1st Page

Journal of Animal Science - Symposia

Physiology and Endocrinology Symposium: FGF21: Insights into mechanism of action from preclinical studies12


This article in JAS

  1. Vol. 92 No. 2, p. 407-413
    Received: Aug 27, 2013
    Accepted: Nov 01, 2013
    Published: November 24, 2014

    3 Corresponding author(s):

  1. P. J. Antonellis,
  2. A. Kharitonenkov and
  3. A. C. Adams 3
  1. Lilly Research Laboratories, Lilly Corporate Center, Indianapolis, IN 46285


Fibroblast growth factor 21 (FGF21) is a multifaceted metabolic regulator which has several potential applications in the treatment of metabolic disease. When administered in vivo, FGF21 exhibits a plethora of actions, modulating metabolic homeostasis in a diverse manner. However, the mechanism and site of action underlying these effects were, until recently, entirely uncertain. Using mouse models lacking either FGF receptor isoform 1 (FGFR1) or βKlotho (KLB), a transmembrane co-factor critical for FGF21 action, our group and others sought to determine the tissue on which FGF21 acts and the receptor complex responsible for mediating its in vivo efficacy. Importantly, when KLB was ablated from all tissues mice were completely refractory to FGF21 action. Therefore, to determine the precise tissue of action we utilized mice with tissue specific deletion of FGFR1 in either adipose tissue or neurons, respectively. Surprisingly, in animals with neuronal FGFR1 loss there was no change in the metabolic activity of FGF21, suggesting a lack of central FGF21 action in the pharmacologic setting. In contrast, we found dramatic attenuation of metabolic efficacy in mice with adipose-specific FGFR1 ablation following either acute or chronic dosing with recombinant FGF21. Furthermore, several recent studies have suggested that the metabolic effects of FGF21 may occur via modulation of adipokines such as adiponectin and leptin. Importantly, the action of FGF21 via adipose tissue results in alterations in both secretion as well as systemic sensitivity to these factors. Therefore, while FGF21 itself does not seem to directly act on the CNS, leptin and other endocrine mediators may serve as intermediary facilitators of FGF21’s secondary central effects downstream of an initial and direct engagement of FGF21 receptor complex in adipose tissue. Further studies are required to delineate the precise mechanistic basis underlying the interplay between peripheral and central FGF21 modes of action in both the physiological and pharmacological settings.


Fibroblast growth factor (FGF) 21 was classified as an FGF based on its structure, as it possesses a canonical FGF domain and exhibits significant sequence homology with other FGF (Kharitonenkov, 2009). Three factors, FGF19, FGF21, and FGF23 comprise an ‘endocrine’ FGF subfamily due to absence of a heparin binding domain and presence in serum (Kurosu and Kuro-o, 2008; Itoh, 2010; Kharitonenkov and Larsen, 2011). In place of heparin, the endocrine FGF utilize Klotho co-factor proteins to permit activation of their respective fibroblast growth factor receptors (FGFR; Ogawa et al., 2007; Kharitonenkov et al., 2008; Kurosu and Kuro-o, 2008). Importantly, while the tissue distribution of the FGFR is widespread, expression of the co-factor β-klotho (KLB) is restricted to metabolically active tissues (Ito et al., 2000; Ogawa et al., 2007; Adams et al., 2012a).

In terms of its own tissue distribution, Fgf21 was identified in the liver (Nishimura et al., 2000), with further expression reported in white adipose tissue (WAT; Wang et al., 2008), brown adipose tissue (Hondares et al., 2010), muscle (Izumiya et al., 2008) and pancreas (Johnson et al., 2009). To date the major tissues responsible for FGF21 secretion remain to be identified. Furthermore, given the extremely wide range of FGF21 levels observed in human serum (Kharitonenkov and Larsen, 2011), debate exists as to the relevance of circulating levels to biological endpoints. This contention is supported by a recent report suggesting that circulating FGF21 may exist, at least in part, in a truncated form (Hager et al., 2013). This is of particular importance because this cleavage impairs FGF21 activity in vitro (Kharitonenkov et al., 2008).

