The authors state that they have no conflicts of interest.
Two hyperphosphatemic patients with mutations in GALNT3 showed low intact FGF23 levels with marked increase of processed C-terminal fragments. FGF23 protein has three O-linked glycans and FGF23 with incomplete glycosylation is susceptible to processing. Silencing GALNT3 resulted in enhanced processing of FGF23. Decreased function of FGF23 by enhanced processing is the cause of hyperphosphatemia in patients with GALNT3 mutation.
Introduction: Hyperostosis–hyperphosphatemia syndrome (HHS) is an autosomal recessive entity manifesting as severe hyperphosphatemia associated with episodic bone pain and radiological findings of cortical hyperostosis and periosteal reaction. Persistent hyperphosphatemia is not counterbalanced by PTH or 1,25-dihydroxyvitamin D, posing a mirror image of hypophosphatemic states attributed to increased fibroblast growth factor (FGF)23 activity.
Materials and Methods: We describe two children with HHS who were found to be homozygous for a mutation in GALNT3 encoding a peptide involved in mucin-type O-glycosylation (ppGaNTase-T3). FGF23 levels were evaluated by two ELISAs and Western blotting. FGF23 protein was analyzed by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry. Effect of silencing GALNT3 was evaluated using siRNA in cells transfected with expression vector for FGF23.
Results: Both patients had low levels of the full-length FGF23 with markedly augmented amounts of the inactive fragments. Biologically active FGF23 has three O-linked glycans. FGF23 with only one or two O-linked glycans is processed into inactive fragments. Decreasing the expression of the GALNT3 gene by RNA interference resulted in enhanced processing of FGF23.
Conclusions: The primary defect in HHS is impairment of glycosylation of FGF23 resulting from mutations in GALNT3 and leading to augmented processing of FGF23. These changes in FGF23 abolish its phosphaturic effect and lead to severe persistent hyperphosphatemia. This study provides the pathogenetic mechanism of the first mucin-type O-glycosylation defect identified.
Hyperostosis–hyperphosphatemia syndrome (HHS) is a rare clinical entity characterized by markedly increased serum phosphate levels in children, associated with recurrent bone pain and radiological findings of cortical hyperostosis and periosteal reaction.(1,2) An additional occasional feature is soft tissue calcifications, a condition referred to as tumoral calcinosis.(3) We describe two children from unrelated consanguineous families of Arab descent who presented with HHS. Severe persistent hyperphosphatemia caused by maximal renal tubular reabsorption of phosphate preceded bone involvement, suggesting a possible pathogenetic role.
Two main pathways have been identified in the synthesis of glycoproteins: N-glycosylation and O-glycosylation. Of these, mucin-type O-glycans with N-acetylgalactosamine (GalNac) as the initiating sugar compose the most prevalent form. Although several inherited diseases with defects in the synthesis of N-linked glycosylation have been reported and known as congenital disorders of glycosylation (CDG),(4,5) the first inherited disorder in mucin-type O-linked glycosylation was only recently identified. Mutations in GALNT3 encoding UDP-N-acetyl-α-d-galactosamine:polypeptide N-acetylgalactosaminyl-transferase 3 (ppGaNTase-T3), a protein involved in mucin-type O-linked glycosylation, were found to be responsible for hyperphosphatemic familial tumoral calcinosis (HFTC).(6) This recessive disease manifests by recurrent painful subcutaneous masses associated with hyperphosphatemia but without bone involvement. However, both our patients with HHS were found to be homozygous for a mutation in the GALNT3 gene, showing that HHS and HFTC are allelic disorders.(7) The common feature of HHS and HFTC is persistent hyperphosphatemia, which is not counterbalanced by either increased PTH or suppressed 1,25-dihydroxyvitamin D [1,25(OH)2D] levels.
