Regulation of Fibroblastic Growth Factor 23 Expression but Not Degradation by PHEX*

Inactivating mutations of Phex cause X-linked hypophosphatemia (XLH) by increasing levels of a circulating phosphaturic factor. FGF23 is a candidate for this phosphaturic factor. Elevated serum FGF23 levels correlate with the degree of hypophosphatemia in XLH, suggesting that loss of Phex function in this disorder results in either diminished degradation and/or increased biosynthesis of FGF23. To establish the mechanisms whereby Phex regulates FGF23, we assessed Phex-dependent hydrolysis of recombinant FGF23 in vitro and measured fgf23 message levels in the Hyp mouse homologue of XLH. In COS-7 cells, overexpression of FGF23 resulted in its degradation into N- and C-terminal fragments by an endogenous decanoyl-Arg-Val-Lys-Arg-chloromethyl ketone-sensitive furin-type convertase. Phex-dependent hydrolysis of full-length FGF23 or its N- and C-terminal fragments could not be demonstrated in the presence or absence of decanoyl-Arg-Val-Lys-Arg-chloromethyl ketone in COS-7 cells expressing Phex and FGF23. In a reticulolysate system, apparent cleavage of FGF23 occurred with wild-type Phex, the inactive Phex-3′M mutant, and vector controls, indicating nonspecific metabolism of FGF23 by contaminating enzymes. These findings suggest that FGF23 is not a direct Phex substrate. In contrast, by real-time reverse transcriptase PCR, the levels of fgf23 transcripts were highest in bone, the predominant site of Phex expression. In addition, Hyp mice displayed a bone-restricted increase in fgf23 transcripts in association with inactivating Phex mutations. Increased expression of fgf23 was also observed in Hyp-derived osteoblasts in culture. These findings suggest that Phex, possibly through the actions of unidentified Phex substrates or other downstream effectors, regulates fgf23 expression as part of a potential hormonal axis between bone and kidney that controls systemic phosphate homeostasis and mineralization.

Inactivating mutations of Phex cause X-linked hypophosphatemia (XLH) by increasing levels of a circulating phosphaturic factor. FGF23 is a candidate for this phosphaturic factor. Elevated serum FGF23 levels correlate with the degree of hypophosphatemia in XLH, suggesting that loss of Phex function in this disorder results in either diminished degradation and/or increased biosynthesis of FGF23. To establish the mechanisms whereby Phex regulates FGF23, we assessed Phex-dependent hydrolysis of recombinant FGF23 in vitro and measured fgf23 message levels in the Hyp mouse homologue of XLH.

In COS-7 cells, overexpression of FGF23 resulted in its degradation into N-and C-terminal fragments by an endogenous decanoyl-Arg-Val-Lys-Arg-chloromethyl ketone-sensitive furin-type convertase. Phex-dependent hydrolysis of full-length FGF23 or its N-and C-terminal fragments could not be demonstrated in the presence or absence of decanoyl-Arg-Val-Lys-Arg-chloromethyl ketone in COS-7 cells expressing Phex and FGF23.
In a reticulolysate system, apparent cleavage of FGF23 occurred with wild-type Phex, the inactive Phex-3M mutant, and vector controls, indicating nonspecific metabolism of FGF23 by contaminating enzymes. These findings suggest that FGF23 is not a direct Phex substrate. In contrast, by real-time reverse transcriptase PCR, the levels of fgf23 transcripts were highest in bone, the predominant site of Phex expression. In addition, Hyp mice displayed a bonerestricted increase in fgf23 transcripts in association with inactivating Phex mutations. Increased expression of fgf23 was also observed in Hyp-derived osteoblasts in culture. These findings suggest that Phex, possibly through the actions of unidentified Phex substrates or other downstream effectors, regulates fgf23 expression as part of a potential hormonal axis between bone and kidney that controls systemic phosphate homeostasis and mineralization.