While significant progress has been made in furthering understanding the molecular basis of FGF21 action, significant knowledge gaps still exist. The present paper discusses advances in elucidating aspects of the mechanistic basis for FGF21 action in vivo.


Mice lacking Fgf21 (FGF21KO) exhibit mild obesity and glucose intolerance in the basal state (Badman et al., 2009). Furthermore, when fed a high fat diet (HFD), FGF21KO animals gain more weight than their wild-type (WT) counterparts. In addition to excess weight gain, their basal glucose intolerance is exacerbated to a state of hyperglycemia and insulin resistance (Adams et al., 2013a). These findings are in agreement with the characterization of an independently-generated strain of Fgf21 knockout mice (Hotta et al., 2009). However, a third FGF21KO strain was reported to be lipodystrophic (Dutchak et al., 2012). Potential factors suggested to underlie these differences in phenotype between the 3 strains include genetic background, breeding strategy, and inclusion of soy in the diet. Therefore, to thoroughly characterize our own strain, we recently re-examined a large cohort of WT and FGF21KO mice (Adams et al., 2013a). Once again we found that when fed either standard chow or HFD, FGF21KO mice weighed significantly more than WT animals. By measuring body composition, we determined that the difference in body weight was due primarily to an increase in adipose tissue mass in the FGF21KO mice, while there was no difference in lean mass or total water, suggestive of either enhanced adipose mass accrual or diminished lipid mobilization. Conceptually, the elevated fat mass observed in FGF21KO mice is concordant with the ability of exogenous FGF21 to ameliorate obesity and its associated metabolic dysfunctions, in large part due to reduction of adipose mass (Coskun et al., 2008).

The observed phenotype of FGF21KO mice is also consistent with our previous reports of an FGF21-overexpressing (FGF21Tg) strain (Kharitonenkov et al., 2005). Indeed, whereas FGF21KO mice are mildly obese when fed either chow or HFD, FGF21Tg mice weigh less and have less fat mass when compared with WT mice (Fig. 1). In addition to being lean, FGF21Tg mice have improved glycemic control (Fig. 2A) due to enhanced insulin sensitivity (Fig. 2B). In fact, FGF21Tg animals are so exquisitely insulin sensitive that circulating insulin levels are undetectable during prolonged fasting (Fig. 2B). In addition, FGF21Tg mice exhibit noticeably reduced serum leptin levels in line with their dramatically reduced adiposity (Fig. 2C). However, a further drop in circulating leptin is still observed in the fasted state indicating that an extremely sensitive and functional leptin system remains in mice with FGF21 overexpression (Fig. 2C). When fed a HFD FGF21Tg mice gain weight more slowly and accrue significantly less adipose tissue mass than WT mice, due to increased energy expenditure and brown adipose tissue mass (Kharitonenkov et al., 2005). In addition to the lean phenotype and resistance to diet induced obesity observed in our own strain of FGF21Tg mice, an independent strain of FGF21-overexpressing mice also display significantly augmented lifespan (Zhang et al., 2012). Interestingly, mice from both FGF21Tg strains are shorter than their WT counterparts and exhibit altered IGF-1 tone, findings that are indicative of common physiology shared between the two independently-created strains (Inagaki et al., 2008). It is known that diminished IGF-1 signaling correlates with enhanced lifespan in mice (Xu et al., 2013), thus the altered IGF1 system in the FGF21Tg mice may play a role in their notable longevity.

Figure 1.
Figure 1.

Metabolic homeostasis is significantly altered in both our FGF21 transgenic (FGF21Tg) and FGF21 knockout (FGF21KO) mice. When fed either standard chow or high fat diet (HFD), our strains of FGF21Tg and FGF21KO mice exhibit opposing body weight phenotypes, with FGF21Tg mice displaying a lean phenotype while FGF21KO mice are mildly obese when compared with their wild-type counterparts. Concordant with these differing levels of adipose tissue, adiponectin secretion is elevated in our FGF21Tg mice and lowered in FGF21KO mice. See online version for figure in color.

