Fibroblast growth factor-21 improves insulin action in nonlactating ewes

Abstract

During metabolically demanding physiological states, ruminants and other mammals coordinate nutrient use among tissues by varying the set point of insulin action. This set point is regulated in part by metabolic hormones with some antagonizing (e.g., growth hormone and TNFα) and others potentiating (e.g., adiponectin) insulin action. Fibroblast growth factor-21 (FGF21) was recently identified as a sensitizing hormone in rodent and primate models of defective insulin action. FGF21 administration, however, failed to improve insulin action in dairy cows during the naturally occurring insulin resistance of lactation, raising the possibility that ruminants as a class of animals or lactation as a physiological state are unresponsive to FGF21. To start addressing this question, we asked whether FGF21 could improve insulin action in nonlactating ewes. Gene expression studies showed that the ovine FGF21 system resembles that of other species, with liver as the major site of FGF21 expression and adipose tissue as a target tissue based on high expression of the FGF21 receptor complex and activation of p44/42 extracellular signal-regulated kinase (ERK1/2) following exogenous FGF21 administration. FGF21 treatment for 13 days reduced plasma glucose and insulin over the entire treatment period and improved glucose disposal during a glucose tolerance test. FGF21 increased plasma adiponectin by day 3 of treatment but had no effect on the plasma concentrations of total, C16:0-, or C18:0-ceramide. Overall, these data confirm that the insulin-sensitizing effects of FGF21 are conserved in ruminants and raise the possibility that lactation is an FGF21-resistant state.

INTRODUCTION

Variation in insulin action is a fundamental mechanism coordinating nutrient partitioning and metabolic health in ruminants. Naturally occurring examples include late pregnancy and early lactation when insulin resistance (IR) drives a growing proportion of the maternal glucose supply to the fetoplacental unit or the mammary gland (1–3). Not surprisingly, the ability to promote IR explains a significant portion of the positive effects of nutritional and pharmacological modulators on productivity in lactating ruminants (e.g., dietary palmitic acid supplementation and exogenous growth hormone administration) (4–6). The glucose partitioning effects of IR, however, can become maladaptive particularly when combined with excessive negative energy balance or obesity and the consequent increase in liver uptake of fatty acids mobilized from adipose tissue (7, 8). Under these conditions, hepatic capacity for disposition of fatty acids via complete oxidation or export as VLDL is exceeded, and fatty acids accumulate in the form of triglyceride (8). In a subset of animals, liver also metabolizes an excessive proportion of fatty acids to ketone bodies resulting in circulating levels in excess of 3 mM and manifestation of clinical diseases such as pregnancy toxemia in late pregnant ewes and ketosis in early lactating dairy cows (7–9).

The set point for whole animal insulin response is regulated by metabolic hormones, with some potentiating (e.g., adiponectin) and others interfering (e.g., growth hormone and TNFα) with insulin action (10–12). A recent development in this context is the identification of fibroblast growth factor-21 (FGF21) as a potent insulin sensitizer (13–15). Unlike most other members of the FGF superfamily, FGF21 acts in an endocrine manner because it lacks the heparin-binding domain (HBD) necessary for retention in producing tissues (16). Another consequence of HBD absence is FGF21 reliance on the coreceptor β-klotho for high-affinity receptor binding (17, 18). Exogenous FGF21 administration not only normalizes basal glucose and insulin in obese rodents or primates but it also improves glucose disposal during glucose or insulin tolerance tests (13, 19–21). These positive effects are usually associated with an increase in the insulin-sensitizing adipokine adiponectin (21–23).

We recently administered FGF21 to early lactating dairy cows, but contrary to expectations, FGF21 treatment failed to improve any indices of insulin action (24). This failure may relate to the lactating state of these animals rather than an intrinsic inability of ruminants to respond to FGF21. Indeed, insulin-sensitizing effects of FGF21 have been described in a variety of other species but always in the nonlactating state (19, 21, 25, 26). Accordingly, we hypothesized that in the absence of lactation, ruminants are similarly capable of responding to the insulin-sensitizing actions of FGF21. We tested this hypothesis in nonlactating, adolescent ewes by first characterizing essential elements of the ovine FGF21 system and second, by assessing whether exogenous FGF21 administration elicits insulin-sensitizing effects.

MATERIALS AND METHODS

Animals and Design

All procedures were conducted with the approval of the Cornell University Institutional Animal Care and Use Committee.

Experiment 1: Description of the FGF21 system in sheep.

Four Finn × Dorset ewes of uniform age (10.6 ± 1.3 mo) and body weight (BW; 52.5 ± 4.1 kg) were used to survey tissue expression of FGF21, β-klotho, and FGF receptor (FGFR) isoforms. Ewes were offered an unlimited amount of a pelleted diet consisting of corn, wheat middlings, soy hulls, soybean meal, and mineral/vitamin supplement in the ratio of 47:21:15:13:4 and containing 173 g crude protein and 2.88 Mcal of metabolizable energy per kilogram of dry matter. Ewes were slaughtered by stunning with a captive bolt and exsanguinated, followed by tissue collection within 30 min. Tissues collected were liver, subcutaneous, omental, and retroperitoneal adipose tissue, kidney, gracilis muscle, and lung. Tissues were snap frozen in liquid nitrogen upon collection and stored at −80°C until further analysis.

