi), siGENOME siRNA sets designed by Dharmacon were used. Chromatin immunoprecipitation (ChIP) analysis was performed with the anti–PPAR-δ antibody K-20 from Santa Cruz Biotechnology.

All 38 myotube donors underwent an oral glucose tolerance test (OGTT) and a hyperinsulinemic-euglycemic clamp. The 108 subjects with plasma ANGPTL4 measurements were characterized by OGTT and a subgroup of 91 subjects also by hyperinsulinemic-euglycemic clamp. All subjects gave informed written consent to the study. The protocol was approved by the local ethics committee. Total, visceral, and nonvisceral fat was determined by magnetic resonance imaging. Intramyocellular and intrahepatic lipids were measured by magnetic resonance spectroscopy. Glucose, insulin, C-peptide, NEFA, glycerol, triglycerides, and ANGPTL4 were quantified by standard laboratory methods.

In an initial attempt to identify novel LCFA target genes related to metabolic syndrome, we treated human myotubes derived from a healthy male German donor with bovine serum albumin (BSA; carrier control), the saturated fatty acid palmitate, or the unsaturated fatty acid linoleate and performed whole-genome gene array analysis. Palmitate treatment repressed 181 and induced 316 genes, and linoleate repressed 30 and induced 104 genes. Thus, LCFAs appear to influence the expression of numerous genes in human myotubes. Notably, only five genes were repressed (Supplementary Table 3, found in the online appendix), and only 35 genes were induced (Table 1) by both palmitate and linoleate. Thus, the majority of LCFA-regulated genes appear to represent LCFA-specific targets, and modulation of their expression might depend on LCFA chain length and/or the degree of saturation. The gene that revealed highest fold-induction by both palmitate and linoleate was ANGPTL4 (Table 1), encoding angiopoietin-like protein 4 (ANGPTL4). The sole purpose of this single nonreplicated experiment was to generate new hypotheses. Therefore, these results cannot be generalized.

ANGPTL4 was described as hepatic fibrinogen/angiopoietin-related protein (13), fasting-induced adipose factor (14), and PPAR-γ target gene related to angiopoietin (15) and was characterized as a secreted protein predominantly produced by WAT, but also at lower levels by other tissues (13–15). The role of ANGPTL4 was extensively explored in mice by injection, targeted gene knockout, and transgenic and retroviral overexpression, and in this way, ANGPTL4 was shown to affect lipid metabolism: by inhibition of lipoprotein lipase (LPL), clearance of VLDL and chylomicrons is blocked and hypertriglyceridemia is provoked (16–20). Furthermore, ANGPTL4 stimulates WAT lipolysis (21) resulting in elevated plasma glycerol and NEFA levels (16,21). Besides hyperlipidemia, ANGPTL4 promotes WAT weight loss and hepatic steatosis (18,21). Importantly, ANPTL4 expression was consistently found upregulated in genetic mouse models of obesity and type 2 diabetes (15). Thus, ANGPTL4 represents a metabolically relevant candidate gene induced by common plasma LCFA species.

To assess whether ANGPTL4 is produced by human myotubes at relevant levels, we measured basal ANGPTL4 mRNA expression by real-time RT-PCR. Untreated human myotubes expressed 20.9 ± 7.2 fg ANGPTL4 mRNA/μg total RNA (mean ± SD; n = 5). This level was not only in the range of that found in a representative human SKM biopsy (12 fg ANGPTL4 mRNA/μg total RNA) but also represents ∼40% of the mRNA level found in subcutaneous WAT (56 fg ANGPTL4 mRNA/μg total RNA), a major site of ANGPTL4 expression.

