Dual actions of fibroblast growth factor 19 on lipid metabolism.

Elevated triglyceride (TG) and cholesterol levels are risk factors for cardiovascular disease and are often associated with diabetes and metabolic syndrome. Recent reports suggest that fibroblast growth factor (FGF)19 and FGF21 can dramatically improve metabolic dysfunction, including hyperglycemia, hypertriglyceridemia, and hypercholesterolemia. Due to their similar receptor specificities and co-receptor requirements, FGF19 and FGF21 share many common properties and have been thought to be interchangeable in metabolic regulation. Here we directly compared how pharmacological administration of recombinant FGF19 or FGF21 proteins affect metabolism in B6.V-Lep(ob)/J leptin-deficient mice. FGF19 and FGF21 equally improved glucose parameters; however, we observed increased serum TG and cholesterol levels after treatment with FGF19 but not with FGF21. Increases in serum TGs were also observed after a 4-day treatment with FGF19 in C57BL6/J mice on a high-fat diet. This is in contrast to many literature reports that showed significant improvements in hyperlipidemia after chronic treatment with FGF19 or FGF21 in high-fat diet models. We propose that FGF19 has lipid-raising and lipid-lowering actions mediated through different FGF receptors and target tissues, and the results described here provide a potential mechanism that may explain the inconsistency in the reported effects of FGF19 on lipid metabolism.

protocol, and treated with various FGF molecules as previously described ( 15 ). Cells were collected 15 min after treatment, snap frozen in liquid nitrogen, and resuspended in lysis buffer, and the total and phosphorylated ERK were measured using an MSD whole cell lysate Phospho-ERK1/2 kit (Meso Scale Discovery) according to the manufacturer's instructions. All experiments were performed in duplicate.

Animals and treatments
All animal housing conditions and research protocols were approved by the Amgen Institutional Animal Care and Use Committee. They were housed in a specifi ed-pathogen free, AAALAC, Intl-accredited facility in ventilated microisolators. Procedures and housing rooms are positively pressured and regulated on a 12:12 dark:light cycle. Six week-old B6.V-Lep ob /J male mice (stock 632; The Jackson Laboratory) were fed standard chow (2020× Teklad global soy protein-free extruded rodent diet; Harlan) and received reverse-osmosis purifi ed water ad libitum. Mice were initially randomized in groups of 10 animals and regrouped (three to four animals per cage) based on their body weight and baseline glucose after a 4 h fasting period. Preconditioned C57BL/6J male mice (stock 380050; Jackson Laboratories) fed a high-fat diet (60 kcal%) for 12 weeks from 6 weeks of age were singly housed and fed a D12492 (Research Diet) diet. Mice were then randomized. Glucose was measured using Accu-chek active glucometers and test strips (Roche Diagnostics). Purifi ed recombinant proteins or vehicle (PBS) were injected intraperitoneally daily at the stated concentrations for the duration as indicated. On the last day, mice were fasted for 4 h. Dosing of vehicle and proteins occurred 1 h before baseline glucose measurement. At termination, mice were euthanized, the blood was collected by cardiac puncture, and liver tissues were harvested. The bile acid pool size was determined as bile acid content of the liver, the gallbladder, and the small intestine and its contents. In the case of CYP7A1 measurements, dosing occurred 5 h before tissue collection. For data in supplementary Fig. I, FGF19 or FGF21 were administered via continuous subcutaneous infusion (150 g/kg/day) via osmotic minipumps (model 1007D; Alzet, Cupertino, CA) implanted in the interscapular region. Livers were collected 3 days after pump implantation.

Serum fractionation
Serum samples were fractionated by size exclusion chromatography (SEC) using an AKTA FPLC (GE Healthcare). Samples (100 l) were loaded onto a Superose 6 10/300 GL column (GE Healthcare) and resolved in PBS containing 1.5 mM EDTA and 0.015% sodium azide at a fl ow rate of 0.8 ml/min. Fractions (300 l) were collected for lipid profi le analyses. SEC fractions (80 l) were incubated in 96-well plates with 120 l analytic reagent (Infi nity Triglyceride by Thermo Scientifi c or Chol-E; Wako Chemicals). Plates were incubated for 30 min at room temperature, and absorbance was measured using a Spectramax 384plus plate reader (Molecular Devices). Lipid content was quantifi ed by comparison to parallel analytic standards from Wako Chemical.