Experiment 2: FGF21 signaling in adipose tissue and pharmacokinetics.

Six Finn × Dorset ewes of uniform age (12.9 ± 0.1 mo), body condition (3.58 ± 0.11), and weight (64.5 ± 2.2 kg) were used. They were fed unlimited amounts of a total mixed ration consisting of corn grain, hay, soyhulls, soybean meal, mineral/vitamin supplement, and corn oil in the ratio of 48:20:17:9:4:2 and containing 136 g crude protein and 2.56 Mcal of metabolizable energy per kilogram of dry matter. The nutrient composition of the total mixed ration was based on standards for ewes of this age and weight (27). Sheep were fitted with bilateral intrajugular catheters (Tygon S-54-HL, 1.02-mm i.d. × 1.78-mm o.d.; Saint-Gobain Performance Plastics, Akron, OH) 1 wk before treatment. They were then assigned to a crossover design with experimental periods of 13 h separated by a 7-day washout period. Treatment consisted of a single intravenous bolus of either excipient solution (Control) or recombinant human FGF21 (FGF21) at the dose of 5 mg/kg metabolic body weight (BW^0.75; MBW). The dose was prepared with the human FGF21 variant LY2405319 with 84% identity to ovine FGF21 and engineered for enhanced in vivo stability (28). It was produced in Pichia pastoris, purified to homogeneity using reversed-phase and anion-exchange chromatography, and prepared as a 33 mg/mL solution in an excipient solution of 10 mM citrate, 150 mM NaCl, pH 7.0 (Eli Lilly, Indianapolis, IN) (29). Blood samples were collected at fixed times (−60, −30, −1, 1, 5, 20, 40, and 60 min and 2, 4, 6, 8, and 12 h relative to the bolus at time 0). Blood was mixed with heparin (15 IU/mL) and centrifuged at 3,000 rpm (2,171 g) for 15 min at 4°C. Resulting plasma was stored in aliquots at −20°C until analyzed for FGF21. In addition, adipose tissue biopsies were taken at −10, 15, and 30 min relative to bolus administration. Biopsies of tailhead adipose tissue were performed, as previously described (30). In brief, the site was surgically prepared, followed by incision through the skin and dissection of adipose tissue. Pre bolus biopsies were obtained from the left side of the tailhead and post bolus biopsies were obtained from the right side of the tailhead. Tissue was snap frozen in liquid nitrogen and stored at −80°C until further analysis.

Experiment 3: Effects of chronic FGF21 administration.

Six Finn × Dorset ewes of uniform age (9.4 ± 0.1 mo), body condition (3.38 ± 0.09), and weight (49.7 ± 1 kg) were selected and moved to individual tie stalls under a controlled environment (20°C, light on 0730–2000 h). Ewes were randomly allocated to a crossover design with treatments consisting of daily subcutaneous injections of excipient solution (control) or the human FGF21 variant LY2405319 (FGF21) at the dose of 15 mg/kg MBW. Treatments lasted 13 days and were separated by a 14-day intervening period. Sheep were fitted with bilateral intrajugular catheters 1 wk before the initiation of the treatment period. Throughout the experiment, ewes were fed unlimited amounts of the total mixed ration used during experiment 2; the ration was delivered as meals offered every 4 h.

On the first day of the experimental period, treatments were administered as a single subcutaneous injection at 0800 h. Blood samples were obtained at fixed times relative to the injection (−60, −30, 0, 5, 15, 30, and 60 min and 2, 4, 6, 8, 12, 16, and 24 h), immediately mixed with heparin (15 IU/mL), and centrifuged at 3,000 rpm (2,171 g) for 15 min at 4°C. Resulting plasma was stored at −20°C until analyzed for metabolites and hormones.

For the remainder of the experimental period, treatments were administered as twice daily injections at 0800 and 2000 h. Blood samples (3 samples over 1 h) were taken from 1000–1100 h on experimental days-1, 3, 6, 9, 12 and processed immediately to plasma. Glucose tolerance tests (GTT) and epinephrine challenges (EC) were performed on days 12 and 13 of treatment, respectively; the bolus of glucose (0.25 g/kg BW, 50% wt/vol, VetOne, Boise, ID) or epinephrine (1.4 μg/kg BW, Anpro Pharmaceutical, Arcadia, CA) was administered at 1030 h. Other variables measured during each experimental period included daily feed intake, beginning and ending BW, and body condition score using a five-point scale [1 = thin, 5 = fat] (31).

