FGF21 Maintains Glucose Homeostasis by Mediating the Cross Talk Between Liver and Brain During Prolonged Fasting

Hepatic gluconeogenesis is tightly controlled by counter-regulatory hormones such as glucagon, cortisol, and insulin, via regulating the expression of key gluconeogenic enzymes, including glucose 6 phosphatase (G6Pase) and phosphoenolpyruvate carboxykinase (PEPCK). Fibroblast growth factor 21 (FGF21), a metabolic regulator mainly secreted from the liver in response to fasting and starvation under the control of the nuclear receptor peroxisome proliferator–activated receptor α (PPARα), plays a critical role in maintaining energy homeostasis and insulin sensitivity in both rodents and nonhuman primates (1–6). A therapeutic dose of FGF21 decreased blood glucose in diabetic animals without causing hypoglycemia (4). FGF21 has also been shown to act as a key downstream effector of PPARα, mediating several metabolic adaptation responses to starvation, including hepatic fatty acid oxidation, ketogenesis, and growth hormone resistance (1,2,7). In addition, FGF21 is implicated in hepatic gluconeogenesis, although it remains controversial whether hepatocytes are a direct action site of FGF21 (8,9). There is an obvious dichotomy between the effects of endogenous FGF21 and pharmacological actions of the recombinant peptide with respect to hepatic metabolism (4,6,9).

FGF21 can cross the blood-brain barrier (10) and is detectable in both human and rodent cerebrospinal fluid (10,11). Continuous intracerebroventricular injection of FGF21 into obese rats increases energy expenditure and insulin sensitivity (12). More recently, FGF21 has been shown to act on the central nervous system to increase systemic glucocorticoid levels, suppress physical activity, and alter circadian behavior (13). Furthermore, FGF21 acts on the hypothalamus to suppress the vasopressin-kisspeptin signaling cascade, thereby mediating starvation-induced infertility of female mice (14). However, the physiological roles of FGF21 and its central actions in regulating glucose metabolism during adaptive starvation responses remain unknown.

In this study, we show that FGF21 is a key metabolic regulator essential for maintaining glucose homeostasis, by sending the starvation signal from the liver to the brain, where it acts on the hypothalamic neurons to induce phosphorylation of the mitogen-activated protein (MAP) kinase extracellular signal–related kinase 1/2 (ERK1/2), which in turn stimulates the production of corticotropin-releasing hormone (CRH) and subsequent release of adrenal corticosterone by activation of cAMP response element binding protein (CREB), thereby leading to enhanced hepatic gluconeogenesis.

All animal experimental protocols were approved by the Animal Ethics Committee of The University of Hong Kong. PPARα knockout (KO) mice in C57BL/6N background were originally obtained from The Jackson Laboratory (Sacramento, CA). FGF21 KO mice in C57BL/6J background were described previously (15). The mice were housed in a room under controlled temperature (23 ± 1°C) with free access to water and standard chow. Male mice at the age of 12–14 weeks were used in all animal experiments. Mice were studied by hyperinsulinemic-euglycemic clamp to assess endogenous glucose production as previously described (16), except that mice were fasted 24 h before the clamp and somatostatin was not infused. Bilateral or sham adrenalectomy was conducted under isoflurane anesthesia 1 week before various fasting experiments. Plasma corticosterone levels were measured to confirm the completeness of adrenalectomy.

Male FGF21 KO and PPARα KO mice and their respective wild-type (WT) littermates at the age of 12–14 weeks were chronically implanted with 26G stainless cannula (Plastics One, Inc., Roanoke, VA) in the cerebral ventricle or the paraventricular nucleus (PVN) region of the brain as previously described (17). A neutralizing monoclonal antibody against FGF receptor 1 (FGFR1, 0.2 μg/g body weight; NBP2–12308, Novus) or CRH antagonist (α-helical CRH_9–41, 0.2 μg/g body weight; C2917, Sigma-Aldrich) was infused by intra-PVN injection before central administration of endotoxin-free recombinant mouse FGF21 (rmFGF21, 0.02 μg/g body weight) (18).