Current data indicate that Phex regulates the production and/or degradation of a systemic phosphaturic hormone, referred to as phosphatonin (15). The presence of phosphatonin in XLH/Hyp was detected by parabiosis experiments in which Hyp mice transferred the phosphaturic phenotype to normal mice (16). Studies in parathyroidectomized Hyp mice eliminated parathyroid hormone as the responsible phosphaturic factor (17). Subsequently, cross-kidney transplant studies confirmed the presence of a circulating factor in Hyp mice and excluded the presence of a primary renal phosphate wasting defect (18). The observation that Phex is most highly expressed in cartilage, bone, and teeth (3, 10, 11, 19 -22), where it is present in osteoblasts, osteocytes, and odontoblasts, raises the possibility that locally produced substrates in the skeleton are involved in the pathogenesis of XLH. In addition, osteoblasts derived from Hyp mice produce a factor capable of inhibiting sodium-dependent phosphate transport in renal tubular cells (23), implicating bone as a possible source of phosphatonin.
Genetic studies of autosomal dominant hypophosphatemia (ADHR) have identified FGF23 as a likely candidate for phosphatonin (24). FGF23 is an ϳ30-kDa (251 amino acids) protein with an N-terminal region containing the FGF homology domain and a novel 71-amino acid C terminus. Several studies have confirmed that full-length FGF23 is a phosphaturic hormone (25)(26)(27) and that cleavage of FGF23 at the RXXR motif generates biologically inactive N-and C-terminal fragments (27). In ADHR mutations of the 176 RXXR 179 site (R176Q, R179W, and R179Q) prevent cleavage and inactivation of FGF23 (24). Tumor-induced osteomalacia (TIO), also called oncogenic osteomalacia, is an acquired hypophosphatemic disorder with phenotypic features similar to ADHR and XLH (28). FGF23 is also secreted from TIO tumors (27,29,30), and removal of the TIO tumor is associated with reductions in circulating FGF23 concentrations and correction of the hypophosphatemia (28,31). In addition, FGF23 is elevated in some subjects with XLH (28,31,32). Serum phosphate concentrations are negatively correlated with circulating FGF23 levels in patients with XLH, suggesting that elevated FGF23 is causing the hypophosphatemia in this disorder as well (31).
The phenotypic similarities among ADHR, XLH, and TIO form the basis for an unproven model to explain their common pathogenesis (33). This model presumes that PHEX degrades active full-length FGF23 into inactive fragments (33) and that ADHR, XLH, and TIO are caused by increased circulating levels of FGF23, which acts as a phosphaturic hormone to inhibit sodium-dependent phosphate uptake in the renal proximal tubule. In this model, FGF23 is increased in ADHR because of mutations in FGF23 that render it resistant to PHEXdependent cleavage, in XLH because of inactivating mutations of PHEX that prevent the normal degradation of FGF23, and in TIO, because overproduction of FGF23 by the tumor overwhelms the degradation capacity of PHEX.
Currently, three fundamental aspects of this model have been established, namely that mutations preventing the metabolism of FGF23 causes ADHR (33), FGF23 possess phosphaturic actions (26,27), and serum concentrations of FGF23 are elevated in some subjects with TIO and XLH (28,31,32). The requirement that FGF23 is a substrate for PHEX, however, has not been tested rigorously. Indeed, the data regarding Phex metabolism of FGF23 are conflicting. One report suggests that recombinant Phex may cleave FGF23 at the RXXR motif or a nearby site (34), but other studies have failed to confirm Phexdependent cleavage of FGF23 (35). In addition, the RXXR motif is the consensus cleavage site for pro-protein convertases (36), and all cell lines and expression systems tested to date for generating recombinant FGF23 contain enzymes capable of metabolizing FGF23 into its N-and C-terminal fragments (27,30), suggesting that the processing enzyme is widely expressed. Also, no studies have investigated the possibility that FGF23 biosynthesis might be increased in XLH/Hyp.
In the current investigations we have generated recombinant FGF23 and Phex and tested whether Phex metabolizes FGF23. In addition, we have examined whether fgf23 expression is increased in the Hyp mouse homologue of XLH.