Analysis of Plasma Metabolites and Hormones

Plasma fatty acids and glucose were analyzed using spectrophotometric methods based on the enzymes acyl coenzyme A–oxidase and glucose oxidase, respectively (32). The plasma concentrations of insulin were determined using a double antibody RIA validated in ruminants (32). The concentrations of plasma glucose, fatty acids, and insulin during the sampling windows were used to calculate the revised quantitative insulin-sensitivity check index {RQUICKI = 1/[log glucose (mg/dL) + log insulin (μU/mL) + log NEFA (mmol/L)]} (33). The plasma glucose response area was calculated between 0 and 90 min for the GTT and between 0 and 60 min for the EC; these values were corrected for baseline concentrations (mean of concentrations at −15, −10 and −5 min for the GTT; –30, –15, –10, –5, and –1 min for the EC). The plasma insulin-response area during the GTT was calculated between 0 and 120 min and was corrected for baseline concentrations (mean of concentrations at −10 and −5 min). The fatty acid response area during the EC was calculated between 0 and 60 min and was corrected for baseline concentrations (mean of concentrations at –30, –15, –10, –5, –1, 120, 130, and 140 min). The plasma concentrations of FGF21 were determined using double antibody assays previously validated in ovine and bovine plasma (32). Plasma samples were analyzed for ceramide concentrations using liquid chromatography and tandem mass spectrometry (LC-MS/MS), according to previously described methods (34, 35). Inter- and intra-assay coefficients of variation were <6% and <5%, respectively, for metabolite assays and <10% and <7%, respectively, for hormone assays.

Pharmacokinetic Analysis

The plasma concentrations of human FGF21 after the intravenous bolus were analyzed using noncompartmental analyses with a commercial software (PK Solutions 2.0, Summit Research Services, Montrose, CO) (36). This software uses a curve-stripping procedure to resolve the concentration-time curves into a series of exponential terms that correspond to the kinetic phases of circulating FGF21. The initial total FGF21 concentration was estimated through linear extrapolation of the first two plasma concentration values to time 0. Area under the concentration-time curve from 0 to 12 h (AUC_{0–12 h}) was estimated using the trapezoid method up to the last measured concentration at 12 h post bolus. The apparent terminal disposition rate constant (λ_z) was determined using linear regression analysis of the terminal portion of the log plasma concentration-time curve. Elimination half-life (t_1/2) and clearance rate were calculated as ln (2)/λ_z and dose/AUC_{0–12 h}, respectively. The volume of distribution (V_d) was estimated as D/AUC_0-∞·λ_z where D is the ratio of dose to body weight. Mean residence time (mRT) was estimated as AUMC_0-∞C/AUC_0-∞ where AUMC_0-∞ is the area under the concentration · time-time curve extrapolated to infinity.

Western Immunoblotting

Adipose tissue was homogenized in 1 mL of lysis buffer (10 mM Tris, pH 7.6, 10 mL/L of Triton X-100, 1 mM EGTA, 150 mM NaCl, 1 mM EDTA) supplemented with commercial proteases and phosphatase inhibitors (Halt phosphatase inhibitor mixture EDTA-free; Thermo Fisher Scientific, Waltham, MA). Homogenates were clarified by centrifugation (10,000 g for 20 min at 4°C). Protein concentrations of cellular extracts were determined using a bicinchoninic acid protein assay kit (Thermo Fisher Scientific). Fixed amounts of protein (25 μg) and Precision Plus Protein Kaleidoscope Standards (Bio-Rad Laboratories, Hercules, CA) were separated on 9% polyacrylamide gels under reducing conditions and transferred onto nitrocellulose membranes (Protran, Schleicher, and Schuell Bioscience, Dassel, Germany). Membranes were immersed in blocking solution (50 mM Tris, pH 7.4, 200 mM NaCl, 1 mL/L of Tween 20, 50 g/L of BSA). For signaling proteins, membranes were incubated with 1:1,000 dilution of primary antibodies against p44/42 MAPK (ERK1/2; Cell Signaling Technology, Danvers, MA; RRID:AB_390779) and threonine 202/tyrosine 204 phosphorylated ERK1/2 (pERK1/2; Cell Signaling Technology, RRID:AB_331646). Signals were developed with a 1:20,000 dilution of IRDye 800 anti-rabbit secondary antibody (LI-COR Biosciences, Lincoln, NE; RRID:AB_621843), followed by visualization and quantification with the LI-COR Odyssey infrared imaging system using the 800-nm channel.

For determination of plasma adiponectin, samples (10 μL of a 1:20 dilution) were mixed with 3.5 µL of 6X Laemmli buffer containing dithiothreitol and mercaptoethanol and boiled for 10 min. They were then separated on 12% polyacrylamide gels and transferred onto nitrocellulose membranes. Membranes were immersed in blocking solution of TBS-T (50 mM Tris, pH 7.4, 200 mM NaCl, 1 mL/L of Tween 20, 50 g/L of nonfat dry milk) for 1 h and then immunodecorated overnight at 4°C in blocking solution and a 1:5,000 dilution of a primary antibody against human adiponectin (a gift from T. Funahashi, Osaka University, Osaka, Japan) validated in cattle (37). Signals were developed with a 1:20,000 dilution of IRDye 800 anti-mouse secondary antibody (LI-COR Biosciences; RRID:AB_621842), followed by visualization and quantification with the LI-COR Odyssey infrared imaging system using the 800-nm channel.

