Nutrient deprivation or starvation frequently correlates with amino acid limitation. Amino acid starvation initiates a signal transduction cascade starting with the activation of the kinase GCN2 (general control non-derepressible 2) phosphorylation of eIF2 (eukaryotic initiation factor 2), global protein synthesis reduction and increased ATF4 (activating transcription factor 4). ATF4 modulates a wide spectrum of genes involved in the adaptation to dietary stress. The hormone FGF21 (fibroblast growth factor 21) is induced during fasting in liver and its expression induces a metabolic state that mimics long-term fasting. Thus FGF21 is critical for the induction of hepatic fat oxidation, ketogenesis and gluconeogenesis, metabolic processes which are essential for the adaptive metabolic response to starvation. In the present study, we have shown that FGF21 is induced by amino acid deprivation in both mouse liver and cultured HepG2 cells. We have identified the human FGF21 gene as a target gene for ATF4 and we have localized two conserved ATF4-binding sequences in the 5′ regulatory region of the human FGF21 gene, which are responsible for the ATF4-dependent transcriptional activation of this gene. These results add FGF21 gene induction to the transcriptional programme initiated by increased levels of ATF4 and offer a new mechanism for the induction of the FGF21 gene expression under nutrient deprivation.
Mammals have developed a wide range of mechanisms to detect and respond to episodes of malnutrition and starvation. Nutrient deprivation or starvation frequently correlates with amino acid limitation. Amino acid starvation initiates a signal transduction cascade starting with the activation of the GCN2 (general control non-derepressible 2) kinase, phosphorylation of eIF2 (eukaryotic initiation factor 2), and increased synthesis of ATF4 (activating transcription factor 4) .
Dietary amino acid availability alters metabolic pathways beyond protein homoeostasis since there is a link between dietary amino acids and lipid metabolism. A GCN2-dependent inhibition of fatty acid synthase activity and expression of lipogenic genes in liver, and increased mobilization of lipid stores, occur in response to leucine deprivation in mice . In addition, increased expression of β-oxidation genes and decreased expression of lipogenic genes and activity of fatty acid synthase in WAT (white adipose tissue), and increased expression of UCP1(uncoupling protein 1) in BAT (brown adipose tissue), has been observed [3,4]. GCN2 triggers the amino acid-response signal transduction pathway when GCN2 kinase activity is activated by its binding to any uncharged tRNA molecule [5,6]. Although global protein synthesis is reduced, the translation of a group of mRNA species is increased as a part of this response. Among these is ATF4 [7,8], a transcription factor that binds to CARE [C/EBP (CCAAT/enhancer-binding protein)/ATF-response element; also named AARE (amino acid-response element)] and modulates a wide spectrum of genes involved in the adaptation to dietary stress . Food deprivation reduces free intracellular amino acid, and increases eIF2α phosphorylation and ATF4 mRNA levels in skeletal muscle .
FGF (fibroblast growth factor) 21 is a member of the FGF family, predominantly produced by the liver, but also by other tissues such as WAT, BAT, skeletal muscle and pancreatic β cells [11–14]. FGF21 expression in liver is under tight control by PPAR (peroxisome-proliferator-activated receptor) α [15–18], it is induced in the liver during fasting and its expression induces a metabolic state that mimics long-term fasting. Thus FGF21 is critical for the induction of hepatic fat oxidation, ketogenesis and gluconeogenesis, which are metabolic processes critical for the adaptive metabolic response to starvation .
In the present study, we have shown that the hormone FGF21 is induced by amino acid deprivation both in mice liver and cultured HepG2 cells. We have identified the human FGF21 gene as a target gene for ATF4 and we have localized two evolutionary conserved ATF4-binding sequence in the 5′ regulatory region of the human FGF21 gene. These sequences are responsible for the ATF4-dependent transcriptional activation of this gene. These results add FGF21 gene induction to the transcriptional programme initiated by increased levels of ATF4 and offer a new mechanism for the induction of the FGF21 gene expression under nutrient deprivation.