Recent study has enhanced our understanding of the pathogenesis of other disorders of phosphate regulation. The dominantly inherited hypophosphatemic rickets (ADHR; OMIM no. 193100) and tumor-induced osteomalacia (TIO) manifest as the combination of significantly increased urinary phosphate wasting leading to hypophosphatemia and severe bone disease characterized by impaired mineralization of bone matrix. Serum 1,25(OH)2D levels remain unexpectedly low or normal. Both conditions have been attributed to increased fibroblast growth factor 23 (FGF23) activity,(8,9) although the possible contribution of other tumor-derived peptides, such as matrix extracellular phosphoglycoprotein (MEPE) and secreted frizzled-related protein (sFRP-4), to the development of TIO have been suggested.(10,11) X-linked dominant hypophosphatemic rickets (XLH; OMIM 307800), which has a similar phenotype, is caused by inactivating mutations of the cell surface metalloprotease phosphate-regulating endopeptidase (phosphate-regulating gene with homologies to endopeptidases on the X chromosome [PHEX]).(12) Increased serum FGF23 levels have been shown in some but not all patients with XLH.(13,14) Although initial studies suggested that FGF23 might be a substrate for PHEX,(15) additional studies failed to confirm this observation.(16)
We have previously shown that the FGF23 protein has a proteolytic processing site, with a 176-RXXR-179 motif, between Arg179 and Ser180. Cleavage of the intact full-length FGF23 peptide abolishes its biological activity: purified full-length recombinant FGF23 but not the processed fragments induced increased urinary phosphate leak that resulted in hypophosphatemia.(9,17) The exact mechanism governing the processing of FGF23 has not been extensively studied. Because recombinant FGF23 was previously shown to have various O-linked, but not N-linked, sugar chains,(17) the pathogenetic role of FGF23 in HHS was evaluated.
MATERIALS AND METHODS
The following study was approved by the institutional review board of Shaare Zedek Medical Center, Jerusalem, and appropriate informed consent was obtained from participants.
Sandwich ELISA for the detection of human FGF23
Serum FGF23 levels were determined by two kinds of commercially available sandwich ELISA kits. The ELISA assay for the intact FGF23 was developed to detect only the uncleaved peptide, at the processing site, using the combination of two monoclonal antibodies (FN-1 and FC-1) that recognize the N-terminal and C-terminal portions, respectively, of the processing site of FGF23 (Kainos, Tokyo, Japan).(13) The second ELISA assay for the C-terminal portion of FGF23 (Immutopics, San Clemente, CA, USA) detects both the full-length and the processed C-terminal fragment of FGF23.(14)
FGF23 and its processed fragments in sera were recovered by immunoprecipitation and studied by Western blot analysis. Three milliliters of plasma derived from both patients (cases 1 and 2), their parents, and a healthy control was immunoprecipitated with either the FN-1 or FC-1 anti- body. The precipitated complexes were subjected to SDS-PAGE followed by Western blot analysis. FC-1 antibody was used to detect the processed C-terminal fragment and the uncleaved intact FGF23. Because FN-1 was unsuitable for Western blot analysis, another monoclonal antibody (FN-2) that recognizes the N-terminal portion of the cleaved FGF23 was used to detect both the N-terminal fragments and the intact uncleaved FGF23. The expression of ppGaNTase-T3 was assessed in cell lysates using anti-ppGaNTase-T3 antibody (a generous gift from Dr Ulla Mandel, Department of Oral Diagnostics, School of Dentistry, Faculty of Health Sciences, University of Copenhagen, Copenhagen, Denmark). The results were visualized by an enhanced chemiluminescence (ECL) detection system (Amersham Pharmacia Biotech, Little Chalfont, UK).
Mutation screening of FGF23 gene
Genomic DNA was extracted from peripheral blood lymphocytes using QIAamp Blood Kit (Qiagen, Hilden, Germany). DNA sequencing was performed by PCR amplification of all coding exons of the FGF23 encoding gene and direct sequencing of the PCR products.
ppGaNTase-T3 protein expression in skin biopsies
Skin biopsies were snap-frozen in liquid nitrogen, and 6-μm cryostat sections were incubated with primary mouse anti-ppGalNAc-T3 monoclonal antibody for 24 h at room temperature and with a secondary FITC-conjugated goat anti-mouse IgG antibody (Zymed) for 45 minutes. The sections were examined under an Axioscop2 upright microscope (Zeiss, Thornwood, NY, USA), and images were processed using Image Pro+ (Media Cybernetics, Silver Spring, MD, USA).