RNA Extraction and Analysis of Gene Expression

Tissue samples were lysed with Qiazol (Qiagen, Valencia, CA), followed by total RNA purification using RNeasy Mini columns and on-column RNase-free DNase treatment (Qiagen). Quality of RNA was determined using the RNA Nano Lab Chip kit and bioanalyzer (Agilent, Palo Alto, CA) with samples having an RNA integrity number average of 7.9 ± 0.1. Reverse transcription reactions were performed with 1 μg of total RNA in a total 20 μL volume with the high-capacity cDNA reverse transcription kit and RNase inhibitor (Applied Biosystems, Foster City, CA). Gene expression was analyzed by quantitative real-time PCR (qPCR) using Power SYBR Green Master Mix (Applied Biosystems). Real-time PCR assays were performed in duplicate with a total 25 μL reaction volume containing 500 nM concentration of each primer and reverse-transcribed mRNA (25 ng except 2.5 ng for the internal standard gene 18S). The sequences of all primers used are given in Supplemental Table S1 (see https://doi.org/10.6084/m9.figshare.16821028.v1). Primer pairs were designed using Primer-BLAST software (http://www.ncbi.nlm.nih.gov/tools/primer-blast/). For all genes (except 18S), each member of the primer pair was located in adjoining exons and shown by blast analysis to anneal exclusively to the target sequence. For all primer pairs, amplification efficiency ranged from 90% to 103% and yielded a single product by melting curve analysis. A relative standard curve based on a serial fourfold dilution of pooled cDNA was used to analyze FGF21 and β-klotho expression across tissues and the effect of FGF21 on EGR1 expression in adipose tissue. Unknown sample expression was determined from the standard curve, adjusted for invariant control expression, and reported as arbitrary unit, as indicated in the figure legends. 18S was used as the invariant control, as its expression did not vary across analyzed tissues. Any mRNA with a cycle number greater than 34 was declared undetectable. To compare the relative expression of FGFR isoforms in liver and adipose tissue, the efficiency-corrected ΔCT method was used with all assays performed at the same detection threshold (38, 39). PCR efficiencies (E) were calculated from the slope of the standard curve using E = 10^(−1/slope), and the quantity of the various FGFR isoforms and 18S RNA calculated as quantity = E^−Ct (38). An arbitrary expression level was obtained by normalizing the efficiency-corrected value for each receptor to 18S expression (38).

Statistical Analysis

All data were analyzed by a mixed model using the fit model procedure of JMP Pro 14.3.0 statistical software (SAS Institute Inc., Cary, NC). Expression data describing the FGF21 system were analyzed with tissue as a fixed effect and animal as the random effect, followed by pairwise comparison using Tukey adjustment. Data collected at a single time point or collated over the treatment period were analyzed using a model accounting for treatment (Control vs. FGF21) and period as fixed effects and animal as the random effect. Data collected over time were analyzed using a mixed model accounting for treatment (Control vs. FGF21), time, period, and their interactions as fixed effects and animal as the random effect; a covariate corresponding to data collected before treatment on day 0 was included in the statistical models only when significant. Statistical significance and tendency were respectively set at P < 0.05 and P < 0.10 for main effects and at P < 0.10 and P < 0.15 for the FGF21 × Time interactions.

RESULTS

Components of the FGF21 System in Ewes

To identify possible sites of FGF21 production, tissues were collected from growing ewes and assayed by qPCR for FGF21 expression (Fig. 1A). The FGF21 cycle threshold number (Ct) was 29.2 ± 0.8 in liver, whereas it was barely below the Ct detection limit of 34 in subcutaneous and retroperitoneal adipose tissue and in kidney and undetectable in omental fat and skeletal muscle. As a consequence, FGF21 expression was 17-fold higher in liver than in any other expressing tissues (Tissue, P = 0.05).

Figure 1.

Expression of the components of the fibroblast growth factor-21 (FGF21) signaling system. Tissues were obtained from growing ewes. A: mRNA abundance of FGF21 and β-klotho across selected tissues. The significant effect of tissue is reported. mRNA levels were measured by quantitative real-time PCR, and arbitrary units were derived from a relative standard curve based on a serial fourfold dilution of pooled cDNA and normalized to 18S expression. Bars with different letters are significantly different at P < 0.05; ND, nondetectable. B: mRNA abundance of β-klotho-interacting FGF receptors in liver and subcutaneous adipose tissue. mRNA levels were measured by quantitative real-time PCR using the efficiency-corrected ΔCT method. An arbitrary expression level was obtained by normalizing the efficiency-corrected value for each receptor to 18S expression. The significant effect of FGF receptor (FGFR) is reported. Bars with different letters differ at P < 0.05. C: combined mRNA expression of β-klotho-interacting FGF receptors in liver and subcutaneous adipose tissue (AT). *P = 0.02. For all panels, bars represent least squares mean ± SE of four ewes. AT, adipose tissue; Om, omental; Rp, retroperitoneal; Sc, subcutaneous.

FGF21 signaling is restricted to tissues expressing the coreceptor β-klotho. β-Klotho expression was highest in retroperitoneal adipose tissue and in liver followed by subcutaneous and omental adipose tissue and finally kidney (Fig. 1A, Tissue, P < 0.001); β-klotho expression was barely detectable in muscle and absent in lungs.

Next, we measured the mRNA abundance of the β-klotho interacting FGF receptors FGFR1c, FGFR2c, and FGFR3c in liver and subcutaneous adipose tissue. The FGF receptor expression profile was specific to each tissue, with over 69% of all receptors accounted by FGFR2c in liver and 98% by FGFR1c in subcutaneous adipose tissue (Fig. 1B). The sum total of β-klotho-interacting FGF receptors was 3.7-fold higher in subcutaneous adipose tissue than in liver (Fig. 1C, P = 0.02). The other fat depots were similar to subcutaneous adipose tissue in terms of FGF receptor profile and sum total of β-klotho-interacting FGF receptors (results not shown). These results identify liver as a major site of FGF21 expression in the sheep and adipose tissue and liver as potential FGF21 target tissues.

FGF21 Signaling in Adipose Tissue and Pharmacokinetics

To obtain direct evidence of FGF21 signaling in adipose tissue, ewes were administered a single intravenous bolus of excipient solution (Control) or human FGF21 at the dose of 5 mg/kg MBW followed by determination of pERK1/2 activation and early growth response 1 (EGR1) expression in adipose tissue biopsies. In tissues collected 15 min after bolus, FGF21 administration was associated with a 3.1-fold increase in immunoreactive pERK1/2, whereas the excipient had no effect (Fig. 2A, FGF21 × Time, P < 0.001). Both treatments increased EGR1 expression 30 min after bolus, but the increase was greater after the FGF21 than excipient bolus (Fig. 2B, FGF21 × Time, P = 0.06). In aggregate, these results confirm adipose tissue as a genuine FGF21 target as well as bioactivity of the human FGF21 variant LY2405319 in the sheep.

Figure 2.

Fibroblast growth factor-21 (FGF21) signaling in subcutaneous adipose tissue. Growing ewes received an intravenous bolus of excipient (Control) or human FGF21 (FGF21) at the dose of 5 mg/kg MBW. A: biopsies of subcutaneous adipose tissue were obtained immediately before intravenous bolus at time 0 and then 15 and 30 min later. Total cellular extracts were prepared and analyzed by Western immunoblotting. Left: threonine 202/tyrosine 204 phosphorylated (p-ERK1/2) and total p44/42 mitogen-activated protein kinase (ERK1/2) abundance in subcutaneous adipose tissue extracts from representative Control and FGF21-treated sheep. Right: ERK1/2 activation (ratio of phosphorylated to total ERK1/2) was calculated and expressed relative to Control at time 0. B: total RNA was isolated from subcutaneous adipose tissue obtained at time 0 and 30 min relative to the intravenous bolus. The abundance of early growth response 1 (EGR1) mRNA was measured by quantitative real-time PCR and arbitrary units were derived from a relative standard curve based on a serial fourfold dilution of pooled cDNA, normalized to 18S expression, and expressed relative to Control values at time 0. C: plasma samples were analyzed for FGF21 after bolus at 0.017, 0.083, 0.33, 0.67, 1, 2, 4, 6, 8, and 12 h for FGF21 and at 0.017, 2, and 12 h for Control. For all panels, data points and bars represent the least squares mean ± SE of six ewes. The effects of FGF21 treatment, time, and the interaction between FGF21 treatment and time (FGF21 × Time) are reported for A and B when significant.

We also assessed the pharmacokinetics of exogenous FGF21 in the sheep. The basal concentration of plasma FGF21 was similar in Control and FGF21-treated sheep (0.8 ± 0.3 vs. 1.3 ± 0.6 ng/mL). Plasma FGF21 reached a peak of 53,803 ng/mL within 1 min of FGF21 administration, followed by a steady decline to 6.5 ng/mL after 12 h (Fig. 2C). Analysis of the concentration-time curve predicted a maximal concentration of 61,451 ng/mL at 0 min and an elimination t_1/2 of 94 min (Table 1). Other relevant pharmacokinetic parameters are reported in Table 1. As expected, plasma FGF21 did not vary significantly in sheep receiving the excipient bolus and averaged 0.9 ng/mL (Fig. 2C).

Table 1.

Pharmacokinetic parameters of FGF21 after a single intravenous bolus injection*

Parameters†	Average ± SE
Ci, ng/mL	61,451 ± 1,747
Vd, mL/kg	163 ± 7
AUC (ng · min/mL)	1,494,352 ± 73,627
Elimination t1/2 (min)	94 ± 4
mRT (min)	37 ± 2
Cl, mL/min/kg	1.2 ± 0.06

*Six female sheep received a single intravenous bolus of human FGF21 at the dose of 5 mg/kg metabolic body weight (BW^0.75).

†Ci, initial concentration; V_d, volume of distribution; AUC, area under the concentration-time curve extrapolated to infinity; t_1/2, half-life; mRT, mean residence time; Cl, clearance rate.

Effects of Chronic FGF21 Treatment on Growth and Metabolic Parameters

We next conducted a 13-day treatment with the excipient solution or human FGF21 at the dose of 15 mg/kg MBW/day. To assess acute FGF21 effects, the first dose was administered as a single injection, followed by frequent blood sampling over the next 24 h. At time 0, plasma FGF21 concentration in both groups of sheep averaged 0.75 ± 0.23 ng/mL and remained at this level in the sheep receiving the excipient solution (Fig. 3A). In contrast, the plasma FGF21 concentration rose within 30 min to a near peak of 1,806 ± 140 ng/mL in the FGF21 group, followed by a decline to 224 ± 83 ng/mL over the next 23 h.

Figure 3.

Acute effects of fibroblast growth factor-21 (FGF21) treatment on plasma variables. Growing ewes were treated with a single subcutaneous injection of excipient (Control) or human FGF21 (FGF21) at the dose of 15 mg/kg MBW. Blood samples were collected over the first 24 h of treatment. A–D: plasma concentration of FGF21, glucose, insulin, and fatty acids. Data points represent the least squares mean ± SE of six ewes. For B–D, effects of FGF21 treatment, time, and the interaction of FGF21 treatment and time (FGF21 × Time) are reported when significant.

The concentration of plasma glucose was lower in FGF21 than excipient-treated sheep, particularly between 8 and 24 h post bolus (Fig. 3B, FGF21 × Time, P = 0.02). The concentration of plasma insulin was lower in FGF21 than excipient-treated sheep 30 min post bolus, followed by a rebound over the next 2.5 h (Fig. 3C, FGF21 × Time, P = 0.07). The concentration of plasma fatty acids declined over the first 12-h post bolus and then rebounded (Fig. 3D, Time, P = 0.02) but was not modulated by treatment.

For the remainder of the experimental period, treatments were administered as twice daily injections. Under this mode of administration, plasma FGF21 averaged 2.2 ± 0.3 and 2,012 ± 55 ng/mL for excipient and FGF21-treated animals over the treatment period when measured 2 h after the first daily bolus (results not shown). Initial BW did not differ between treatment (Control, 52.7 ± 1.3 kg; FGF21, 51.3 ± 1.3 kg), and FGF21 had no effect on voluntary feed intake (Control, 1.13 ± 0.04 kg; FGF21, 1.09 ± 0.04 kg) or weight gain (Control, 1.83 ± 0.68 kg; FGF21, 1.67 ± 0.68 kg).

FGF21 caused a reduction in plasma glucose throughout the entire treatment period (Fig. 4A, P = 0.008) as well as a depression in plasma insulin with treatments differing the most on days 6 and 9 of the experimental period (Fig. 4A, FGF21 × Time, P = 0.04). FGF21 treatment increased RQUICKI values in a constant manner between days 3 and 12 of treatment (0.36 vs. 0.43, Control vs. FGF21, P = 0.008), suggesting improved insulin action. To assess the possibility that these FGF21 effects represented insulin sensitization, sheep were administered a glucose tolerance test on day 12 of treatment. FGF21 treatment resulted in a slightly lower plasma glucose profile and tended to cause a lower glucose response area (Fig. 4B, FGF21, P = 0.06). FGF21 also lowered the plasma insulin profile but did not lead to a significant reduction in the response area (Fig. 4B).

Figure 4.

Effects of chronic fibroblast growth factor-21 (FGF21) treatment on indices of insulin action. Growing ewes received daily subcutaneous injection of excipient (Control) or human FGF21 (FGF21) at the dose of 15 mg/kg MBW/day. Treatments were administered as a single injection on the first day of treatment and as twice daily injections thereafter. A: plasma samples collected every 3 days during the treatment period were analyzed for glucose and insulin concentrations. Data points represent the LSM ± SE of six ewes. Effects of FGF21 treatment, time, and the interaction of FGF21 treatment and time (FGF21 × Time) are reported when significant. B: a glucose tolerance test was performed on day 12 of treatment. Blood samples were collected during the glucose tolerance test, and plasma was analyzed for glucose and insulin concentrations. Plasma concentrations and response areas are shown on the left for glucose (response area calculated over 90 min) and on the right for insulin (response area calculated over 120 min). Data represent least squares mean ± SE of six ewes; †P = 0.06.

Next, we assessed the possibility that the positive effects of FGF21 on insulin action are associated with changes in plasma adiponectin and ceramide levels. FGF21 increased plasma adiponectin by twofold over the treatment period, whereas excipient treatment had no effect (Fig. 5A, P = 0.04). To determine the timing of this effect, plasma adiponectin was analyzed every 3 or 4 days; the positive effect of FGF21 became apparent on day 3 of treatment and continued to increase until reaching a plateau on day 9 of treatment (Fig. 5B, FGF21 × Time, P = 0.04). In contrast, FGF21 had no effect by the end of treatment on individual ceramide species such as C16:0- or C24:0-ceramide or total ceramides (Fig. 5C).

Figure 5.

Effects of chronic fibroblast growth factor-21 (FGF21) treatment on plasma adiponectin and ceramides. Growing ewes received daily subcutaneous injection of excipient (Control) or human FGF21 (FGF21) at the dose of 15 mg/kg MBW/day. Treatments were administered as a single injection on the first day of treatment and as twice daily injections thereafter. Plasma samples were analyzed for adiponectin by Western immunoblotting and for ceramides using liquid chromatography and tandem mass spectrometry. A: Left: plasma samples collected before (day 0) and at the term of the treatment period (day 13) were analyzed in duplicate by Western immunoblotting for adiponectin. Right: the adiponectin signal was calculated and expressed relative to day 0. *P = 0.04. B: changes in plasma adiponectin over the treatment period. Effects of FGF21 treatment, time, and the interaction of FGF21 treatment and time (FGF21 × Time) are reported when significant. C: plasma concentrations of C16:0-, C24:0-, and total ceramides. For all panels, data points and bars represent least squares mean ± SE of six ewes.