Preparation of Phex-expressing Membrane Protein-We confirmed the activity of Phex against oligopeptide substrates using recombinant Phex expressed in Sf9 membranes. Crude Sf9 membranes expressing Phex were solubilized with 1% n-dodecyl-␤-D-maltoside, collected after centrifugation at 10,000 ϫ g for 15 min and stored at Ϫ70°C in multiple aliquots. The protein content of each sample was determined by the NanoOrange TM protein quantitation kit (Molecular Probes, Eugene, OR).
Cell Culture and Transfections-We evaluated Phex-dependent hydrolysis of recombinant FGF23 in co-transfection and co-culture models. COS-7 cells were cultured in Dulbecco's modified Eagle's medium with 10% fetal bovine serum. We transiently transfected pFLAG-FGF23-His, pFLAG-FGF23R179Q-His, pFLAG-C-FGF23-His, pFLAG-N-FGF23-His, pcDNA3.1-Phex-V5-His, or corresponding vector control plasmids in COS-7 cells. All transfections were performed with the FuGENE 6 transfection reagent (Roche Applied Science) following the manufacturer's protocols. In addition, COS-7 cells were co-transfected with FG23 expression constructs (1 g) and either pcDNA3.1-Phex-V5-His or pcDNA3.1/V5-His vector (1 g). A FuGENE 6 to DNA ratio of 3 l:1 g was maintained for all experiments. The cells and conditioned media were harvested 48 h post-transfection. For the generation of stable cell lines overexpressing FGF23, COS-7 cells were transfected with pFLAG-FGF23-His and cultured in the presence of G418 (500 g/ml) for 14 days. TMOb-Nl and TMOb-Hyp immortalized cells derived from normal and Hyp mice calvaria (38) were grown for periods of up to 14 days in ␣-minimum essential medium containing 10% fetal bovine serum (Atlas Biologicals, Fort Collins, CO) containing 5 mM ␤-glycerophosphate and 25 g/ml ascorbic acid as reported previously (38).
Assessment of FGF23 Hydrolysis in Culture-To assess endogenous FGF23 hydrolysis in the absence of Phex, we collected total cell lysates and conditioned media from COS-7 cells after transient and stable transfection with FGF23 expression constructs. To assess Phexdependent hydrolysis of FGF23, we collected conditioned media from COS-7 cells following co-transfected with pFLAG-FGF23-His and pcDNA3.1-Phex-V5-His constructs. In addition, we plated COS-7 cells producing FGF23 (1 ϫ 10 5 cells/well) with COS-7 cells expressing Phex (1 ϫ 10 5 cells/well) or cells transfected with the control vector and incubated for 24 h in Dulbecco's modified Eagle's medium with 10% fetal bovine serum. The co-cultures were washed with Hanks' balanced salt solution, and conditioned media were collected after incubation for 48 h in serum-free Dulbecco's modified Eagle's medium. The conditioned medium was centrifuged at 7000 ϫ g for 10 min to remove cells and cellular debris, and FGF23 hydrolysis was monitored by Western blot analysis (see below). For the inhibitor studies, COS-7 cells stably transfected with pFLAG-FGF23-His were seeded in 6-well plates. After 48 h the medium was replaced with serum-free media containing different concentrations of the furin inhibitor dec-RVKR-cmk (39, 40) (BIOMOL Research Laboratories, Inc., Plymouth Meeting, PA) or equivalent amounts of Me 2 SO vehicle. The cells were cultured for an additional 24 h and then the conditioned medium was collected for Western blot analysis.
Assessing Phex Enzyme Activity in Vitro-We assessed Phex activity using either the Z-Ala-Ala-Leu-p-nitroanilide (ZAAL-pNA) chromogenic substrate (35) or a fAFF1-TC-LIB phage clone containing the VPQSDS peptide sequence cloned between an epitope (YGGFL) tether and a truncated form of M13 gene (40,41). ZAAL-pNA was purchased from Bachem Biosciences, Inc. (King of Prussia, PA). Leucine aminopeptidase was purchased from Sigma-Aldrich. We used membrane fractions from the Sf9 cells (50 g) expressing vector or the V5-His epitopetagged Phex constructs. Protein samples were incubated with 50 M ZAAL-pNA in 100 l of 100 mM MES, pH 6.5, for 1 h at 37°C. After completion of the initial incubation, the reaction mixture was further incubated with 0.4 milliunits of leucine aminopeptidase for 20 min at 37°C. The reaction was stopped by the addition of 100 mM EDTA, and the absorbency was measured at 405 nm after centrifugation. In some studies membrane fractions were preincubated with inhibitors EDTA, phosphoramidon, or dec-RVKR-cmk at the stated concentrations for 30 min before the addition of ZAAL-pNA.