Serum triglyceride, cholesterol, and bile acids measurement
Serum samples were diluted 1:5 in water. Diluted sample (4 l) was added to each well in a 384-well plate. To the sample, 36 l of triglyceride reagent or cholesterol reagent (Infi nity Triglycerides or Cholesterol Liquid Stable Reagent; Thermo Scientifi c) was added. Samples and standards were incubated at room temperature for 15min, and absorbance was read at 500 nm. Triglyceride and cholesterol values were calculated from the standard curve. Bile acids were measured using the Total Bile biased toward FGFR1c/ ␤ Klotho and with novel mimetic proteins that specifi cally activate only the FGFR1c/ ␤ Klotho receptor complex demonstrate the essential role this receptor complex plays in mediating the effects of FGF19 and FGF21 on regulating the glucose and lipid metabolism as well as energy expenditure ( 14 ).
Despite extensive similarities in their in vivo and in vitro pharmacology, differences exist between FGF19 and FGF21. FGF19 can also signal through FGFR4, but FGF21 cannot (4,15). The ability of FGF19 to activate FGFR4, the predominant receptor expressed in the liver, has been linked to its regulation of bile acid homeostasis. It has been proposed that FGF19 is an FXR target gene ( 16 ). Postprandial increases in intestinal bile acid levels activate FXR in intestinal epithelia, which in turn induce expression and secretion of FGF19. FGF19 then acts as the enterohepatic signal to suppress hepatic expression of cholesterol 7 ␣ -hydroxylase (CYP7A1), the enzyme responsible for the rate-limiting step of bile acid synthesis. This occurs in the liver via activation of hepatic FGFR4, hence completing a negative feedback loop on bile acid synthesis ( 16 ). In addition, FGF19 has been shown to affect expression of other Cyp enzymes that may lead to alterations in bile acid pool composition ( 17 ). However, given that the elimination of cholesterol is primarily through conversion to bile acids and subsequently excreted, FGF19 treatment would be expected not only to reduce bile acid levels but also to increase cholesterol levels. In addition, bile acids are known to affect TG homeostasis. A reduction in bile acid levels due to CYP7A1 defi ciency or from treatment with bile acid-binding resins in humans leads to elevated TG levels (18)(19)(20)(21). Therefore, the expected effects of increased cholesterol and TG levels due to reduction or alteration in bile acid pool size and composition by treatment with FGF19 may appear to be inconsistent with the observed reduction of these parameters in FGF19 transgenic mice and in mice treated chronically with recombinant FGF19 protein ( 8,22 ).
In the present study, we compared the effects of FGF19, FGF21, and a variant of FGF19 that only activates FGFR4 on metabolism in rodent models. Our results revealed a new action of FGF19 and suggest that FGF19 has a dual function in lipid metabolism.

Expression and purifi cation of recombinant FGF19 and FGF21 proteins
Recombinant FGF19, FGF21, and FGF19dCTD proteins used in this study were expressed and purifi ed from Escherichia coli as previously described ( 13,15  our previous report demonstrating that activation of FGFR4 alone does not improve glucose parameters ( 15 ).
Triglyceride and cholesterol concentrations were also measured at the end of the 1 week study. Mice from the FGF19 treatment group displayed increased serum triglyceride and cholesterol levels after the 1 week treatment, whereas the PBS and FGF21 treatments resulted in no change in either of these parameters ( Fig. 1C, D ). FGF19 is known to suppress CYP7A1 expression in the liver through activation of FGFR4. Consistent with this, mice treated with FGF19 displayed reduced liver CYP7A1 expression, whereas no signifi cant changes were observed with FGF21 treatment because FGF21 does not activate FGFR4 (supplementary Fig. I ).
Serum lipoproteins VLDL, LDL, and HDL were separated by SEC, and triglyceride and cholesterol content of lipoprotein fractions was quantifi ed ( Fig. 2 ). In serum from FGF19-treated mice, higher levels of triglyceride were detected in VLDL fractions ( Fig. 2A ). VLDL carries the majority of the triglyceride in serum. Furthermore, cholesterol content was increased in all lipoprotein fractions, including VLDL, LDL, and HDL, in the FGF19 treatment group ( Fig. 2B ).
To test if the increase in plasma TG observed in the ob/ob model could be resolved in a longer-term treatment, a 2 week study was carried out with daily injection of 1 mg/kg FGF19 protein in 7 week-old male ob/ob mice. Similar to results shown in Fig. 1 , FGF19 injection induced signifi cant reduction in plasma glucose levels at week 1 and week 2 after initiation of the treatment, and there is also a signifi cant improvement in the ability of FGF19-treated mice to dispose glucose in an oral glucose tolerance test ( Fig. 3 ). However, similar to the 1 week treatment shown in Fig. 1 , FGF19 Acids Assay Kit (Calorimetric; BQ Kits, San Diego, CA) according to the manufacturer's instructions. The bile acid pool size was determined from the total extract of liver, gallbladder, and the small intestine and its contents and was expressed as µmol of bile acid/100 g of body weight.

