Activation of PPARα Ameliorates Hepatic Insulin Resistance and Steatosis in High Fructose–Fed Mice Despite Increased Endoplasmic Reticulum Stress

Male C57BL/6J mice (aged 14 weeks) from the Animal Resources Centre (Perth, Australia) were kept at 22 ± 1°C on a 12-h light/dark cycle. After 1 week of acclimatization, mice were fed ad libitum for up to 15 days with either a chow (CH) or an HFru diet. The CH diet comprised 70% calories from starch, ∼10% calories from fat, and ∼20% calories from protein (Specialty Feeds, Glen Forrest, Western Australia), and the HFru diet comprised 35% fructose, 35% starch, ∼10% fat, and ∼17% protein. The detailed recipe for the HFru diet is described in our previous studies (16,21). FB (Sigma-Aldrich, Castle Hill, New South Wales, Australia) was provided as a supplement by mixing it into the diets at a concentration of 100 mg/kg/day. All experiments were approved by the Animal Ethics Committees of RMIT University (#1012).

Body weight and food intake were measured daily. The whole-body metabolic rate was measured at 22°C using an indirect calorimeter (Comprehensive Laboratory Animal Monitoring System; Columbus Instruments, Columbus, OH) as described previously (22) between 5 and 8 days after the administration of FB. Mice were fasted for 5–7 h before being killed; tissues of interest were collected and freeze-clamped immediately. Epididymal fat mass was weighed using an analytical balance. Liver triglycerides were extracted by the method of Folch and determined by a Peridochrom Triglyceride GPO-PAP kit (Roche Diagnostics Australia) as previously described (16). In separate experiments, glucose tolerance tests (GTTs) (glucose 2.5 g/kg i.p.) were performed on the 5–7-h fasted mice using a glucometer (Accu-Chek Performa Nano; Roche, Victoria, Australia), and blood samples were collected at 0, 30, and 60 min for plasma insulin measurement. For the assessment of insulin signaling in the liver, the mice were injected with insulin 2 U/kg i.p. 20 min before tissue collection.

FA oxidation was measured in liver homogenates using methods described previously (23). Briefly, the liver homogenate (50 μL) was incubated at 30°C for 90 min in a reaction mixture (pH 7.4) containing (in mmol/L) 0.2 [¹⁴C]palmitate (0.5 μCi); 2 L-carnitine; and 0.05 CoA ± 0.02 etomoxir, a specific inhibitor of FA oxidation in mitochondria (24). The reaction was stopped by ice-cold 1 mol/L perchloric acid. CO₂ produced from the reaction was captured in 1 mol/L NaOH. ¹⁴C counts in the acid-soluble fraction were combined with the CO₂ values to give the total palmitate oxidation rate. Hepatic DNL was determined by measuring the incorporation of [³H]H₂O into triglyceride (TG) in the liver as described previously (16,25).

Approximately 20 mg of frozen liver tissue was homogenized in 175 mmol/L KCl and 1.98 mmol/L EDTA-containing buffer (pH 7.4) with a glass homogenizer before being subjected to three freeze-thaw cycles. Citrate synthase and β3-hydroxyacyl-CoA dehydrogenase (β-HAD) activities were determined as described previously (16,21) with a Flexstation 3 plate reader (Molecular Devices, Sunnyvale, CA).