Animals and diets
Eight-week-old male C57BL/6J mice were fed on either a control diet (nutritionally complete diet) or a (−)leu diet (diet devoid of the essential amino acid leucine) for 7 days. Food intake and body weight were recorded daily. Sample collections and quantification of FGF21 in serum were carried out as described in . Both diets were obtained from Research Diets. All of the experimental protocols with mice were performed with approval of the animal ethics committee of the University of Barcelona (Barcelona, Spain).
Cell culture and treatment conditions
HepG2 cells were cultured in fresh MEM (minimal essential medium) and 10% (v/v) fetal bovine serum for 16 h before initiating all treatments. Wy14643 and HisOH (histidinol) were purchased from Sigma–Aldrich. MG132 was obtained from Calbiochem. Cells were transfected as described previously . Adenovirus expressing PPARα was a gift from F. Villarroya (University of Barcelona).
RNA isolation and relative quantitative RT (reverse transcription)–PCR
Levels of HMGCS2 (3-hydroxy-3-methylglutaryl-CoA synthase 2) and FGF21 mRNA were determined as described previously . cDNA was synthesized from 1 μg of total RNA by MMLV (Moloney murine leukaemia virus) reverse transcriptase (Invitrogen) with random hexamers (Roche Diagnostics) according to the manufacturer's instructions. TaqMan Gene Expression Master Mix and TaqMan Gene Expression Assays (Invitrogen/Applied Biosystems) were used for the PCR step. Amplification and detection were performed using the Step-OnePlus Real-Time PCR System. Each mRNA from a single sample was measured in duplicate. For HepG2 cells, human gene probes were used: FGF21, Hs00173927_m1; HMGCS2, Hs00985427_m1. For mice experiments, a mouse probe (Fgf21, Mm00840165_g1) was used. Relative mRNA abundance was obtained by normalizing to 18S levels (Applied Biosystems). To measure the transcriptional activity from the human FGF21 gene, oligonucleotides derived from FGF21 intron 2 and exon 3 were used to measure the short-lived unspliced transcript (hnRNA, heterogeneous nuclear RNA) . Real-time quantitative PCR was performed by using a SYBR Green I-containing PCR mixture (Applied Biosystems), following the manufacturer's recommendations. Sequences of primers were 5′-CCTGGATCCTGGGTCTTACA-3′ in intron 2 and 5′-CGGTGTGGGGACTTGTTC-3′ in exon 3.
The FGF21 promoter (nucleotides −768/+115)–luciferase plasmid was generated in the pGL3-basic plasmid (Promega). 5′ Deletions were generated by PCR from human genomic DNA. The mutations were made by site-directed mutagenesis [QIAquick mutagenesis kit (Qiagen)] by replacing the sequences of the AARE1 and AARE2 core for CAGATGGAC. Human ATF4 expression vector (pRK-ATF4)  was from Addgene (plasmid 26114).
Western blot analysis
Western blot assays for HepG2 and liver cell extracts were carried out using antibodies against ATF4 (sc-200, Santa Cruz Biotechnology) and actin (A2066, Sigma).
ChIP (chromatin immunoprecipitation) analysis
Cross-linked chromatin from HepG2 cells was sonicated using a Bioruptor® Next Gen (Diagenode). Real-time quantitative PCR was performed by using a SYBR Green I-containing mixture with specific primers to amplify the human FGF21 AARE1 and AARE2, and exon 1 as a negative control. Relative occupancy of the immunoprecipitated factor at a locus was calculated using the formula 2(CTIgG−CTATF4), where CTIgG and CTATF4 are mean threshold cycles of PCR from negative-control ChIP (non-immune IgG) and target ChIP (anti-ATF4 antibody) respectively. The sequence of the primers used were F1, 5′-AGCCAACCTGTCTTCCCTCT-3′, and R1, 5′-ATGCTCAGACCCTGGACATC-3′, for AARE1; F2, 5′-GCTTGAGACCCCAGATCCTT-3′, and R2, 5′-CATTTGGCAGGAGCTACAGA-3′, for AARE2; and 5′-GGACTGTGGGTTTCTGTGCT-3′ and 5′-ATCTCCAGGTGGGCTTCTGT-3′ for the unrelated control in exon 1.