FGF19 treatment unexpectedly increased plasma TG and cholesterol levels in ob/ob mice
Previous reports described a reduction in plasma TG and cholesterol with chronic FGF19 exposure in transgenic animals or long-term treatment with recombinant protein in high-fat diet mouse models. However, in our studies of the effects of FGF19 on metabolism in leptindefi cient ob/ob mice, we unexpectedly observed increases in plasma TG and cholesterol levels ( Fig. 1 ). Ob/ob mice were sorted by body weight and serum glucose levels and then injected intraperitoneally with a PBS control, 1 mg/kg FGF19, or 1 mg/kg FGF21 daily for 7 days. Perhaps due to the relatively short treatment, no signifi cant differences in body weight were observed among the three groups ( Fig. 1A ). However, consistent with results from numerous publications, fasting serum glucose levels were signifi cantly reduced in mice injected with FGF19 or FGF21 as compared with the PBS control group at the end of the 1 week treatment ( Fig. 1B ). These data confi rmed previous studies that showed that FGF19 and FGF21 were nearly interchangeable in terms of glucose metabolism and that activation of FGFR4 by FGF19 did not appear to add benefi t toward glucose regulation. This is also consistent with obesity (DIO) mouse models ( 8,17 ), we wondered if our observations in ob/ob mice were model specifi c. In our previous studies in DIO mice, we also reported a reduction in plasma TG levels after daily injection of FGF19 for 2 weeks ( 14 ). Therefore, we explored whether there could be a different acute effect of FGF19 on plasma TG levels in the DIO mice.
Male C57BL/6J mice were put on a high-fat diet at 4 weeks of age. After 14 weeks on the high-fat diet, the animals were divided into groups (n = 12) based on body weight, glucose, and triglyceride. Mice were injected intraperitoneally daily with PBS or 1 mg/kg FGF19 for 10 days. Body weight and glucose were measured. Serum samples were collected by tail bleeding for measuring triglyceride and total cholesterol level. In this study, FGF19 treatment did not induce signifi cant body weight changes and induced a trend (with one signifi cant time point) in reducing fasting plasma glucose levels during the treatment ( Fig. 4A, B ). Insulin levels were signifi cantly reduced with the FGF19treated group, indicating improvements in insulin sen sitivity ( Fig. 4C ). Similar to ob/ob mice, the FGF19 treatment induced a signifi cant increase in plasma TG levels, but only transiently. The increase peaked at day 4 after beginning FGF19 injection and returned to no signifi cance by day 10 ( Fig. 4D ). Cholesterol levels were not signifi cantly affected at these time points (supplementary Fig. II ). These results suggest that the observed increase in plasma TG levels by FGF19 treatment is perhaps not model specifi c.

Selective activation of liver FGFR4 increased plasma TG and cholesterol levels without affecting glucose metabolism
Bile acid metabolism can affect triglyceride and cholesterol homeostasis. Therefore, we wanted to determine if the observed increase in TG and total cholesterol with FGF19 treatment could be at least in part due to its ability to regulate bile acid production via activation of liver FGFR4. treatment also increased plasma TG and cholesterol levels at the end of the 2 week treatment ( Fig. 3D, E ).

