FGF21 (fibroblast growth factor 21), first described as a main fasting-responsive molecule in the liver, has been shown to act as a true metabolic regulator in additional tissues, including muscle and adipose tissues. In the present study, we found that the expression and secretion of FGF21 was very rapidly increased following lactate exposure in adipocytes. Using different pharmacological and knockout mice models, we demonstrated that lactate regulates Fgf21 expression through a NADH/NAD-independent pathway, but requires active p38-MAPK (mitogen activated protein kinase) signalling. We also demonstrated that this effect is not restricted to lactate as additional metabolites including pyruvate and ketone bodies also activated the FGF21 stress response. FGF21 release by adipose cells in response to an excess of intermediate metabolites may represent a physiological mechanism by which the sensing of environmental metabolic conditions results in the release of FGF21 to improve metabolic adaptations.
White adipocytes, which are localized in WAT (white adipose tissue), store and release fuel according to the metabolic needs of the organism . Besides their metabolic function, adipocytes are considered true endocrine cells that secrete a wide range of molecules commonly named adipokines, mediating interactions of adipose tissues with many physiological functions of the organism [2,3]. Brown adipocytes are specialized in the non-shivering thermogenesis process and energy dissipation thanks to the specific expression of the mitochondrial UCP1 (uncoupling protein 1) [4,5]. Since the first discovery of the presence of UCP1-expressing adipocytes among white fat pads [6–8], a large body of literature documents investigation of this phenomenon named browning . Although sharing many phenotypic and functional features with brown adipocytes, including thermogenic potential and substrate dissipation , these cells are different from classical brown adipocytes and have been recently named beige or brite (brown in white) adipocytes [11,12]. Although established for several years for white adipocytes, the endocrine activity of brown/brite adipocytes has only recently been demonstrated and participates to their beneficial effect in the overall energy expenditure .
FGF21 (fibroblast growth factor 21) is a metabolic regulator that triggers adaptive responses to limit metabolic stresses and acts through endocrine and autocrine/paracrine mechanisms . Although FGF21 expression is increased in liver and muscle during fasting and mitochondrial stress respectively [15,16], its expression is up-regulated in BAT (brown adipose tissue) and WAT during cold exposure [17,18]. Brown and brite adipocytes express and release FGF21 , which contributes to the regulation of Ucp1 expression [17,18,20]. Pharmacological injections of FGF21 display anti-obesogenic and anti-diabetic actions , although the requirement of browning for FGF21’s beneficial metabolic effect has been questioned .
Recently, we showed that brite adipocytes development is induced by specific metabolites. Unexpectedly, we found that lactate, the end product of glycolysis considered for many years a simple glycolytic waste product, induced the browning remodelling of white adipocytes, via a redox-dependent pathway . These data are in line with several recent papers highlighting the role of lactate, not only as an energetic substrate, but also as a signalling molecule driving the fate and function of different cell types [24–27].
To explore further the biological effects of lactate on adipocyte biology, we searched for early lactate-responsive genes. Among them, we found that the expression of Fgf21 was very rapidly increased following lactate exposure. This occurred through a redox-independent/p38-MAPK (mitogen activated protein kinase)-dependent pathway. The present study also reveals that Fgf21 induction is not restricted to lactate, as additional metabolites including pyruvate and ketone bodies also up-regulate the Fgf21 pathway, highlighting a new and general mode of regulation of this metabolic growth factor by an excess of metabolites.
MATERIALS AND METHODS
All studies were carried out using male C57Bl/6J mice obtained from the Harlan Laboratory. Fgf21−/− mice were obtained from the Mutant Mouse Regional Resource Center (B6N; 129S5-Fgf21tm1Lex/Mmucd). Ppara−/− mice were obtained from Jackson Laboratory. Sirt1+/− mice were kindly provided by Frederick W. Alt. Sirt3−/− mice (strain name: B6; 129S5-SIRT3Gt(neo)218Lex) were obtained from the Mutant Mouse Regional Resource Center. Animals were housed in a controlled environment (12-h light/12-h dark cycles at 21°C) with unrestricted access to water and a standard chow diet (UAR), in a pathogen-free animal facility (IFR150) and were killed by cervical dislocation. All experimental procedures were conducted in compliance with French Ministry of Agriculture regulations for animal experimentation.