In vitro transcription and translation
pcDNA3 empty vector, ATF4 and C/EBPβ were transcribed and translated by using commercially available kits according to the manufacturer's instructions (Promega).
Labelling probe with digoxigenin
Each oligonucleotide (0.5 nmol) was annealed by heating at 65°C for 10 min and slowly cooled to room temperature (22°C) in a buffer of 10 mM Tris/HCl (pH 8), 1 mM EDTA and 150 mM NaCl. The double-stranded oligonucleotide was then labelled using the reagents provided in the DIG Gel Shift Kit, 2nd Generation (Roche).
EMSA (electrophoretic mobility-shift assay)
An aliquot of 2 μl of each factor synthesized in vitro was pre-incubated on ice for 10 min in 25 mM Hepes (pH 7.9), 60 mM KCl, 5% glycerol, 0.75 mM dithiothreitol, 0.1 mM EDTA, 2.5 mM MgCl2, 1 μg of polylysine and 1 μg of poly(dI-dC)·(dI-dC). The total amount of protein was kept constant in each reaction through the addition of pcDNA3 empty vector. When indicated, a 50-fold excess of unlabelled AARE1 or AARE2, or mutAARE1 or mutAARE2, was added to the reaction mixture. Next, 40 fmol of digoxigenin-labelled probe was added and the incubation was continued for 15 min at room temperature. The final volume for all the reactions was 20 μl. Samples were electrophoresed at 4°C on a 6% polyacrylamide gel in 0.5% TBE buffer [45 mM Tris, 45 mM boric acid and 1 mM EDTA (pH 8.0)], and the DNA was transferred on to a positively charged nylon membrane. The DNA was then cross-linked to the membrane using a UV Stratalinker® (Stratagene). Immunological detection was performed following the manufacturer's instructions, and exposing the membrane to an imaging device. The EMSA oligonucleotides used were as follows. AARE1 (wild-type): forward, 5′-TGAAAGAAACACCAGGATTGCATCAGGGAGGAGGAGGCTG-3′, and reverse, 5′-CAGCCTCCTCCTCCCTGATGCAATCCTGGTGTTTCTTTCA-3′; mutAARE1: forward, 5′-TGAAAGAAACACCAGGcagatggacGGGAGGAGGAGGCTG-3′, and reverse, 5′CAGCCTCCTCCTCCCgtccatctgCCTGGTGTTTCTTTCA-3′; AARE2 (wild-type): forward, 5′-ATTGAAAGGACCCCAGGTTACATCATCCATTCAGGCTGC-3′, and reverse, 5′-GCAGCCTGAATGGATGATGTAACCTGGGGTCCTTTCAAT3′, mutAARE2: forward, 5′-ATTGAAAGGACCCCAGcagatggacTCCATTCAGGCTG-3′, and reverse, 5′-GCAGCCTGAATGGAgtccatctgCTGGGGTCCTTTCAAT-3′. Mutagenic residues are in lower-case type.
siRNA (small interfering RNA) transfection
The human ATF4 siRNA (L-005125-00), siControl non-targeting siRNA (D-001210-01) and DharmaFECT 4 transfection reagent were purchased from Dharmacon. Transfection was performed according to the manufacturer's instructions. After HisOH treatment, total RNA and nuclear protein extracts were isolated and analysed by real-time PCR and immunoblotting respectively.
Measurement of serum FGF21
A mouse FGF21 ELISA kit was obtained from Millipore for the quantification of FGF21 in mouse serum. The assay was conducted according to the manufacturer's protocol. Briefly, a calibration curve was constructed by plotting the difference of absorbance values at 450 and 590 nm against the FGF21 concentrations of the calibrators, and concentrations of unknown samples (performed in duplicate) were determined using this calibration curve.
All results are expressed as means±S.E.M. Significant differences were assessed using a two-tailed Student's t test. P<0.05 was considered statistically significant.
