Mechanism of FGF23 processing in fibrous dysplasia

Authors

  • Nisan Bhattacharyya,

    1. Skeletal Clinical Studies Unit, Craniofacial and Skeletal Diseases Branch, National Institute of Dental and Craniofacial Research, National Institutes of Health, Bethesda, MD, USA
    Search for more papers by this author
  • Malgorzata Wiench,

    1. Laboratory of Receptor Biology and Gene Expression, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA
    Search for more papers by this author
  • Claudia Dumitrescu,

    1. Skeletal Clinical Studies Unit, Craniofacial and Skeletal Diseases Branch, National Institute of Dental and Craniofacial Research, National Institutes of Health, Bethesda, MD, USA
    Search for more papers by this author
  • Brian M Connolly,

    1. Oral and Pharyngeal Cancer Branch, National Institute of Dental and Craniofacial Research, National Institutes of Health, Bethesda, MD, USA
    Search for more papers by this author
  • Thomas H Bugge,

    1. Oral and Pharyngeal Cancer Branch, National Institute of Dental and Craniofacial Research, National Institutes of Health, Bethesda, MD, USA
    Search for more papers by this author
  • Himatkumar V Patel,

    1. Section on Biological Chemistry, National Institute of Diabetes and Kidney Diseases, National Institutes of Health, Bethesda, MD, USA
    Search for more papers by this author
  • Rachel I Gafni,

    1. Skeletal Clinical Studies Unit, Craniofacial and Skeletal Diseases Branch, National Institute of Dental and Craniofacial Research, National Institutes of Health, Bethesda, MD, USA
    Search for more papers by this author
  • Natasha Cherman,

    1. Skeletal Clinical Studies Unit, Craniofacial and Skeletal Diseases Branch, National Institute of Dental and Craniofacial Research, National Institutes of Health, Bethesda, MD, USA
    Search for more papers by this author
  • Monique Cho,

    1. Kidney Disease Branch, National Institute of Diabetes and Kidney Diseases, National Institutes of Health, Bethesda, MD, USA
    Search for more papers by this author
  • Gordon L Hager,

    1. Laboratory of Receptor Biology and Gene Expression, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA
    Search for more papers by this author
  • Michael T Collins

    Corresponding author
    1. Skeletal Clinical Studies Unit, Craniofacial and Skeletal Diseases Branch, National Institute of Dental and Craniofacial Research, National Institutes of Health, Bethesda, MD, USA
    • Skeletal Clinical Studies Unit, Craniofacial and Skeletal Diseases Branch, National Institute of Dental and Craniofacial Research, National Institutes of Health, Building 30, Room 218, Bethesda, MD 20892, USA.
    Search for more papers by this author

Abstract

Fibroblast growth factor-23 (FGF23) is a phosphate- and vitamin D-regulating hormone derived from osteoblasts/osteocytes that circulates in both active (intact, iFGF23) and inactive (C-terminal, cFGF23) forms. O-glycosylation by O-glycosyl transferase N-acetylgalactosaminyltransferase 3 (ppGalNAcT3) and differential cleavage by furin have been shown to be involved in regulating the ratio of active to inactive FGF23. Elevated iFGF23 levels are observed in a number of hypophosphatemic disorders, such as X-linked, autosomal recessive, and autosomal dominant hypophosphatemic rickets, whereas low iFGF23 levels are found in the hyperphosphatemic disorder familial tumoral calcinosis/hyperphosphatemic hyperostosis syndrome. Fibrous dysplasia of bone (FD) is associated with increased total FGF23 levels (cFGF23 + iFGF23); however, classic hypophosphatemic rickets is uncommon. Our results suggest that it can be explained by increased FGF23 cleavage leading to an increase in inactive cFGF23 relative to active iFGF23. Given the fact that FD is caused by activating mutations in the small G-protein Gsα that results in increased cyclic adenosine monophosphate (cAMP) levels, we postulated that there may be altered FGF23 cleavage in FD and that the mechanism may involve alterations in cAMP levels and ppGalNacT3 and furin activities. Analysis of blood specimens from patients with FD confirmed that the elevated total FGF23 levels are the result of proportionally increased cFGF23 levels, consistent with less glycosylation and enhanced cleavage by furin. Analysis of primary cell lines of normal and mutation-harboring bone marrow stromal cells (BMSCs) from patients with FD demonstrated that BMSCs harboring the causative Gsα mutation had higher cAMP levels, lower ppGalNAcT3, and higher furin activity. These data support the model wherein glycosylation by ppGalNAcT3 inhibits FGF23 cleavage by furin and suggest that FGF23 processing is a regulated process that controls overall FGF23 activity in FD patients. © 2012 American Society for Bone and Mineral Research.