For dot blot analysis of Phex activity, we used a phage clone containing the hexapeptide VPQSDS sequences demonstrated previously (41)(42)(43) to be optimal for Phex-dependent hydrolysis. The VPQSDS-expressing phage was amplified overnight in Escherichia coli K91 cells and precipitated by adding 20 l of 20% polyethylene glycol in 2.5 M NaCl to 100 l of phage supernatant as described previously (42). After incubating on ice for 30 min, the precipitated phage were microfuged for 5 min and resuspended in 10 l of TBS (50 mM Tris-HCl, pH 7.4, 150 mM NaCl). The phage suspension was incubated with Sf9 membranes (30 g) expressing Phex-WT or vector control membranes at 37°C in a total volume of 100 l. At various time points, 30-l aliquots of the reaction was removed and stopped by adding TBS buffer containing 5 mM EDTA. Aliquots were transferred on nitrocellulose membrane using the Bio-dot system (Bio-Rad) and immunoblotted with anti-dynorphin monoclonal antibody 3-E7 (Gramsch Laboratories, Schwabhausen, Germany). Washed filters were probed with a 1:5000 dilution of goat anti-mouse IgG horseradish peroxidase conjugate and stained as directed using the Western enhanced chemiluminescence kit (Amersham Biosciences).
RNA Isolation-We isolated RNA from various mouse tissues by grinding snap frozen tissues in liquid nitrogen and then extracting total RNA using Tri reagent (Molecular Research Center, Inc., Cincinnati, OH). RNA samples, pretreated with DNase, were further cleaned using an RNeasy spin column (Qiagen, Inc., Valencia CA), and the yield was quantified using a Ribogreen RNA quantitation kit (Molecular Probes, Eugene, OR).
Quantitative Real-time PCR-1.5 g of total RNA was denatured for 5 min at 65°C in the presence of 0.5 mol of random hexamer, snap cooled in ice water, and then reverse transcribed in 100 l using the TaqMan reverse transcription kit (Applied Biosystems, Foster City, CA) according to the manufacturer's protocol. PCR reactions contained 100 ng of template (cDNA or RNA), 300 nM each forward and reverse primer, and 1ϫ SybrGreen PCR Master Mix (Applied Biosystems, Foster City, CA) in 50 l. Samples were amplified for 40 cycles in an ABI Prism 7700 Sequence Detection System (Applied Biosystems, Foster City, CA) with an initial melt at 95°C for 10 min followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. PCR product accumulation was monitored at multiple points during each cycle by measuring the increase in fluorescence caused by the binding of SybrGreen I to doublestranded DNA. The partial cycle at which a statistically significant increase in fgf23 product was first detected (threshold cycle; Ct) and normalized to the Ct for cyclophilin. A passive reference dye, ROX, was used to normalize for variations in volume and/or dye concentration between sample wells. Post-amplification melting curves were performed to confirm that a single PCR product was produced in each reaction. The contribution of contaminating genomic DNA to the observed product was determined from the Ct given by the RNA template. This quantity was usually less than 0.1%. The primer sequences used include FGF23 forward primer 5Ј-ACTTGTCGCAGAAGCATC and FGF23 reverse primer 5Ј-GTGGGCGAACAGTGTAGAA and cyclophilin A forward primer 5Ј-CTGCACTGCCAAGACTGAAT and cyclophilin A reverse primer 5Ј-CCACAATGTTCATGCCTTCT.
Statistics-We evaluated differences between groups by one-way analysis of variance for multiple comparison or by Student's t test for comparison between two groups. All values are expressed as mean Ϯ S.E. The relationship between gene expressions was assessed by regression analysis. All computations were performed using the Statgraphic statistical graphics system (STSC, Inc., Rockville, MD).

RESULTS
FGF23 Is Processed at the RXXR Site-To assess FGF23 hydrolysis, full-length, N-and C-terminal fragments and a cleavage-resistant form of FGF23 containing the R179Q muta-  tion at the RXXR site (Fig. 1A) were expressed in COS-7 cells using pFLAG-CMV-3 expression vector (Fig. 1A). In total cell lysates derived from COS-7 cells expressing full-length recombinant FGF23, we observed a single band that migrates at 32 kDa (Fig. 1B). In conditioned media derived from these cells, however, the secreted 32-kDa full-length FGF23 protein was present in lower amounts, and we observed the presence of 22-kDa N-terminal and 16-kDa C-terminal fragments as the predominant products. The presence of N-and C-terminal fragments in conditioned media and the absence of these fragments in cell lysates suggests hydrolysis of FGF23 occurred during secretion, possibly by an ectoenzyme present on the surface of COS-7 cells (Fig. 1B). Similar hydrolysis of mouse fgf23 was observed in COS-7 cells (data not shown). Consistent with other reports, the R179Q mutation prevented cleavage of FGF23 in COS-7 cells (Fig. 1B).
A Chloromethyl Ketone-based Inhibitor of Furin-type Convertase Blocks FGF23 Cleavage-To determine whether the observed cleavage of FGF23 in COS-7 cells was mediated by endogenous pro-protein convertases, we evaluated whether the furin inhibitor dec-RVKR-cmk prevented cleavage of FGF23. dec-RVKR-cmk (5-50 M) added to the culture media resulted in a dose-dependent inhibition of FGF23 processing in COS-7 cells (Fig. 2), indicating that endogenous furin-type convertase are responsible for FGF23 metabolism.
FGF23 Is Not Cleaved by Phex in Mammalian Cells-To examine whether Phex metabolizes FGF23, we co-expressed Phex and FGF23 in COS-7 cells and assessed cleavage by Western blot analysis (Fig. 3). The co-expression of Phex and full-length FGF23 in COS-7 cells did not alter the ability of the endogenous furin-like enzyme to cleave FGF23 into its N-and C-terminal fragments (Fig. 3A). In addition, co-expression of Phex with the cleavage-resistant R179Q FGF23 mutant did not result in any degradation products. To investigate the possibility that FGF23 might require processing before Phex-mediated hydrolysis, we co-expressed Phex with an expression vector containing either the N-terminal or C-terminal FGF23 fragments in COS-7 cells. Phex also failed to hydrolyze either the N-or C-terminal FGF23 fragments, suggesting that the puta-tive Phex cleavage sites identified in FGF23 (43) are not accessible to the recombinant Phex enzyme under the conditions studied. To exclude the possibility that both Phex and an endogenous furin-like enzyme were cleaving FGF23 at the RXXR site, we co-transfected COS-7 cells with full-length FGF23 and Phex in the presence and absence of the furin inhibitor dec-RVKR-cmk (Fig. 3B). Consistent with the prior findings ( Fig.  1), the addition of dec-RVKR-cmk blocked the degradation of FGF23 in COS-7 cells (Fig. 3B). In the presence of dec-RVKRcmk, Phex did not result in cleavage of FGF23. Finally, to exclude the possibility that co-expression of Phex and FGF23 in the same cell might somehow limit enzyme-substrate interactions, we co-cultured COS-7 cells expressing Phex with COS-7 cells producing FGF23. Similar to co-expression studies (Fig.  3A), we found no evidence for Phex-dependent cleavage of FGF23 in co-culture experiments (Fig. 3C).
To confirm the activity of the recombinant Phex, we tested the ability of cell membrane fractions expressing Phex to hydrolyze two different oligopeptide substrates reported to be cleaved by Phex (35,41). We found that membranes expressing rPhex cleaved the hexapeptide VPQSDS, whereas membranes from vector controls did not cleave this peptide (Fig. 4A). In addition, membranes expressing rPhex hydrolyzed the substrate ZAAL-pNA (Fig. 4B). To rule out the possibility that dec-RVKR-cmk inhibition of Phex might account for the negative results of Phex-dependent cleavage of FGF23, we tested the ability of this inhibitor on Phex-dependent cleavage of the synthetic substrate ZAAL-pNA (Fig. 4, B and C). dec-RVKRcmk did not block Phex-dependent cleavage of ZAAL-pNA (Fig.  4B), making it unlikely that dec-RVKR-cmk prevented Phexdependent cleavage of FGF23 in the preceding experiments (Fig. 3). In addition, we found that Phex-dependent cleavage of ZAAL-pNA could be blocked by EDTA but not by phosphoramidon (Fig. 4C).
Nonspecific Cleavage of FGF23 Occurs in Vitro-To re-examine prior reports that rPhex hydrolyzes FGF23 in vitro (34), we generated recombinant Phex and FGF23 using a reticulolysate system and incubated the resulting proteins in vitro to assess cleavage (Fig. 5). For these studies mouse full-length (Phex- WT) and C-terminal truncated Phex (Phex-3ЈM), human wildtype or R179Q mutant FGF23, and pcDNA 3.1/V5-His vector were transcribed in vitro and translated in rabbit reticulocyte lysates. We found that incubation with reticulolysates from wild-type Phex, the inactive Phex-3ЈM mutant, and vector alone resulted in apparent cleavage of FGF23 as evidenced by the reduction in the intensity of the band for both the wild-type (Fig. 5A) and the FGF23R179Q mutation (Fig. 5B). The failure to observe Phex-specific cleavage products indicates that the reduction in band intensity is because of the presence of contaminating enzymes in the reticulolysate system.

Fgf23 Expression Is Increased in Hyp Mouse and Correlates with Increased Mepe Expression in
Bone-Because we were unable to confirm Phex-dependent metabolism of FGF23, we explored whether fgf23 expression might be increased in the Hyp mouse homologue of XLH that has a 3Ј deletion of Phex. First, we examined fgf23 expression by real-time PCR in the normal tissues, including bone, where Phex is predominately expressed (Fig. 6). In normal mice, the rank order of fgf23 expression in normal mouse tissues was bone Ͼ thymus Ͼ brain Ͼ heart Ͼ skeletal muscle Ͼ spleen Ͼ skin Ͼ lung Ͼ testes, with nearly undetectable levels in liver and kidney (Fig.   6A). Next, we compared fgf23 expression in different skeletal sites in normal and Hyp mice (Fig. 6B). We found that normal mice express fgf23 in calvaria, mandible, and diaphysis but extremely low levels in bone marrow. Hyp mice expressed markedly increased fgf23 levels in the calvaria, mandible, and diaphysis compared with normal mice. The increase in fgf23 in Hyp mice was limited to bone, because no increases of fgf23 were observed in bone marrow (Fig. 6B) or kidney, lung, and liver (data not shown) derived from Hyp mice.
To compare the increase of fgf23 with mepe, another bonerelated gene reported to be increased in Hyp mice (44), we measured fgf23 and mepe expression in calvaria from Hyp and normal control mice (Fig. 7, A and B). Although the overall abundance of mepe transcripts were higher than fgf23, the magnitude of the increase in fgf23 was greater than that observed for mepe (Fig. 7). Of potentially greater significance was the very strong correlation between fgf23 and mepe (Fig. 7C), suggesting that fgf23 might regulate mepe expression or that both are regulated by a common upstream factor resulting from Phex deficiency.
Fgf23 Is Expressed in Hyp-derived Osteoblasts Cell during Maturation-We previously isolated and characterized immortalized osteoblast cell lines derived from normal wild-type and Hyp mice, respectively designated TMOb-Nl and TMOb-Hyp (38). Both TMOb-Nl and TMOb-Hyp osteoblasts undergo a temporal sequence of maturation, but TMOB-Hyp, which lack a functional Phex, display an intrinsic mineralization defect in culture. We examined fgf23 expression by real-time PCR in TMOb-Hyp and TMOb-Nl osteoblasts during a 14-day culture period during which these cells undergo maturation (Fig. 8). We observed low levels of expression of fgf23 in TMOb-Nl osteoblasts that were not significantly altered during the maturational sequence from 5 through 14 days of culture. In contrast, in TMOb-Hyp osteoblasts we observed progressive culture duration-dependent increase in fgf23 expression that peaked at day 10 of culture and then declined during the latter stages of differentiation as these cells transitioned to mature osteoblasts.