FGF21 expression is induced by the 26S proteasome inhibitor MG132
Considering the central role of PPARα in metabolic homoeostasis, we investigated how the turnover of PPARα affected the expression of its target genes. We checked the mRNA levels of FGF21 and HMGCS2, two prototypical PPARα target genes [15–18,23], in HepG2 cells infected with adenovirus expressing PPARα and exposed to DMSO or to the 26S proteasome inhibitor MG132. As expected, MG132 treatment blocked the PPARα-dependent expression of HMGCS2 (Figure 1A), indicating that the transcriptional activity of PPARα is increased by protein degradation . Contrary to what we had predicted, the expression of FGF21 was strongly increased by the MG132 treatment in a time-dependent manner (Figure 1B). The opposite effects of MG132 treatment on the expression of HMGCS2 and FGF21 genes points to a different mechanism controlling the expression of the FGF21 gene in response to the inhibition of the proteasome activity.
Amino acid starvation induces FGF21 expression
We hypothesized that proteasome inhibition in HepG2 cells could decrease the pool of free amino acids. This hypothesis prompted us to test whether FGF21 expression is induced during amino acid starvation. For this purpose, we treated HepG2 cells with HisOH, a potent and reversible inhibitor of protein synthesis that acts by decreasing the activation of histidine, mimicking amino acid starvation . Using real-time PCR, we measured the FGF21 mRNA levels in HisOH-treated HepG2 cells, and observed that amino acid deprivation produced a time-dependent induction of FGF21 mRNA (Figure 2A). To test whether this induction was due to an increase in FGF21 gene transcription, we measured the FGF21 primary transcript (hnRNA) levels; HisOH treatment clearly induced FGF21 hnRNA levels in a time-dependent manner, and this induction was maximal after 4 h of treatment (Figure 2B).
FGF21 is an ATF4 target gene
Proteasome inhibition leads to an increase in eIF2α phosphorylation, a significant reduction in protein synthesis and a concomitant induction of ATF4 expression . To test whether the HisOH-induced increase in FGF21 expression was due to ATF4, we first analysed the effect of HisOH treatment on ATF4 protein levels in HepG2 cells. The analysis of total HepG2 protein extracts by Western blotting showed that HisOH induced an increase in the ATF4 protein levels soon after 2 h of treatment (Figure 2C). Next, we analysed the sequence of the 5′-flanking region of the human FGF21 gene, looking for putative ATF4-binding sites. We found two putative AAREs starting at positions −152 and −610 upstream of the transcription start site. These sites matched the consensus for the CAREs and are conserved among several mammalian species (Figure 3A). In order to test whether FGF21 gene transcription was induced by ATF4, we made several constructs with the luciferase gene as a reporter (Figure 3B) and transfected HepG2 cells with those constructs and an expression vector for human ATF4. The expression of ATF4 induced the wild-type reporter in a concentration-dependent manner (Figure 3C). This induction was totally obliterated either when the AARE1 was mutated (mut1) or when both elements were deleted (delta2). Induction was diminished when AARE2 was mutated (mut2) (Figure 3D). These results identify both AARE sequences as ATF4-responsive elements in the FGF21 human gene.
FGF21 is an ATF4 target gene
ATF4 binds to both AAREs in
To analyse further the functionality of this sequence, we confirmed the in vivo binding by ChIP experiments. As shown in Figure 4(B), the chromatin binding of ATF4 was greatly increased in both ATF4-responsive sequences in HisOH-treated cells. In order to test the ability of ATF4 to bind to both AAREs, we performed EMSA using digoxigenin-labelled oligonucleotides containing the ATF4 composite sites (AAREs) of the human FGF21 gene (Figure 4C). In vitro translated ATF4 and C/EBPβ were incubated with labelled probes containing the wild-type AARE1 (lanes 1–7) and AARE2 (lanes 8–14) sequences. The specificity of this interaction was demonstrated by competition experiments with a 50-fold molar excess of unlabelled wild-type (lanes 5 and 12) or mutant (lanes 6 and 13) probes. The presence of ATF4 in the complex was confirmed by a supershift experiment with an anti-ATF4 antibody (lanes 7 and 14).