To confirm the gene array results and to explore the influence of other plasma LCFAs on ANGPTL4 expression, we treated human myotubes with BSA or selected major plasma LCFA species and subsequently quantified the cellular ANGPTL4 mRNA contents by real-time RT-PCR (normalized to 28S rRNA). ANGPTL4 mRNA expression was induced 10- to 50-fold by all LCFAs tested (Fig. 1A). The ANGPTL4 mRNA levels found after treatment with palmitate, stearate, palmitoleate, oleate, linoleate, or a combination of palmitate and linoleate were significantly different from their respective BSA controls (Fig. 1A). The ANGPTL4 mRNA contents detected after treatment with a combination of palmitate and linoleate did not significantly differ from those obtained with palmitate or linoleate alone. Thus, treatment with palmitate or linoleate alone might already be sufficient to reach the maximum response of LCFA-inducible ANGPTL4 expression. Moreover, ANGPTL4 induction depended neither on the chain length nor on the saturated/unsaturated nature of the LCFA but, rather, represented a general LCFA effect. To verify these LCFA-mediated gene regulations at the protein level, we used the first commercially available ELISA kit and measured intracellular ANGPTL4 contents after 20 h LCFA treatment. All LCFAs increased intracellular ANGPTL4 protein 1.5- to 2.3-fold, with palmitate, oleate, and linoleate reaching statistical significance (Fig. 1B). The observed differences between mRNA and protein induction rates are a well-known phenomenon that most probably reflects that gene transcription rate and mRNA half-life are not necessarily tightly linked to translation efficiency and protein half-life. In addition, we tried to measure and visualize secreted ANGPTL4 protein. However, the high BSA concentrations used for LCFA treatment interfered with the detection of secreted ANGPTL4. To see whether human myotube ANGPTL4 expression was modulated by glucose, we performed a dose-response experiment with 2.5, 5, 10, and 25 mmol/l glucose (added to glucose-free medium). After 20 h, there was no significant glucose effect on basal myotube ANGPTL4 expression (P = 0.9, ANOVA; n = 4).

Because ANGPTL4 was described as a fasting-induced factor (14,15,22) and plasma LCFAs are known to be elevated during the fasting state due to unsuppressed WAT lipolysis, these results turn plasma LCFAs into potential candidate mediators of fasting-induced ANGPTL4 production.

ANGPTL4 is a known target gene of PPAR-α and -δ in liver (14,22) and PPAR-γ in WAT (14,15), and a PPAR response element was identified in intron 3 of the ANGPTL4 gene (23). To see whether LCFAs enhance myotube ANGPTL4 expression via induction of the genes encoding PPAR-α, -γ, or -δ (PPARA, PPARG, or PPARD, respectively), we quantified the respective mRNA levels before and after LCFA treatment. In untreated human myotubes, the PPARD mRNA contents were ∼10-fold higher than those of PPARA and PPARG (data not shown). Neither PPARA nor PPARG mRNA contents were significantly altered by LCFA treatment when compared with their respective BSA controls (P ≥ 0.2, both, ANOVA; n = 5). Furthermore, all LCFAs shown above to induce ANGPTL4 (Fig. 1A) did not induce PPARD expression, with only one exception: stearate increased PPARD mRNA contents 2.6-fold (P < 0.0001, ANOVA; post hoc P < 0.05; n = 5). This stearate-specific effect, however, cannot explain the general effect of LCFA on ANGPTL4 expression. To explore whether stearate's inductive effect is translated to the protein level, we performed immunoblot analysis. None of the LCFAs tested, including stearate, provoked relevant changes in PPAR-δ protein after 20 h of treatment (n = 3; data not shown). Therefore, we suggest that LCFAs act as PPAR ligands or activators (intracellularly metabolized to ligands), but not as regulators of PPAR expression/production.

To examine whether LCFAs are able to activate one of the three PPAR isoforms in human myotubes, we also quantified the mRNA levels of PPAR target genes, i.e., PDK4 (encoding pyruvate dehydrogenase kinase 4) as a target of PPAR-δ and -α (24), CD36 (encoding fatty acid translocase) and UCP3 (encoding uncoupling protein 3) as PPAR-α–specific targets (25,26), and PPARG as a target of PPAR-γ itself (27). Compared with their respective BSA controls, none of the LCFAs modulated the expression of CD36, UCP3, or PPARG in these cells (P ≥ 0.4, all, ANOVA; n ≥ 4). By contrast, all LCFA induced PDK4 mRNA expression 7- to 22-fold, and the effects of palmitate, palmitoleate, oleate, linoleate, and the combination of palmitate and linoleate were statistically significant (Fig. 1C). This led us to assume that the LCFAs tested in this study selectively activate PPAR-δ in human myotubes, at least at the concentration tested (0.5 mmol/l).