Primary culture of adipose-derived stem/stromal cells and adipocyte differentiation
SVFs (stromal vascular fractions) of white inguinal fat pads from 6–8-week-old wild-type, Fgf21−/−, Ppara−/−, Sirt3−/− or Sirt1+/− heterozygote mice were obtained as described in . Following centrifugation, SVF cells resuspended in culture medium (αMEM plus 0.25 unit/ml amphotericin, 100 units/ml penicillin, 100 mg/ml streptomycin, biotin, ascorbic acid, panthotenic acid and 10% newborn calf serum) were plated (10000 cells/cm2) and rinsed with PBS 3 h after plating. Remaining adherent adipose-derived stem/stromal cells were grown to confluence in complete medium. Once they reached confluence, cells were exposed to the adipogenic cocktail containing 5 μg/ml insulin, 2 ng/ml T3 (3,3′,5-tri-iodothyronine), 33.3 nM dexamethazone, 10 μg/ml transferrin and 1 μM rosiglitazone in complete medium. Adipocytes differentiated for 8–10 days were treated with different compounds at time and concentrations as indicated in the Figure legends.
Western blot analysis
Proteins were extracted and Western blotting was performed as described in . Sources of antibodies are anti-(p)-p38-MAPK (Cell Signaling Technology, 9216S, diluted 1/2000), anti-(p)-HSP27 (heat-shock protein 27) (Cell Signaling Technology, 2401, diluted 1/1000) and anti-β-actin (Sigma; A5541, diluted 1/5000).
RNA extraction and quantitative PCR
Total RNA from culture cells was isolated using an RNeasy kit (Qiagen). For quantitative real-time PCR analysis, 1000 ng of total RNA was reverse-transcribed using the High Capacity cDNA Reverse Transcription Kit (Life Technologies/Applied Biosystems), SYBR® Green PCR master mix (Life Technologies/Applied Biosystems) and 300 nM primers on an Applied Biosystems StepOne instrument. Relative gene expression was calculated by the ∆∆CT method and normalized to 36b4. Primers sequences are the following: 36b4 forward, 5′-AGTCGGAGGAATCAGATGAGGAT-3′, and 36b4 reverse, 5′-GGCTGACTTGGTTGCTTTGG-3′; Fgf21 forward, 5′-TACACAGATGACGACCAAGA-3′, and Fgf21 reverse, 5′-GGCTTCAGACTGGTACACAT-3′, Ucp1 forward, 5′-GACCGACGGCCTTTTTCAA-3′, and Ucp1 reverse, 5′-AAAGCACACAAACATGATGACGTT-3′; Atf4 forward, 5′-AGCAAAACAAGACAGCAGCC-3′, and Atf4 reverse, 5′-ACTCTCTTCTTCCCCCTTGC-3′; Chop-10 forward, 5′-GCCAACAGAGGTCACACGC-3′, and Chop-10 reverse, 5′-CCTGGGCCATAGAACTCTGAC-3′; Grp78 forward, 5′-GTGTGTGAGACCAGAACCGT-3′, and Grp78 reverse, 5′-GCAGTCAGGCAGGAGTCTTA-3′.
FGF21 protein levels in mouse adipocyte cell culture medium were determined using a specific ELISA (Phoenix Secretomics).
All results are expressed as means±S.E.M. from at least three individual experiments. Unpaired Student's t test was used to calculate final P-values. Significance is given as *P<0.05, **P<0.01 and ***P<0.001.
Lactate caused a rapid up-regulation of Fgf21 expression, unrelated to the further induction of Ucp1
As shown in Figure 1(A), lactate sharply up-regulated Fgf21 mRNA levels in primary differentiated adipocytes, with a significant induction observed after 4 h of treatment. This up-regulation preceded the rise of Ucp1 mRNA levels which is significant after 24 h of lactate exposure. This lactate-dependent regulation is specific to Fgf21 because no change or even decreased expression was observed for additional secreted molecules (adiponectin and leptin respectively; results not shown). As FGF21 was described to regulate Ucp1 gene expression and browning in response to cold or adrenergic stimulation [17,18,20], we asked whether FGF21 could be involved in lactate-induced Ucp1 expression. To address this issue, we used cells isolated from Fgf21-deficient mice. Lactate treatment for 24 h increased Ucp1 mRNA levels in the same manner in Fgf21−/− adipocytes as in wild-type cells (Figure 1B). This clearly demonstrates that Fgf21 is not required for the acute effect of lactate on Ucp1 expression.