Serum levels of insulin and FGF21 were quantified with immunoassays according to the manufacturer’s instructions (Antibody and Immunoassay Services, The University of Hong Kong). Mouse FGF21 assay was based on an affinity-purified rabbit polyclonal antibody specific to mouse FGF21 and did not cross-react with other members of the FGF family. The intra- and interassay variations were 4.2 and 7.6%, respectively. Mouse insulin assay was based on a monoclonal antibody pair raised using recombinant mouse insulin as an antigen, with intra- and interassay variations of 5.3 and 6.2%, respectively. Serum levels of corticosterone (Enzo Life Sciences) and glucagon (Millipore) and plasma levels of adrenocorticotropic hormone (ACTH; MD Bioproducts) and CRH (Phoenix Pharmaceuticals) were determined using immunoassays according to the manufacturer's instructions. For analysis of adrenaline and noradrenaline in the liver, frozen tissue was extracted with 0.1 N HCl and analyses performed according to the supplier's instructions (Labor Diagnostika Nord, Nordhorn, Germany).

The livers of male 12-week-old C57BL/6 mice were isolated, and primary hepatocytes were prepared from male Wistar rats (200 g) and used to determine glucose production as previously described (19). The mouse hypothalamic explants were prepared as previously described (20). The hypothalamic slices were pretreated without or with PD98059 for 30 min, followed by treatment with rmFGF21 for different periods.

Total RNA was extracted from mouse tissues and rat hepatocytes, cDNA was synthesized, and the mRNA expression levels were quantified with real-time PCR as previously described (21).

Total cellular proteins in liver and hypothalamic lysates were separated by SDS-PAGE and probed with a rabbit polyclonal antibody against peroxisome proliferator–activated receptor-γ coactivator 1α (PGC-1α; Abcam), FGF21 (18), phosphorylated ERK (p-ERK), total ERK (t-ERK), p-CREB, t-CREB (Cell Signaling Technology), FGFR1, βKlotho (Santa Cruz), or β-actin (Sigma-Aldrich).

Paraformaldehyde-fixed brains were frozen in liquid nitrogen and stored at −80°C until sectioning with a cryostat (CM1950, Leica) at −20°C. Slides were washed and then blocked in 10% goat serum with 3% BSA in TBS for 1 h at room temperature, followed by incubation with primary antibody overnight at 4°C. On the 2nd day, the slides were further incubated with Alexa Fluor 488 or 594–labeled mouse-specific secondary antibodies (Invitrogen) for another hour at room temperature. Images were captured with a fluorescence microscope (IX71; Olympus; QImaging).

All analyses were performed with Statistical Package for Social Sciences version 11.5 for Windows (SPSS, Chicago, IL). Comparison between groups was performed using one-way ANOVA followed by least-significant difference post hoc test. In all statistical comparisons, a P value <0.05 was used to indicate a statistically significant difference.

Blood glucose levels were comparable between FGF21 KO and WT mice in the fed state and within 6 h of food deprivation (Fig. 1A). However, during the prolonged fasting period (24–48 h), FGF21 KO mice exhibited a much more rapid decline in blood glucose levels and more severe hypoglycemia as compared with WT mice (Fig. 1A). Moreover, both the mRNA and protein expression levels of hepatic PGC-1α, a transcriptional coactivator that plays a key role in mediating the fasting-induced expression of gluconeogenic genes (22), were comparable between FGF21 KO mice and WT littermates in the fed state (Fig. 1A and B). However, the amplitude of fasting-induced elevation in PGC-1α expression was significantly decreased as compared with WT mice, and this change was accompanied by a marked impairment in the fasting-induced expression of G6Pase and PEPCK (Fig. 1A and B). The rate of glucose production in FGF21 KO mice, as estimated by pyruvate tolerance test, was much lower than that in WT littermates in fasted state (Fig. 1C). We further assessed whole-body glucose homeostasis in euglycemic glucose clamp studies. During constant hyperinsulinemia in mice after fasting, a higher glucose infusion rate was required to maintain a normal glucose level in KO mice than in WT mice (Fig. 1D). This difference was primarily accounted for by a significant reduction in glucose production in KO mice, although whole-body glucose disposal was similar in WT and KO mice (Fig. 1D). Treatment of mouse liver explants with rmFGF21 for up to 24 h had no obvious effect on the expression of gluconeogenic genes (PGC-1α, PEPCK, and G6Pase) and glucose production, although the glucocorticoid receptor agonist dexamethasone led to a significant induction of the gluconeogenic program (Supplementary Fig. 1A and B). Similarly, dexamethasone, but not FGF21, increased the expression of the gluconeogenic genes and glucose production in primary rat hepatocytes (Supplementary Fig. 1C and D). Furthermore, there was no difference in circulating levels of insulin and glucagon nor the levels of hepatic adrenaline and noradrenaline between FGF21 KO mice and WT littermates in either the fed or fasted state (Supplementary Fig. 2A–D), suggesting that insulin, glucagon, and hepatic sympathetic nerves are not the downstream mediators of FGF21 in the maintenance of fasting glucose homeostasis.