Determination of mucin-type O-linked glycosylation
Recombinant mutant FGF23 (R176Q, R179Q) proteins, which were resistant to the cleavage between Arg179 and Ser180, were purified from conditioned medium from cultures of Chinese hamster ovary (CHO) cells stably expressing these proteins as previously described.(17) The three major polypeptides were separated by 10–20% gradient SDS-PAGE under reducing condition and stained with Coomassie Brilliant blue. Each band was excised from the gel and incubated with trypsin. The tryptic fragments were subjected to molecular mass analysis by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF/MS; Voyager-DE/STR; Applied Biosystems). Unmodified tryptic fragments were identified by their observed molecular mass unit corresponding to a theoretical mass unit of tryptic peptides from FGF23. The tryptic fragments modified by mucin-type O-linked glycosylation were determined by the signals with a 365- or 656-mass unit increase per site, which corresponds to a Gal-GalNAc or sialylated Gal-GalNAc residue, respectively.
siRNA for GALNT3 and FGF23 expression
Silencer predesigned siRNA (Ambion, Austin, TX, USA) was introduced into HOS-TE85 cells using siPORT NeoFX (Ambion). Thirty picomoles of three kinds of siRNAs for GALNT3 (ID 111460, 111461, and 14834) were used for 8 × 104 cells in a 24-well plate. The same amount of three negative control siRNAs (4611, 4713, and 4615) was used for control. Twenty-four hours later, expression vector for FGF23 was introduced using lipofectamine reagent (Invitrogen, Tokyo, Japan), and conditioned media was harvested 48 h later. ELISA assays and Western blotting were performed as described above.
Statistical significance was evaluated by Student's t-test.
The clinical characteristics of a 14-year-old girl (case 1) and a 12-year-old boy (case 2), from Arab descent, diagnosed with HHS, were recently described.(7) In brief, they were noted to have severe hyperphosphatemia (7.4– 10.0 mg/dl) at the ages of 7 and 9 years, respectively, with normal serum calcium, urea, creatinine, alkaline phosphatase, and 1,25(OH)2D levels (Table 1). The tubular reabsorption rate of phosphate (TRP) and the maximum rate of tubular reabsorption corrected for glomerular filtration rate (TmPO4/GFR) were markedly elevated. Serum PTH concentrations (N-tact PTH) were within reference range and not elevated. They had repeated self-remitting episodes of localized bone pain at different locations with radiological findings of cortical hyperostosis and severe periosteal reaction. The girl had a single episode of subcutaneous calcification (tumoral calcinosis) in her calf muscles at the age of 13 years. Their growth and development were normal.
Table Table 1.. Laboratory Data of Patients With HHS
Analysis of FGF23 protein in patients
Serum FGF23 levels were evaluated in both patients and their respective parents using two ELISA assays: one detects only uncleaved biologically active FGF23 and the other recognizes its C terminus. The latter detects both the intact and the processed C-terminal portion. Evaluating two serum samples from each patient by both methods repeatedly yielded discrepant results: whereas the intact FGF23 levels were low-normal (12.1, 17.9 and 7.0, 10.1 pg/ml, respectively, with a reference of 10–50 pg/ml), the values by the assay that recognizes both the intact and the C-terminal fragments were extremely high (4700, 11,760 and 1478, 1600 RU/ml, respectively, with a reference <150 RU/ml; Table 2). The respective parents showed serum levels that were comparable with the control samples by both assays. These results indicate that patients' sera samples contain large amounts of processed FGF23 protein in the presence of rather low levels of full-length FGF23, suggesting that there is increased production and processing of the intact biologically active FGF23 in patients with HHS.
Table Table 2.. Serum FGF23 Levels: Full-Length and Inactive Fragments
To further study this assumption, Western blotting using antibodies that recognize the C-terminal and N-terminal fragments of FGF23 (as well as the intact peptide) was performed. Both fragments lack significant biological activity determined by their inability to induce urinary phosphate leak.(17) Our study showed that patient plasma contained only barely detectable levels of the full-length FGF23 by Western blotting, whereas the amounts of the processed C-terminal and N-terminal portions were elevated (Fig. 1). The molecular size of the fragments (16.5 kDa for the N-terminal portion and 10 kDa for the C-terminal portion) was the same as that of FGF23 fragments produced by CHO cells in vitro(17) and suggests that processing between Arg179 and Ser180 was markedly augmented in these samples. Plasma samples derived from parents or control individuals, assayed by this method, contained almost undetectable amount of FGF23 fragments. Sequencing analysis of the FGF23 gene showed that there was no mutation in these patients. Therefore, the primary defect of these patients seemed to be enhanced intracellular processing and increased production of FGF23.