Lactate causes early up-regulation of Fgf21 that does not mediate lactate-induced Ucp1 expression
Lactate-induced Fgf21 regulation is redox-independent
Lactate-induced Fgf21 expression was abrogated by the monocarboxylate transporter inhibitor CHC [2-cyano-3-(4-hydroxyphenyl)-2-propenoic acid], known to reduce lactate import in different types of cells [29,30] (Figure 2A). This indicates that the lactate effect on Fgf21 expression requires its import into adipocytes. As investigated previously for Ucp1 , we asked whether intracellular redox changes triggered Fgf21 up-regulation. Because the reversible reaction catalysed by lactate dehydrogenase is at equilibrium, the NADH,H+/NAD+ ratio is imposed by the relative quantity of substrates, i.e. lactate and pyruvate (Figure 2B). Similarly to lactate, pyruvate increased Fgf21 expression in a dose-dependent manner (Figure 2C). Furthermore, both β-hydroxybutyrate and acetoacetate (that increase or decrease redox state respectively by their transformation by β-hydroxybutyrate dehydrogenase; Figure 2B) increase Fgf21 mRNA levels (Figure 2C). Together, these data demonstrate that Fgf21 expression is not governed by redox state changes in contrast with Ucp1 for which only lactate and βHB strongly up-regulated its mRNA levels  (Figure 2D). The up-regulation of Fgf21 gene expression by the different intermediate metabolites led to a concomitant increase in the release of FGF21 protein into the culture medium (Figure 2E) that demonstrates efficient translation and secretion of FGF21 by adipocytes. Altogether, these findings highlight that different metabolites in excess converge through a common metabolic response in adipocytes supported by the stress-responsive molecule FGF21.
Lactate-induced Fgf21 expression is redox-independent
PPARα/γ, sirtuins, ATF4 and endoplasmic reticulum pathways are not required for lactate-induced Fgf21 regulation
Next, we investigated the molecular pathways mediating Fgf21 regulation and focused on known regulators of the Fgf21 gene in others systems, such as PPARα (peroxisome-proliferator-activated receptor α) , sirtuins , ER (endoplasmic reticulum) stress and ATF4 (activating transcription factor 4) signalling  and PPARγ . As shown in Figure 3(A), lactate induced a similar increase in Fgf21 mRNA levels in wild-type and Ppara−/− adipocytes, suggesting that this transcription factor does not mediate lactate-induced Fgf21 up-regulation. Similarly, Fgf21 expression was the same in lactate-treated Sirt1+/− and Sirt3−/− adipocytes as in their respective wild-type cells (Figures 3B and 3C respectively), excluding any role for these sirtuins. It has been shown that ER stress and mitochondrial stress induce Fgf21 expression through ATF4-dependent signalling . Consistent with this finding, we indeed found that treatment of white adipocytes with antimycin, a mitochondrial respiratory chain inhibitor, increased Fgf21 expression, this being associated with an up-regulation of the mRNA levels encoding ATF4 and its downstream target CHOP10 [C/EBP (CCAAT/enhancer-binding protein)-homologous protein 10] (Figure 3D). In contrast, no change in Atf4 or Chop10 mRNA levels was associated with lactate-induced Fgf21 up-regulation (Figure 3D), suggesting that lactate regulates Fgf21 expression through a different pathway. Whereas ER stress induced by brefeldin A treatment (monitored through the up-regulation of Grp78 expression) was associated with increased Fgf21 expression, the lack of any increase in Grp78 mRNA levels in lactate-treated cells (a significant reduction was even observed) excluded any role for ER stress in lactate-induced Fgf21 regulation (Figure 3E). For PPARγ signalling, whereas basal Fgf21 mRNA levels were higher in the presence of the PPARγ ligand rosiglitazone than under rosiglitazone-free conditions (Figure 3F), the lactate-induced Fgf21 expression was identical in the presence or in the absence of rosiglitazone. These data, together with the fact that the PPARγ antagonist GW9962 did not affect lactate-induced Fgf21 expression (results not shown), rule out a role for PPARγ in lactate-induced Fgf21 expression. Altogether, these experiments excluded a role for any of these pathways in lactate-induced Fgf21 expression.
Induction of Fgf21 by lactate is not mediated by PPARα, sirtuins, ATF4 and ER stress pathways
p38-MAPK triggers lactate-induced Fgf21 up-regulation
We then tested the involvement of the p38-MAPK signalling pathway shown previously to control Fgf21 expression in adipocytes . Acute treatment (15 min) with lactate sharply increased the phosphorylation status of p38-MAPK (Figure 4A). The effect of lactate on p38-MAPK phosphorylation was dose-dependent and only transient as no further difference in p38-MAPK phosphorylation was observed after 60 min of treatment (Figure 4B). The treatment of white adipocytes with the p38-MAPK inhibitor SB203580 abrogated lactate-induced Fgf21 expression (Figure 4C). Consistent with this finding, Figure 4(D) shows that lactate strongly increased the amount of phosphorylated HSP27, a direct target of activated p38-MAPK when the p38-MAPK inhibitor SB203580 abrogated lactate-induced phosphorylation of HSP27, validating its efficiency. Together, these data demonstrate that lactate controls Fgf21 expression through activation of the stress-responsive p38-MAPK pathway. In contrast, although p38-MAPK is an important regulator of Ucp1, it does not seem to be involved in lactate-induced Ucp1 expression, as treatment with the p38-MAPK inhibitor did not abrogate lactate-induced Ucp1 expression (Figure 4E). One can postulate that the transient increase in p38-MAPK phosphorylation does not constitute a strong enough signal to contribute to lactate-induced Ucp1 expression which appears only after 24 h of treatment (Figure 1A).