Figure 2.
Figure 2.

FGF21 transgenic (FGF21Tg) mice exhibit improved glycemic control and enhanced sensitivity to the hormones insulin and leptin. Under fed conditions, FGF21Tg mice have lower basal levels of leptin (Panel A), insulin (Panel B), and serum glucose (Panel C). When challenged with an overnight fast FGF21Tg mice are able to maintain appropriate glucose levels (Panel C) despite extremely low levels of insulin (Panel B). Additionally, FGF21Tg mice are capable of proper leptin regulation when fasted (Panel A). Blood samples from fed and 24-h fasted 12 wk old, male wild-type and FGF21Tg mice (n = 6 per group) were collected on ice before storage of plasma at –80°C. Insulin (Crystal Chem Inc., Downers Grove, IL), leptin (Crystal Chem Inc.) and total adiponectin (BioVendor Inc., Karasek, Czech Republic) were measured by specific ELISA. Glucose was measured using a Hitachi 912 clinical chemistry analyzer (Roche Diagnostics, Indianapolis, IN).


Treatment with FGF21 Leads to a Profound Shift in Metabolic Tone

Pharmacological administration of recombinant FGF21 has been demonstrated to have profound metabolic effects in multiple species ranging from rodents to non-human primates (Kharitonenkov et al., 2007; Coskun et al., 2008; Mu et al., 2012; Adams and Kharitonenkov, 2013; Camacho et al., 2013). Acute FGF21 signaling is observed in liver, adipose tissue and pancreas, with FGF21 activating the immediate early gene response as early as 5 min posttreatment (Johnson et al., 2009; Fisher et al., 2011; Adams et al., 2013b). Physiologically, this activation is coupled to a reduction in serum glucose, insulin, triglycerides, and FFA, as well as an increase in adiponectin secretion (Adams et al., 2013c). Chronic exposure to FGF21 leads to sustained and progressive reductions in serum glucose, insulin and triglycerides, as well as a dose-dependent decrease in body mass (Kharitonenkov et al., 2005; Coskun et al., 2008). As in FGF21Tg mice, pharmacological treatment with FGF21 induces a dramatic elevation in energy expenditure. Recently, it has been demonstrated that in a chronic treatment setting FGF21 induces the expression of thermogenic genes, such as uncoupling protein 1 and cell death-inducing DFFA-like effector A in brown adipose tissue (Chartoumpekis et al., 2011; Fisher, 2012). Furthermore, FGF21 is able to induce ‘browning’ of WAT depots by activating these same genes leading to the appearance of brown-like adipocytes within WAT (Kharitonenkov et al., 2005; Fisher et al., 2012); however, the contribution of this browning effect to the pharmacologic efficacy of FGF21 remains to be clarified.

The Co-Factor KLB is Vital for FGF21 Action in Vitro and in Vivo

As stated earlier, unlike canonical FGF, FGF21 does not utilize heparin to bind to its receptor. In place of heparin, FGF21 interacts with the transmembrane protein KLB to facilitate binding to FGFR, an interaction that our group and others have shown is absolutely necessary for activation of the FGF21 receptor complex (Ogawa et al., 2007; Kharitonenkov et al., 2008; Yie et al., 2009). When the FGF21 protein is truncated artificially at its C terminal KLB interaction site, potency is greatly diminished in vitro (Kharitonenkov et al., 2008). However, the requirement of KLB in propagation of FGF21 metabolic actions in animals was only recently demonstrated, with earlier reports suggesting that FGF21 is able to function in a KLB-independent manner both in vitro (Zhang et al., 2006) and in vivo (Tomiyama et al., 2010). Furthermore, an initial publication on mice lacking KLB (KLBKO) claimed that FGF21 signaling was fully preserved in KLBKO mice, a potential controversy suggestive of either poor translation of in vitro findings to the in vivo condition, or confounding methodological discrepancies (Tomiyama et al., 2010). To further address the need for KLB as a member of the FGF21 receptor complex, we generated a strain of KLBKO mice and examined FGF21 action in vivo (Adams et al., 2012a).