ATF4 binds to the
ATF4 mediates the amino-acid-starvation-induced increase in FGF21 expression
To test whether the induction of FGF21 produced by amino acid starvation was mediated by ATF4, we treated HepG2 cells with HisOH while simultaneously depleting ATF4 by transfecting an ATF4-targeting siRNA. The analysis by real-time PCR showed that the FGF21 mRNA levels after 4 h of HisOH treatment were significantly lower when ATF4 was depleted compared with the control siRNA-treated cells (Figure 5). We have also investigated the effect of other stress stimuli on the expression of FGF21. As expected, treatment of HepG2 cells with tunicamycin increased the ATF4 protein levels and consequently FGF21 mRNA levels; this increase was significantly diminished by the siRNA-mediated interference of ATF4 (results not shown).
Effect of siRNA-mediated ATF4 knockdown on endogenous
Leucine deprivation increases
FGF21 mRNA levels in mouse liver
To analyse the effect of amino acid deprivation on FGF21 expression in vivo, we fed mice on a (−)leu diet or a control diet for 7 days. The analysis by real-time PCR showed that the Fgf21 mRNA levels were greatly increased in liver from mice fed on a leucine-deprived diet compared with the control-fed animals (Figure 6A). The circulating FGF21 levels quantified by ELISA were also, consequently, increased in the serum of leucine-deprived animals, paralleling hepatic gene expression (Figure 6B). The circulating levels that were reached, approximately 3.5 ng/μl, where even higher than those observed during starvation [15,20]. As expected, ATF4 protein levels were induced in liver under leucine deprivation (Figure 6C).
Leucine deprivation induces FGF21 serum levels and mRNA expression in liver
FGF21 is a member of the endocrine FGF subfamily produced by the liver, but also by other tissues such as WAT, BAT, skeletal muscle and pancreas, which plays a role in the adaptation to metabolic states that require increased fatty acid oxidation. The expression of FGF21 is controlled by several transcriptional activators such as PPARα in the liver [15–18], and PPARγ in the adipose tissue [27–29], and it is negatively regulated by PGC-1α (PPARγ co-activator-1α). The PGC-1α-mediated reduction of FGF21 expression is dependent on Rev-Erbα and the synthesis of haem, a ligand of Rev-Erbα . In the present study, we have shown that FGF21 is induced by amino acid deprivation in both mouse liver and cultured HepG2 cells. Furthermore, the results of the present study identify the human FGF21 gene as a target gene for ATF4, and we have localized two evolutionarily conserved functional ATF4-binding sequences in the 5′ regulatory region of the human FGF21 gene that are responsible for the ATF4-dependent transcriptional activation of this gene. Our results show that, in accordance with results published previously [31,32], ATF4 binds in vitro to both sites as a heterodimer with C/EBPβ. Comparison of FGF21 mRNA and pre-mRNA levels suggests that both transcription and mRNA stabilization contribute to the ATF4-mediated induction of FGF21 expression, as has been described previously for ATF3, a known ATF4-activated gene [33,34]. The decline in FGF21 transcription after 4 h of amino acid deprivation could be caused by the increased expression of ATF3 and other ATF4-activated genes, such as C/EBPβ and CHOP (C/EBP-homologous protein), that act as counter-regulatory signals and lead to a self-limiting cycle of ATF4-dependent transcription [35,36]. The ATF4-dependent induction of FGF21 shown in the present study suggests a new explanation of the effect of ATF4 on the regulation of obesity, glucose homoeostasis and energy expenditure . Amino acid and other nutrient depletion, through the activation of several kinases, induce eIF2α phosphorylation, selective translation of some stress-responsive transcripts, including that of ATF4, and certain autophagy genes . We have investigated the effect on FGF21 expression of other stress stimuli that, like nutrient stress, are also able to induce autophagy. Our results show that FGF21 expression is also induced by ER (endoplasmic reticulum) stress. As expected, treatment of HepG2 cells with tunicamycin increased ATF4 protein levels and, consequently, FGF21 mRNA levels.