Lactate induced Fgf21 up-regulation via p38-MAPK activation
The present study demonstrates that lactate, which is accumulated when the oxidative capacities of the cells are overwhelmed, signals to FGF21 which is a relevant activator of glucose consumption  and carbohydrate and lipid oxidative processes [19,34]. In addition to the up-regulation of Ucp1 expression, this constitutes an additional and different mechanism by which lactate would increase the oxidative capacity of adipocytes. As hypoxia is known to occur in adipose tissue during obesity , we could propose that low oxygen pressure generates a lactate-dependent signal to FGF21, to favour oxidation processes in adipocytes, in the context of dysregulated lipid and carbohydrate metabolism. In contrast with Ucp1, the up-regulation of Fgf21 expression occurs through a NADH/NAD-independent manner, but requires an active p38-MAPK signalling pathway. This finding is consistent with a recent study which demonstrates that lactate activates a p38-MAPK pathway in germ cells, probably through reactive oxygen species production . The involvement of p38-MAPK in the regulation of Fgf21 is also totally consistent with the previously described role of this kinase in the control of Fgf21 expression under thermogenic activation . Importantly, we demonstrated that FGF21 up-regulation is not restricted to lactate, as additional intermediate metabolites such as pyruvate and ketone bodies, whose accumulation reflect impairment in metabolic fluxes, also increased adipose Fgf21 expression. We therefore propose that the induction of the stress-responsive FGF21 pathway constitutes a general adaptive mechanism triggered by an excess of intermediates metabolites.
The correlation between serum FGF21 and lactate levels in humans  suggests that additional cell types, including hepatocytes which represent the primary source of plasmatic FGF21 , could respond to increased circulating lactate levels through the secretion of FGF21. During starvation (when lactate and ketone bodies are high), up-regulation of FGF21 secretion would finely tune glucose consumption with oxidative processes as a way to counterbalance the oxidative deficit which is at the origin of intermediate metabolites production. The consequence would be subsequent decreased lactate and ketone body production, slowing down FGF21-dependent glucose uptake and avoiding any possible hypoglycaemia event, as described previously . Previous studies have demonstrated that muscle cells with impairment in mitochondrial activity show an activated FGF21 response [16,38], mediated by p38-MAPK activation . One can reasonably imagine that the increased lactate production under such mitochondrial stress conditions may be involved, at least partly, in the regulation of muscular Fgf21 expression. The present study, together with our previous study , emphasizes the diversity and the complexity of lactate effects and the need to better understand its consequences on the physiology of adipose cells. It also highlights the role of intermediates metabolites that constitute true signalling molecules enabling cell adaptation to specific metabolic conditions. Moreover, although the induction of FGF21 may have a major autocrine role, it cannot be ruled out that adipocytes release FGF21 to systemic circulation as a way to influence overall organism metabolism to cope with an excess of intermediate metabolites.
Yannick Jeanson and Audrey Carrière designed the experiments, researched and analysed data. Francesc Ribas, Anne Galinier, Emmanuelle Arnaud, Marion Ducos, Mireille André and Vanessa Chenouard researched and discussed data. Francesc Villarroya, Audrey Carrière and Louis Casteilla discussed data, wrote and reviewed the paper. Yannick Jeanson, Audrey Carrière and Louis Casteilla are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
We are grateful to the I2MC/UMR1048, Plateforme GeT du Génopole Toulouse. We especially thank J. Caturla and S. Berger-Müller (STROMALab) for excellent technical assistance. We also thank Frederick W. Alt (Howard Hughes Medical Institute, Harvard Medical School, Boston) for the gift of the Sirt1+/− mice.
This work was supported by the European Union Framework Programme 7 projects: DIABAT [grant number HEALTH-F2-2011-278373] and METABOSTEM [grant number PCIG9-GA-2011-293720].
activating transcription factor 4
C/EBP (CCAAT/enhancer-binding protein)-homologous protein 10
fibroblast growth factor 21
heat-shock protein 27
stromal vascular fraction
uncoupling protein 1
white adipose tissue