Analysis of GALNT3 gene product
Both patients were found to be homozygous for the G→A transition at cDNA position 1524 + 1 (starting from the ATG) in the GALNT3 gene, encoding ppGaNTase-T3, leading to disruption of the intron 7 donor splice site consensus sequence.(7) Because ppGaNTase-T3 is strongly expressed in stratified epithelia, we used an antibody directed against ppGaNTase-T3 and stained skin biopsies obtained from a patient and a control individual. This experiment showed that expression of ppGaNTase-T3 was completely absent in the epidermis of our patient (case 1) but was observed in the control (Fig. 2).
Analysis of mutant FGF23 protein
Our assumption was that reduced ppGaNTase-T3 expression would result in defects in glycosylation and subsequently in augmented processing of the FGF23 molecule. A number of experiments were carried out to address this hypothesis. When human wildtype FGF23 was expressed in CHO cells, and conditioned medium was analyzed by Western blotting using FN-2 antibody, the full-length mature FGF23 protein of 32.5 kDa and smaller products around 16.5 kDa were observed (Fig. 3). On the other hand, when the mutant cleavage-resistant FGF23 protein with R176Q or R179Q substitutions was expressed under the same conditions, processed fragments were barely detectable. Furthermore, at least three heterogeneous bands, at approximately the same size of the full-length mature FGF23 protein, appeared instead. The bands of −30–32.5 kDa observed in the mutant FGF23 protein were purified and digested with trypsin and subjected to molecular mass analysis by MALDI-TOF/MS. All three bands contained tryptic fragments derived from both the N-terminal and C-terminal portions of the FGF23 protein (amino acids from 25 to 48 and 246 to 251; Fig. 3), indicating that they all had a full-length mature polypeptide chain. The heterogeneity of these bands was attributed to the differences in the number of O-linked sugar chains. The protein band with the smallest molecular weight had only one tryptic fragment with one O-linked glycosylation (199–228). This O-linked glycosylated fragment was observed in all three bands. The second and the largest bands had an additional O-linked sugar chain in the fragment (162–175) at Thr175, whereas only the largest band had a third O-linked sugar chain in the tryptic fragment (176–187). These results suggest that FGF23 is prone to serially undergo O-linked glycosylation: first at amino acids between 199 and 228, then between 162 and 175, and finally between 176 and 187. Given the fact that protein bands with slightly lower molecular weight than 32.5 kDa with one or two O-linked sugar chains could not be observed when wildtype FGF23 was expressed and that these bands were only detectable when cleavage resistant mutant FGF23 was introduced, it was hypothesized that full-length FGF23 with only one or two sugar chains is susceptible to degradation between Arg179 and Ser180. In other words, it is plausible that the ultimate O-glycosylation around the cleavage site (between Arg179 and Ser180) prevents the processing of FGF23 protein, because the O-glycosylation in the 176–187 region was detected only in the largest band with a molecular weight of 32.5 kDa. To further address this issue, we generated mutant FGF23 proteins with single amino acid substitutions at various positions including the potential O-glycosylation sites, Thr178 and Ser180. Alanine substitutions with Arg176 or Arg179 led to cleavage resistance as observed in the mutant proteins with R176Q and R179Q (Fig. 4). Interestingly, the substitution of Thr178 with alanine resulted in the absence of the full-length mature form of FGF23, whereas the mutant protein with S180A retained the full-length mature form with 32.5 kDa. These findings indicate that O-glycosylation at Thr178 of FGF23 is critical to maintain the native full-length molecule in a completely glycosylated form without undergoing cleavage between Arg179 and Ser180. In turn, deficiency in O-glycosylation of this specific residue likely leads to degradation and consequently to absence of the full-length FGF23 of −32.5 kDa.
Effect of ppGaNTase-T3 on FGF23 protein
Patients with HHS have only minute amounts of the intact full-length FGF23 peptide, which renders performing direct assessment of its glycosylation status quite impossible. We therefore took a different approach implementing the RNA interference method. Screening the expression of the GALNT3 gene by RT-PCR indicated that the human osteoblastic cell line HOS-TE85 expressed this gene, which enabled us to use it for the following experiments. Because there was no human cell line available that produces sufficient amount of FGF23 protein, we introduced expression vector for FGF23 into HOS-TE85 cells and analyzed the involvement of ppGaNTase-T3 in the processing of FGF23 protein. Western blotting using anti-ppGaNTase-T3 antibody confirmed that siRNA for GALNT3 decreased, although not completely, the expression of ppGaNTase-T3 in these cells. Western blotting of conditioned media showed that the amounts of full-length FGF23 decreased, whereas the processed C-terminal fragments increased compared with cells transfected with control siRNA (Fig. 5). Densitometric analysis of these three repeated experiments indicated that C-terminal fragment of FGF23 from cells with siRNA for GALNT3 significantly increased to 124.3 ± 14.3% (SD) of cells with control siRNA. Consistent with these results by Western blotting, measurement of FGF23 in the conditioned media of cells transfected with siRNAs for GALNT3 showed that full-length FGF23 significantly decreased to 60.9 ± 6.1% (average of three experiments; p < 0.01) of cells with control siRNA. In contrast, the C-terminal assay, which measures both the full-length and the C-terminal fragments, showed that the amount of FGF23 in the conditioned media of cells with the siRNAs for GALNT3 was 96.2 ± 17.5% compared with control cells. These results indicate that ppGaNTase-T3 is necessary for maintaining full-length FGF23 and suppression of ppGaNTase-T3 activity is associated with not only decreased levels of full-length but also augmented processing resulting in increased amount of the C-terminal fragments of FGF23.
The syndrome of hyperostosis and hyperphosphatemia is characterized by persistent hyperphosphatemia associated with recurrent bone pain. Hyperostosis and periosteal reaction subside spontaneously over weeks to months only to recur in the same or different location. Tumoral calcinosis, documented in one of our patients, probably represents a continuum of the same syndrome in individuals with markedly elevated calcium-phosphate product. In fact, mutations in the GALNT3 gene were reported in patients with both HHS and familial tumoral calcinosis.(7)
Our study showed a discrepancy between the amount of full-length FGF23 and its fragments: There were low-normal levels of the biologically active protein as determined by full-length ELISA assay and almost undetectable full-length FGF23 in a Western blot. In contrast, Western blot indicated the presence of huge amounts of the C- and N- terminal fragments, which are undetectable in Western blots of healthy individuals. ELISA for C-terminal FGF23 also showed increased amounts of FGF23 in these patients. It has recently been shown that elevated serum FGF23 concentrations are found in individuals with end-stage renal disease (ESRD),(14,18) suggesting a compensatory mechanism aimed at enhancing urinary phosphate excretion and decreasing serum phosphate levels. Extremely high levels of FGF23 fragments determined by the C-terminal assay in our patients and the presence of FGF23 fragments shown by Western blotting, also suggests that production of FGF23 is enhanced to counterbalance hyperphosphatemia but fail to do so because of inappropriately low intact FGF23 levels.