Finally, we asked whether FGF21 treatment altered the lipolytic tone. FGF21 administration did not have any effects on plasma fatty acids over the treatment period (Fig. 6A) but tended to increase the plasma fatty acid response area to an epinephrine challenge on day 13 of treatment (Fig. 6B, P = 0.07), suggesting that it increased β-adrenergic responsiveness of adipose tissue.

Figure 6.

Effects of chronic fibroblast growth factor-21 (FGF21) treatment on plasma fatty acids and epinephrine response. Growing ewes received daily subcutaneous injection of excipient (Control) or human FGF21 (FGF21) at the dose of 15 mg/kg MBW/day. Treatments were administered as a single injection on the first day of treatment and as twice daily injections thereafter. A: plasma samples collected every 3 days during the treatment period were analyzed for the concentration of fatty acids. B: plasma fatty acid responses during an epinephrine challenge test performed on day 13 of treatment. Response areas are shown on the right and were calculated over 60 min. †P = 0.07. For all panels, data points and bars represent least squares mean ± SE of six ewes.

DISCUSSION

Exogenous FGF21 administration consistently reduces plasma insulin and glucose as well as improves other indices of insulin action in rodent and primate models (15, 40). Work in these species has been focused on the therapeutic potential of FGF21 and accordingly has involved most frequently obesity and type 2 diabetic disease models such as diet-induced obese mice and obese and diabetic primates (13, 19–21). To the best of our knowledge, FGF21 has not been administered to lactating animals with the single exception of our recent study in dairy cows (24, 30). Unexpectedly, FGF21 treatment failed to elicit any insulin-sensitizing effects in this physiological state, prompting us to reexamine its action in the absence of lactation in the sheep, a smaller but closely related ruminant.

First, we used mRNA analysis to characterize the ovine FGF21 system in peripheral tissues. Our data establish liver as the predominant peripheral tissue expressing FGF21 and adipose tissue as a secondary site with barely detectable expression. This pattern of FGF21 expression is nearly identical to those previously described in cattle (41) and humans (42) but contrasts with that of mice where basal expression is nearly as high in adipose tissue as in liver (43). Next, we measured expression of the coreceptor β-klotho to identify potential FGF21 target tissues. β-klotho is required for high-affinity FGF21 binding with FGF receptors because unlike most other members of the FGF superfamily, it lacks the proteoglycan-interacting HBD (16). In mice and cattle, β-klotho expression is many fold higher in adipose tissue and liver than any other peripheral tissues (39, 41). The pattern of β-klotho expression in the present study also identifies adipose tissue and liver as two potential FGF21 target tissues in the sheep. Finally, we asked whether these two tissues express meaningful levels of FGFR1c, FGFR2c, and FGFR3c with the dual capability of interacting with β-klotho and mediating FGF21 actions (17, 18). Importantly, recent work suggests that among this FGFR subset, FGFR1c mediates virtually all FGF21 actions, as shown by blunted effects in FGF21-treated Fgfr1 knockout mice and by recapitulation of FGF21 effects in mice and primates treated with FGFR1/β-klotho-activating antibodies (44–46). Our results in the sheep show that FGFR1c expression accounts for 98% of all FGFR in adipose tissue and is 12-fold more abundant in adipose tissue than liver, in full agreement with data we recently reported in cattle (41). When considered in aggregate, the β-klotho and FGFR expression data identify adipose tissue as a major FGF21 target tissue in the sheep.

The gold standard method to confirm target tissues is documentation of ERK1/2 phosphorylation and increased expression of the early response genes C-FOS or EGR1 immediately after FGF21 administration (13, 25). Using this approach, we identified adipose tissue as an FGF21 target tissue in the sheep, in complete agreement with results previously obtained in mice and dairy cows (13, 25, 30, 47). Unfortunately, we could not assess FGF21 signaling in liver because of difficulties in obtaining liver biopsies in the sheep. However, we note that hepatic FGF21 signaling has been reported to be absent or barely detectable in the mouse (47, 48). Moreover, we previously reported negligible FGFR1c expression combined with the absence of FGF21 signaling events in bovine liver (30). Accordingly, with similarly low FGFR1c expression, liver appears as an unlikely direct FGF21 target in the sheep. This experiment also allowed an initial characterization of FGF21 kinetics in the sheep. This work revealed a half-life of 94 min. Intravenous administration of human FGF21 in cattle, monkeys, and mice yielded half-lives of 194, 120, and 90 min (24, 25, 49).