Glucocorticoids are fasting-inducible hormones that play a central role in activating gluconeogenic genes and sustaining glucose homeostasis during starvation. Our recent clinical study showed that FGF21 and cortisol display a similar circadian rhythm in humans (23). Therefore, we next investigated the functional relationship between FGF21, corticosterone, and its upstream regulators in mice. In WT mice, serum levels of corticosterone were progressively increased in response to fasting (Fig. 1E). However, the magnitude of fasting-induced elevation of serum corticosterone in FGF21 KO mice was much smaller than WT mice (Fig. 1E). Likewise, fasting for 24 h caused a marked induction in hypothalamic expression of CRH (Fig. 1F) and pituitary release of ACTH (Fig. 1H), whereas such fasting-induced expression of these two corticotropic hormones was severely impaired in FGF21 KO mice (Fig. 1F and H), suggesting that FGF21 deficiency causes defective fasting-induced release of corticosterone possibly by impairing the activation of the hypothalamic-pituitary-adrenal (HPA) axis. On the contrary, the expression of arginine vasopressin (AVP) was comparable between WT and FGF21 KO mice under both fed and fasted states (Fig. 1G), suggesting that these hormones do not contribute to the hypoglycemia phenotypes in FGF21 KO mice. Moreover, the difference in fasting blood glucose between WT and KO mice totally disappeared after the production of corticosterone was blocked by adrenalectomy in mice (Fig. 1I), suggesting that corticosterone mediates the effects of FGF21 in regulating fasting glucose homeostasis.

We next investigated whether the fasting hypoglycemia phenotype in FGF21 KO mice can be reversed by supplementation with corticosterone. Injection of the synthetic glucocorticoid dexamethasone into FGF21 KO mice reversed fasting hypoglycemia to a level comparable to WT mice (Supplementary Fig. 3A). The dexamethasone-mediated normalization of fasting glucose homeostasis was associated with significantly elevated expression of PGC-1α and its target genes PEPCK and G6Pase (Supplementary Fig. 3B–D).

Consistent with a previous report (10), we found that FGF21 was present in the hypothalamus and its hypothalamic level was markedly elevated in response to fasting (Fig. 2A), despite that there was not any detectable FGF21 mRNA expression in this tissue even in the fasted state (Fig. 2A). Furthermore, intraperitoneal injection of exogenous rmFGF21 led to its marked accumulation in the hypothalamus region (Fig. 2B), confirming that FGF21 can cross the blood-brain barrier into the hypothalamus.

To investigate whether FGF21 induces fasting gluconeogenesis through its central actions, rmFGF21 was infused directly into the cerebroventricle of FGF21 KO mice by intracerebroventricular injection via a cannula preimplanted 1 week before the treatment. The results showed that the central infusion of rmFGF21 at 0.02 μg/g body weight, a dose at which there was still no detectable FGF21 in the peripheral blood of FGF21 KO mice, led to a significant elevation of blood glucose levels as compared with the vehicle-treated mice (Fig. 2C). The central administration of rmFGF21 also induced the expression of hepatic PGC-1α, PEPCK, and G6Pase (Fig. 2C) and increased the hypothalamic level of CRH (Fig. 2D) and pituitary release of ACTH (Fig. 2F) and serum corticosterone (Fig. 2G) but had no effect on hypothalamic expression of AVP (Fig. 2E). Furthermore, direct incubation of mouse hypothalamic slices with rmFGF21 was sufficient to induce the CRH expression in a dose-dependent manner (Fig. 2H), whereas the central effects of FGF21 on stimulation of corticosterone release and elevation of blood glucose levels were blunted by intra-PVN administration of the CRH antagonist α-helical CRH_9–41 (Fig. 2I and J). On the contrary, intravenous injection of a low dose of rmFGF21 (0.02 μg/g), which was insufficient to cause any notable elevation of FGF21 in the brain of FGF21 KO mice, had no obvious effect on either blood glucose level or the expression of the gluconeogenic genes in the liver or the hormones of the HPA axis (CRH, ACTH, and corticosterone) (Supplementary Fig. 4A–E).