Liver and red quadriceps muscle samples were homogenized in ice-cold lysis buffer (pH 7.5) containing (in mmol/L) 50 Tris, 150 NaCl, 1% Triton X-100, 10 NaP, 100 NaF, 2 Na₃VO₄, 1 EDTA, 1 EGTA, and 10% glycerol supplemented with protease inhibitor cocktail and phosphates inhibitor cocktail (Sigma Aldrich) and DL-dithiothreitol. Protein samples were then denatured in SDS sample buffer (125 mmol/L Tris-HCl, pH 6.8, 50% glycerol, 2% SDS, 5% β-mercaptoethanol, and 0.01% bromophenol blue). The insulin signal transduction was assessed by total and phosphorylated (Ser473) Akt and total and phosphorylated (Ser219) glycogen synthase kinase 3β (GSK-3β) (Cell Signaling Technology). Key lipogenic enzymes were determined by Western blotting with specific antibodies, including sterol regulatory element-binding protein (SREBP1c) (Santa Cruz Biotechnology), phosphorylated and total acetyl-CoA carboxylase (ACC) (Cell Signaling), fatty acid synthase (FAS) (Cell Signaling), and stearoyl-CoA desaturase 1 (SCD-1) (Cell Signaling). Oxidative capacity included acyl-CoA oxidase 1 (ACOX1) (Santa Cruz). ER stress included phosphorylated (Thr980) PERK (p-PERK) (Cell Signaling), total and phosphorylated (Ser51) eIF2α (Cell Signaling), growth arrest and DNA damage-inducible protein 34 (GADD34) (Cell Signaling), CCAAT/enhancer-binding protein (C/EBP) homologous protein (CHOP) (Santa Cruz), phosphorylated (Ser724) IRE1 (Abcam), XBP1 (Santa Cruz), and activating transcription factor 6 α (ATF6α) (Santa Cruz). Serine/threonine kinases included phosphorylated (Ser176/177) IKKα/IKKβ (Cell Signaling), IκBα, and phosphorylated (Thr183/Tyr185) JNK (Cell Signaling). Immunolabeled bands were quantified by densitometry, and representative blots are shown.

DAG and ceramide were extracted and quantified according to the methods of Preiss et al. (26). Lipids were extracted from liver homogenates using chloroform:methanol:PBS + 0.2% SDS (1:2:0.8). DAG kinase and ³²P-labeled ATP (0.55 GBq/mmol cold ATP) were added to the lysates preincubated with cardiolipin/octylglucoside, and the reaction was stopped after 2.5 h by the addition of chloroform:methanol (2:1). Samples were spotted onto thin-layer chromatography plates and developed. ³²P-labeled phosphatidic acid and ceramide-1-phosphate were identified by autoradiography, dried, scraped from the thin-layer chromatography plate, and counted in a liquid scintillation analyzer (LS6500; Beckman Coulter Inc.).

Data are presented as mean ± SE. One-way ANOVA was used for comparison of the relevant groups. When significant differences were found, the Turkey-Kramer multiple comparisons test was applied. Differences at P < 0.05 were considered to be statistically significant.

HFru feeding resulted in significant increases in calorie intake (∼38%), whole-body Vo₂ (∼8%) and respiratory exchange ratio (RER), body weight gain (1.2 g), and adiposity (67% in epididymal fat mass, P < 0.01) compared with untreated CH-fed mice (CH-Veh) (Table 1). In CH-fed mice, FB treatment had no significant effects on body weight gain, adiposity, caloric intake, or RER except for a 14% increase in Vo₂ (P < 0.05 vs. CH-Veh). In HFru-fed mice, FB significantly (P < 0.05 for all) increased Vo₂ (8%), reduced the RER, and completely diminished HFru-induced body weight gain and adiposity. FB lowered blood glucose and insulin levels in the HFru-fed mice (both P < 0.01 HFru-FB) and, hence, led to an improved homeostasis model assessment insulin resistance (HOMA-IR) index. The untreated HFru-fed mice (HFru-Veh) showed glucose intolerance compared with CH-Veh (Fig. 1A and B). FB treatment completely normalized the glucose tolerance seen in the HFru-fed mice to the levels of the CH-fed mice and reduced the requirement for plasma insulin level (Fig. 1C).

In skeletal muscle, the insulin-stimulated phosphorylation of Akt (all P < 0.001 vs. corresponding basal) was unaltered by diet or FB treatment (Supplementary Fig. 1). In contrast, HFru feeding blunted the insulin-stimulated phosphorylation of Akt (by 53%, P < 0.01 vs. CH) and its downstream target GSK3β (by 60%, P < 0.001 vs. CH) in the liver, which were fully restored by the treatment of FB (Fig. 2A and B), indicating that HFru feeding results in impairment of hepatic insulin signaling and that treatment with FB is effective in restoring hepatic insulin sensitivity.

Lipid accumulation in the liver (i.e., hepatic steatosis) is believed to be closely linked to insulin resistance (1); hence, we examined the effects of FB treatment on hepatic lipid content. As expected, the HFru feeding induced a marked increase in hepatic TG levels (2.7-fold, P < 0.01 vs. CH), which was ameliorated by treatment with FB (Fig. 3A), although the fasting plasma TG levels were unaffected between the HFru-Veh and HFru-FB groups (Table 1). In agreement with an increased TG level, hepatic DAG content of the HFru-fed mice was also elevated by 53% (P < 0.05 vs. CH), which was normalized by treatment with FB (Fig. 3B). The total content of ceramide was attenuated by 57% (P < 0.001 vs. CH) by HFru feeding but was restored (P n.s. vs. CH-Veh and CH-FB) by treatment with FB (Fig. 3C). These data suggest that the implication of DAG in the apparent hepatic insulin resistance resulted from HFru feeding.

Because enhanced FA oxidation is one of the key events resulting from the activation of PPARα by FB in the liver (27–29), we measured molecular markers of oxidative capacity in the livers of mice. The expression of peroxisomal ACOX1, a direct downstream effector of PPARα activation that catalyses the first step of peroxisomal β-oxidation of FAs (30), was markedly upregulated in response to treatment with FB (Fig. 4A). Moreover, the phosphorylation of ACC, which regulates the mitochondrial β-oxidation of FAs, was markedly elevated (eightfold, P < 0.001 vs. CH-Veh and HFru-Veh) in response to FB treatment in the livers of HFru-fed mice (Fig. 4B). In line with an increased oxidative capacity, the activity of β-HAD, which catalyses the third step of mitochondrial β-oxidation, was augmented by 2.4-fold (P < 0.01 vs. CH-Veh and HFru-Veh) with FB treatment in the HFru-fed mice (Fig. 4C). The activity of citrate synthase was significantly enhanced (by 19%, P < 0.01 vs. CH) with HFru feeding independently of PPARα activation (Fig. 4D), indicating that PPARα activation specifically enhances the oxidative capacity of the liver without affecting mitochondrial content under HFru feeding conditions. Hepatic FA oxidation was increased (∼60%) by treatment with FB in the HFru-fed mice, and this resulted in an increase in the component resistant to the inhibition by etomoxir (Fig. 4E).

Having established that treatment with FB was effective in eliminating hepatic lipid accumulation and restoring insulin signaling, we next examined the effects of FB on the three major UPR pathways. The phosphorylation of IRE1 (Fig. 5A), spliced form of XBP1 (sXBP1) (Fig. 5B), phosphorylation of PERK and eIF2α (Fig. 5C and D), and expression of CHOP (Fig. 5E) were markedly enhanced by PPARα activation, regardless of the feeding conditions. In addition, the expression of GADD34, a well-characterized phosphatase of eIF2α (9), was concomitantly downregulated in response to PPARα activation (Fig. 5F). As expected, HFru feeding significantly increased the phosphorylated form of IRE1 (P < 0.05 vs. CH) (Fig. 5A) and eIF2α (P < 0.01 vs. CH) (Fig. 5D). No changes were detected in the maturation of ATF6 as a result of HFru feeding or FB treatment (Supplementary Fig. 2).