The clinical findings of these patients resemble the recently published characteristics of the FGF23-null mice that show significant increments in serum phosphate levels with increased renal tubular phosphate reabsorption caused by augmented expression and activity of the Na-Pi-2a transporter.(19,20) Increased serum 1,25(OH)2D levels resulting at least in part from enhanced expression of 25-hydroxyvitamin D-1α–hydroxylase (1αOHase) mRNA levels were also noted. Because we could not detect any mutation in the FGF23-encoding gene, the primary defect of these patients seemed to be enhanced intracellular processing of FGF23 at the 176-RXXR-179 motif. The expression of ppGaNTase-T3 that initiates O-glycosylation, a prevalent form of post-translational modification, was found to be absent in the epidermis of our patient as assessed by immunostaining. FGF23 has several O-linked glycosylation sites; not all of them have been precisely defined. Our results show an association between the degree of glycosylation of FGF23 and the propensity to undergo processing: whereas only the fully glycosylated molecule is resistant to processing, underglycosylated proteins are susceptible to degradation. Furthermore, decreasing the expression of GALNT3 in a human osteoblastic cell line by RNA interference led to decreased amounts of the intact biologically active FGF23 caused by increased processing and resulting in elevated levels of the C-terminal fragments. Western blotting indicated that C-terminal fragments of these patients have the same size as that produced by CHO cells. Because the molecular weight of the C-terminal fragment peptide is −7.5 kDa and less than that observed by Western blotting, it suggests that the C-terminal fragments in HHS patients are glycosylated to the same extent as those produced by CHO cells. Furthermore, although N-terminal fragments of FGF23 produced by CHO cells seem to be heterogeneous,(17) the size of N-terminal fragments of these patients was similar to that of the fragment with the highest molecular weight by CHO cells. These results suggest that processed fragments of FGF23 but not full-length FGF23 in the patients have no defect in glycosylation. There are several members of the ppGaNTase family and these members have redundant substrate specificity. This redundancy may explain the fact that there is only one known genetic disease of mucin-type O-linked glycosylation and mutations in GALNT3 gene result in rather mild phenotype with derangements in only phosphate and bone metabolism in contrast to severe lethal diseases usually seen in defects in N-linked glycosylation. These results together with those displayed in Figs. 1 and 4 suggest that ppGaNTase-T3 activity is specifically involved in the glycosylation of Thr178.
Our observations indicate that the primary defect in HHS patients is a defect in glycosylation of FGF23 resulting from reduced expression of ppGaNTase-T3, caused by mutations in GALNT3, and leading to augmented processing at the cleavage site. These changes in FGF23 would abolish its phosphaturic effect and lead to severe persistent hyperphosphatemia. Injection of recombinant FGF23 was recently found to result in decreased serum phosphate levels simultaneously with a marked reduction in the expression of Na-Pi-2a co-transporter in renal brush border membranes and without any increase in PTH levels.(21) Furthermore, FGF23 was shown to significantly improve hyperphosphatemia toward the normal range in parathyroidectomized rats.(21) Our observation together with those derived from the FGF23-null mice model and the recent in vivo studies lend further credence to the primary and essential role of FGF-23 in phosphate homeostasis through a PTH-independent pathway. This was substantiated most recently by several reports that showed that mutations in the FGF23 gene are responsible for FHTC.(22–24) In these patients with FHTC, there was a similar pattern of rather low FGF23 levels by the full-length assay and quite high FGF23 by the C-terminal assay. Therefore, it has been shown that there are at least two genes that are involved in genetic hyperphosphatemic diseases, namely GALNT3 and FGF23.
It has been recently documented that secretion of full-length FGF23 requires ppGaNTase-T3.(25) These authors have also shown that recombinant ppGaNTase-T3 can mediate O-glycosylation of Thr178, and this glycosylation prevents the processing of a peptide derived from FGF23 in vitro.(25) These results are consistent with ours, showing the importance of O-glycosylation by ppGaNTase-T3 in preventing the processing of FGF23. In addition, we have shown here that ppGaNTase-T3 activity actually modulate the processing of FGF23 in cells expressing FGF23 and we propose that ppGaNTase-T3 is specifically involved in the synthesis of one of three O-linked sugar chains in FGF23. Together with the clinical data and the analysis of samples shown here, these studies confirm the involvement of ppGaNTase-T3 in post-translational modification of FGF23.
Our results provide the pathogenetic mechanism underlying the first documented disease with defects in mucin-type O-glycosylation. Because there is an adequate tubular response to PTH, its potential use as a therapeutic agent should be considered. However, a better approach would be to develop a safe recombinant FGF23 preparation that could initially benefit our patients and subsequently be applied to various hyperphosphatemic states.
The authors are indebted to Drs O Topaz, G Richard, and E Sprecher for performing the immunostaining of the skin biopsies. This work was supported in part by grants from the Ministry of Education, Culture, Sports, Science and Technology, and from the Ministry of Health, Labour and Welfare, Japan.