Central injection of rmFGF21 did not increase the levels of noradrenaline or adrenaline in the liver (Supplementary Fig. 5A and B), suggesting that sympathetic nerves are not involved in FGF21-mediated hepatic glucose production. Plasma ketone bodies and hepatic expression of key ketogenic genes and several genes governing lipid metabolism in the liver were not altered by central treatment of rmFGF21 (Supplementary Fig. 5C and D), indicating that the effects of FGF21 on ketogenesis and lipid metabolism are not attributed to its central actions.

To further confirm the role of the HPA axis in FGF21-mediated gluconeogenesis, FGF21 KO mice were pretreated with the glucocorticoid receptor antagonist RU486 or subjected to adrenalectomy prior to intracerebroventricular injection of rmFGF21. The results showed that the central effects of rmFGF21 on elevation of blood glucose and induction of the hepatic gluconeogenic genes were largely abrogated by either pharmacological blockage of the glucocorticoid receptor or depletion of endogenous corticosterone by adrenalectomy (Fig. 2K–M).

The metabolic actions of FGF21 are mediated by FGFR1 in complex with βKlotho (24). FGFR1 is highly expressed in hypothalamic areas such as PVN and supraoptic nuclei (25). Both real-time PCR and Western blot analysis showed that the expression level of FGFR1 in the hypothalamus was comparable to that in white adipose tissue, a major peripheral target of FGF21 (Fig. 3A). FGFR1 is abundantly expressed in the PVN of the hypothalamus (Fig. 3B), and a large portion of FGFR1 colocalized with CRH neurons (Fig. 3B). In addition, both mRNA and protein expressions of βKlotho were detectable in the hypothalamus, although its expression level was much lower than that in adipose tissues and livers (Fig. 3C). Specifically, βKlotho was detectable in the PVN of the hypothalamus (Fig. 3D).

To interrogate the roles of hypothalamic FGFR1 in FGF21-induced hepatic gluconeogenisis via the HPA axis, we performed bilateral intra-PVN injection to directly deliver the neutralizing antibody into PVN. The intra-PVN injection of anti-FGFR1 antibody largely blocked rmFGF21-induced activation of ERK1/2 in the hypothalamus (Fig. 4). The central actions of rmFGF21 on elevation of blood glucose and induction of hepatic gluconeogenic gene expression were abrogated by a single intra-PVN administration of a neutralizing monoclonal antibody against FGFR1 (Fig. 5A and B). Furthermore, blockage of hypothalamic FGFR1 by its neutralizing antibody also negated the stimulatory effects of central administration with rmFGF21 on hypothalamic CRH expression and serum levels of ACTH and corticosterone (Fig. 5C–F). On the other hand, intra-PVN administration of the anti-FGFR1 neutralizing antibody had no obvious effect on glucose levels, hepatic gluconeogenic gene expression, and the HPA activity in FGF21 KO mice without FGF21 treatment (Fig. 5A–F). Physiologically, a single intra-PVN administration of the anti-FGFR1 antibody lowered fasting blood glucose and reduced the expression of hepatic gluconeogenic genes in WT mice to a level comparable to those in KO mice (Fig. 5G and H). Furthermore, fasting-induced elevations in hypothalamic CRH expression and serum levels of corticosterone were significantly diminished by the neutralizing antibody-mediated blockage of hypothalamic FGFR1 (Fig. 5I and J).

During fasting, the transcription factor PPARα is activated in the liver, which in turn promotes hepatic expression of FGF21 and its release into the circulation (1–3). Therefore, we investigated whether PPARα activation induces corticosterone production by induction of FGF21. Chronic treatment of WT mice with the PPARα agonist fenofibrate led to a progressive elevation in serum levels of FGF21 as well as corticosterone (Fig. 6A and B). However, the magnitude of the fenofibrate-induced increase of serum corticosterone was significantly attenuated in FGF21 KO mice (Fig. 6B). Likewise, fenofibrate-induced hypothalamic expression of CRH was also impaired in FGF21 KO mice as compared with WT controls (Fig. 6C). Furthermore, fasting-induced elevation of serum corticosterone levels and hypothalamic expression of CRH in PPARα KO mice were markedly reduced as compared with WT littermates (Fig. 6D and E). On the other hand, the impairments in the fasting-induced production of corticosterone and CRH were partially reversed by intracerebroventricular injection of rmFGF21 (Fig. 6D and E).