Acute FGF19 treatment also increased plasma TG levels in diet-induced obesity mice
Because the increase in plasma TG levels we observed after FGF19 treatment was in contrast to the previously reported lowering effects of FGF19 on plasma TG in diet-induced Values are means ± SEM (n = 10). ** P < 0.01, *** P < 0.001, and **** P < 0.0001 ( t -test). activation of FXR ( 23 ) and that bile acid-binding resins increase VLDL triglyceride content ( 20 ). Furthermore, the addition of bile acids to cultured rat and human hepatocytes decreased VLDL secretion ( 24 ), indicating a reciprocal relationship between the bile acid pool and triglyceride content. Bile acid is also the end product of cholesterol catabolism. A great proportion of hepatic cholesterol is used in bile acid synthesis catalyzed by CYP7A1; thus, defi ciency in CYP7A1 leads to elevated total cholesterol ( 21 ). Therefore, inhibition of bile acid synthesis would be expected to cause an increase in plasma TG and cholesterol levels.
FGF19 is a unique endocrine FGF molecule, and its effects on bile acid metabolism have been well established. FGF19 activates FGFR4, which is predominantly expressed in the liver, and activation of this receptor by FGF19 has been shown to reduce the bile acid pool size ( 16,25 ) and affect bile acid composition ( 17 ). This pathway has been proposed as the enterohepatic signal that mediates postprandial inhibition of bile acid synthesis ( 16,25 ). Given the connections between bile acid, TG, and cholesterol homeostasis as discussed above, FGF19 treatment would also be expected to increase plasma TG and cholesterol levels. This is what we observed in our study. Treatment with FGF19 for 7 days in the ob/ob mouse model caused a reduction in bile acid and plasma glucose levels and an increase in TG and total cholesterol levels ( Figs. 1, 2, and  5 ). A variant of FGF19, FGF19dCTD, which specifi cally activates only FGFR4, recapitulated FGF19's effects on bile acids, TG, and cholesterol levels but, distinct from FGF19, had no effect on glucose ( Fig. 5 ). This confi rms that the FGF19-induced changes in lipid profi le is due to activation of liver FGFR4, whereas glucose regulation is mediated through a different FGFR. However, our fi nding that FGF19 We previously generated an FGF19 variant, FGF19dCTD, that selectively activates FGFR4 but not FGFR1c, -2c, or -3c ( Fig. 5A ) ( 15 ). This molecule allowed us to isolate FGFR4 activation and assess its regulation of bile acid, TG, and cholesterol metabolism without interference from simultaneous activation of other FGFRs. Similar to the experiment described in Fig. 1 , no signifi cant differences in body weight were observed in ob/ob mice treated with 5 mg/kg FGF19 or 5 mg/kg FGF19dCTD daily for 7 days as compared with the PBS vehicle group ( Fig. 5B ). However, in contrast to treatment with FGF19 or FGF21, FGF19dCTD did not display reduced plasma glucose levels ( Fig. 5C ). This is consistent with our previous study ( 15 ) and suggests that activation of FGFR4 alone is not suffi cient to improve glucose metabolism.
Consistent with its ability to activate FGFR4, FGF19dCTD treatment inhibited liver CYP7A1 expression and reduced bile acid pool size similar to native FGF19 ( Fig. 5D, 5E ; supplementary Fig. III ). Like native FGF19, treatment with FGF19dCTD increased plasma TG and total cholesterol levels ( Fig. 5F, G ). Therefore, selective activation of FGFR4 by FGF19dCTD separated FGF19's effects on lipid metabolism from its regulation of glucose metabolism, indicating that FGFR4 is primarily responsible for the inhibition of bile acid production and increased TG and total cholesterol levels associated with FGF19 treatment over the 7 day study.