FGF21 research in rodents and primates has focused on its therapeutic potential for type 2 diabetes, an effort initially motivated by the in vitro observation that FGF21 stimulates insulin-independent glucose uptake in adipocytes (13). In vivo, FGF21 lowers plasma glucose and insulin within 1 h of injection in rodent models of obesity, including the ob/ob and diet-induced obese mice; these effects are sustained during chronic administration (13, 20). Using dynamic tests such as GTT, these FGF21 effects have been traced to improvement in insulin-mediated glucose disposal. Importantly, the insulin-sensitizing effects of chronic FGF21 treatment are associated with loss of fat and body weight, effects attributed predominantly to increased energy expenditure in rodents and to reduced feed intake in primates (20, 21, 29). Our observation of improved insulin action with chronic FGF21 treatment (i.e., reduced plasma glucose and insulin, increased RQUICKI values, and improved glucose disposal during GTT) agree with previous results in primates and rodents but differ in that they occurred in the absence of any reduction in intake or body weight. Although we did not measure body composition in this experiment, reduced adiposity also appears unlikely to explain improved insulin action by FGF21, as the treatment period was relatively short and the sheep had normal body condition. Overall, these data show that the insulin-sensitizing effects of FGF21 are conserved in ruminants and raise the possibility that lactation prevented such effects in our previous study in dairy cattle (30).

The adiponectin-ceramide axis has been proposed as a mediator of the positive FGF21 effects on insulin action on the basis of three observations. First, adiponectin and ceramide exert opposite effects on insulin action. Specifically, adiponectin is an insulin-sensitizing hormone, with overexpression improving insulin action in mice and its absence having the opposite effect (50–52). In contrast, animals unable to produce C16:0-ceramide in liver or adipose tissue or C18:0-ceramide in skeletal muscle have improved insulin action (53, 54). Indirect evidence for ceramide interfering with insulin action also exists in fasted or lactating dairy cows where plasma C16:0- or C24:0-ceramides are positively associated with insulin resistance (55, 56). Second, the adiponectin receptors ADIPOR1 and ADIPOR2 are endowed with ceramidase activity and their activation reduces tissue and plasma ceramide levels (57–59). Third, chronic FGF21 administration increases plasma adiponectin in mice and reciprocally reduces hepatic and plasma ceramide levels (22, 23); in the absence of adiponectin, much of the insulin-sensitizing effects of FGF21 are lost in parallel with unaltered ceramide levels (22). In partial agreement with these observations, we observed a significant increase in plasma adiponectin with chronic FGF21 treatment but unchanged plasma C16:0- and C24:0-ceramide, and total ceramides. This discrepancy may relate to FGF21 effects on ceramides depending on an excessive supply of dietary or endogenously derived circulating fatty acids (22). Overall, our data show that positive effects of FGF21 on carbohydrate metabolism in ruminants do not require a reduction in plasma ceramide levels.

Effects of FGF21 on lipolysis have been reported to vary according to energy status, with FGF21-stimulating lipolysis in the fed state but inhibiting it during fasting (60). Under fed conditions, mice overexpressing FGF21 exhibit increased lipolytic activity in adipose tissue in addition to elevated circulating fatty acids when fed (61). A lipolytic effect also was observed in alcohol-treated fed mice where elevated plasma FGF21 increases circulating catecholamines and plasma fatty acids (62). The animals used in the present study were in positive energy balance, and FGF21 had no effect on plasma fatty acids under basal condition but in agreement with lipolytic effects seen in fed rodents (60), tended to increase epinephrine-stimulated release of fatty acids from adipose tissue.

Perspectives and Significance

FGF21 improves insulin action in rodents and primate models of obesity and type 2 diabetes but failed to exert similar effects in dairy cattle during the naturally occurring insulin resistance of lactation. We now show that FGF21 is a potent insulin sensitizer in nonlactating sheep, demonstrating conservation of FGF21 actions in ruminants and raising the possibility that lactation is an FGF21-resistant state. Future studies are needed to assess the role of factors associated with early lactation in ruminants, such as negative energy balance and increased growth hormone secretion (4, 63). Understanding the basis for FGF21 resistance during lactation could offer insights on physiological mechanisms limiting FGF21 actions.

SUPPLEMENTAL DATA

Supplemental Table S1: https://doi.org/10.6084/m9.figshare.16821028.v1.

GRANTS

This work was supported in part by Hatch/Multistate projects under Awards 1010605 and 1017053 (to Y. R. Boisclair).

DISCLOSURES

A. Butterfield and J.W. Perfield II are paid employees of Eli Lilly and Company and may own company stock or possess stock options. None of the other authors has any conflicts of interest, financial or otherwise, to disclose.

AUTHOR CONTRIBUTIONS

C.L.L., S.L.G., and Y.R.B. conceived and designed research; C.L.L., S.L.G., M.M.M., J.W.P., A.B., M.M., N.J.H., and J.W.M. performed experiments; C.L.L., S.L.G., and M.M.M. analyzed data; C.L.L. and Y.R.B. interpreted results of experiments; C.L.L., S.L.G., and M.M.M. prepared figures; C.L.L. and Y.R.B. drafted manuscript; C.L.L., S.L.G., M.M.M., J.W.P., A.B., M.M., N.J.H., J.W.M., and Y.R.B. edited and revised manuscript; C.L.L., S.L.G., M.M.M., J.W.P., A.B., M.M., N.J.H., J.W.M., and Y.R.B. approved final version of manuscript.