Because both activation of PPARα (28) and UPR signaling (7) can promote DNL in the liver through the action of SREBP1c (31), we examined the expression of SREBP1c and key enzymes involve in this process. The Western blot analysis revealed upregulations of the mature form of SREBP1c (mSREBP1c) (threefold, P < 0.05 vs. CH-Veh), ACC (3.5-fold, P < 0.01 vs. CH-Veh), FAS (2.5-fold, P < 0.001 vs. CH), and SCD1 (14.5-fold, P < 0.01 vs. CH-Veh) in the liver of the HFru-fed mice (Fig. 6A–D). PPARα activation in the CH-fed mice stimulated the expression of mSREBP1c, ACC, FAS, and SCD1 to levels comparable with that of the HFru-fed mice. PPARα activation in conjunction with HFru feeding elicited a further increase in the expression of mSREBP1c (sixfold) and SCD1 (38-fold, both P < 0.001 vs. HFru-Veh) but not ACC (2.8-fold, P < 0.01 vs. CH) or FAS (2.9-fold, P < 0.001 vs. CH, both not different vs. HFru-Veh). In keeping with the upregulated lipogenic enzymes, hepatic DNL was significantly increased (∼38%) by PPARα activation in CH-fed mice, and this increase was maintained in HFru-FB mice (Fig. 6E). These data suggested the FB-induced UPR signaling may enhance the lipogenic capacity of liver independent of the effects of dietary fructose.

Production of deleterious lipids (lipotoxicity) through DNL and activation of serine/threonine kinases are key consequences of UPR signaling, which interferes with insulin signal transduction at various points (15). As shown in Fig. 3B, treatment with FB can correct the elevated DAG content induced by HFru feeding. Meanwhile, activation of JNK and IKK are well-demonstrated consequences of UPR signaling, resulting in the impairment of insulin signal transduction (3). HFru feeding did not result in a significant induction of JNK (Fig. 7A) or IKK (Fig. 7B), and the expression of IκBα (Fig. 7C), the downstream target of IKK (32), remained unaffected, which is consistent with our previous observation (16). Despite the significant induction of the two specific arms of UPR pathways, the phosphorylation status of these kinases remained unaffected in response to FB treatment (Fig. 7A–C). These data suggest that PPARα activation is effective in eliminating lipotoxicity and that the FB-induced UPR signaling did not result in the activation of these stress kinases.

The current study has established a hepatic ER stress model independent of lipid accumulation in the liver with the use of the PPARα activator FB in HFru-fed mice. This model enabled us to examine the effects of ER stress on hepatic insulin sensitivity devoid of the influence of hepatic steatosis. The data show that PPARα activation completely eliminated HFru-induced hepatic steatosis and insulin resistance without altering JNK and IKK in the face of marked dual activation of the IRE1/XBP1 and PERK/eIF2α branches of the UPR pathways. These findings indicate that hepatic steatosis, but not JNK, is required for ER stress to cause insulin resistance. To the best of our knowledge, this report is the first to demonstrate that PPARα activation induces UPR signaling while ameliorating hepatic insulin resistance.

PPARα is a key transcriptional regulator for lipid metabolism, and it can be endogenously activated by FAs as well as pharmacologically activated by agonists like FB (17). FB is a specific agonist of PPARα commonly used to treat dyslipidemia and hypercholesterolemia in humans (33). These beneficial effects are attributed to the PPARα-driven peroxisomal and mitochondrial β-oxidation and microsomal ω-oxidation of FA, with the liver being a major site of action (34). The data show that FB treatment was effective in activating PPARα in vivo as evidenced by the increased expression of ACOX1 (also known as palmitoyl-CoA oxidase) (35), which is a direct target of PPARα. The concomitant increase in Vo₂ and the induction of the phosphorylated form of ACC and β-HAD activity along with the augmented FA oxidation in the liver are indicative of an enhanced oxidative capacity and energy expenditure, which are consistent with the reported effects of PPARα activation (34,36). In line with the upregulation of ACOX1 expression, FB-induced increase in hepatic FA oxidation can be attributed to the enhanced peroxisomal oxidation that is not inhabitable by etomoxir, which blocks the entry of long-chain FAs into mitochondria for oxidation. It has been reported that peroxisomal oxidation breakdown of (very) long-chain FAs into medium- and short-chain FAs further oxidation in mitochondria (36). Unlike long-chain FAs, the short- and medium-chain FAs do not rely on carnitine palmitoyltransferase 1 to enter the mitochondria (36), which may explain, at least in part, the increased oxygen consumption observed at the whole-body level. It is likely that the reduced body weight and adiposity observed only in HFru-FB mice is a result of enhanced peroxisomal oxidation, which was not evident in CH-FB mice.