A dietary amino acid imbalance alters metabolism beyond protein homoeostasis. The results of the present study suggest that FGF21 could be a link between amino acid imbalance and the metabolic cell response to nutrient deprivation. For example, a protein-free diet resulted in a decrease in serum cholesterol levels and decreased expression of genes involved in cholesterol biosynthesis in rat liver . Nevertheless, the cholesterol levels and the expression of cholesterol synthesis genes were unchanged in the liver of mice deprived of leucine . However, it has been described that expression of lipogenic genes, including SREBP1c (sterol-regulatory-element-binding protein 1c) and FASN (fatty acid synthase), and the activity of fatty acid synthase in the liver are repressed, and that lipid stores in adipose tissue are mobilized in mice upon leucine deprivation. It has recently been shown that the expression of FGF21 repressed the transcription of SREBP1c and decreased the amount of mature SREBP1c in HepG2 cells . All of these results and those of present study suggest that FGF21 would mediate the reduction in lipogenesis observed under leucine deprivation.
The effect on the expression of lipogenic genes and activity of fatty acid synthase has also been described in WAT, where leucine deprivation, in addition, increases energy expenditure, lipolysis and expression of β-oxidation genes and increases the expression of UCP1 in BAT . These effects could be mediated by the induction of FGF21, a hepatic hormone whose action is necessary for the appropriate metabolism of lipids when fatty acids are the major fuel source. Mice maintained on a leucine-deficient diet for 7 days showed a dramatic reduction in abdominal fat mass  caused by increased fat mobilization and suppressed fatty acid synthesis in WAT as well as increased energy expenditure . Our data demonstrate that the increased hepatic production of FGF21 under this nutritional deprivation may contribute to this effect.
It has been proposed recently that leucine deprivation improves whole-body insulin sensitivity and insulin signalling in liver by sequentially activating GCN2 and decreasing mTOR (mammalian target of rapamycin)/S6K1 (S6 kinase 1) signalling . Furthermore, activation of AMPK (AMP-activated protein kinase), under these circumstances, also contributes to increased insulin sensitivity . The activation of GCN2 under amino acid starvation would lead to an increase in the expression of ATF4 which, as we have shown in the present study, will produce an increase in the expression and circulating levels of FGF21, which plays a crucial role in mediating the adaptive response of the liver to nutritional deprivation , contributing to the regulation of fatty acid oxidation, ketogenesis, TCA (tricarboxylic acid) cycle flux and carbohydrate metabolism. In addition, FGF21 regulates energy expenditure in adipocytes through activation of AMPK and SIRT1 (sirtuin 1) activities . These effects may thus improve insulin sensitivity. Our results suggest that FGF21 mediates the induction of AMPK observed under leucine deprivation, although further work is needed to prove this point.
In conclusion, our results expand our current knowledge of how FGF21 expression is regulated in liver, adding FGF21 gene induction to the transcriptional programme initiated by increased levels of ATF4 and offering a new mechanism for the induction of the FGF21 gene expression in response to nutrient availability.
amino acid-response element
AMP-activated protein kinase
activating transcription factor 4
brown adipose tissue
eukaryotic initiation factor 2
electrophoretic mobility-shift assay
fibroblast growth factor
general control non-derepressible 2
3-hydroxy-3-methylglutaryl-CoA synthase 2
heterogeneous nuclear RNA
minimal essential medium
small interfering RNA
sterol-regulatory-element-binding protein 1c
uncoupling protein 1
white adipose tissue
Ana Luísa De Sousa-Coelho planned the experiments, undertook the experimentation, analysed the experimental data and wrote the paper; Pedro Marrero and Diego Haro designed the research, analysed the experimental data and wrote the paper.
We thank Dr Yihong Ye (Laboratory of Molecular Biology, National Institutes of Health, Bethesda, MD, U.S.A.) for the pRK-ATF4 plasmid.
This work was supported by the Ministerio de Educación y Ciencia [grant numbers SAF2010-15217 (to D.H.) and BFU2007-67322/BMC (to P.F.M.)]; and the Ajut de Suport als Grups de Recerca de Catalunya [grant number 2009 SGR163]. A.L.D.S.C. was supported by the Fundação para a Ciência e a Tecnologia (FCT) from the Portuguese Government.