Following FGF21 treatment in WT mice, target engagement in WAT is observed in the form of activation of the immediate early genes early growth response protein 1 (EGR1) and cFOS, a response which is absent in our KLBKO mice (Adams et al., 2012a). This result directly opposes the conclusion of Tomiyama et al. (2010) but is consistent with the later studies of Yang et al. (2012). We suggest that the differences between these data sets are possibly due to the source of protein utilized. Tomiyama et al. (2010) employed non-purified supernatants from FGF21 cDNA transfected cells while later studies, including our own, utilized purified FGF21 protein (Adams et al., 2012a; Yang et al., 2012).

The absence of target engagement at the molecular level in the KLBKO mice also translates to a lack of subsequent metabolic regulation. Indeed, in our own experiments there was a profound FGF21-induced shift in energy homeostasis observed only in WT animals. Furthermore, FGF21-mediated effects on weight loss, glycemia, and lipid homeostasis were evident in WT mice but all were absent in the KLBKO animals. Of note, the metabolic effects of FGF21 in WT mice were largely reflected at the transcriptional level by significant induction of ‘browning’ related genes in WAT, which as discussed previously, may be coupled to the increased energy expenditure elicited by FGF21 treatment. In the KLB null animals, these effects were not observed, concordant with a lack of increased caloric expenditure (Adams et al., 2012a). Thus, lack of WAT activation at the signaling level due to KLB ablation translates to a complete inability of FGF21 to impact its downstream metabolic endpoints.

In parallel with our work, another group has recently reported on the metabolic phenotype of both whole body and adipose-specific KLB null mice (Ding et al., 2012) and reached similar conclusions. In their report, Ding et al. (2012) describe attenuated signaling in KLBKO mice following FGF21 treatment, similar in magnitude to that observed in our own KLBKO strain. An important point when interpreting their data is that to create a situation of excess FGF21, KLBKO mice were bred to FGF21 transgenic (FGF21Tg) mice, as opposed to injections of recombinant FGF21 utilized by our group and others (Adams et al., 2012a; Yang et al., 2012). However, despite these differences in experimental methodology, the conclusions of Ding et al. (2012) are concordant with our own observations. The summation of the recent data from these mouse models generated by several groups reinforces the notion that the presence of KLB is absolutely required for FGF21 action in vivo.

Adipose FGFR1 is Critical for the Metabolic Action FGF21

We initially reported bioactivity of FGF21 in a glucose uptake assay utilizing 3T3L1 adipocytes, a cell line that primarily expresses FGFR1, the main FGF receptor isoform present in WAT (Kharitonenkov et al., 2005). We later went on to demonstrate in vitro that FGFR1, in the context of KLB expression, is sufficient to facilitate both FGF21 and FGF19 signaling (Adams et al., 2012b). Given these facts and the knowledge that KLB is expressed in large quantities in adipose tissue, we sought to evaluate the contribution of this organ to the metabolic effects of FGF21. To assess the requirement of FGFR1 in adipose tissue for mediating both signaling and metabolic actions of FGF21 and FGF19 action in vivo, we utilized FGFR1lox/lox AP2CRE (FR1KO) mice that have a conditional deletion of FGFR1 in adipose tissue (Yang et al., 2012).

In adipose tissue of FR1KO mice, the FGF21-induced transcription of EGR1 was reduced, whereas induction in pancreas, a tissue that also highly expresses FGFR1, remained unaltered (Adams et al., 2013b). Unexpectedly, the immediate early gene response in the liver, another previously-speculated direct target of FGF21 was also significantly attenuated, indicating that some hepatic actions of FGF21 may be mediated indirectly via activation of adipose tissue. Following a single bolus dose of recombinant FGF21 in WT animals, we observed the known spectrum of FGF21 activity including lowering of plasma glucose, insulin, triglycerides, and FFA concentrations (Adams et al., 2013b). However, with the notable exception of FFA lowering, these effects were absent in FR1KO animals. Furthermore, a rapid elevation of adiponectin was observed in WT mice but was absent in FR1KO mice (Adams et al., 2013b).

Following our initial acute experiments we went on to assess the impact of adipose FGFR1 deletion on the response to chronic FGF21 dosing. As with our initial acute experiments, we observed the normal breadth of FGF21 action in the WT DIO animals including weight loss, increased energy expenditure, improved glycemia, and ameliorated lipid levels in circulation. However, the majority of these effects were lost in the FR1KO mice confirming that activation of adipose FGFR1/KLB is absolutely required for the metabolic effects of FGF21 action in both the acute and chronic setting. Supporting this adipose-centric hypothesis are recent reports regarding the treatment of lipodystrophic mice with recombinant FGF21, which display a dramatically diminished response to treatment, likely due to a paucity of the target organ for FGF21 action (Veniant et al., 2012). In addition to whole body deletion, Ding et al. (2012) also conditionally deleted KLB from adipose. Importantly, animals with fat-specific KLB ablation lacked the insulin sensitization and lipid lowering effects of FGF21 (Ding et al., 2012). Additionally, the normal increase in adiponectin observed with FGF21 treatment was also absent in adipose-specific KLB null mice. This indicates that adiponectin may represent a biomarker for FGF21 target engagement of FGFR1/KLB in adipose tissue. Additionally, mice with adipose specific deletion of FGF1 present with marked metabolic deficiencies further confirming the important role FGF signaling in adipose for the maintenance of normal glucose homeostasis (Jonker et al., 2012).


Emerging evidence indicates that adipose tissue-derived secretory factors are likely to be critical drivers of the metabolic improvements elicited by FGF21. Adiponectin and leptin are both prominent adipokines that may permit downstream responses to FGF21. Adiponectin significantly ameliorates hyperglycemia, insulin resistance, inflammation, and lipotoxic damage (Turer and Scherer, 2012). While our FGF21-transgenic mice (Kharitonenkov et al., 2005) only show a 10% increase in total circulating adiponectin, they exhibit a more striking elevation of high molecular weight adiponectin concentrations (Holland et al., 2013). Furthermore, in WT animals, administration of recombinant FGF21 increases concentrations of both high and low molecular weight adiponectin (Holland et al., 2013). These findings are corroborated by recent reports from other groups demonstrating that adiponectin is elevated in an independently-derived strain of FGF21-transgenic mice (Ding et al., 2012). When compared with WT mice, our FGF21KO animals (Badman et al., 2009) on a high-fat diet show significantly impaired adiponectin production (Holland et al., 2013). Thus, FGF21 is likely critical for maintaining basal adiponectin levels, particularly during states of metabolic dysfunction.

Importantly, acute FGF21 signaling is not compromised in adiponectin null (Adn–/–) mice when compared with their WT counterparts. However, only in WT animals with intact adiponectin signaling is FGF21 able to trigger a cascade of metabolic events leading to correction of hyperglycemia and hyperinsulinemia (Holland et al., 2013; Lin, 2013). Correlated with the lack of improvement in glycemia and insulin sensitivity, there was a lack of ceramide lowering in FGF21 treated Adn–/– animals. Supportive of partitioning of the metabolic endpoints of FGF21, weight loss, elevated energy expenditure, and reduced circulating lipids were yet evident in FGF21-treated Adn–/– mice (Holland et al., 2013; Lin, 2013). Taken as a whole, these data support the contention that FGF21 is a potent regulator of adiponectin secretion and demonstrate that FGF21 critically depends on adiponectin to exert its glycemic and insulin sensitizing effects in vivo.


If one closely examines previous data on FGF21 activity in vivo, another interesting facet of its action becomes readily apparent, specifically, FGF21 seems to act in part not only by modulating circulating levels of other hormones but also the expression of their respective receptors. Indeed, in our initial manuscript describing the metabolic effects of FGF21 administration to diet induced obese mice, we reported that one of the most significantly induced genes was not a downstream effector, as we had imagined, but the leptin receptor (Coskun et al., 2008). At the time, this was noted as a possible mechanism via which FGF21 may act but was not investigated further until recently. If one couples the finding of decreased leptin levels in both FGF21Tg mice and in animals treated with recombinant FGF21 with the observation of increased leptin receptor expression, a picture of causal rather than correlative hormonal sensitization begins to form. Supportive of this contention is the attenuation of body weight reduction observed in leptin deficient animals treated with FGF21 (Coskun et al., 2008). Additional studies have shown that the receptors for other factors modulated by FGF21, such as insulin (Coskun et al., 2008) and adiponectin (A. C. Adams, unpublished data), are also up-regulated by chronic FGF21 treatment.


The favorable effects of FGF21 make it an attractive target for the treatment of type 2 diabetes; however, the native protein presents significant challenges as a pharmaceutical agent. Indeed, human wild-type FGF21 has a relatively short half-life in the blood stream, lasting approximately 1 h, and is a subject to significant inactivating proteolytic cleavage (Kharitonenkov et al., 2007). In an attempt to increase both the efficacy and time of action of FGF21 several groups have reported on novel engineered variants of the wild-type protein. A long-acting analog was generated by fusing a fragment crystallizable region (Fc) to a variant of human FGF21. This molecule showed similar or greater efficacy to WT FGF21 in both mice and monkeys while being administered less often (Veniant et al., 2012). Site-specific pegylation is also able to confer potency and increased half-life (Huang et al., 2011; Mu et al., 2012). Indeed, it was demonstrated that a single low dose of pegylated FGF21 was able to normalize whole body glucose in insulin resistant mice (Camacho et al., 2013). Another more stable FGF21 variant produced and validated, in part, by our group at Eli Lilly and Company (LY2405319), was generated by the addition of a disulfide bond as well as the removal of amino acids susceptible to cleavage when this variant is produced in yeast (Kharitonenkov et al., 2013). Chronic administration of LY2405319 was able to improve metabolic homeostasis in various rodent models and diabetic rhesus monkeys, causing a reduction in body weight as well as plasma glucose, triglycerides, and cholesterol (Kharitonenkov et al., 2013). In a landmark first study in human subjects presenting with type 2 diabetes, treatment with LY2405319 led not only to weight loss and improved lipid profiles but also to heightened insulin sensitivity (Gaich et al., 2013). Taken as a whole these results demonstrate the feasibility of developing FGF21 analogs that can be used to translate the various metabolic effects of FGF21 from preclinical species into man.


Data from our group and others using several novel mouse lines indicates that FGFR1/KLB in adipose tissue is the key receptor complex and site of action for the majority of the metabolic effects of exogenously-administered FGF21 in living animals. Furthermore, recent studies have begun to explore events downstream of FGFR1 activation in adipose tissue, such as secretion of adipokines, which represent the next step in FGF21’s in vivo action. Thus, we propose that fat is the site of action via which FGF21 is able to modulate the production and secretion of endocrine factors that produce synergistic, integrated metabolic outputs. Finally, while the framework we present here is aimed to provide a better understanding of the in vivo physiology of FGF21, significant work remains to further deconvolute the complex biology of this novel factor. Taken as a whole the recent data we detail here represent a significant advance in the understanding of the in vivo mode of action of FGF21, which spans a variety of mechanistic levels from initial receptor activation events to the integration of its varied downstream metabolic endpoints.




Be the first to comment.

Please log in to post a comment.
*Society members, certified professionals, and authors are permitted to comment.