To further investigate the role of PPAR isoforms in human myotube ANGPTL4 expression, we treated the cells with the PPAR-α–specific fibrates Wy-14,643 and fenofibrate, the PPAR-γ–specific thiazolidinediones troglitazone and rosiglitazone, and the PPAR-δ–specific activator GW501516. As presented in Fig. 2A, ANGPTL4 mRNA expression was markedly induced by GW501516 and moderately induced by troglitazone (17-fold and 4-fold, respectively). By contrast, ANGPTL4 mRNA expression was neither altered by rosiglitazone, a thiazolidinedione with 10-fold higher affinity for PPAR-γ than troglitazone, nor by any of the fibrates tested. Only GW501516 treatment reached a level of ANGPTL4 induction comparable to that found with LCFAs. Furthermore, ANGPTL4 induction by GW501516 followed the same kinetics as the other PPAR-δ target gene PDK4 (Figs. 2B and 2C). We also tested GW501516's effect on intracellular ANGPTL4 protein contents using ELISA. GW501516 treatment (1 μmol/l, 20 h) increased intracellular ANGPTL4 protein 2.5-fold over DMSO control (365.6 ± 52.9 vs. 908.2 ± 154.0 pg/mg; P = 0.0158, t test; n = 4). These data strengthened our suggestion of PPAR-δ being a crucial mediator of LCFA-induced ANGPTL4 expression in human myotubes.

In addition, 48-h treatment of human myotubes with GW501516 resulted in continuous accumulation of ANGPTL4 protein in the culture supernatant (Fig. 2D). Hence, human ANGPTL4 can be added to the novel and growing list of muscle-derived secreted proteins with metabolic functions (myokines), which also comprises interleukin-6 (28), interleukin-15 (29), and musclin (30). Besides full-length ANGPTL4 (∼70 kDa), a 50-kDa COOH-terminal and a 26-kDa NH₂-terminal fragment (both with biological activity) were reported to circulate in the bloodstream (18,23,31). Therefore, we asked whether ANGPTL4 secreted into the supernatant by GW501516-treated human myotubes is proteolytically cleaved. Immunoblotting revealed that both full-length ANGPTL4 and the COOH-terminal fragment accumulated in the culture supernatant during the treatment period (Fig. 2E). The NH₂-terminal fragment could not be detected with antibodies from different suppliers directed against full-length or the NH₂-terminal part of ANGPTL4 and, thus, seems to be rapidly degraded. In conclusion, ANGPTL4 secreted by human myotubes is cleaved and biologically activated.

To evidence that LCFA-induced myocyte ANGPTL4 expression depends on PPAR-δ, we knocked down PPARD expression by RNAi. This was done in murine C2C12 myocytes because human myotubes could not be efficiently transfected. First, we treated C2C12 myocytes with representative saturated fatty acid and unsaturated fatty acid species. As presented in Fig. 3A, all LCFAs tested increased the C2C12 ANGPTL4 mRNA contents when compared with their respective BSA controls, and the effects of stearate, oleate, and linoleate reached the level of significance. Thus, LCFA-induced ANGPTL4 expression was not restricted to human myotubes but was reproduced in a murine SKM cell line. Moreover, as in human myotubes, GW501516 treatment induced ANGPTL4 mRNA expression up to 20-fold in C2C12 cells (Fig. 3B). As depicted in Fig. 3C, transfection with siRNA directed against PPARD, but not with control siRNA directed against bacterial luciferase, reduced C2C12 PPARD mRNA contents by 82%. After PPARD knockdown, oleate-induced ANGPTL4 expression was significantly impaired (2.4-fold in the presence vs. 6.8-fold in the absence of PPARD siRNA); moreover, the oleate effect was no longer significant (Fig. 3D). Furthermore, control transfection with siRNA directed against bacterial luciferase still allowed marked ANGPTL4 induction by oleate (Fig. 3D). Thus, induction of ANGPTL4 expression in C2C12 myocytes by the representative LCFA oleate requires PPAR-δ. Importantly, C2C12 PPARA expression ranged at the detection limit of the real-time PCR method, even upon utilization of sensitivity-enhancing hybridization probes. Therefore, we conclude that PPARA is not expressed in these cells and is not involved in LCFA-induced ANGPTL4 expression. Transfection of the cells with siRNA directed against PPARG, but not with control siRNA, reduced C2C12 PPARG expression by 88%. However, this manipulation had no significant impact on oleate-induced ANGPTL4 expression (P = 1.0, t test; n = 3). Hence, LCFA effects on ANGPTL4 expression are specifically mediated by PPAR-δ.