Consistent with previous reports (26), we found that PPARα KO mice exhibited similar blood glucose levels with WT controls in the fed state but developed severe hypoglycemia after 16 h of fasting (Fig. 6F). The hypoglycemic phenotype of PPARα KO mice further deteriorated after prolonged fasting (24 and 48 h). On the other hand, intracerebroventricular administration of rmFGF21 into PPARα KO mice not only led to a significant alleviation of fasting hypoglycemia but also reversed the impairment of fasting-induced expression of several gluconeogenic genes (PGC-1α, PEPCK, and G6Pase) (Fig. 6F and G). Likewise, replenishment of PPARα KO mice with dexamethasone also resulted in a partial reversal of hypoglycemia and decreased gluconeogenic gene expression in fasted PPARα KO mice, suggesting that PPARα maintains fasting glucose homeostasis via activation of the FGF21-glucocorticoid axis (Fig. 6F and G).

CREB is a key transcription factor responsible for transactivation of the CRH gene in PVN neurons (27). The MAP kinase ERK1/2 activates CREB by phosphorylating the transcription factor at serine 133 (28). Notably, FGF21 induces phosphorylation and activation of ERK1/2 both in adipocytes (4) and in whole hypothalamus of mice (29). Therefore, we next investigated the roles of ERK1/2 and CREB in mediating the central actions of FGF21 in mice. Consistent with previous reports (30), we found that phosphorylation of both ERK1/2 and CREB was significantly enhanced by fasting for 24 h in WT mice (Fig. 7A). However, fasting-induced phosphorylation of ERK1/2 and CREB was impaired in FGF21 KO mice (Fig. 7A).

Injection of rmFGF21 into the PVN region of FGF21 KO mice induced phosphorylation of both ERK1/2 and CREB, whereas such a stimulatory effect of rmFGF21 was completely blocked by pretreatment with the ERK inhibitor PD98059 in the hypothalamic PVN region (Fig. 7B and Supplementary Fig. 6). Likewise, the central effects of rmFGF21 on induction of hypothalamic CRH expression, elevation of serum corticosterone, and blood glucose levels, and stimulation of hepatic gluconeogenic gene expression, were all abrogated by intra-PVN injection of PD98059 (Fig. 7C–F). Furthermore, direct incubation of mouse hypothalamic slices with rmFGF21 was sufficient to induce the phosphorylation of ERK and CREB (Fig. 7G) and to increase the mRNA expression and protein release of CRH (Fig. 7H and I). However, all these effects of rmFGF21 on hypothalamic slices were blocked by PD98059 (Fig. 7G–I). Taken together, these findings suggest that FGF21 stimulates hypothalamic CRH production via the ERK1/2-CREB signaling cascade, which in turn triggers the release of corticosterone for induction of gluconeogenesis (Fig. 8).

Appropriate counter-regulatory hormone responses to hypoglycemia are critical for maintaining blood glucose levels within a narrow range. However, the molecular events that coordinate this process remain poorly defined. In both rodents and humans, fasting-induced hepatic production of FGF21 is mediated by PPARα, which is a master regulator coordinating metabolic adaptions to fasting and starvation (1–3). In addition to its central role in controlling hepatic fatty acid metabolism, PPARα is also a critical mediator of fasting-induced hepatic gluconeogenesis in mice (26,31). PPARα KO mice exhibit severe fasting hypoglycemia despite normoglycemia in the fed state (26). However, PPARα is not a direct transcriptional activator of the key gluconeogenic genes, suggesting that it induces hepatic gluconeogenesis via an indirect mechanism(s) (32). In the current study, we found that both PPARα and FGF21 KO mice exhibited a similar degree of hypoglycemia due to defective hepatic gluconeogenesis during prolonged fasting, whereas hypoglycemia in fasted PPARα KO mice and FGF21 KO mice can be rectified by central administration of FGF21, suggesting an obligatory role of the PPARα-FGF21 axis in inducing hepatic gluconeogenesis during metabolic adaptation to prolonged fasting. In support of this notion, transgenic expression of FGF21 was sufficient to induce hepatic gluconeogenesis in the fed state to a level normally attained during prolonged fasting, whereas FGF21 KO mice showed impaired hepatic gluconeogenesis after 24-h fasting (9). By contrast, Badman et al. (33) found that fasting glucose levels were comparable between WT and FGF21 KO mice generated in their laboratory. This discrepancy may be due to different ages of mice or different fasting schedules. Nevertheless, they also showed impaired gluconeogenic gene induction (PGC-1α and PEPCK) in their FGF21 KO mice, which is consistent with our present study.