The restored HOMA-IR resulting from the lowered fasting blood glucose and insulin levels together with the restored hepatic insulin signal transduction in the HFru-fed mice by FB suggest improved insulin sensitivity in these mice. This interpretation is also supported by the striking reduction in insulin secretion in CH-FB mice while maintaining unaltered glucose clearance. The insulin-sensitizing effect of FB observed in the current study is consistent with our previous report of the insulin-sensitizing effect of PPARα activation in high fat–fed insulin resistant rats as determined by the hyperinsulinemic-euglycemic clamp (37).

The ER plays a pivotal role in protein processing to maintain cellular homeostasis under physiological conditions through the three canonical branches of UPR signaling pathways: PERK/elF2α, IRE1/XBP1, and ATF6. The initialing proteins PERK, IRE, and ATF6 all have sensors facing the ER lumen, and they can be activated under ER stress, such as the accumulation of misfolded proteins (9,15). The activated IRE1/XBP1 pathway has been suggested to promote DNL in the liver, leading to the production of lipids (4,7). In addition, both IRE1 (10) and PERK (38) have been suggested to activate JNK and IKK. These mechanisms, acting in concert or alone, are sufficient to impair insulin signaling in the liver. However, it has been difficult to separate the effect of activated UPR pathways on hepatic insulin signaling in vivo from the influence of lipid accumulation. Moreover, FB is a lipid-lowering drug commonly used in humans, and fructose consumption is closely related to the epidemic of obesity and fatty liver (6,39). Thus, the approach of administration of FB to insulin resistant mice induced by HFru feeding would not only allow us to dissect such an integral relationship, but also to provide insight into the mechanism relevant to human conditions.

Both DAG and ceramide are key lipid intermediaries that link hepatic steatosis to insulin resistance (4). The results show that liver DAG content was higher in HFru-fed mice (as a result of increased DNL), and this is consistent with a previous report in HFru-fed mice (40). DAG can activate protein kinases C ε, which in turn phosphorylates IRS1 at serine 307 to disrupt tyrosine phosphorylation of IRS1 (41). This could blunt IRS-mediated phosphorylation of its downstream signaling target, such as Akt (42). While maintaining elevated DNL (as indicated by mSREBP1c, ACC, FAS, and SCD1 and the incorporation of [³H]H₂O into TG) induced by an HFru diet, FB outpaced DNL by a much greater effect to accelerate FA oxidation (as indicated by the 14–16% increase in Vo₂ and 3.9-fold, 8-fold, and 2.1-fold increases in ACOX1, p-ACC, and β-HAD, respectively), hence eliminating the accumulation of TG and DAG. This may offer an explanation for the improved insulin signaling by FB. In HFru-fed rat, hepatic ceramide has been reported to be increased (43), and this lipid metabolite can suppress the phosphorylation of Akt through protein phosphatase 2A (44,45). In the current study, ceramide is unlikely to be a contributor for the blunted insulin signaling because its level in HFru-FB mice was similar to the level in CH-fed mice. However, the precise role of ceramide requires further investigation because the cellular location may be a key determinant of its effect on insulin sensitivity (46).