DISCUSSION
Bile acid metabolism can affect triglyceride and cholesterol homeostasis. It has been shown that bile acid treatment reduces serum triglyceride concentrations through  in these studies. In contrast, with the short duration of treatment with FGF19, the lipid-raising action played a more dominant role due to its faster kinetics. This is seen in our study in DIO mice where the TG increase was seen only on day 4 and to a lesser extent on day 7 and then returned to baseline ( Fig. 4C ). Although the exact mechanism for the TG increase is unclear, it is possible that it is mediated through FXR or other bile acid receptors by the alteration in bile acid pool composition or size. However, our fi ndings appear to resolve the inconsistency between the observed hypolipidemic effects upon long-term FGF19 treatment as well as the expected hyperlipidemic effects from FGF19 action on liver bile acid metabolism. ob/ob mice seem more sensitive to the hyperlipidemic effects of FGF19 because the plasma TG and cholesterol levels remained elevated even 2 weeks after FGF19 treatment ( Fig. 3 ). Recent results suggest that although leptin is not required for FGF21's antidiabetic actions, leptin may exert additive or synergistic effects to FGF21 on glucose and body weight lowering and restores FGF21 sensitivity in lipodystrophyic mice ( 30 ). Whether there is also a cross talk between FGF19 and leptin pathways and whether leptin defi ciency exacerbates the hyperlipidemic effects of FGF19 is an interesting question to explore in the future.
In conclusion, due to its ability to specifi cally activate multiple FGF receptors, FGF19 can have two separate and opposite effects on lipid metabolism. Unlike previous studies that suggested that FGF19 might be solely benefi cial for treatment of hyperlipidemia, our data reveal another consequence of FGF19 action on lipid metabolism. By comparison, lacking an ability to signal through FGFR4, FGF21 is probably a "cleaner" molecule regarding glucose and lipid metabolism. Finally, it is worthwhile to speculate that antagonizing FGFR4 may provide benefi t for treating hyperlipidemia.
The authors thank Murielle Veniant and Scott Simonet for helpful discussions and critical reading of the manuscript and Joanne Greenberg for editing this manuscript. causes an increase in plasma TG and cholesterol has never been reported in the literature. On the contrary, improvements in hyperglycemia and hyperlipidemia have been reported in mice receiving chronic treatment with FGF19 protein and in FGF19 transgenic mice under high-fat diet conditions ( 8,22 ). FGF21 is a unique FGF molecule that belongs to the same subfamily as FGF19 ( 26 ). Chronic exposure to FGF21 in transgenic mice or with prolonged injection of recombinant protein led to pharmacology identical to that of similar chronic exposure of FGF19: both improved glucose metabolism and reduced plasma TG, cholesterol, and body weight ( 9,27 ). Recent evidence suggests that these beneficial effects observed with FGF21 or FGF19 treatment are not mediated through FGFR4 but distinctly through the FGFR1c/ ␤ Klotho complex. For example, treatment of mice with an FGF molecule biased toward the FGFR1c/ ␤ Klotho receptor complex or FGFR1 agonist antibody can recapitulate metabolic phenotypes induced by FGF19 and FGF21, including improvement of hyperglycemia and hyperlipidemia ( 14,28 ). These effects were absent in lipoatrophic animals ( 28 ), suggesting FGFR1 in adipose tissue is the main target of FGF21 action. Activation of FGFR1 in adipose tissue causes increases in metabolic rate and energy expenditure, partly due to brown adipose tissue activation ( 28 ). The observed improvement in hyperlipidemia may result from a combination of directly increased LDL receptor expression in hepatocytes ( 29 ) and perhaps indirectly through increased energy expenditure. The increased mobilization of lipids through oxidative metabolism ( 22,28 ) may lead to lower lipid content in liver and plasma.
The metabolic effects of FGF19 and FGF21 are often thought to be interchangeable. However, our study shows that FGF19 can raise serum triglyceride and cholesterol concentrations in pharmacological studies at supraphysiologic levels, but not FGF21. In retrospect, this is not surprising because the receptor specifi cities of these two molecules are not identical; only FGF19 can activate FGFR4 in the liver and suppress bile acid synthesis. Therefore, to reconcile increased plasma TG and cholesterol levels we observed after FGF19 treatment with the hypolipidemic effects previously reported, we believe FGF19 has two opposing actions on lipid metabolism: one lipid raising action mediated through the activation of FGFR4 and modulation of bile acid synthesis in the liver, and a second lipid-lowering action through FGFR1c primarily in adipose and perhaps other nonhepatic tissues ( Fig. 6 ). Lipid homeostasis thus results from the combination of these two actions. The kinetics of these two actions may be different. FGFR4-mediated CYP7A1 suppression can be observed within hours after FGF19 treatment ( 15 ), whereas increased metabolic rate may take longer to manifest. Prior studies in DIO mice may have missed this observation because measurements of serum lipids were made only after chronic treatment. At this point, body lipid content may have already decreased due to increased energy expenditure along with increased uptake and subsequent metabolism through alternative pathways. Therefore, the lipid-lowering action of FGF19 became predominant over the opposing lipid-raising action Fig. 6. Schematic representation of the proposed role of FGF19 and FGF21 on lipid metabolism mediated through different FGF receptors. Activation of liver FGFR4 by FGF19 may reduce bile acid levels and alter bile acid pool composition, which potentially leads to increased TG and total cholesterol levels. Activation of FGFR1c by FGF19 or FGF21 may lead to improved glucose metabolism and reduce TG and total cholesterol levels. Plasma lipid homeostasis is determined by the combined actions of these two pathways. Supplemental Material can be found at: