Fibroblast growth factor 21 participates in adaptation to endoplasmic reticulum stress and attenuates obesity-induced hepatic metabolic stress
Fibroblast growth factor 21 (FGF21) is an endocrine hormone that exhibits anti-diabetic and anti-obesity activity. FGF21 expression is increased in patients with and mouse models of obesity or nonalcoholic fatty liver disease (NAFLD). However, the functional role and molecular mechanism of FGF21 induction in obesity or NAFLD are not clear. As endoplasmic reticulum (ER) stress is triggered in obesity and NAFLD, we investigated whether ER stress affects FGF21 expression or whether FGF21 induction acts as a mechanism of the unfolded protein response (UPR) adaptation to ER stress induced by chemical stressors or obesity.
Hepatocytes or mouse embryonic fibroblasts deficient in UPR signalling pathways and liver-specific eIF2α mutant mice were employed to investigate the in vitro and in vivo effects of ER stress on FGF21 expression, respectively. The in vivo importance of FGF21 induction by ER stress and obesity was determined using inducible Fgf21-transgenic mice and Fgf21-null mice with or without leptin deficiency.
We found that ER stressors induced FGF21 expression, which was dependent on a PKR-like ER kinase–eukaryotic translation factor 2α–activating transcription factor 4 pathway both in vitro and in vivo. Fgf21-null mice exhibited increased expression of ER stress marker genes and augmented hepatic lipid accumulation after tunicamycin treatment. However, these changes were attenuated in inducible Fgf21-transgenic mice. We also observed that Fgf21-null mice with leptin deficiency displayed increased hepatic ER stress response and liver injury, accompanied by deteriorated metabolic variables.
Our results suggest that FGF21 plays an important role in the adaptive response to ER stress- or obesity-induced hepatic metabolic stress.
KeywordsATF4 eIF2α ER stress FGF21 Insulin resistance Lipid accumulation Liver injury Obesity
AMP-activated protein kinase
Activating transcription factor 4
Activating transcription factor 6α
Brown adipose tissue
Binding immunoglobulin protein
CCAAT/enhancer binding protein homologous protein
Eukaryotic translation factor 2α
Fibroblast growth factor 21
Insulin tolerance test
Mouse embryonic fibroblasts
Nonalcoholic fatty liver disease
PKR-like endoplasmic reticulum kinase
Pyruvate tolerance test
Unfolded protein response
White adipose tissue
X-box binding protein 1
Spliced X-box binding protein 1
Fibroblast growth factor 21 (FGF21) is an endocrine hormone produced predominantly in the liver but also in white adipose tissue (WAT), brown adipose tissue (BAT), pancreas and skeletal muscle [1, 2]. In starvation, FGF21 promotes lipolysis, beta oxidation or ketogenesis , indicating that FGF21 is a critical regulator of lipid homeostasis in adaptation to starvation. FGF21 can improve deteriorated metabolic variables in obese diabetic humans and in animal models of obesity and diabetes [4, 5, 6, 7]. In addition, FGF21 plays a protective role in toxin-induced tissue injury [8, 9].
Protein folding occurs in the endoplasmic reticulum (ER). Perturbations of ER homeostasis cause accumulation of misfolded proteins in the ER lumen, triggering the unfolded protein response (UPR), an adaptive programme to resolve misfolded protein accumulation in the ER . The UPR is regulated through three ER transmembrane sensors: inositol-requiring 1α (IRE1α), activating transcription factor 6α (ATF6α) and PKR-like ER kinase (PERK) . X-box binding protein 1 (XBP1) and eukaryotic translation factor 2α (eIF2α)-activating transcription factor 4 (ATF4) participate in various adaptive responses to ER stress downstream of IRE1α and PERK, respectively. Unresolved ER stress leads to the development and progression of various diseases such as obesity and diabetes [11, 12].
Recent studies reported that FGF21 is increased in the liver in mouse models of and humans with obesity or nonalcoholic fatty liver disease (NAFLD) [13, 14, 15]. However, little is known about the functional role and molecular mechanism of FGF21 induction in these conditions. As ER stress is elevated in obesity , we investigated the relationship between ER stress and FGF21 induction. We found that FGF21 is induced by ER stress in a PERK-eIF2α-ATF4-dependent manner. In addition, we observed that FGF21 deletion accelerates ER stress-induced hepatic injury or lipid accumulation, and exacerbates obesity-induced ER stress and metabolic deterioration. These results suggest that FGF21 may play a role in the adaptive response to ER stress induced by a pharmacological ER stressor or obesity.
Fgf21+/+ and Fgf21−/− mice have been described elsewhere . Fgf21+/− mice were crossed with ob/w mice (Jackson Laboratory, Bar Harbor, ME, USA) to generate Fgf21+/+ob/ob and Fgf21−/−ob/ob mice. Mice with a liver-specific defect in eIF2α phosphorylation (Eif2αA/A/fTg/Alfp-Cre) were generated by breeding Eif2αA/A/fTg mice  with Eif2αS/A/Alfp-Cre mice. Liver-specific inducible Fgf21-transgenic (Apoe-rtTA*M2/TetO-Fgf21) mice were generated using a tetracycline-inducible system. Fgf21-null and inducible Fgf21-transgenic mice were maintained in a specific pathogen-free facility of Samsung Biomedical Research Institute. Eif2α mutant mice were maintained under a specific pathogen-free condition in the laboratory animal care facility of the University of Ulsan. All animal experiments were conducted in accordance with the guidelines of the Institutional Animal Care and Use Committee of Samsung Biomedical Research Institute or University of Ulsan. See Electronic Supplementary Material (ESM) Methods for details.
GTT, insulin tolerance test and pyruvate tolerance test
GTTs and pyruvate tolerance tests (PTTs) were performed in overnight-fasted mice with intraperitoneal injection of glucose (1 g/kg) and pyruvate (1 g/kg), respectively. An insulin tolerance test (ITT) was conducted in 6 h-fasted mice with an intraperitoneal injection of insulin (1 U/kg). Blood glucose levels were measured with an Accu-Check glucometer (Roche, Mannheim, Germany).
Ex vivo glucose-stimulated insulin secretion assay
Pancreatic islets were isolated from female Fgf21+/+ob/ob and Fgf21−/−ob/ob mice by collagenase digestion and Biocoll (Biochrom AG, Berlin, Germany) gradient centrifugation. See ESM Methods for details.
Plasmid constructs and reagents
See ESM Methods for details.
Generation of adenovirus
Adenoviruses expressing ATF4 or Flag-XBP1s were generated by homologous recombination between a linearised transfer vector (pAd-Track-ATF4 or pAd-Track-Flag-XBP1s) and an adenoviral backbone vector (pAd-Easy).
FaO, AML12, HepG2 cells, primary mouse hepatocytes and mouse embryonic fibroblasts (MEFs) deficient in UPR genes were maintained at 37°C in a humid atmosphere of 5% CO2. See ESM Methods for details.
See ESM Methods for details.
See ESM Methods for details.
Real-time RT-PCR was conducted using SYBR Green Master Mix (Takara, Otsu, Shiga, Japan) and gene-specific primers (ESM Table 1) in ABI Prism 7000 (Applied Biosystems, Foster City, CA, USA). Relative expression values of specific genes were normalised to L32 mRNA. See ESM Methods for details.
See ESM Methods for details.
Histology and staining analysis
Liver and pancreas tissues were fixed with 10% neutral buffered formalin and 4% paraformaldehyde to make paraffin- and optimal cutting temperature-embedded blocks, respectively. See ESM Methods for details.
Blood chemistry and metabolite analysis
See ESM Methods for details.
All values are expressed as mean ± SEM. Statistical significance was tested with the unpaired two-tailed Student's t test using GraphPad Prism Version 5.02 Software (La Jolla, CA, USA). A p value of less than 0.05 was considered significant.
ER stress induces FGF21 expression through PERK–eIF2α–ATF4 pathway in vitro
We next investigated the mechanisms by which ER stress increases FGF21 expression using MEFs deficient in ER stress sensors. Fgf21 induction by ER stressors was much weaker in Perk−/− MEFs compared with control MEFs (Fig. 1e), while ER stressor-induced Fgf21 expression in Ire1α−/−, Atf6α−/− or Xbp1−/− MEFs was comparable with that in control MEFs (ESM Fig. 1a–c). Furthermore, adenovirus-mediated overexpression of XBP1s had no effect on FGF21 expression (ESM Fig. 1d). We next tested whether eIF2α and ATF4 downstream of PERK induce FGF21 expression. As expected, two ER stressors increased Fgf21 mRNA levels in Eif2αS/S (wild-type) and Atf4+/+ MEFs, while Fgf21 induction was markedly attenuated in Eif2αA/A MEFs (harbouring S51A mutation) and Atf4−/− MEFs (Fig. 1f, g). ATF4 knockdown consistently suppressed tunicamycin-induced Fgf21 expression (Fig. 1h, i), while ATF4 overexpression increased FGF21 mRNA expression and Fgf21 promoter activity (Fig. 1j, k). However, XBP1s or ATF6α overexpression had no effect on luciferase activity (Fig. 1k). Moreover, mutations of putative ATF4-response elements abolished the increase in Fgf21 promoter activity by ER stressors (Fig. 1l), indicating that two ATF4-response elements are required for ER stress-induced FGF21 expression. Together, these findings suggest that the PERK–eIF2α–ATF4 pathway is critical for ER stress-induced FGF21 expression in vitro.
Hepatic eIF2α–ATF4 pathway is required for ER stress-induced FGF21 expression in vivo
To study the importance of the eIF2α–ATF4 axis in FGF21 induction by ER stress in vivo, we generated liver-specific Eif2αA/A mice (Eif2αA/A/fTg/Alfp-Cre) with a homozygous Ser51Ala mutation in Eif2α alleles of the liver (Fig. 2d). Phosphorylation of eIF2α in the liver was not observed in Eif2αA/A/fTg/Alfp-Cre mice (Fig. 2e), confirming a defect in hepatic eIF2α phosphorylation in these mice. Importantly, we observed that hepatic expression of FGF21 or ATF4 and serum FGF21 level were markedly decreased in liver-specific Eif2αA/A mice compared with control heterozygous Eif2αS/A/fTg mice under ER stress condition (Fig. 2e–g). These data indicate that the eIF2α–ATF4 pathway is critical for FGF21 induction in response to ER stress in vivo.
FGF21 suppresses ER stress response and alleviates ER stress-induced hepatic injury and lipid accumulation
As ER stress can induce hepatic lipid accumulation , we next studied the physiological role of FGF21 in ER stress-mediated hepatosteatosis. Importantly, Fgf21−/− mice exhibited elevated hepatic triacylglycerol (TG) accumulation compared with Fgf21+/+ mice after tunicamycin administration (Fig. 3d, e). When we analysed changes in the expression of genes involved in lipid metabolism, the expression of fatty acid and TG synthesis genes such as Scd1, Adrp (also known as Plin2) and Dgat2 appeared to be increased in the liver of Fgf21−/− mice compared with that of Fgf21+/+ mice under ER stress condition (Fig. 3a). By contrast, we did not observe differences in the expression of genes such as Pparγ, Srebp1c (also known as Srebf1) and Acc1 (Fig. 3a). The expression of Cd36, which is involved in lipid uptake, was also significantly increased in the liver of Fgf21−/− mice compared with Fgf21+/+ mice (Fig. 3a). Additionally, we observed a tendency towards increased expression of the genes involved in beta oxidation, including Pparα and Mcad (also known as Acadm) in the liver of Fgf21−/− mice compared with control mice after tunicamycin administration (ESM Fig. 3b), which is likely to be an adaptive response to excessive TG accumulation caused by FGF21 deletion. These results indicate that enhanced hepatic lipid synthesis and uptake may contribute to increased hepatic TG accumulation in Fgf21−/− mice under ER stress condition.
We next investigated the effect of exogenous FGF21 on tunicamycin-induced ER stress and hepatic lipid accumulation using liver-specific inducible Fgf21-transgenic mice. As expected, inducible Fgf21-transgenic mice had increased serum FGF21 levels and elevated hepatic Fgf21 gene expression when fed chow diet containing doxycycline (Fig. 3f, ESM Fig. 4). Importantly, these mice showed attenuated expression of UPR genes and decreased hepatic lipid accumulation after tunicamycin administration (Fig. 3g, h). Together, these findings suggest that FGF21 expression is an adaptive response to ER stress to alleviate excessive liver injury and lipid accumulation.
Obesity-mediated FGF21 induction alleviates lipid-induced ER stress, liver injury and metabolic deterioration
Given the role of ER stress in obesity-related metabolic diseases , we next studied the role of FGF21 on obesity-mediated metabolic derangement. Notably, both male and female Fgf21−/−ob/ob mice displayed increased random-fed or fasting glucose levels and worsened glucose tolerance/insulin resistance compared with their respective control Fgf21+/+ob/ob mice without changes in body weight (Fig. 4g–j, ESM Fig. 5a–c), although the differences in fasting glucose level in female mice were marginal. However, we did not observe any difference in metabolic profile between nonobese Fgf21+/+ and Fgf21−/− mice of either sex (ESM Fig. 6), suggesting that FGF21 deletion exacerbates metabolic variables only in the metabolically stressed condition.
Despite deteriorated glucose tolerance and insulin resistance, serum insulin level, beta cell mass, beta cell apoptosis/proliferation and glucose-stimulated insulin secretion from islets were not different between Fgf21−/−ob/ob and Fgf21+/+ob/ob mice (Fig. 4k–m, ESM Fig. 5d–f), suggesting that beta cell dysfunction is not involved in metabolic deterioration of Fgf21−/−ob/ob mice. As ER stress contributes to the increased gluconeogenesis in obesity and diabetes , we conducted a PTT in Fgf21−/−ob/ob mice. As expected, both male and female Fgf21−/−ob/ob mice exhibited impaired pyruvate tolerance compared with Fgf21+/+ob/ob mice, accompanied by elevated expression of gluconeogenesis-related genes. However, differences in the PTTs of male mice were marginal (Fig. 4n–p), implying that deteriorated glucose tolerance is probably due to increased hepatic glucose production. Nonetheless, hepatic lipid content and serum levels of metabolites, such as TG, cholesterol and NEFA, were not different between the two groups in both sexes (ESM Fig. 7a–d). Together, these data suggest that obesity-mediated FGF21 induction plays a protective role in obesity-induced ER stress and metabolic deterioration.
Recent studies showed that FGF21 expression is increased in the liver of patients with and mouse models of obesity and NAFLD [13, 14, 15]. However, the molecular mechanism of FGF21 induction and its pathophysiological function in these conditions is unclear. Here, we demonstrated that a PERK–eIF2α–ATF4 pathway is critical for ER stress-induced FGF21 expression in vitro and in vivo. Additionally, we showed a protective role of FGF21 in chemical- or obesity-induced hepatic ER stress.
It has been reported that FGF21 is induced by various stresses, including glucose starvation, cold and autophagy deficiency, which may play an important role in adaptation to these stresses [3, 20, 21]. Additionally, a recent paper has reported that FGF21 levels are increased in hepatocytes and in mouse liver after treatment with an ER stressor . However, the physiological or pathological significance of ER stress-mediated FGF21 induction is unknown. Furthermore, molecular mechanisms of in vitro FGF21 induction by ER stress are unclear, although the effects of ATF4 overexpression on FGF21 mRNA expression and FGF21 promoter activity have been evaluated [22, 23]. Here, we demonstrated the importance of a PERK–eIF2α–ATF4 pathway in ER stress-induced Fgf21 expression in vivo and in vitro, and the physiological role of endogenous FGF21 induction caused by ER stress. In addition to ATF4, CHOP, the downstream target of ATF4, has been shown to be partially involved in ER stress-induced Fgf21 expression in primary hepatocytes . However, we did not observe differences in Fgf21 expression between Chop+/+ and Chop−/− MEFs under ER stress conditions (ESM Fig. 8). During the course of the revision of our paper, the IRE1α–XBP1 pathway has been reported to mediate ER stress-induced FGF21 expression , which is in contrast to our findings showing no significant role of IRE1α and XBP1 in FGF21 induction by ER stress. There is no clear explanation regarding these discrepancies, but they might be attributed to the differences in cell lines (hepatocytes vs MEFs) or experimental procedures.
Fgf21−/− mice exhibited an aggravated ER stress response and hepatic cell death compared with Fgf21+/+ mice after tunicamycin administration, while inducible Fgf21-transgenic mice had an attenuated ER stress response, indicating that FGF21 plays a protective role against ER stress. However, we did not further address the mechanisms by which FGF21 alleviates ER stress or ER stress-induced liver injury in the present study. Several possible mechanisms may be proposed. First, downstream target proteins of FGF21 may participate in FGF21-mediated attenuation of ER stress. It has been reported that FGF21 activates the AMP-activated protein kinase (AMPK)–Sirtuin1 (SIRT1) pathway , and AMPK or SIRT1 has a protective role against ER stress [26, 27]. We, therefore, hypothesise that AMPK or SIRT1 may mediate the effects of FGF21 on ER stress or hepatic injury due to ER stress. Another possibility is that reduced ER stress is a result of FGF21-induced suppression of lipid accumulation, since excessive lipid overloading can cause ER stress in the liver . Given the protective role of FGF21 against oxidative stress in cardiac damage , reduced oxidative stress may be a mechanism by which FGF21 alleviates ER stress-induced hepatic injury. Intriguingly, FGF21 has been reported to exert metabolic effects via its actions in other target organs such as brain and adipose tissue [30, 31, 32]. In particular, increased adiponectin levels derived from adipose tissue contribute to the metabolic improvement elicited by FGF21 [31, 32]. However, we did not observe changes of circulating adiponectin levels or adipose tissue Adipoq mRNA level in tunicamycin-treated mice (ESM Fig. 9a, b). In addition, serum adiponectin level in Fgf21−/− mice was not different compared with Fgf21+/+ mice after tunicamycin administration (ESM Fig. 9c). Thus, we could exclude the contribution of adiponectin in the adaptive effect of FGF21 in response to ER stress. However, we still cannot eliminate the possibility that other changes in non-hepatic tissues contribute to the amelioration of hepatic ER stress by FGF21. Further studies will be necessary to elucidate the mechanisms underlying the protective role of FGF21 against ER stress.