The hypothalamus has emerged as a key player in the regulation of glucose homeostasis, by sensing hormones and nutrients to initiate metabolic responses (34). Hormones such as insulin, glucagon, and leptin have been shown to regulate hepatic glucose metabolism through both central and peripheral actions (34,35). In this study, we provide several lines of evidence demonstrating that the stimulatory effect of FGF21 on hepatic gluconeogenesis is mainly mediated by its central actions on hypothalamic neurons. First, fasting hypoglycemia and impaired expression of hepatic gluconeogenic genes in both FGF21 KO and PPARα KO mice can be reversed by intracerebroventricular injection of FGF21 at a low dose that is not detectable in the peripheral blood. Second, FGF21-induced corticosterone release and hepatic gluconeogenesis can be completely blocked by central administration of a neutralizing antibody against its receptor FGFR1 or a pharmacological inhibitor of ERK1/2. Furthermore, both our animal and ex vivo studies showed that FGF21 stimulates CRH expression in the hypothalamus via ERK1/2-mediated activation of CREB, suggesting that CRH neurons are the direct target of FGF21.

Consistent with our observations in PPARα KO mice, administration of the PPARα agonist fenofibrate has been shown to elevate fasting serum levels of corticosterone, enhance hepatic gluconeogenesis, and increase blood glucose levels (36). These findings support the notion that PPARα coordinates metabolic adaptation responses to starvation through a bipartite mechanism: direct transactivation of the genes involved in hepatic fatty acid oxidation and ketogenesis (26,31) and indirect activation of the HPA axis via induction of FGF21. The synergic effect of this dual action of PPARα is to provide glycerol, fatty acids, ketones, and alanine as sources for gluconeogenesis to maintain glucose homeostasis.

In peripheral tissues, adipocytes have been identified as a main target that confers the pleiotropic metabolic actions of FGF21 (4,19). In white adipocytes, adiponectin expression and secretion is induced by FGF21 and mediates several metabolic benefits of FGF21 treatment (18,37). These findings together with our present report further support the notion that FGF21 exerts its diverse metabolic effects indirectly, by modulating the production of other hormones such as adiponectin and corticosterone. The central and peripheral actions of FGF21 on activation of the HPA axis and its peripheral actions on promoting adipose secretion of adiponectin appear to have opposite effects on hepatic glucose production. However, these two actions of FGF21 may occur under different physiological conditions. FGF21 secretion in adipose tissue is elevated in the fed state, which in turn serves as an autocrine signal to induce PPARγ activity and adiponectin secretion, thereby leading to enhanced hepatic insulin sensitivity (18). Conversely, the central effect of FGF21 on hepatic glucose production consequent to increased release of corticosterone occurs only during prolonged fasting. Such a dual effect of FGF21 may enable the control of blood glucose within a narrow range in different nutritional states.

In conclusion, our results demonstrate FGF21 as a critical hormonal regulator of glucose homeostasis during prolonged fasting, by coupling hepatic PPARα activation to corticosterone release via stimulation of the HPA axis, thereby leading to enhanced hepatic gluconeogenesis (Fig. 8). This novel liver-PPARα-FGF21-brain-corticosterone circuit can fine tune the cross talk between brain and liver to avoid hypoglycemia during nutrient deprivation or other adverse clinical events. Interestingly, a novel “hepato-adrenal syndrome” has recently been suggested, demonstrating a defect in HPA activation in patients with severe liver disease (38). Hypoglycemia is a common clinical problem in patients with a defect in HPA regulation such as hypopituitarism and Addison disease. It is also frequently seen in diabetic patients treated with insulin or other medications improperly. Unraveling the liver-adrenal circuit with FGF21 as key player may open new avenues for the treatment of these metabolic diseases with hypoglycemia unawareness.

Acknowledgments. The authors thank N. Itoh and M. Konishi (University of Kyoto) for FGF21 KO mice.

Funding. This work is supported by the National Basic Research Program of China (2011CB504004), the General Research Fund (HKU 783413 and 784111 to A.X.), the Collaborative Research Fund (HKU2/CRF/12R) from the Research Grants Council of Hong Kong, the Matching Fund for the State Key Laboratory of Pharmaceutical Biotechnology from The University of Hong Kong, and the Qatar National Research Fund (NPRP 6-428-3-113).

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

Author Contributions. Q.L. designed and performed the experiments, analyzed the data, and wrote the manuscript. L.Z. and J.Z. performed the experiments. Y.W., S.R.B., C.R.T., and H.D. provided technical assistance, reagents, and valuable advice for this study. K.S.L.L. supervised the study and edited the manuscript. A.X. conceived the concept, supervised the project, and wrote the manuscript. Q.L. 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 the accuracy of the data analysis.