The other key mechanism for ER stress-induced insulin resistance is the activation of JNK and associated stress kinases. Sustained ER stress has been shown to cause hepatic insulin resistance through the induction of JNK and IKK (8,13), and all three canonical arms of the UPR pathways are capable of activating JNK and IKK signaling under conditions of severe ER stress (3). Consistent with our previous finding (16), HFru feeding was accompanied by the presence of ER stress. Despite further activation of the IRE1/XBP1 and PERK/eIF2α signaling by the activation of PPARα with FB, the unaltered phosphorylation of JNK and IKK or IκBα content argues against their role in the improved insulin-signaling properties in the liver. In addition, cellular ceramide is known to be implicated in the upregulation of IKK and JNK (44). The fact that neither ceramide nor these stress-related kinases were upregulated by FB is also consistent with our interpretation of the reduction in DAGs as a more likely mechanism for the alleviation of hepatic insulin resistance by the activation of PPARα. Of interest, Jurczak et al. (40) recently demonstrated alleviation of hepatic DAG accumulation in mice with conditional knockout of XBP1. The absence of XBP1 can reverse fructose-induced insulin resistance despite the presence of ER stress and JNK activation, which supports the notion of DAGs being the major culprit for hepatic insulin resistance induced by HFru feeding.

It has been suggested that mild ER stress may enhance hepatic insulin signaling and protect against lipotoxicity through the induction of an adaptive UPR (47). Mice carrying liver-specific deletion of IRE1 showed overt steatosis when challenged with ER stress inducers (48), whereas genetic ablation of either ER stress-sensing or ER quality-control molecules also resulted in the development of hepatic steatosis (49). Furthermore, IRE1 has been reported to be capable of repressing the expression of key metabolic transcriptional regulators, including C/EBPβ, C/EBPδ, PPARγ, and enzymes involved in TG biosynthesis (49), suggesting that UPR might be an important mechanism for mitigating steatosis. The current study results highlight the need for further investigation to examine whether specific UPR signaling may in fact contribute to the PPARα-mediated effects on insulin sensitivity.

Although attenuated body weight gain in the HFru-fed mice induced by FB may cloud our interpretation at first glance, the pivotal role of lipids (but not the activated UPR per se) in ER stress–associated insulin resistance is also demonstrated in CH-FB mice without body weight change (compared with CH-fed mice). Despite similar dual activation of both IRE1/XBP1 and PERK/eIF2α pathways in the CH-FB mice, insulin-mediated phosphorylation of Akt and GSK3β remained intact in the absence of lipid accumulation. This interpretation is consistent with a previous study that showed reduced liver lipids as the underlying mechanism of improved hepatic insulin sensitivity during body weight loss in patients with type 2 diabetes (50).

In summary, the data indicate that lipid (particularly DAG) accumulation, but not the activation of JNK or IKK, is required for ER stress to cause hepatic insulin resistance and glucose intolerance during HFru consumption. Increased peroxisomal oxidation of FAs and energy expenditure are likely to underpin the observed reduction in hepatic steatosis and insulin resistance in HFru-FB mice despite marked increases in UPR signaling and DNL. Therefore, activation of PPARα with FB ameliorates HFru-induced hepatic insulin resistance by eliminating lipid deposition by blocking its link with the ER (Fig. 8). The findings also suggest a need for further investigation into whether activation of specific UPR pathways may in fact contribute to the therapeutic effects of fibrate drugs, which are commonly used in humans.

This study was supported by the National Health and Medical Research Council of Australia Program Grant 535921 (allocation to J.-M.Y.) and Australian Research Council (DP 110102396 to J.-M.Y.). M.J.W. is a Senior Research Fellow of National Health and Medical Research Council.

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

S.M.H.C. designed the study, contributed to the research data, and wrote the manuscript. R.-Q.S. contributed to the research data and wrote the manuscript. X.-Y.Z., Z.-H.C., and H.W. contributed to the research data. M.J.W. contributed to the research data and provided reagents, materials, and analysis tools. J.-M.Y. conceived and designed the study, provided reagents, materials, and analysis tools, and wrote the manuscript. J.-M.Y. 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 accuracy of the data analysis.

The authors thank Dr. Juan Carlos Molero for his comments and Xiu Zhou and Songpei Li (RMIT University) for their technical assistance.