Introduction

Inorganic phosphate (Pi) levels in human circulation are regulated by several factors, including parathyroid hormone (PTH), vitamin D, and the phosphate- and vitamin D-regulating hormone fibroblast growth factor-23 (FGF23). Various hyperphosphatemic and hypophosphatemic disorders have been shown to be directly related to altered FGF23 levels in the circulation.1–4 FGF23 is known to regulate blood Pi by decreasing the expression of the Pi transporter genes, Npt2A and Npt2C, in renal proximal tubule cells, thereby reducing renal Pi reabsorption.5 FGF23 also plays a role in vitamin D metabolism by inhibiting the expression of renal proximal tubule 1-α hydroxylase, thus decreasing the production of the active vitamin D metabolite, 1,25 (OH)2 vitamin D,6 and intestinal Pi absorption.5

FGF23 circulates in a physiologically active, intact form (∼32 to 35 kDa, iFGF23) and in an inactive, c-terminal form (∼15 to 17 kDa, cFGF23).7 Elevated total FGF23 levels (cFGF23 + iFGF23) are seen in several human hypophosphatemic disorders, such as X-linked hypophosphatemia (XLH),8 autosomal dominant and autosomal recessive hypophosphatemic rickets (ADHR and ARHR),9, 10 tumor-induced osteomalacia (TIO),8 and fibrous dysplasia of bone (FD).11 Conversely, low levels of iFGF23 and elevated levels of cFGF23 are found in familial tumoral calcinosis (FTC) and hyperphosphatemic hyperostosis syndrome (HHS).12 FTC and HHS can result from homozygous mutations in the O-linked glycosyl transferase N-acetylgalactosaminyltransferase 3 (ppGalNAcT3), a member of the galactosaminyl transferase family that is known to transfer N-acetylgalactomine residues to specific serine/threonine residues.13 It has been shown that mutations in ppGalNAcT3 lead to a loss of FGF23 glycosylation and enhanced cleavage of FGF23, resulting in very low levels of iFGF23 and correspondingly high cFGF23 levels.14 On the other hand, patients with ADHR who carry mutations in FGF23 within the conserved furin cleavage motif, 176RHTR179,15, 16 have elevated blood iFGF23 levels.9, 17 This increase in iFGF23 has been proposed to be the result of impaired furin-mediated degradation of iFGF23 to cFGF23. Furthermore, a study performed in HEK 293 cells demonstrated that FGF23 cleavage can be inhibited by Dec-RVKR-CMK, a selective inhibitor of furin.18 However, to date there is no direct evidence demonstrating the role of furin in the processing of FGF23.

Using an assay that measures total FGF23 (iFGF23 + cFGF23) and does not distinguish between iFGF23 and cFGF23, we previously demonstrated that a majority of patients with FD had elevated plasma total FGF23 levels.11 However, only a minority of FD patients had frank hypophosphatemia.19 These data suggest that, similar to patients with FTC, patients with FD may have a disproportionate increase in circulatory cFGF23. We hypothesized that this elevation in cFGF23/iFGF23 ratio could be because of greater furin activity and/or less furin-blocking glycosylation of FGF23 by ppGalNAcT3. Consistent with this hypothesis is the recent report of a subject with Jansen's metaphyseal chondrodysplasia with very high total (i + c) FGF23 levels and iFGF23 levels that were less elevated relative to total FGF23.20 Given that FD is the result of activating mutations in Gsα and Jansen's metaphyseal chondrodysplasia is the result of an activating mutation in PTH/PTHrP receptor, both of which lead to elevated cyclic adenosine monophosphate (cAMP), we postulated that FGF23 processing may be a process regulated by cAMP via its effect on ppGalNAcT3 and/or furin activities. To test this hypothesis, we investigated the levels of total FGF23 and iFGF23 in subjects with FD and disease controls, as well as the ppGalNAcT3 and furin activity in normal and mutant bone marrow stromal cells (BMSCs), and an FGF23 overproducing cell line.

Materials and Methods

Patients

Complete clinical characterization and confirmation of diagnosis of subjects with FD (n = 14), TIO (n = 12), XLH (n = 8), and renal failure (n = 11) who were enrolled serially were performed as part of ongoing IRB-approved studies of FD and other skeletal diseases at the Mark O. Hatfield Clinical Center at the National Institutes of Health. Informed consent was obtained from all subjects or their guardians. All subjects with FD were diagnosed in the clinical context of McCune-Albright syndrome, all subjects with XLH were confirmed by molecular diagnosis of mutations in PHEX (phosphate-regulating gene with homologies to endopeptidases on the X chromosome), and all subjects with TIO had FGF23-secreting tumors. All subjects with FD, TIO, and XLH had normal renal function. Subjects with XLH and TIO were euparathyroid at the time of analysis. Subjects characterized as having renal failure were 11 consecutive subjects at various stages of posttransplantation and with varying levels of renal function, in terms of creatinine clearance. Serum samples were obtained the morning after an overnight fast and stored at −80°C until used for ELISA analysis. FD is a mosaic disease in which there are Gsα-bearing (mutant) and wild-type (WT) cells. Mutant and WT BMSC cell lines were created from a bone biopsy sample from a patient with FD by cloning mutant and WT cells. The biopsy was collected as part of the same ongoing IRB-approved study.

Detection of FGF23 in human plasma by ELISA

iFGF23 and total (i + c) FGF23 was measured using commercially available ELISA kits following the manufacturers' instructions (Kainos Laboratories, Tokyo, Japan, and Immutopics, San Clemente, CA, USA, respectively).

Bone marrow stromal cells

A bone biopsy sample from a FD patient was used to prepare BMSCs.20 Cells were grown in α-minimal essential medium (α-MEM) containing 20% fetal bovine serum (FBS, Atlanta Biologicals, Lawrenceville, GA, USA), L-glutamine (Glutamax, GIBCO, Carlsbad, CA, USA), and penicillin-streptomycin mix. Because FD is the result of a somatic mutation in Gsα, patients demonstrate a mosaic pattern of tissue involvement. As such, only some bones are involved, and only some bone cells carry the mutation. To be sure cells under investigation were pure populations of either WT or mutation-bearing cells (FD cells), cells were subcloned and the Gsα sequence of individual colonies tested to ensure the cells were pure populations of WT or FD cells. Several colonies were isolated and grown separately. Colonies were selected on the basis of the Gsα sequence—WT (R201R) or FD (R201H)—and maintained as previously described.21 DNA from multiple colonies was PCR amplified and sequenced for subsequent Gsα genotyping. For further studies, two purified colonies were selected that carried the genotypes for either both wild-type alleles (CGT, WT Gsα) or a heterozygous monoallelic (CGT/CAT; R201H Gsα) mutation of GNAS (Supplemental Fig. S1). FD BMSCs grew relatively slower than the WT BMSCs and exhibited more fibroblastic-like phenotype.

Gsα genotyping by DNA sequencing

DNAs from individual colonies were isolated and examined for Gsα mutations (GNAS gene). Total DNA was isolated from individual BMSC colonies and used for amplification reactions to enrich for the Gsα-specific genomic regions by using two specific oligonucleotides: 5'-TGACTATGTGCCGAGCGATC-3' and 5'-CCACGTCAAACATGCTGGTG-3'. PCR products were gel purified and were subsequently sequenced.

cAMP ELISA

To confirm that cAMP levels were altered in Gsα-bearing mutants, cAMP levels were assessed by ELISA. WT and FD BMSCs were grown up to 70% to 80% confluence, and the medium was changed to serum-free medium overnight to avoid any hormone-dependent increase in cAMP. Cells were treated with fresh serum-free medium containing 10 µM 3-isobutyl-1-methylxanthine (IBMX), a phosphodiesterase inhibitor (Sigma, St. Louis, MO, USA). Conditioned medium as well as whole cell extracts were prepared from these cells after 1 hour. cAMP levels from cell-conditioned medium or whole cell extracts were measured using the cAMP Screen-Direct Chemiluminiscent ELISA (Applied Biosystems, Foster City, CA, USA) according to the manufacturer's instructions.

RT-PCR

Relevant transcripts of proteins involved in FGF23 processing were assessed by RT-PCR. Total RNA from WT and FD BMSCs were isolated by RNeasy Mini Kit (Qiagen Biosciences, Gaithersburg, MD, USA). Samples were treated with RNase-free DNase set (Qiagen) to avoid any chromosomal DNA contamination. Equal amounts of total RNA was taken for first-strand synthesis using Cloned AMV First Strand cDNA Synthesis Kit (Invitrogen, Carlsbad, CA, USA). Amplification reactions were performed to detect FGF23, ppGalNAc-T3, furin, and GAPDH using gene-specific primers (Supplemental Table S1). Reaction products were run on 3% agarose gel and photographed.

Cytoimmunochemistry and laser confocal microscopy

To assess for the presence and subcellular colocalization of FGF23, GalNAc-T3, and furin, cytoimmunochemistry and laser confocal microscopy were performed. BMSCs (1 × 105/24-well plate) or HEKF cells were grown on glass coverslips (Fisher Scientific, Pittsburgh, PA, USA) in complete media for 24 hours. Before microscopy, cells were washed in phosphate-buffered saline (PBS) three times for 5 minutes each and then fixed with 4% paraformaldehyde (Electron Microscopy Sciences, Hatfield, PA, USA) in PBS at room temperature for 15 minutes. Cells were washed with PBS several times and subsequently permeabilized with 0.5% Triton X-100 in PBS at room temperature for 15 minutes. Cells were washed with PBS several times and blocked using 5% FBS in PBS for 4 to 6 hours at room temperature. Samples were incubated overnight at 4°C with primary antibodies (1:400 dilution of goat anti-human FGF23 and either mouse anti-human ppGalNAcT3 or mouse anti-human furin). Affinity-purified goat anti-human FGF23 antibody (raised against amino acid residues 225 to 244) was a gift from Immutopics. Cells were washed with PBS several times and incubated in 4% BSA, 0.1% Tween 20 in PBS with secondary anti-goat IgG FITC and anti-mouse Texas Red antibodies (Jackson Immunoresearch Laboratories, West Grove, PA, USA). The coverslips were DAPI-stained and mounted on slides using ProLong Gold antifade reagent with DAPI (Molecular Probes, Eugene, OR, USA). The samples were imaged using Zeiss 510 Meta Laser-scanning confocal microscope (Carl Zeiss, Inc., Thornwood, NY, USA).

In vitro ppGalNAc-T3 assay using HIV gp120 peptide

To examine the effect of cAMP on ppGalNAc-T3 enzyme activity, ppGalNAc-T3 activity was measured in an established in vitro system. The ppGalNAc-T3-specific synthetic peptide HIV gp120 (RGPGRAFVTIGKIGNMR)22 and the control substrate EA2 (PTTDSTTPAPTTK, corresponding to the tandem repeat sequence of rat submandibular gland apomucin, detailed in Albone and colleagues23) were synthesized by the Facility for Biotechnology Resources, Center for Biologics Evaluation and Research (Bethesda, MD, USA). Reactions were carried out at 37°C in 25-µl mixture containing 500 µM gp 120 or EA2 peptides and 51 mM UDP-[14C]GalNAc (7.8 mCi/mmol, PerkinElmer Life Sciences, Waltham, MA, USA) in 10 mM MnCl2, 40 mM sodium cacodylate (pH 6.6), 40 mM β-mercaptoethanol, and 0.1% Triton X-100 as described.24 The reactions were initiated by adding 12.5 µg total cell extract protein for each sample. Reactions with cell extracts were carried out in the presence of 20 mM CDP-choline to inhibit endogenous non-ppGalNAcT-mediated hydrolysis activity. Control reactions were also performed using purified enzymes, human ppGalNAcTs from P. pastoris (0.1 pmol; 6.3 ng of ppGalNAcT1, 5.7 ng of ppGalNAcT2, and 6.9 ng of ppGalNAcT3). The mixtures were incubated for either 30 or 90 minutes at 37°C such that no more than 10% of the limiting substrate was converted to product. The reactions were stopped by addition of 0.1% trifluoroacetic acid in water (80 µL), and glycopeptides were separated using Sep-Pak Vac 1 cc (50 mg) C18 cartridges (Waters Corp., Milford, MA, USA) as described.25 Briefly, the Sep-Pak cartridges were equilibrated with 0.1% trifluoroacetic acid in water, and glycopeptides were bound in 0.1% trifluoroacetic acid, washed with 3 mL of 5% methanol in 0.1% trifluoroacetic acid in water. The bound glycopeptides were then eluted with 0.5 mL of 70% methanol in 0.1% trifluoroacetic acid in water. To this elutant, 4 mL of scintillation fluid was added, mixed, and counted by scintillation counter to determine the incorporation of GalNAc into the peptides. Purified human ppGalNAcT1 (T1), ppGalNAcT2 (T2), and ppGalNAcT3 (T3) were used as controls for each selected peptide substrate. All assays were performed in duplicate.

Furin activity

To examine the effect of various cAMP levels on furin activity, furin proteolytic activity was measured under various conditions by a previously described method.26 Briefly, live cells were incubated with furin-specific fluorescent substrate, and changes in fluorescence were measured. Cells were seeded at a density of 2 × 104 cells/well on 96-well plates precoated with 100 µL of 100 µg/mL poly L-lysine (Sigma) and incubated overnight at 37°C. The following day, cells were rinsed with serum-free media and treated as indicated in fresh media containing 2% serum for 24 hours. The reaction was performed at 37°C in 0.25 mL phenol red-free DMEM containing 26 nM (2 mg/mL) of furin-specific anthrax protective antigen (PrAg) variant, PrAg-33 (164RKKR167 of PrAg changed to RAAR), and 90 nM (5 mg/mL) anthrax lethal factor LF/b-Lac fusion protein. The wild-type PrAg as well as cleavage-resistant PrAg-U7 were used as negative and positive control, respectively. PrAg and LF/b-Lac-treated cells were incubated for 90 minutes at 37°C in a tissue culture incubator. Cells were washed twice at room temperature in phenol red-free DMEM, and the plate was transferred to room temperature. Twenty milliliters of 6 X loading solution or substrate loading solution was added to each chamber well containing 100 mL of room-temperature phenol red-free DMEM. Samples were incubated in the dark for 4 hours at room temperature to allow the loading of cells with CCF2/AM. After incubation, cells were examined using an inverted fluorescence microscope (Zeiss Axioplan, Carl Zeiss, Jena, Germany). For acquisition of blue fluorescence with this microscope, the excitation filter HQ405/20 nm bandpass, dichroic 425DCXR, and emitter filter HQ460/40 nm bandpass were used. For acquisition of green fluorescence with this microscope, the HQ405/20 nm bandpass, dichroic 425DCXR, and emitter filter HQ530/30 nm bandpass were used (Chroma Technology, Rockingham, VT, USA).

FGF23-producing cell line (HEKF)

To study FGF23 processing in a FGF23-secreting cell line, the FGF23-secreting HEKF cell line was created. HEK 293 cells were grown in α-MEM containing 10% fetal bovine serum (Hyclone-Thermo, Waltham, MA, USA), antibiotic mixture, L-glutamine, HEPES buffer, nonessential amino acid mixture, and sodium pyruvate (GIBCO) in 5% CO2 at 37°C. Cells were transfected with either an hFGF23 construct, an expression plasmid DNA construct that carries the full-length human FGF23 cDNA subcloned in pCMV4A, or the control empty vector by using Lipofectamine 2000 and Plus reagents (Invitrogen). Cells overexpressing FGF23 were enriched by selecting for Geneticin resistance by using 1 mg/mL of Geneticin G418 (GIBCO) for 3 to 4 weeks. Condition media from isolated colonies were tested for FGF23 levels by ELISA, and several of the highest-expressing cells were selected for further experiments.

To assess for the effect of increased cAMP on FGF23 processing, HEKF cells were transiently transfected with the constructs containing WT and the disease-causing R201C and R201H mutations of Gsα, which have been previously described.27 Briefly, after 24 hours of transfection, cells were rinsed with a media containing 2% FBS and were incubated in the same media for another 24 hours. Conditioned media were collected from each sample and were concentrated (Amicon Ultra, Millipore, Billerica, MA, USA) for Western blot purposes.

Statistical analyses

GraphPad Prism (La Jolla, CA, USA) software was used for statistical analyses. Differences between the means of unmatched groups were assessed by one-way ANOVA using Dunn's multiple comparison test. Differences between paired samples were assessed by Student's t test.

Results

iFGF23 and total FGF23 levels (iFGF23 + cFGF23) in FD patients and controls

The total FGF23 (i + c) and iFGF23 levels in plasma from a group of subjects with FD (n = 14), TIO (n = 12), XLH (n = 8), renal failure (n = 11), and normal controls (n = 10) were measured by two separate ELISA assays that detect total FGF23 or iFGF23 only. Consistent with greater processing of iFGF23 to cFGF23 in FD, the ratio of total FGF23 to iFGF23 was higher in subjects with FD (1.7 ± 0.46, mean ± 1 SD) compared with various disease states and normal controls (TIO 0.97 ± 0.24, XLH 1.0 ± 0.25, and renal failure 0.79 ± 1.2, and normal controls 0.99 ± 0.26, p < 0.05 for FD versus all other groups but XLH), whereas there was no difference in the ratio of total to iFGF23 between any of the control groups (Fig. 1).

Figure 1.

The ratio of total FGF23 to intact FGF23 in human plasma. (A) The ratios of total FGF23 (intact + C-terminal FGF23) to intact FGF23 (iFGF23) are shown in subjects with fibrous dysplasia (FD), other disease states, and controls. Other disease states include tumor-induced osteomalacia (TIO), X-linked hypophosphatemia (XLH), and renal failure (RF), as well as normal control subjects (Normal). Significant (p < 0.05) and nonsignificant (p = NS) are indicated. Of note, the difference between the FD group and the XLH group was not significant. (B) The total (intact + C-terminal) FGF23 values, the relationship between the two as defined by the regression line, and the slope and p value of the line are shown for the subjects with fibrous dysplasia. (C) The total and intact FGF23 values, the relationship between the two as defined by the regression line, and the slope and p value of the line are shown for all other subjects with TIO, XLH, RF, and Normal (NV). FGF23 was measured by two different ELISA assays; one that detects total FGF23 (C-terminal + intact FGF23) and one that detects only iFGF23. Differences between groups were assessed by one-way ANOVA using Dunn's multiple comparison test, and linear regression was assessed by best fit.

BMSC cAMP levels

To confirm that the activating mutations in the Gsα resulted in increased levels of cellular cAMP, levels of cAMP from conditioned media and whole cell lysates were measured in the presence of the phosphodiesterase inhibitor IBMX in both WT and FD BMSCs, which were derived from patients with FD by subcloning colonies of cells with WT or mutant Gsα sequence. Results indicated that FD cell lysates contained higher cAMP compared with WT control (Fig. 2A).

Figure 2.

cAMP levels, FGF23, O-linked glycosyl transferase N-acetylgalactosaminyltransferase 3 (ppGalNAcT3), and furin expression in human bone marrow stromal cells (BMSCs). (A) Media and cell lysate cAMP levels were measured by ELISA in BMSCs that had been derived from a subject with fibrous dysplasia (FD). Because patients with FD are mosaics for mutations in Gsα, primary cultures of cells from a bone specimen were subcloned, and pure colonies bearing WT Gsα sequence R201R or FD-causing Gsα sequence R201H (FD) were studied. cAMP levels were higher in lysates from mutant cells (error bar = 1 SD, p < 0.05). (B) Products of RT-PCR amplification reactions performed with total RNA isolated from cultured BMSCs from the WT and FD cells are shown and demonstrate the presence of FGF23, ppGaNAcT3, and furin transcripts in BMSCs.

FGF23, ppGalNAcT3, and furin gene expression

To demonstrate that WT and FD BMSCs express the enzymatic machinery for FGF23 processing, total RNA was isolated from BMSCs and gene expression of FGF23, ppGalNAcT3, and furin was assessed by RT-PCR and qPCR. The results showed that all genes crucial for FGF23 processing were expressed in both FD-patient WT and FD BMSCs and that no major differences in gene expression levels were observed by RT-PCR (Fig. 2B) or qPCR (data not shown). Although primary cultures of human BMSCs express FGF23, GalNAcT3, and furin RNA, neither i + cFGF23 nor iFGF23 could be detected in the media by ELISA. This was in spite of efforts that included manipulating culture conditions by combinatorial addition and concentration manipulations of multiple reagents, including the differentiating agent β-glycerol phosphate, 1,25 (OH)2 vitamin D, calcium, and phosphate salts (data not shown). This was also in spite of the fact FGF23 protein could be detected in cells by confocal microscopy (Fig. 3) and by Western blotting of cell lysates (data not shown). However, this is not the result of technical problems because we were able to detect robust FGF23 production in vitro by a stably transfected HEK cell line that expresses FGF23, confirming that the inability to detect FGF23 production in the media of BMSCs is intrinsic to the BMSCs. This likely reflects the fact that in vitro BMSCs are perhaps a less differentiated cell that does not secretes FGF23.

Figure 3.

Colocalization of FGF23 and O-linked glycosyl transferase N-acetylgalactosaminyltransferase 3 (ppGalNAcT3) or furin in normal (WT) and fibrous dysplasia bone marrow stromal cells. Confocal microscopy was performed for FGF23, ppGalNAcT3 (A,B), and furin (C,D) in wild-type (WT; A,C) and fibrous dysplasia (FD; B,D) bone marrow stromal cells. Because patients with FD are mosaics for mutations in Gsα, primary cultures of cells from a bone specimen were subcloned, and pure colonies bearing WT Gsα sequence R201R or FD-causing Gsα sequence R201H (FD) were studied. Cells were grown on glass coverslips, fixed, and immunostained. Throughout, FGF23 is stained in green, ppGalNAcT3 or furin in red, and nuclei in blue (DAPI). Merged images are indicated. FGF23 is colocalized with ppGalNAcT3/furin in a perinuclear, Golgi-associated pattern. Representative data from two separate experiments are shown.

Because osteocytes are believed to be the primary source of FGF23 production in vivo, we attempted to investigate the FGF23 processing in the mouse osteocyte-like cell line MLO-Y4.28 Although Western blotting demonstrated that MLO-Y4 cells produce FGF23, mouse FGF23 migrated faster than human FGF23 and failed to show any processing to cFGF23, suggesting that in mice FGF23 may be differently processed. In addition, in spite of the fact that FGF23 could be detected by Western blot, it could not be detected by the intact FGF23 assay, which is known to detect mouse FGF23 (data not shown). In addition, when MLO-Y4 cells were treated with dibutyryl cAMP (dbcAMP) or forskolin, no detectable cFGF23 was observed by Western blot (data not shown), indicating that cAMP-mediated processing of FGF23 could be different in this mouse cell line. Together these data suggest that this cell line is not an adequate model for the study of FGF23 processing.

Colocalization of FGF23, ppGalNAcT3, and furin in BMSCs

FGF23, GalNAcT3, and furin protein expression and appropriate subcellular localization were confirmed by confocal microscopy (Fig. 3). Results show that FGF23 and ppGalNAcT3 (Fig. 3A for WT and Fig. 3B for FD BMSCs) were specifically colocalized in the perinuclear regions. In addition, FGF23 and furin co-occupy the same subcellular region in these two cell types (WT BMSC is shown in Fig. 3C and FD BMSC in Fig. 3D). The results also indicate that, compared with ppGalNAcT3, furin localization was observed uniformly in the cytoplasm; however, no major differences in the fluorescence level were observed in these two cell types.

ppGalNAcT3 and furin assays

To examine the mechanism of increased FGF23 processing observed in FD cells, we measured the enzymatic activities of ppGalNAcT3 and furin in the WT (colonies of R201R-bearing cells) and FD (colonies of R201H-bearing cells) BMSCs. Equal amounts of whole cell lysates from WT and FD BMSCs were used in an in vitro ppGalNAcT3 enzymatic assay utilizing enzyme-specific peptide substrates. Decreased ppGalNAcT3 enzymatic activity was observed in mutant whole cell lysates relative to WT BMSCs (Fig. 4A). To demonstrate that the decrease in ppGalNAc activity was specific for ppGalNAcT3, we also examined ppGalNAcT1 and ppGalNAcT2 enzyme activities using purified recombinant human ppGalNAcT1 and T2 with T1- and T2-specific substrates in FD cells (Supplemental Fig. S2). The results clearly demonstrate that only purified ppGalNAcT3 is able to transfer radioactivity on to gp120, a ppGalNAcT3-specific substrate (Fig. 4A, right panel).

Figure 4.

O-linked glycosyl transferase N-acetylgalactosaminyltransferase 3 (ppGalNAcT3) and furin enzymatic activities in wild-type and fibrous dysplasia bone marrow stromal cells. (A) ppGalNAcT3 enzymatic activity was measured in vitro from equal amounts of wild-type (WT) and fibrous dysplasia (FD) cell lysates using a ppGalNAcT3-specific HIV gp 120-specific peptide as a substrate. Because patients with FD are mosaics for mutations in Gsα, primary cultures of cells from a bone specimen were subcloned, and pure colonies bearing WT Gsα sequence R201R or FD-causing Gsα sequence R201H (FD) were studied. FD BMSCs demonstrated significantly less ppGalNAcT3 activity than WT cells (p < 0.05), consistent with underglycosylation leading to increased furin susceptibility and FGF23 processing. Substrate specificity for HIV gp120 peptide was tested by using purified ppGalNacT1, T2, and T3 (right panel) (see Supplemental Fig. S2 for additional controls). (B) Furin enzymatic activity was measured in equal numbers of WT and FD BMSCs using a fluorogenic substrate. FD cells had significantly greater furin activity than WT cells (p < 0.05). Furin activity in WT cells was increased to that of FD cells by treatment with the nonhydrolyzable analogue of cAMP, dibutylryl cAMP (dbcAMP), or adenylyl cyclase activator forskolin.

Furin enzyme activity was measured in an in vitro furin assay using a furin-specific fluorogenic substrate. Results indicated that there was a marked increase in furin activity in FD versus WT BMSCs (Fig. 4B). cAMP dependence of the increase in furin activity was confirmed by treatment with dbcAMP and forskolin (Fig. 4B).

cAMP, FGF23, ppGalNAcT3, and furin in HEKF cells

To further assess cAMP-mediated processing of FGF23 in cells that are able to secrete FGF23, we investigated cAMP, iFGF23, and cFGF23 levels, as well as ppGalNAcT3 and furin activity in HEKF cells that had been transiently transfected with WT or the FD-causing mutations of Gsα (R201C or R201H). An equivalent expression of c-Myc Gsα protein level was observed in each cell extract (data not shown). First, an increase in cAMP levels was confirmed in HEKF cells expressing additional WT or mutant Gsα (Fig. 5A). Next, iFGF23 and i + cFGF23 was measured in cell culture supernatants (Fig. 5B). Although iFGF23 and i + cFGF23 levels were increased in cells transfected with WT, R201C, and R201H Gsα-expressing cells, there was not a relative increase in i + cFGF23 levels compared with iFGF23 in R201C or R201H versus WT Gsα-overexpressing cells (Fig. 5C). However, there was a relatively higher total FGF23 (i + cFGF23) in HEKF cells overexpressing Gsα in general, as reflected by the fact that the slope of the regression line comparing total FGF23 (i + cFGF23) versus iFGF23 was significantly greater than 1 (2.6) and highly significant (r2 = 0.99, p < 0.001, Fig. 5D), which was maintained with increasing cAMP. In addition, both ppGalNAcT3 and furin activity were relatively high in untransfected HEKF cells (Fig. 6A,B) compared, for example, with hBMSCs (Fig. 4A,B), which may account for the elevated total FGF23 to iFGF23 ratio at baseline. With additional Gsα (WT transfectants) and especially the addition of mutant Gsα (R201C or R201H), there was a graded decrease in T3 activity (Fig. 6A). In terms of furin activity, basal furin activity in HEKF cells was high, relative to BMSCs (Fig. 4B), but furin activity increased with the addition of mutant Gsα constructs relative to WT Gsα (Fig. 6B).

Figure 5.

FGF23 production by HEKF cells transfected with wild-type or mutant Gsα. FGF23-producing HEKF cells were transiently transfected with either wild-type (WT) or fibrous dysplasia-causing Gsα mutants (R201C or R201H) to assess for the effect of increased cAMP caused by these mutations on FGF23 processing. (A) Cells transfected with mutated Gsα constructs had significantly higher levels of cAMP. (B) Transfected cells produced very high levels of FGF23 with relatively higher levels of total (intact + C-terminal FGF23) in all transfected cells. (C) However, the relative levels of total FGF23 to intact (iFGF23) were not higher in the cells bearing the mutated Gsα constructs, nor were the levels of total FGF23 different between cells transfected with WT versus mutant Gsα constructs. (D) Regression analysis of the relationship between total FGF23 versus intact FGF23 revealed there were significantly higher levels of total FGF23 than intact FGF23, consistent with greater processing of intact FGF23 in HEKF cells overexpressing Gsα and cAMP.

Figure 6.

O-linked glycosyl transferase N-acetylgalactosaminyltransferase 3 (ppGalNAcT3) and furin enzymatic activities in HEKF cells expressing activating Gsα mutants. (A) ppGalNAcT3 enzymatic activity was measured in vitro from equal amounts of cell lysates from parental HEKF cells and HEKF cells transfected with equal levels of wild-type (WT) or constructs bearing the fibrous dysplasia-causing activating mutations of Gsα (R201C and R201H). The ppGalNAcT3-specific HIV gp 120-specific peptide was used as a substrate. All cells transfected with additional Gsα demonstrated significantly less ppGalNAcT3 activity than parental HEK cells (p < 0.05), and cells with activating mutations demonstrated less activity than cells transfected with additional WT Gsα, consistent with underglycosylation leading to increased furin susceptibility and FGF23 processing. (B) Furin enzymatic activity was measured in equal numbers of HEKF and HEKF with Gs constructs using a fluorogenic substrate. Cells with either the R201C- or R201H-activating mutations of Gsα had significantly greater furin activity than cells with WT Gsα (p < 0.05).

Discussion

The physiologic importance of distinguishing iFGF23 from cFGF23 in total FGF23 assays is demonstrated in patients with FTC, who have very high levels of cFGF23, little or no iFGF23, and who display a metabolic phenotype consistent with loss of FGF23 action. As such, regulation of FGF23 processing, by which iFGF23 is cleaved and inactivated, represents a potential focus for control of overall FGF23 activity. Our previous observation that subjects with extensive MAS/FD have markedly elevated levels of total FGF23 (cFGF23 + iFGF23)11 but rarely have frank rickets19 led us to investigate the possibility that there may be relatively greater cFGF23 compared with iFGF23 in subjects with FD. Here, we confirmed this hypothesis and investigated the mechanism underlying an altered FGF23 processing. Because FD is caused by activating mutations in the cAMP-regulating protein, Gsα, which leads to elevated levels of cAMP, we postulated that alterations in FGF23 processing may be cAMP-mediated. This mechanism would also explain the findings in a recently reported case of Jansen's metaphyseal chondrodysplasia, which is caused by activating mutations in the PTH/PTHrP receptor also leading to elevations in cAMP, and in which the patient demonstrated a similar pattern of relatively higher i + cFGF23 versus iFGF23 (i + cFGF23 852.7 RU/mL [normal < 150] and iFGF23 100 pg/mL [normal < 55]).20 Normally, the ratio of i + cFGF23 (in RU/mL) to iFGF23 (in pg/mL) is approximately 1:1.

Because it has been shown that posttranslational processing of FGF23 involves O-glycosylation by ppGalNAcT3 and cleavage by furin1 and that glycosylation protects against furin-mediated degradation from active iFGF23 to inactive cFGF23, we hypothesized that there is decreased ppGalNAcT3 activity and increased furin activity in FD. There is precedent for similar coordinated O-glycosylation/furin protein processing in the work of Smenov and colleagues, in which they demonstrated that furin degradation of pro brain naturetic peptide is inhibited by O-glycosylation in HEK293 cells.29

We tested this hypothesis in primary cultures of WT and mutant human BMSCs and found that there was increased cAMP in mutant BMSCs (Fig. 2A) and a coordinated decrease in ppGalNAcT3 activity and increased furin activity (Fig. 4A,B). The interaction between FGF23, ppGAlNAcT3, and furin is further supported by their subcellular colocalization in a perinuclear, Golgi-associated pattern (Fig. 3).

However, the use of hBMSCs to investigate this process was to some extent limited in that hBMSCs were not able to secrete FGF23 into the media, in spite of the fact FGF23, ppGAlNAcT3, and furin expression were observed. Therefore, we studied this in an FGF23-overproducing cell line we created, HEKF cells. These cells express FGF23, ppGAlNAcT3, and furin and secrete very high levels of FGF23 into the media. In HEKF cells that were transiently transfected with WT and FD-causing Gsα mutants, which increased cAMP levels, there was an increase in the ratio of i + cFGF23 relative to iFGF23 secreted into the media. This was accompanied by a decrease in ppGAlNAcT3 enzymatic activity and an increase in furin enzymatic activity, confirming the role of coordinated regulation of ppGAlNAcT3 and furin enzyme activity in cAMP-mediated FGF23 processing.

In addition to what we have demonstrated here in patients with FD, other lines of evidence suggest that FGF23 processing is involved in the overall physiologic regulation of FGF23 activity in phosphate and vitamin D homeostasis. For example, in TIO only cFGF23 is found in the circulation immediately after iFGF23-secreting tumors are removed.30 This presumed cFGF23 “overproduction” may represent a physiologic process by which the body attempts to dampen the effect of high levels of iFGF23 secreted by TIO tumors by secreting only cFGF23, which can potentially block the action of iFGF23. This concept that cFGF23 fragments may block the action of iFGF23 is supported by the work of Goetz and colleagues. They showed that the isolated C-terminal tail of FGF23 inhibited the action of intact FGF23 in an in vivo rodent system.31 In fact, we have observed this phenomenon in patients with FTC, wherein the iFGF23 level can be well in the normal range (50 pg/mL), but the i + cFGF23 level is >20 times higher (1800 RU/mL) and the patient had little or no evidence of FGF23 action (unpublished data), in spite of normal iFGF23 levels. Further support of the concept of regulation of FGF23 processing as a mechanism for altering the i + cFGF23/iFGF23 ratio and thus the overall action of FGF23 may be observed in patients with renal failure, wherein only iFGF23 and no cFGF23 is produced.32 This may represent a physiologic response to the high blood phosphorus levels present in renal failure and a futile effort by the body to clear this via a failing kidney. Additional support for altered FGF23 processing as a means of regulating overall FGF23 activity is provided by the fact that in cord blood FGF23 exists almost exclusively as cFGF23.33 This lack of intact, functional FGF23 may reflect an effort by the fetal/maternal unit to maintain the high levels of phosphorus needed in the developing fetus. Recent work in ADHR also points to a role for the regulation of FGF23 processing as an important physiologic process. In an iron-deficient state, humans with ADHR and an ADHR model mouse respond by increasing FGF23 blood levels.34 Interestingly, in response to iron deficiency, the WT control of the ADHR mice showed a dramatic increase in cFGF23 in the blood but maintained normal blood iFGF23 and phosphorus levels. This finding is paralleled in normal humans, wherein there is a significant inverse correlation between cFGF23 levels and serum iron levels but no such relationship with iFGF23 levels.17 All of these observations point to the regulation of FGF23 processing as an important process in the maintenance of mineral homeostasis.

In conclusion, these data demonstrate the molecular mechanisms that underlie the observations of differential processing of FGF23 that takes place in FD, and identify FGF23 processing as a focus for the physiologic regulation of FGF23 activity.

Disclosures

All the authors state that they have no conflicts of interest.

Acknowledgements

We thank Dr Larry Fisher and Dr Pamela Gehron Robey (NIDCR, NIH) for their critical reading of the manuscript, and Dr Tatiana Karpova and Dr Jim McNally (NCI, NIH) for their assistance with the confocal microscopy.

Authors' roles: Study design: NB and MTC. Clinical care: CD, RIG, MC, and MTC. Conduct of experimental studies: NB, MW, CD, BMC, THB, HVP, RIG, NC, and GLH. Data analysis: NB, MW, CD, BMC, THB, HVP, RIG, NC, and MTC. Data interpretation: NB, MW, CD, BMC, THB, HVP, RIG, NC, and MTC. Drafting manuscript: NB and MTC. Revising manuscript content: NB, MW, CD, BMC, THB, HVP, RIG, NC, and MTC. Approving final version of manuscript: NB, MW, CD, BMC, THB, HVP, RIG, NC, GLH, and MTC. NB and MTC take responsibility for the integrity of the data analysis.

Ancillary