As ER stress contributes to obesity-associated metabolic disease  and FGF21 expression is increased in patients with and mouse models of obesity [13, 14, 15], we studied the pathophysiological function of FGF21 in ER stress and metabolic alterations caused by obesity using ob mice. While the metabolic effects of FGF21 deletion on diet-induced obesity models were variable [33, 34], FGF21 deletion exacerbated hepatic ER stress and worsened glucose tolerance/insulin resistance in ob mice. However, alterations of these metabolic variables were not associated with beta cell failure. Our findings are inconsistent with previous reports showing beneficial effects of FGF21 on beta cell function and impairment of FGF21 signalling in islets of obese mice [35, 36]. This discrepancy is probably due to differences between exogenous FGF21 administration vs endogenous FGF21. Although the role of FGF21 in the regulation of gluconeogenesis is controversial , we observed elevated gluconeogenesis in Fgf21−/−ob/ob mice, which is consistent with previous reports showing suppressive effects of FGF21 on gluconeogenesis in obese conditions . Together, our results suggest that endogenous FGF21 is important in alleviating ER stress and metabolic deterioration due to obesity.
In conclusion, our results indicate that FGF21 is induced in response to ER stress through the eIF2α–ATF4 axis, which serves as a compensatory mechanism to attenuate ER stress-induced liver injury and hepatic lipid accumulation. We also showed that FGF21 induction plays a protective role in obesity-related ER stress and metabolic deterioration. Thus, our findings provide new insights into the role of FGF21 in ER stress response and suggest an innovative therapeutic strategy for the treatment of ER stress-associated diseases such as NAFLD, obesity and diabetes.
We thank D. Ron (University of Cambridge, Cambridge, UK) for Atf4 and Perk MEFs, R. J. Kaufman (Sanford Burnham Medical Research Institute, La Jolla, CA, USA) for Ire1α, Atf6α, Chop, Eif2α and Xbp1 MEFs and L. H. Glimcher (Weill Cornell Medical College, New York, NY, USA) for Xbp1 MEFs.
This work was supported by the Basic Science Research Program through the National Research Foundation of Korea funded by the Ministry of Education (2013R1A6A3A04065825 to KHK and 2014R1A6A1030318 to SHB) and the Global Research Laboratory Grant of the National Research Foundation of Korea (K21004000003-10A0500-00310 to M-SL). M-SL received the National Research Foundation of Korea grant funded by the Korea government (MSIP) (2014M3A9D8034459).
Duality of interest
The authors declare that there is no duality of interest associated with this manuscript.
KHK and M-SL are the guarantors of the work. SHK, KHK and M-SL conceived and designed the experiments. SHK, KHK, H-KK, M-JK, SHB, MK and NI contributed to the acquisition of data. SHK, KHK and M-SL analysed and interpreted the data. SHK, KHK, SHB and M-SL wrote the manuscript. All authors have revised the manuscript critically for important intellectual content and approved the final version to be published.
- 8.Ye D, Wang Y, Li H, et al (2014) FGF21 protects against acetaminophen-induced hepatotoxicity by potentiating PGC-1alpha-mediated antioxidant capacity in mice. Hepatology 60:977–989Google Scholar
- 34.Adams AC, Coskun T, Cheng CC, LS OF, Dubois SL, Kharitonenkov A (2013) Fibroblast growth factor 21 is not required for the antidiabetic actions of the thiazoladinediones. Mol Metab 2:205-214Google Scholar