To ultimately prove that PPAR-δ is activated by LCFAs and binds to the ANGPTL4 gene, we performed ChIP analysis in C2C12 cells treated for 6 h with oleate. In the anti–PPAR-δ immunoprecipitate, markedly more ANGPTL4 intron 2 DNA (harboring the PPRE) could be detected in oleate-treated cells compared with BSA-treated cells (Fig. 3E). This demonstrates that PPAR-δ is activated and recruited to the PPRE of the ANGPTL4 gene upon LCFA treatment.

ANGPTL4 is reported to enhance WAT lipolysis via induction of adipose triglyceride lipase (21). We therefore asked whether triglyceride lipases of SKM cells, which are required for breakdown of intramyocellular triglycerides, are also under the control of muscle-derived ANGPTL4 (mANGPTL4). Untreated C2C12 myocytes expressed 46.3 ± 10.6 fg PNPLA2 mRNA (encoding adipose triglyceride lipase) and 11.8 ± 2.0 fg LIPE mRNA (encoding hormone-sensitive lipase)/μg total RNA (means ± SD; n = 3). Treatment of C2C12 cells with GW501516 provoked significant increments over time of LIPE and PNPLA2 mRNA (Fig. 4A and B) compared with DMSO (carrier control). However, the kinetics of both gene regulations was completely different from that of the aforementioned PPAR-δ target genes PDK4 and ANGPTL4 in that these lipases showed delayed induction (compare Fig. 2B and C with Fig. 4A and B). This suggests that LIPE and PNPLA2 are not direct PPAR-δ target genes.

The GW501516 effect on PNPLA2 expression was only obvious at very late time points (beyond 24 h after start of treatment) when DMSO alone also revealed some gene regulatory effects. Using RNAi, we therefore assessed here the potential autocrine/paracrine role of myocyte ANGPTL4 expression during the moderately delayed stage of LIPE expression (at 20 h of treatment). As depicted in Fig. 4C, transfection with siRNA directed against ANGPTL4, but not with control siRNA directed against bacterial luciferase, reduced the GW501516-induced ANGPTL4 mRNA contents by 90%. The ANGPTL4 knockdown significantly impaired GW501516-mediated LIPE induction (1.6-fold in the presence vs. 2.6-fold in the absence of ANGPTL4 siRNA); in addition, the GW501516 effect was no longer significant (Fig. 4D). Again, control transfection with bacterial luciferase siRNA did not significantly alter the GW501516 effect on LIPE expression (Fig. 4D).

Here, we show that pharmacological PPAR-δ activation upregulates LIPE expression in C2C12 SKM cells and provide preliminary evidence for a role of mANGPTL4 in this gene regulatory event. LIPE induction is supposed to enhance intramyocellular lipolysis and to increase endogenous fatty acyl-CoA, the preferred substrate of SKM oxidative metabolism stimulated by PPAR-δ agonists (32). Moreover, this finding could point to an autocrine/paracrine function of mANGPTL4 in SKM. To corroborate the role of mANGPTL4 in the breakdown of intramyocellular lipids, further studies, e.g., in muscle-specific PPAR-δ gain- and loss-of-function animal models, are clearly needed. Because C2C12 cells do not store measurable amounts of triglycerides (data not shown), other muscle cell models are required to study ANGPTL4's lipolytic effect at the cellular level.

Because our in vitro data demonstrated a close relationship between ANGPTL4 and lipid metabolism, we assessed this protein's role in lipid metabolism, insulin sensitivity, and insulin secretion in humans in vivo. To this end, we quantified plasma ANGPTL4 in 108 thoroughly phenotyped participants of the Tuebingen Lifestyle Intervention Program (TULIP) (33–36), a cohort characterized by a wide range of age, BMI, body fat content, insulin sensitivity, and insulin secretion (clinical characteristics are presented in Supplementary Table 2) using ELISA.

The mean plasma ANGPTL4 concentration measured was 1.73 ± 0.11 ng/ml (means ± SE; range 0.37–8.00 ng/ml). The plasma ANGPTL4 levels were not correlated with sex or age (P = 0.9, both), and this is in good agreement with recently published data (37). There were no significant correlations with BMI; waist-to-hip ratio; body fat measured by bioelectrical impedance; total, visceral, and nonvisceral fat mass measured by magnetic resonance imaging; or plasma levels of adiponectin and leptin (P ≥ 0.06, all, after appropriate adjustments). However, stratification of the cohort into lean (BMI <27 kg/m², 19.5–26.9 kg/m²; n = 30) and obese (BMI ≥30 kg/m², 30.0–48.4 kg/m²; n = 39) subjects revealed significantly elevated ANGPTL4 levels in the obese subgroup (P = 0.0172, t test). Thus, plasma ANGPTL4 is influenced by body adiposity reflecting ANGPTL4 production by WAT (13–15). Ectopic (intrahepatic and intramyocellular) lipids measured by magnetic resonance spectroscopy were neither associated with plasma ANGPTL4 in the overall cohort (P ≥ 0.6, all; adjusted for sex, age, and BMI) nor in the lean and obese subgroups (P ≥ 0.07, all; adjusted for sex and age). Thus, circulating ANGPTL4 does not appear to be involved in lipid breakdown in human muscle. This, however, does not exclude a role of locally produced mANGPTL4 in muscle lipolysis.

As to metabolic traits, plasma ANGPTL4 levels did not reveal significant associations with plasma glucose or insulin concentrations (in the fasting state as well as during OGTT), with OGTT- and hyperinsulinemic-euglycemic clamp–derived indexes of insulin sensitivity, or with indexes of insulin secretion, such as plasma C-peptide levels at 30 min of OGTT and OGTT-derived first-phase insulin secretion in the overall cohort (P ≥ 0.24, all, after appropriate adjustments) or in the lean and obese subgroups (P ≥ 0.3, all, after appropriate adjustments). Plasma triglyceride concentrations were also not significantly associated with ANGPTL4 before (P = 0.14; adjusted for sex, age, and BMI) and after stratification into lean and obese subjects (P ≥ 0.3, all; adjusted for sex and age). Hence, the hypertriglyceridemic action of ANGPTL4 detected in mice due to LPL inhibition (16–20) could not be confirmed in this study. However, a very recent genetic study identified a rare mutation in the ANGPTL4 gene, E40K, in European Americans, which was associated with lower plasma triglyceride levels clearly pointing to a role of ANGPTL4 in VLDL/chylomicron metabolism in humans (38).

Importantly, plasma ANGPTL4 levels were positively associated with plasma NEFAs in the fasting state (Fig. 5A), possibly reflecting NEFA-mediated ANGPTL4 production. In addition, plasma ANGPTL4 was positively correlated with plasma NEFA levels during OGTT (Fig. 5B), an estimate of WAT lipolysis. Even though these associations cannot constitute causality, these results clearly point to a close relationship between plasma NEFA and ANGPTL4 levels in humans and, in part, confirm earlier findings in mice.

The human myotubes used in this study underwent an extended cell culture protocol including isolation from biopsies, expansion, splitting and freezing, storage in liquid nitrogen and thawing, growth to subconfluence, and finally in vitro differentiation. As a result, these cells no longer reflect the metabolic setting in vivo and have lost acquired phenotypes, such as insulin resistance (39). However, some genetically or epigenetically determined features, e.g., susceptibility toward saturated fatty acids (40) and basal expression of genes (41,42), are maintained and show remarkable inter-individual variation. Therefore, human myotubes represent a model to study the unidirectional effects of genes and their individual expression levels on metabolic parameters of the donors, as discussed earlier (42).

To assess the metabolic role of mANGPTL4 in humans in vivo, we determined basal ANGPTL4 mRNA expression in myotubes from 38 nondiabetic donors (for clinical characteristics, see Staiger et al. [41]). From these subjects, plasma glycerol measurements were available in addition to plasma NEFAs. In the myotubes, basal ANGPTL4 expression revealed a strong positive correlation with basal PPARD expression (Fig. 5C) reflecting the close relationship between PPAR-δ and its target gene.

By correlational analysis with the donors’ in vivo parameters, the myotube ANGPTL4 mRNA contents were not associated with sex or age (P ≥ 0.5, both). Furthermore, there were no significant correlations with BMI, waist-to-hip ratio, body fat content, plasma adiponectin and leptin levels, plasma glucose or insulin concentrations (in the fasting state as well as during OGTT), or indexes of insulin sensitivity or insulin secretion (P ≥ 0.10, all, after appropriate adjustments). Intramyocellular lipid measurements were not available from the donors. Therefore, we have currently no proof of our in vitro finding suggesting involvement of mANGPTL4 in muscle triglyceride breakdown. As to lipid parameters, plasma fasting concentrations of triglycerides, NEFA, and glycerol were not significantly associated with myotube ANGPTL4 expression (P ≥ 0.06, all; adjusted for sex, age, and BMI). However, a significant correlation between myotube ANGPTL4 expression and the area under the curve of glycerol during OGTT, an estimate of WAT lipolysis, was detected (Fig. 5D). This not only confirms the results obtained with plasma ANGPTL4 measurements but also indicates that mANGPTL4 is of systemic importance and enhances WAT lipolysis in humans.

To further substantiate the systemic role of mANGPTL4, we additionally measured the plasma ANGPTL4 concentrations of the myotube donors. Importantly, myotube ANGPTL4 expression was significantly correlated with the donors’ plasma ANGPTL4 levels (Fig. 5E). This provides evidence that mANGPTL4 production contributes to circulating ANGPTL4 in humans. The myotube cultures were derived from normal-weight healthy young subjects. Thus, the interesting issue of whether myotubes from patients with metabolic disease (obesity, type 2 diabetes) show altered expression/secretion of ANGPTL4 remains to be clarified, and future studies will shed further light on this question.

In summary, we show here that ANGPTL4 is produced and secreted by human myotubes and is subject to gene regulation by major plasma LCFAs. Furthermore, we provide evidence for an involvement of PPAR-δ in LCFA-induced muscle ANGPTL4 expression. These findings could be of physiological relevance in states of increased β-oxidation due to enhanced muscle PPAR-δ activity and/or PPARD expression, such as fasting (43) and exercise (44–46). In this context, it is conceivable that promotion of WAT lipolysis via PPAR-δ–mediated mANGPTL4 production represents a mechanism that prevents too strong decrements of plasma NEFA levels and, in this way, ensures ongoing fuel supply of SKM (Fig. 6). Such a feed-forward mechanism would favor the efficient use of stored lipids, as opposed to glucose, during periods of increased energy demand. In this scenario, mANGPTL4 would be an important player of crosstalk between SKM and WAT and would explain the loss of WAT mass observed after pharmacological PPAR-δ activation and/or transgenic PPARD overexpression in mice (44,47,48). Clearly, further studies are needed to confirm this hypothesis. If our preliminary in vitro finding that mANGPTL4 is involved in muscle LIPE expression holds also for the in vivo situation, a synergism of PPAR-δ actions can be derived in which PPAR-δ activation, in an ANGPTL4-dependent manner, stimulates breakdown of intramyocellular lipids to fatty acyl-CoA and, in an ANGPTL4-independent manner, promotes acyl-CoA oxidation via induction of β-oxidative enzymes (Fig. 6). Finally, this is to our knowledge the first report demonstrating systemic relevance of mANGPTL4 in humans.

This study was supported by the German Research Foundation (KFO 114/1-2) and the European Community (EUGENE2, LSHM-CT-2004-512013).

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

We thank all study participants for their cooperation and acknowledge the excellent technical assistance of Anna Bury, Heike Luz, and Alke Guirguis.