Targeted Deletion of Fibroblast Growth Factor 23 Rescues Metabolic Dysregulation of Diet-induced Obesity in Female Mice

Abstract

Fibroblast growth factor 23 (FGF23) is a bone-secreted protein widely recognized as a critical regulator of skeletal and mineral metabolism. However, little is known about the nonskeletal production of FGF23 and its role in tissues other than bone. Growing evidence indicates that circulating FGF23 levels rise with a high-fat diet (HFD) and they are positively correlated with body mass index (BMI) in humans. In the present study, we show for the first time that increased circulating FGF23 levels in obese humans correlate with increased expression of adipose Fgf23 and both positively correlate with BMI. To understand the role of adipose-derived Fgf23, we generated adipocyte-specific Fgf23 knockout mice (^AdipoqFgf23^Δfl/Δfl) using the adiponectin-Cre driver, which targets mature white, beige, and brown adipocytes. Our data show that targeted ablation of Fgf23 in adipocytes prevents HFD-fed female mice from gaining body weight and fat mass while preserving lean mass but has no effect on male mice, indicating the presence of sexual dimorphism. These effects are observed in the absence of changes in food and energy intake. Adipose Fgf23 inactivation also prevents dyslipidemia, hyperglycemia, and hepatic steatosis in female mice. Moreover, these changes are associated with decreased respiratory exchange ratio and increased brown fat Ucp1 expression in knockout mice compared to HFD-fed control mice (Fgf23^fl/fl). In conclusion, this is the first study highlighting that targeted inactivation of Fgf23 is a promising therapeutic strategy for weight loss and lean mass preservation in humans.

Obesity has taken epidemic proportions worldwide, and its impact on health extends across multiple organ systems and diseases. Obesity is a major risk factor for many chronic metabolic conditions including heart disease, renal disease, type 2 diabetes (1), and some forms of cancer (2) and is associated with premature death and reduced life quality. Currently, obesity is managed either by bariatric surgery combined with lifestyle interventions such as diet and physical activity regimens or medical therapy. Current pharmacologic options for the treatment of obesity include glucagon-like peptide-1 (GLP-1) receptor agonists, GLP-1 in combination with glucose-dependent insulinotropic polypeptide (GIP) dual agonists, or GLP-1/GIP/glucagon triple receptor agonists, which result in significant weight loss (3-5). However, an area of concern is the loss of lean body mass associated with the use of GLP-1 receptor agonists, leading to sarcopenia, poorer muscular strength, and higher risk for fragility (6). Moreover, other areas of concern include the presence of gastrointestinal side effects related to GLP/GIP usage (6), weight regain and related cardiometabolic complications after their withdrawal (7), and their undetermined ability to chronically sustain weight loss. Thus, there remains a very significant medical need for better nonsurgical treatments by identifying new therapeutic targets to treat obesity.

Fibroblast growth factor 23 (FGF23) is a bone-secreted protein that exerts a profound effect over bone mineralization (8-11). Elevated FGF23 is unquestionably closely linked to bone mineralization defects, low bone mineral density, and greater risk of fractures (12-17). However, outside of the skeleton, FGF23 is present in serum, and several lines of evidence raise the possibility that circulating FGF23 exerts functions beyond those associated with bone mineralization and acts systemically to influence metabolic function. Moreover, circulating FGF23 levels rise dramatically in chronic kidney disease (CKD) patients as their renal function declines, and have been linked to greater cardiovascular risk, vascular calcifications, cardiac hypertrophy, and mortality (18-20). Despite being widely recognized as an important risk factor for cardiovascular disease and CKD progression, our understanding of the detrimental effects of elevated FGF23 on individual organs/systems of the body and how they contribute to pathological conditions remains limited. Over recent years, studies have shown that FGF23 levels rise with a high-fat diet (HFD) (21), they are associated with high-fat mass (22), and they decline after bariatric surgery (23).

These data suggest that elevated FGF23 pathologically contributes to the development of obesity. However, it is not clear whether this regulation of fat metabolism is mediated by FGF23 produced remotely by distant organs (eg, bone) or locally by adipocytes. Regardless of whether the actions of FGF23 are direct or indirect, the critical question is whether targeted deletion of Fgf23 can reduce body weight and fat mass and attenuate obesity-associated complications. Herein, we show for the first time that mice fed HFD and obese humans have increased circulating FGF23 levels that positively correlate with increased expression of Fgf23 in adipose tissue. Therefore, to understand the physiological role of adipose-derived Fgf23, we generated adipocyte-specific Fgf23 knockout mice (^AdipoqFgf23^Δfl/Δfl) and compared them to mice with intact Fgf23 (controls) after both being exposed to a HFD. Changes in body composition, energy metabolism, and liver lipid homeostasis were evaluated. Our findings reveal an obesity-resistant phenotype in the absence of locally produced adipose Fgf23.

This is the first study highlighting novel and critical paracrine/autocrine actions of FGF23 on adipocytes, suggesting that FGF23 influences whole-body metabolism in obesity and demonstrating the therapeutic potential of selective inactivation of FGF23 for weight loss and lean mass preservation in humans.

Materials and Methods

Animals

Fgf23 conditional knockout mice in which the Fgf23 gene is selectively inactivated only in adipocytes were generated by crossing floxed-Fgf23 (Fgf23^fl/fl) mice provided by Dr. W. Chang (Department of Medicine, UCSF/NCIRE-SFVAMC) with adiponectin-Cre (Adipoq-Cre) mice (Jackson Laboratory, Stock # 028020), which express a Cre recombinase under the control of the mouse adiponectin promoter. Floxed-Fgf23 (Fgf23^fl/fl) mice were made by introducing 2 loxP sequences flanking exon 1 of the mouse Fgf23 gene in C57BL/6J background. The Cre transgenes were maintained only in male breeders to avoid transfer of Cre maternally and spontaneous germline recombination.

For genotyping, the following primers were used: Adipoq forward (GCA AAA CAG GCT CTA GCG TTC G) and reverse (CTG TTT CAC TAT CCA GGT TAC GG); Fgf23 forward (ATT GCC AGT TTA GTT CCC TG) and reverse (CCC CCT ACC CCC ATA CAC AA). Adipocyte-specific Fgf23-deficient mice (^AdipoqFgf23^Δfl/Δfl) were compared with littermate control mice (Fgf23^fl/fl). Starting at 8 weeks of age, mice were fed either a control diet (10 kcal% fat, D12450J, Research Diets) or a HFD (60 kcal% fat, D12492i, Research Diets) ad libitum for 24 weeks and allowed free access to water. The Institutional Animal Care and Use Committee at New York University (NYU) approved the animal studies. Body weight was measured every week. Whole-body composition was measured by in vivo micro-computed tomography.

Food Intake

Mice were individually housed for 5 days, and food consumption was calculated by measuring the difference between the remnant food (day 5) and the preweighed served food (day 0). All measurements were converted to calories to compare between control diet and HFD (24, 25). Energy uptake was calculated by multiplying the consumed food in grams with the calories per gram of the respective type of diet (26).

Indirect Calorimetry

Respiratory gas utilization in ^AdipoqFgf23^Δfl/Δfl mice and Fgf23^fl/fl mice fed HFD was measured via indirect calorimetry using an 8-cage open respirometry system (TSE PhenoMaster, TSE Systems GmbH, Germany), as previously described (27). Mice were weighed and individually housed in Techniplast home cages located within a temperature- and humidity-controlled climate chamber (22 ± 0.5 °C, 50 ± 1% relative humidity). The light cycle was set to 12:12 (lights on at 06:30), and the air flow rate for each cage was set to 0.35 L/min (with 0.25 L/min diverted to the gas sensors during the sampling period: 190-second line purge and 10-second active sample). Oxygen consumption (vO2 = mL/hour), carbon dioxide production (vCO2 = mL/hour), and cumulative measures of food consumption (g), water intake (ml), and horizontal and vertical activity (beam break counts, X + Y + Z axes) were recorded at 30-minute intervals. Energy expenditure (EE = kcal/h = 3.941 × vO2 + 1.106 × vCO2, ignoring urinary nitrogen) was calculated using Weir's equation (28) and respiratory exchange ratio (RER) was computed as vO2/vCO2. Body temperature was acquired continuously (250-2000-ms sampling intervals) via an encapsulated temperature biosensor (UID Temperature Programmable Microchip; UCT-2112) implanted into the intrascapular subcutaneous tissue. Mice were acclimated to the metabolic home cages for at least 24 hours before data were used for analysis. Data were acquired and exported with TSE PhenoMaster software V8.1.4.14156 (TSE Systems GmbH, Germany, https://www.tse-systems.com/products/phenomaster). Data were analyzed by analysis of covariancewith genotype and sex as factors, with body weight included as a covariate wherever significant.

Blood, Serum and Tissue Collection

At the end of the experiment, mice were immediately necropsied after euthanasia and blood was collected by cardiac puncture in separate Microtainer® serum separator tubes (Becton, Dickinson and Company, Franklin Lakes, NJ) and centrifuged at 1800 × g for 15 minutes. Perigonadal and inguinal white adipose tissue (WAT), interscapular brown adipose tissue (BAT), and liver were collected, snap-frozen in liquid nitrogen, and stored at −80 °C until further use.

Serum Biochemistry

Serum FGF23 levels were measured using mouse FGF23 Intact (QuidelOrtho Cat# 60-6800, RRID:AB_2813726) and C-terminal (QuidelOrtho Cat# 60-6300, RRID:AB_3073876) ELISA assays or human FGF23 Intact ELISA (QuidelOrtho Cat# 60-6600, RRID:AB_2891250). Cholesterol was measured using Cell Biolabs' Total Cholesterol Assay Kit (VWR). Triglyceride and free fatty acid levels were measured using the Triglyceride Colorimetric Assay Kit (Cayman Chemical/Thomas Scientific) and the EnzyChrom™ Free Fatty Acid Assay Kit (BioAssay Systems/Thomas Scientific), respectively. Glucose and insulin levels were measured after 6 hours of fasting using the Mouse Glucose Assay Kit and the Ultra Sensitive Mouse Insulin ELISA Kit, respectively (Crystal Chem Cat# 90080, RRID:AB_2783626).

Adipose Tissue Histology

Adipose tissues and liver segments were fixed in 10% formalin for 24 hours, changed to PBS, and embedded in paraffin, and 4-um-thick sections were stained with hematoxylin and eosin (H&E). Images were obtained at 40 × magnification.

RNA Isolation, Reverse Transcription, and Real-time Quantitative PCR Analysis

Total RNA was extracted from adipose tissues and liver using Trizol (Ambion; Life Technologies, Carlsbad, CA) according to the manufacturer's protocol (Molecular Research Center, Cincinnati, OH). cDNA was synthesized using the High Capacity cDNA Reverse Transcription Kit, as described by the manufacturer (Applied Biosystems; Thermo Fisher Scientific, Waltham, MA), and amplified by quantitative PCR using the PerfeCTa SYBR Green SuperMix (Quanta Biosciences, Gaithersburg, MD). All primers used in this study are listed in Supplementary Table S1 (29). mRNA levels were normalized to the housekeeping gene (Hprt) in the same reverse transcribed (RT) sample. The relative transcript expression of a gene is given as ΔCt = Ct_target−Ct_reference. The fold change in gene expression, as compared to control mice, was determined as 2^−ΔΔCt values (ΔΔCt = ΔCt_treated−ΔCt_control).

NanoString Expression Profiling

The expression of genes involved in fatty acid synthesis and oxidation, inflammation, and oxidative stress was assessed. Three hundred ng of total RNA was hybridized with barcoded reporter probes and captured probes following manufacturer's instructions and quantified on an nCounter Max analysis system. Transcript counts were normalized to housekeeping genes including Abcf1, Dnajc14, G6pd2, Gapdh, Rpl19, Sdha, and Tbp using nSolver 4.0 software (NanoString Technologies, Seattle, WA).

Western Blot Analysis

Adipose tissues were (Abcam Cat# ab23841, RRID:AB_2213764) lysed in radioimmunoprecipitation assay buffer with phosphatase and proteinase inhibitors (Thermo Fisher Scientific), and protein lysate concentration was measured using the BCA Protein Assay Kit (Thermo Fisher Scientific). Equal concentrations of protein were loaded and electrophoresed on 4% to 15% TGX Criterion gels (BioRad), transferred to polyvinylidene difluoride membranes, and immunoblotted. Antibodies used were the following: FGF23 (Thermo Fisher Scientific Cat# PA5-89609, RRID:AB_2805684), UCP1 (Abcam Cat# ab23841, RRID:AB_2213764), and β-actin (Santa Cruz Biotechnology Cat# sc-47778, RRID:AB_626632), followed by horseradish peroxidase-linked secondary antibodies (Santa Cruz Biotechnology Cat# sc-2004, RRID:AB_631746 and Santa Cruz Biotechnology Cat# sc-2005, RRID:AB_631736). SuperSignal West Pico Chemiluminescent substrate (Thermo Scientific) was used to develop blots. Band intensity was measured using National Institutes of Health ImageJ software.

3T3-L1 Cell Culture and Differentiation

3T3-L1 mouse fibroblasts (ATCC® CL-173™), a widely used murine cell line for studying adipocyte differentiation, were cultured in high glucose DMEM medium, supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin. To induce adipocyte differentiation, the 3T3-L1 cells were allowed to grow to confluence and then incubated in a differentiation medium (day 0) that contained 1 μg/mL insulin, 0.5 mM IBMX, 1 μM dexamethasone, and 0.2 μM pioglitazone for 2 days. Cells were then switched to high-glucose DMEM containing 10% fetal bovine serum, 1% penicillin/streptomycin, 1 μg/mL insulin, and 0.2 μM pioglitazone and replaced every 2 days for the next 6 days. At day 8, differentiated 3T3-L1 cells were either stained with Oil Red O or used for quantitative RT-PCR analysis. PBS (vehicle) or FGF23 at a concentration of 100 ng/mL was added to the differentiation medium from day 0 to day 8.

Oil Red O Staining

After cell differentiation, cells were washed in PBS and fixed with 10% buffered formalin solution for 15 minutes. Fixed cells were washed with MiliQ water and were air dried. Oil Red O (Sigma Cat# O1391) solution made in isopropanol was added to each well and the cells were incubated for 30 minutes. After the removal of Oil Red O solution, cells were washed 4 times with MiliQ water and imaged.

Human Subjects

Blood samples were collected from 23 women with a body mass index (BMI) range of 20.7 to 52.6 kg/m² and ages between 26 and 69 years who were recruited either in the Department of Nutritional Sciences at Rutgers University (the Osteoporosis, Weight Loss and Endocrinology repository) or in the Leon H. Charney Division of Cardiology at the NYU Langone Medical Center to undergo bariatric surgery. All eligible volunteers underwent biochemical and physical screening including a comprehensive chemistry panel, complete blood count, and physical examination to ensure they were healthy and had no evidence of undiagnosed diseases. Subjects signed an informed consent approved by the Rutgers University or NYU Langone Medical Center Institutional Review Boards for the protection of human subjects in research before the initiation of the study protocol.

Deidentified Human Adipose Tissue Analysis

For subsequent confirmation, discarded subcutaneous adipose tissues from 9 deidentified female participants with weight excess or obesity (BMI range 23.6-39.3 kg/m²) were examined. Discarded deidentified tissues are collected under nonhuman subject research certification per NYU Instituational Review Board policy.

Statistics

All data were analyzed by two-way ANOVA to determine the effect of diet (control diet vs HFD diet) and genotype/adipocyte-specific deletion of Fgf23 [control vs knockout (KO), Fgf23^fl/fl vs AdnCre-Fgf23^fl/fl, respectively] using GraphPad Prism version 8.0 for Windows (GraphPad Software, San Diego, CA). Bonferroni correction was used to adjust for multiple pairwise comparison. All data were expressed at mean ± SD. P-values less than .05 were considered significant.

Results

FGF23 is Expressed in White and Brown Adipocytes.

Although osteocytes and osteoblasts in bone are the main source of Fgf23 production (10, 11, 30, 31), ablation of Fgf23 in osteoblastic lineage cells reduced circulating FGF23 by only about 50% (32, 33), indicating that there are other significant sources of Fgf23. Indeed, lower levels of Fgf23 expression have been reported in other tissues, such as the brain, thymus, heart, and liver (10). Our detailed analysis revealed that Fgf23 mRNA and protein are expressed both in white and brown mouse adipose tissues at different anatomical sites, including inguinal white adipose tissues (iWAT), perigonadal white adipose tissues (pgWAT), and interscapular brown adipose tissue (iBAT) (Fig. 1A and 1B). Interestingly, FGF receptors and Klotho, the coreceptor for FGF23, are also coexpressed in mouse white and brown adipose tissues, with Fgfr1 being the most highly expressed (Fig. 1C).

Figure 1.

Expression of Fgf23 and FGF receptors in different types of adipose tissue from 8-week-old C57BL/6J mice. (A) Real-time quantitative RT-PCR showing expression of Fgf23 in mouse white and brown adipose tissues (n = 3). The geometric mean of 2 reference genes (β-actin and Atpf1) was used for internal normalization. Data are represented as relative expression (2^−ΔCt) of Fgf23 normalized to the geometric mean of the 2 reference genes. (B) Western blot image of FGF23 protein expression in adipose tissue depots from C57BL/6J mice (n = 3). (C) Real-time quantitative RT-PCR showing expression of Fgfrs and Klotho in C57BL/6J mouse white and brown adipose tissues. Data are represented as fold change (Δ) relative to Rpl4 (n = 3).

Abbreviations: BAT, brown adipose tissue; FGF, fibroblast growth factor; iBAT, interscapular brown adipose tissue; ingWAT, inguinal white adipose tissue; WAT, white adipose tissue.

High FGF23 Promotes Adipogenesis In Vitro

To investigate whether FGF23 affects adipogenesis in a cell-autonomous manner, we treated 3T3-L1 cells with FGF23 or PBS (vehicle) during differentiation. Specifically, 3T3-L1 cells were cultured in adipocyte differentiation medium to which FGF23 or PBS was added from day 0 to day 8 of differentiation. Our data show that FGF23 promotes adipogenic differentiation of 3T3-L1 cells compared with the vehicle-treated group, as indicated by increased Oil Red O staining (Fig. 2A) and increased expression of aP2 and Pparγ (Fig. 2B).

Figure 2.

FGF23 increases lipid accumulation in differentiated 3T3-L1 cells. (A) Oil Red O staining of differentiated 3T3-L1 cells at day 8 of differentiation after treatment with 100 ng/mL of FGF23 or PBS (vehicle) from day 0 to day 8. (B) Real-time quantitative RT-PCR showing expression of aP2 and Pparγ in differentiated 3T3-L1. Data are normalized to Rpl4 and represented as fold change (Δ) relative to control (vehicle) (n = 3).

Abbreviation: FGF23, fibroblast growth factor 23.

HFD Increases Circulating Levels and Adipose Tissue Expression of Fgf23

Studies have shown that circulating FGF23 levels increase with HFDs and are correlated with increased fat mass in humans and rodents (21-23). We confirmed the increase in serum FGF23 levels in wild-type female mice fed a HFD (Fig. 3A). Importantly, this elevation in serum FGF23 levels is accompanied by an increase in the levels of Fgf23 mRNA in inguinal adipose tissue after a HFD (Fig. 3B). These data indicate a possible role for Fgf23 in obesity.

Figure 3.

Circulating levels and AT Fgf23 expression in wild-type mice after high-fat diet. (A) Serum levels of intact FGF23 measured by ELISA; (B) real-time quantitative RT-PCR showing expression of Fgf23 in perigonadal and inguinal white AT, interscapular brown AT, and bone, in normal mice fed normal or high-fat diet for 24 weeks. Data are normalized to Tbp (n = 6). Re-do normalization for mRNA data.

Abbreviations: AT, adipose tissue; FGF23, fibroblast growth factor 23.

Moreover, to assess the relevance of FGF23 to human obesity, we evaluated FGF23 levels in serum samples collected from deidentified individuals with BMI 20.7 to 52.6 kg/m² and the expression of Fgf23 in subcutaneous adipose tissues from bariatric surgery participants with BMI ranging between 23.6 and 39.3 kg/m². Our data show that circulating FGF23 (Fig. 4A) and adipose tissue Fgf23 transcript levels (Fig. 4B) positively correlate with BMI.

Figure 4.

Circulating and adipose tissue levels of FGF23 in humans are increased as BMI increases. (A) Serum levels of intact FGF23 measured by ELISA in lean (BMI <25), overweight (BMI 25-29.9), and obese (BMI >30) female individuals (n = 23). (B) Linear regression analysis of Fgf23 expression in subcutaneous adipose tissues of clinical female participants with excess weight or obesity (n = 9), with significance assessed by Pearson's correlation. Curved lines indicate 95% confidence intervals. n represents number of biologically independent human participants.

Abbreviations: BMI, body mass index; FGF23, fibroblast growth factor 23.

Effect of HFD on Fat Mass in Mice Lacking Fgf23 in Adipocytes

To determine the effects of FGF23 in fat in response to obesity in vivo, we generated ^AdipoqFgf23^Δfl/Δfl mice by breeding Fgf23^fl/fl mice carrying loxP sequences flanking exon 1 of the mouse Fgf23 gene (33) with Adipoq-Cre mice expressing a Cre transgene under the control of the mouse adiponectin promoter. Deletion of Fgf23 in white and brown adipose tissues was successfully achieved, as shown in Fig. 5A, and Fgf23 gene knockout efficiency at the mRNA level is shown in Fig. 5B.

Figure 5.

Validation of ^AdipoqFgf23^Δfl/Δfl and Fgf23^fl/fl mice. (A) Representative agarose gel image showing amplicons of the Fgf23 excision product in WAT, perigonadal and inguinal WAT, and iBAT in control and KO mice. Excision product size: 370 bp. (B) Real-time quantitative RT-PCR of FGF23 expression, showing significant knockdown of Fgf23 mRNA in inguinal and perigonadal WAT and iBAT. mRNA data are normalized to Tbp and represented as fold change (Δ) relative to control (n = 6).

Abbreviations: Cont, control (Fgf23^fl/fl); FGF23, fibroblast growth factor 23; iBAT, interscapular brown adipose tissue; KO, knockout (^AdipoqFgf23^Δfl/Δfl); WAT, white adipose tissue.

Male and female ^AdipoqFgf23^Δfl/Δfl mice and control littermates (floxed mice without Cre expression) were fed either HFD or normal-fat diet (NFD) for 24 weeks. Data obtained from female mice are shown in the main manuscript, whereas data obtained from male mice are reported in the Supplementary Material (29). Body weight measurements were taken weekly. Fat mass was measured by EchoMRI in live animals, and fat pads were dissected and weighed postmortem. As shown in Fig. 6A to 6D, female Fgf23^fl/fl (control) mice fed a HFD gained substantial body weight and fat mass, whereas female ^AdipoqFgf23^Δfl/Δfl mice fed a HFD had significantly decreased body weight and body fat mass compared to controls. The differences in body weight between female control and KO mice became evident after 18 weeks of HFD, and the body weight of ^AdipoqFgf23^Δfl/Δfl mice remained lower until the end of the HFD period (24 weeks). Importantly, targeted deletion of Fgf23 in adipocytes prevented the loss of lean body mass observed in controls (Fig. 6E). Moreover, the weights of all 3 fat depots (iWAT, pgWAT, and iBAT) collected postmortem were significantly lower in HFD-fed female ^AdipoqFgf23^Δfl/Δfl mice compared to Fgf23^fl/fl (Fig. 6F-6H). Histological analysis of iWAT, pgWAT, and iBAT depots stained with H&E showed decreased adipocyte size in iWAT and pgWAT and reduced lipid accumulation in iBAT in female ^AdipoqFgf23^Δfl/Δfl mice fed a HFD compared to Fgf23^fl/fl mice (Fig. 6I).

Figure 6.

Body weight and fat mass of female ^AdipoqFgf23^Δfl/Δfl and Fgf23^fl/fl mice fed a HFD. Female ^AdipoqFgf23^Δfl/Δfl and Fgf23^fl/fl Cont mice were fed a HFD or NFD for 24 weeks starting at 8 weeks of age. (A) Weekly body weight measurements (g) of ^AdipoqFgf23^Δfl/Δfl and Fgf23^fl/fl Cont mice fed a HFD. (B-E) Representation of body weight and whole-body fat mass measured before HFD (0 weeks) and after 24 weeks of either HFD or NFD using EchoMRI. (B) Body weight (g); (C) whole body fat weight (g); (D) percent of fat to body weight; and (E) percent of lean mass to body weight. (F-H) Weights of dissected fat pad depots were taken posteuthanasia of ^AdipoqFgf23^Δfl/Δfl and Fgf23^fl/fl Cont mice after 24 weeks of HFD or NFD. (F) iWAT; (G) pgWAT; and (H) iBAT. Data are presented as mean ± SD. (Fgf23^fl/fl Cont mice: n = 13-18 per group; ^AdipoqFgf23^Δfl/Δfl: n = 7-16 per group). (I) Representative images of hematoxylin and eosin stained sections of iWAT, pgWAT, and iBAT from female ^AdipoqFgf23^Δfl/Δfl and Fgf23^fl/fl Cont mice fed HFD for 24 weeks (n = 4-6).

Abbreviations: Cont, control (Fgf23^fl/fl); HFD, high-fat diet; iBAT, interscapular brown adipose tissue; iWAT, inguinal white adipose tissue; KO, knockout (^AdipoqFgf23^Δfl/Δfl); NFD, normal-fat diet; pgWAT, perigonadal white adipose tissue; Wks, weeks.

Interestingly, the effects observed in female KO mice were not detected in male KO mice. ^AdipoqFgf23^Δfl/Δfl and Fgf23^fl/fl mice fed a HFD or NFD for 24 weeks had comparable gains in body weight and fat mass (Supplementary Fig. S1A-1D) (29) and decrease in lean body mass (Supplementary Fig. S1E) (29). Similarly, there was no difference in the weights of iWAT and pgWAT collected from male ^AdipoqFgf23^Δfl/Δfl and Fgf23^fl/fl mice after 24 weeks of HFD or NFD (Supplementary Fig. S1F and S1G) (29), although the weight of iBAT appeared to be lower in HFD-fed male ^AdipoqFgf23^Δfl/Δfl compared to Fgf23^fl/fl mice (Supplementary Fig. S1H) (29). The distinct differences in body weight and body fat between female and male mice suggest an important role of adipocytic Fgf23 in mediating sexually dimorphic effects of HFD.

Loss of Fgf23 in Adipocytes Improves Lipid and Glucose Metabolism in HFD State

Given the decreased adipose tissue weight and lipid accumulation in female ^AdipoqFgf23^Δfl/Δfl mice fed a HFD, we assessed whether the phefnotypical changes observed were associated with molecular alterations in the expression of genes involved in lipid homeostasis. Our gene expression studies using nCounter technology demonstrated that female ^AdipoqFgf23^Δfl/Δfl mice fed a HFD have significantly reduced expression of genes associated with fatty acid (FA) synthesis and storage (Fig. 7A-7E) and increased expression of genes associated with FA oxidation (FAO) in iWAT (Fig. 7F-I).

Figure 7.

Expression of genes associated with lipid homeostasis in iWAT. Expression of genes associated with lipid homeostasis was assessed using nCounter technology in iWAT from Fgf23^fl/fl (control) and ^AdipoqFgf23^Δfl/Δfl (KO) female mice fed a HFD for 24 weeks. (A-E) Genes associated with FA synthesis and storage, (F-I) Genes associated with FA oxidation. (A) ACC (acetyl CoA carboxylase), (B) CD36, (C) FAS (FA synthase), (D) Pparγ, (E) Srebp1 (sterol regulatory element binding protein 1), (F) AMPK (AMP-activated protein kinase), (G) Cpt1 (carnitine palmitoyl transferase), (H) Mcad (medium chain acetyl CoA dehydrogenase), and (I) Pgc1α (peroxisome proliferator-activated receptor gamma coactivator 1-alpha). (n = 5-7). Data are presented as mean ± SD.

Abbreviations: Cont, control (Fgf23^fl/fl); FA, fatty acid; HFD, high-fat diet; iWAT, inguinal white adipose tissue; KO, knockout (^AdipoqFgf23^Δfl/Δfl).

Moreover, loss of Fgf23 in adipocytes in female mice fed a HFD was accompanied by decreased iWAT inflammation and oxidative stress as is shown by reduced expression of inflammatory markers (Fig. 8A and 8B) and oxidative stress genes (Fig. 8C) and increased expression of the antioxidant enzyme superoxide dismutase (Sod1) (Fig. 8D) compared with controls (Fgf23^fl/fl).

Figure 8.

Expression of genes associated with inflammation and oxidative stress. Expression of genes associated with inflammation and oxidative stress was assessed using nCounter technology in iWAT from Fgf23^fl/fl (Cont) and ^AdipoqFgf23^Δfl/Δfl (KO) female mice fed a HFD for 24 weeks. (A-B) Genes associated with inflammation; (C-D) genes associated oxidative stress. (A) IFNγ; (B) IL6; (C) Nox4 (NADPH oxidase 4); (D) Sod1 (superoxide dismutase) (n = 5-7). Data are presented as mean ± SD.

Abbreviations: Cont, control (Fgf23^fl/fl); HFD, high-fat diet; iWAT, inguinal white adipose tissue; KO, knockout (^AdipoqFgf23^Δfl/Δfl).

Analysis of serum parameters revealed that female ^AdipoqFgf23^Δfl/Δfl mice on HFD have significantly decreased serum cholesterol, triglycerides (TGs), and free FAs (Fig. 9A-9C) and lower blood glucose and insulin levels (Fig. 9D) than control mice.

Figure 9.

Serum biochemistry of female mice with deletion of Fgf23 in adipocytes. Fgf23^fl/fl (Cont) and ^AdipoqFgf23^Δfl/Δfl (KO) female mice were fed a HFD or NFD. Serum total cholesterol (mg/dL) (A), triglycerides (mg/dL) (B), free fatty acids (μM) (C), fasting blood glucose (mg/dL) (D), and fasting plasma insulin (pmol/L) (E) were measured using commercial kits. (n = 6-9). Data are presented as mean ± SD.

Abbreviations: Cont, control (Fgf23^fl/fl); HFD, high-fat diet; KO, knockout (^AdipoqFgf23^Δfl/Δfl); NFD, normal-fat diet.

Moreover, HFD resulted in significantly higher liver fat accumulation in female Fgf23^fl/fl mice compared to ^AdipoqFgf23^Δfl/Δfl mice. H&E staining of liver sections showed a significant increase in lipid accumulation (Fig. 10A) with elevated hepatic TG levels (Fig. 10B) and higher liver weight (Fig. 10C) in HFD-fed female Fgf23^fl/fl mice compared with ^AdipoqFgf23^Δfl/Δfl mice.

Figure 10.

Hepatic lipid accumulation is reduced in female mice with deletion of Fgf23 in adipocytes. Fgf23^fl/fl (Cont) and ^AdipoqFgf23^Δfl/Δfl (KO) female mice were fed a HFD or NFD for 24 weeks. (A) Representative images of hematoxylin and eosin staining for liver sections from HFD-fed Fgf23^fl/fl (Cont) and ^AdipoqFgf23^Δfl/Δfl (KO) female mice. (B) Liver triglyceride levels (mg/g of liver) and (C) liver weight (g). (n = 4-18). Data are presented as mean ± SD.

Abbreviations: Cont, control (Fgf23^fl/fl); HFD, high-fat diet; KO, knockout (^AdipoqFgf23^Δfl/Δfl); NFD, normal-fat diet.

Given that obesity is often associated with fatty liver, we assessed whether the liver of female KO mice is protected from lipid deposition in the HFD state. As shown in Fig. 11, deletion of Fgf23 in adipocytes resulted in lower levels of expression of genes related to FA uptake (Fig. 11A and 11B), de novo lipogenesis (Fig. 11C), lipid accumulation (Fig. 11D), and inflammation (Fig. 11F), compared to controls. In addition, loss of Fgf23 in adipocytes resulted in upregulation of hepatic Pgc1α in female ^AdipoqFgf23^Δfl/Δfl mice fed a HFD (Fig. 10E). Elevation in hepatic PGC-1α expression results in increased FA oxidation and reduced hepatic TG content and secretion (34).

Figure 11.

Hepatic expression of genes associated with lipid metabolism and inflammation in female mice with deletion of Fgf23 in adipocytes. Real-time quantitative RT-PCR showing expression of (A) Pparγ, (B) CD36, (C) Fas, (D) Perilipin, (E) TNF-α , and (F) Pgc-1α in liver of Fgf23^fl/fl and ^AdipoqFgf23^Δfl/Δfl female mice fed HFD or NFD for 24 weeks. (n = 6). Data are presented as mean ± SD. Data are normalized to Rpl4 and represented as fold change (Δ) relative to control.

Abbreviations: Cont, control (Fgf23^fl/fl); HFD, high-fat diet; KO, knockout (^AdipoqFgf23^Δfl/Δfl); NFD, normal-fat diet.

Taken together, our findings that female KO mice exhibit reductions in weight gain and inflammation of the adipose tissue, as well as a decrease in serum and liver lipids, demonstrate that adipocyte selective deletion of Fgf23 confers a protective effect against HFD-induced obesity and its detrimental consequences.

Mice With Adipocyte Selective Deletion of Fgf23 Have Increased iBAT Ucp1 Levels and Decreased RER

In obesity, there is excess accumulation of WAT and reduction of metabolically active beige and brown adipocytes in favor of white adipocytes and a decline in BAT activation (35, 36). Activating brown and beige adipocytes is a potential therapeutic approach to protect against obesity. Our data show that there is no difference in food and energy intake between ^AdipoqFgf23^Δfl/Δfl and control female mice fed a NFD or a HFD (Fig. 12A and 12B). Moreover, loss of Fgf23 in adipocytes increases Ucp1 protein expression in iBAT in female ^AdipoqFgf23^Δfl/Δfl mice fed a HFD compared to HFD-fed controls (Fig. 12C and 12D), suggesting that adipose Fgf23 plays a role in thermogenesis.

Figure 12.

iBAT Ucp1 protein expression is increased in female mice with deletion of adipose Fgf23 whereas food and energy intake are unchanged. (A) Average food intake and (B) energy uptake of female ^AdipoqFgf23^Δfl/Δfl and Fgf23^fl/fl mice fed a NFD or HFD for 24 weeks. (n = 6 per group). Energy uptake was calculated by multiplying the consumed food in grams with the calories per gram of the respective type of diet. (C) Representative Western blot image showing expression of Ucp1 in iBAT of Cont and KO mice fed a NFD or HFD for 24 weeks. (D) Quantitation of Ucp1 protein expression normalized to β-actin expression. (n = 6 per group). Data are presented as mean ± SD.

Abbreviations: AVG, average; HFD, high-fat diet; iBAT, interscapular brown adipose tissue; KO, knockout (^AdipoqFgf23^Δfl/Δfl); NFD, normal-fat diet.

Our data also demonstrate that ^AdipoqFgf23^Δfl/Δfl fed a HFD exhibit decreased respiratory exchange ratio compared to controls (Fig. 13A and 13B), indicating increased lipid oxidation and further suggesting that the effect of adipocyte selective deletion of Fgf23 on body composition is mediated, at least partly, through enhanced lipid metabolism. Energy expenditure (Fig. 13C and 13D), physical activity (XT + YT) (Fig. 13E and 13F), oxygen consumption (Fig. 13G and 13H), and carbon dioxide production (Fig. 13I and 13J) were not significantly changed as compared to control mice.

Figure 13.

Effects of adipocyte selective deletion of Fgf23 on energy homeostasis. Indirect calorimetry measurements across 72 hours using TSE PhenoMaster were performed in ^AdipoqFgf23^Δfl/Δfl mice (n = 8) and control (Fgf23^fl/fl) (n = 7) fed a HFD for 24 weeks. (A) Average RER ( = vO2/vCO2) across the 72-hour period; (B) scatter plot of mean RER by initial body weight for each genotype with best linear fits; (C) average EE (kcal heat produced per hour); (D) scatter plot of mean EE by initial body weight for each genotype with best linear fits; (E) physical activity as measured by the average number of beam break counts in different dimensions (XT + YT); (F) scatter plot of total activity by initial body weight for each genotype with best linear fits; (G) average VO₂ consumed (ml per hour); (H) scatter plot of mean VO₂ by initial body weight for each genotype with best linear fits; (I) total VCO₂ produced (ml per hour); (J) scatter plot of mean VCO₂ by initial body weight for each genotype with best linear fits. Data are expressed as mean ± SD. Data in panels A, C, E, G, and I were analyzed using a linear mixed effects model, including genotype and sex as categorical variables and body weight as a covariate. Data in panels B, D, F, H, and J were analyzed using simple linear regression to compare intercepts and slopes.

Abbreviations: EE, energy expenditure; HFD, high-fat diet; KO, knockout (^AdipoqFgf23^Δfl/Δfl); RER, respiratory exchange ratio.

Discussion

This is the first in vivo study to show that in the obese state circulating and adipose tissue levels of FGF23 increase and positively correlate with BMI, and targeted deletion of Fgf23 in adipocytes protects from diet-induced weight gain and metabolic dysregulation, such as dyslipidemia and hyperglycemia, while preventing loss of lean mass, increasing lipolysis, and decreasing RER. Specifically, our main findings show that, under long-term HFD, elimination of Fgf23 in adipocytes (^AdipoqFgf23^Δflox/Δflo; KO) in mice (1) results in resistance in body weight gain, (2) does not have anorexigenic properties and does not affect food intake, (3) prevents dyslipidemia and hyperglycemia, and (4) leads to healthier WAT and liver with decreased lipid accumulation and inflammation, compared to control mice (Fgf23^fl/fl). The reduction in fat in the KO mice is accompanied by (1) preservation of lean body mass, (2) decrease in FA synthesis and increase in FAO, (3) increased Ucp1 expression in BAT, and (4) decreased RER. There is no difference between KO and control mice fed a NFD in any of the parameters examined and regardless of sex. Thus, our findings clearly demonstrate that deletion of Fgf23 in adipocytes has long-term protective effects on body mass and metabolic parameters in mice exposed to high dietary fat, proposing that elevated FGF23 has a pathogenic function in the coordination of metabolism and development of obesity.

The finding that lean body mass is preserved even though body weight and fat mass decrease indicates that targeted deletion of Fgf23 in adipocytes not only causes fat selective weight loss but also prevents loss of lean mass. This is a novel and critical finding as current medications for weight loss are responsible for both fat and lean mass reductions, potentially affecting metabolic health and decreasing the body's energy expenditure. Thus, our work is supportive of a candidate weight loss approach that would achieve weight loss while maintaining lean muscle mass and preserving energy expenditure.

Cellular lipid metabolism depends on a delicate balance between FA uptake, de novo synthesis, FAO/catabolism, and secretion. However, a mismatch between these processes occurs in obesity and leads to excessive intracellular storage of FAs. High FGF23 is associated with lower high-density lipoprotein and higher serum TGs in individuals with higher BMI and fat mass (22), as well as with the presence and severity of hepatic steatosis (37). In agreement with published data, HFD led to significantly increased FA synthesis and decreased FAO in WAT and liver and higher blood and liver lipids in Fgf23^fl/fl mice. However, female ^AdipoqFgf23^Δfl/Δfl mice fed a HFD have lower serum cholesterol, TGs, and free FAs; decreased liver lipid accumulation; significantly reduced expression of genes associated with FA synthesis and storage; and increased expression of genes associated with FAO in WAT and liver. The latter is coupled with upregulation of hepatic Pgc1α and it is consistent with published data showing that elevation in hepatic PGC-1α expression results in increased FAO and reduced hepatic TG content and secretion (34). Activation of AMPK facilitates phosphorylation and inactivation of ACC, resulting in reduction in FA synthesis and increased FA uptake and FAO by mitochondria (38, 39). In diet-induced obesity, AMPK activity is reduced, leading to decreased phosphorylation and increased activity of its substrate acetyl-CoA carboxylase (ACC), thereby reducing FAO and increasing FA synthesis (38). It has been shown that AMPK activation downregulates, whereas AMPK inhibition induces FGF23 production (40). Our findings show that inactivation of Fgf23 in adipocytes under HFD conditions results in higher AMPK RNA expression in WAT and lower expression of ACC, supporting a lipogenic role for adipocytic Fgf23. Importantly, selective inactivation of Fgf23 in adipocytes favors lipolysis and reverses the effect of HFD on AMPK activity.

Moreover, loss of Fgf23 in adipocytes results in “healthier” WAT and liver with decreased inflammation, compared with controls (Fgf23^fl/fl), which is in agreement with our previous work showing that inhibition of FGF23 signaling alleviates inflammation (41, 42). Inflammation is a known stimulator of FGF23 (43, 44). Expanded adipose tissue in the obese state is a source of leptin and inflammatory cytokines (eg, TNFα, IL-6) that stimulate directly and indirectly FGF23 production (45-47). Given that inflammatory cells can produce FGF23, its production by these cells may contribute to the regulation of the inflammatory response. Furthermore, we cannot rule out the involvement of aldosterone in the actions of FGF23 with regards to inflammation. In obesity, adipose tissue releases leptin, which stimulates aldosterone secretion. Aldosterone induces inflammation by activating mineralocorticoid receptors, which are expressed in several types of cells including adipocytes and inflammatory cells to regulate blood pressure and electrolyte and fluid homeostasis. Aldosterone has also been shown to directly upregulate FGF23 in CKD (48, 49).

Our data also show that adipocyte Fgf23 affects liver lipid metabolism. However, our study does not demonstrate whether it directly affects hepatic lipid accumulation. To address this, further in vitro studies of primary hepatocytes or HepG2 cells are needed to distinguish direct vs indirect effects of FGF23 on liver lipid metabolism. HepG2 cells, in particular, are a valuable cell line of physiological relevance because they are derived from human liver tissue and they also represent the liver as an ectopic site for lipid deposition.

Obesity is also a major risk for developing insulin resistance, a hallmark of type 2 diabetes. Several studies suggest an inverse correlation between FGF23 levels and insulin sensitivity in obesity (50-53). Our data show that blood glucose and insulin levels were significantly elevated in control mice fed a HFD but they were down to normal levels in ^AdipoqFgf23^Δfl/Δfl mice.

In the present study we also demonstrate that there are no differences in food intake between ^AdipoqFgf23^Δfl/Δfl and Fgf23^fl/fl mice. However, Ucp1 protein expression in iBAT is increased in female ^AdipoqFgf23^Δfl/Δfl mice fed avHFD, and RER is significantly decreased compared to HFD-fed controls, suggesting that adipose Fgf23 deficiency results in increased metabolic activity and increased utilization of fat as a metabolic substrate. The decrease in RER is consistent with the reduction in circulating TGs, indicating increased TG clearance and further suggesting that, in the absence of adipocytic Fgf23, the thermogenic activity of BAT is enhanced and fatty acids serve as fuels to be oxidized for thermoregulation. These findings strongly suggest that selective deletion of Fgf23 in adipocytes is a promising strategy to treat hyperlipidemia and improve metabolic health.

Brown adipocytes are activated in response to certain stimuli such as cold exposure or stimulation of the β3-AR signaling pathway (36, 54, 55). Increased thermogenic potential is associated with increased adipose tissue browning. In obesity, there is excess accumulation of WAT, a reduction of metabolically active beige and brown adipocytes in favor of white adipocytes, and a decline in BAT activation (56, 57). Activating brown and beige adipocytes is a potential therapeutic approach to protect against obesity. Further studies are needed to determine whether BAT activation in ^AdipoqFgf23^Δfl/Δfl mice is the result of cell-autonomous actions of adipose Fgf23 or dependent on the stimulation of β-adrenergic signaling.

In our study we used the adiponectin Cre mouse because, in contrast to other “adipocyte” Cre lines, the adiponectin promoter-driven mouse model expression of the transgene is even across the different WAT and BAT depots and is restricted only to differentiated adipocytes and is not detected in the stromal vascular fraction of adipose tissue, macrophages, or any other cells.

However, because Adipoq-Cre is expressed in both white and brown adipocytes, in the present study we cannot differentiate the effect of Fgf23 in these 2 cell types. To address this, we generated mice in which Fgf23 is specifically deleted in brown and beige adipocytes using Ucp1-Cre. The use of Adipoq-Cre and Ucp1-Cre would allow us to distinguish the effect of Fgf23 in differentiated white and brown adipocytes under HFD conditions. However, to understand the impact of Fgf23 on brown adipocyte differentiation, further studies are needed, such as using the Myf5-Cre, which is only active in the early brown adipocyte precursors and not in mature brown adipocytes. Moreover, future work can establish whether FGF23 has a direct role on WAT expansion and browning and address the possibility that the reduction in body weight and fat in ^AdipoqFgf23^Δfl/Δfl mice may be due to different response to cold.

FGF21 has been shown to have a protective role in obesity as it reduces body weight and whole-body fat mass in diet-induced obesity mice due to marked increases in thermogenesis, total energy expenditure, and physical activity levels (58-60). FGF21 also reduces blood glucose, insulin, and lipid levels and reverses hepatic steatosis (58). To determine whether compensatory upregulation of FGF21 in ^AdipoqFgf23^Δfl/Δfl mice may be responsible for the lean phenotype of the adipose-deficient Fgf23 mice, we measured FGF21 serum levels in our mice. We did not see any significant differences in FGF21 serum levels between control and KO mice fed either NFD or HFD, suggesting that the phenotype of ^AdipoqFgf23^Δfl/Δfl mice is independent of FGF21 actions.

The effects we see upon deletion of Fgf23 in obesity are sexually dimorphic and observed only in female mice. Sexual dimorphism in the metabolic responses to dietary challenges and weight gain/loss has been reported in several studies, and it is either inherent to each sex or due to effects of sex steroids. Sexual dimorphism in obesity is attributed to structural and functional brain differences regulating food intake and energy metabolism (61-63), estrogen-dependent sympathetic innervation promoting recruitment of brown adipocytes in WAT (64), or differences in the gut microbiome (65-68). Our data suggest that the actions of adipocyte Fgf23 are sex-dependent. While investigating these possibilities is not within the scope of the present study, further studies, including gonadectomy and administration of sex hormones, are needed to provide a better understanding of the observed sexual dimorphism in the fat phenotype of our mice.

Although global deletion of Fgf23 has detrimental effects on overall health affecting many organ systems, tissue-specific KO studies have been very successful in analyzing the role of locally produced Fgf23. Successful use of FGF23 antagonists and neutralizing antibodies for the treatment of chronic conditions has been demonstrated by us and others. Our previous studies showed that pharmacologic inhibition of FGF23 signaling alleviates the pathological states of inflammation, iron deficiency, and anemia in mice (41, 42). Chronic inflammation and iron deficiency are well-characterized features of obesity (69, 70), suggesting that lowering or inhibiting FGF23 could be used as a novel therapeutic option to alleviate obesity and associated conditions. Moreover, the use of FGF23 neutralizing antibodies and blocking peptides has therapeutic potential in diseases such as X-linked hypophosphatemia, characterized by rickets and osteomalacia due to high FGF23, as they improve bone quality and mineralization and correct hypophosphatemia (71-74). Recent studies showed an increased prevalence of obesity in children with X-linked hypophosphatemia, attributing it to low phosphorus status caused by elevated FGF23 (75, 76).

Importantly, obesity has been identified as a major independent risk factor for heart disease and a powerful predictor of cardiovascular morbidity and mortality, especially in patients with CKD (1). Ventricular hypertrophy is the predominant and most specific alteration in the heart of obese individuals who develop cardiomyopathy with increased wall thickness and cavity volume of the left ventricle (77). High circulating FGF23 levels have been linked to greater cardiovascular risk, higher vascular and aortic calcifications, cardiac hypertrophy, and mortality, particularly in CKD patients (18-20). A recent study has shown that blocking FGF23 signaling in mice with high endogenous circulating FGF23 significantly improved heart and kidney function (78). We are currently investigating the potential protective role of FGF23 bioactivity attenuation against cardiac hypertrophy in obesity.

Collectively, our findings demonstrate the potential of targeted Fgf23 inhibition for treatment and prevention of obesity and associated conditions such as dyslipidemia, hyperglycemia, inflammation, hepatic steatosis, and cardiac hypertrophy. Therefore, we have identified a new player that will help us to understand the basis of chronic conditions such as obesity associated with metabolic dysregulation. This will open up new pathways for the use of antagonists as novel therapeutic targets that modulate FGF23 production that would lead to improvements in the current therapeutic approaches.

Funding

This work was supported by National Institutes of Health (NIH) grant RF1AG075742 (W.C.), National Institutes of Health (NIH) grant R01DK122259 (W.C.), Biomedical Laboratory Research and Development, VA Office of Research and Development grant I01BX005851 (W.C.), and Biomedical Laboratory Research and Development, VA Office of Research and Development grant IK6BX004835 (W.C.); American Heart Association 17-SFRN33490004 (J.O.A.), National Institutes of Health (NIH) K08 DK117064 (J.O.A.), and Doris Duke Charitable Foundation (J.O.A.).

Disclosures

The authors have nothing to disclose.

Data Availability

Original data generated and analyzed during this study are included in this published article or in the data repositories listed in References.