DISCUSSION
A single enzyme-substrate hypothesis has been proposed to explain the role of PHEX and FGF23 in the pathogenesis of XLH (33). This model presumes that the intact FGF23 protein is the physiologically relevant substrate for Phex and that decreased degradation of FGF23 by mutated PHEX accounts for an increase in the circulating levels of this phosphaturic hormone in XLH. The current investigations, however, do not support this hypothesis. Instead, we show that FGF23 is cleaved by a furin-like enzyme at the RXXR site (see Figs. 1 and 2) rather than by Phex (Fig. 3). In contrast, we fail to demonstrate Phex-dependent cleavage of intact FGF23 or its N-and C-terminal fragments using recombinant Phex and FGF23 proteins expressed in mammalian cells (Fig. 3) or using proteins synthesized in vitro (Fig. 5). The inability to show Phex-dependent cleavage of FGF23 is not because of lack of activity of the recombinant Phex enzyme, because we show activity against synthetic oligopeptide substrates (Fig. 4), as reported previously (41,43). Instead of Phex-dependent degradation, we show that inactivating mutations of Phex result in increased expression of FGF23 transcripts in the bone and cultured osteoblasts of the Hyp mouse homologue of XLH (see Figs. 6 -8), indicating that Phex may somehow regulate the biosynthesis of FGF23.
Of these findings, the demonstration that the skeleton is the predominant site of fgf23 expression in normal mice and that fgf23 transcripts are markedly increased in the skeleton of Hyp mice (Fig. 6) are potentially the most important. Prior studies reported fgf23 expression in low abundance in several normal tissues similar to our findings ( Fig. 6A) but did not examine fgf23 expression in bone or the effect of Phex deficiency on the levels of fgf23 transcripts (27). In Hyp mice lacking a functional Phex, increased levels of fgf23 transcripts were observed in mandible, calvaria, and diaphysis of the long bone (Fig. 6B). The increase in fgf23 message in Hyp mice was limited to the skeletal sites, because levels of fgf23 expression in bone marrow and other tissues of Hyp mice were similar to levels of fgf23 found in normal mice. Although the present studies did not define the cell type(s) in bone tissue that expresses fgf23, it is likely that osteoblasts and/or osteocytes are the source of fgf23 expression in bone, because differential expression of fgf23 between normal and Hyp mice was not observed in bone marrow (Fig. 6B). In addition, recent studies in patients with the McCune-Albright syndrome have identified FGF23 production by fibrous dysplasia osteoprogenitors and normal bone-forming cells in vivo and in vitro (45). Consistent with the predominant expression of fgf23 in osteoblasts, additional results demonstrate fgf23 transcripts in osteoblast cultures (Fig. 8). More importantly, similar to the increased expression of fgf23 in Hyp bone, fgf23 expression was greater in Hyp compared with normal osteoblasts. The increase in fgf23 in Hyp-derived osteoblasts was most evident at day 10 of culture (Fig. 8), at the transition between the maturation of pre-osteoblasts to mature osteoblasts when Phex expression is normally up-regulated The asterisk denotes a significant difference at p Ͻ 0.05. C, regression between fgf23 and mepe in normal and Hyp calvaria. A significant correlation between fgf23 and mepe levels was observed (p ϭ 0.002). (38). Our prior studies failed to detect fgf23 transcripts in the same osteoblastic cell lines (35), possibly because of the use of less sensitive detection methods and examination of osteoblasts at stages of lowest fgf23 expression. Although fgf23 transcripts are detected in osteoblast cultures, the level of expression in culture is lower than in skeletal tissue. The lower expression in cell lines might be explained by the loss of in vivo stimuli required for maintenance of FGF23 expression or by the expression of fgf23 in other cell types in bone. Thus, the skeleton appears to be the major site of FGF23 expression in health and disease.
Nevertheless, the study of age-and sex-matched Hyp mice and their wild-type littermates more clearly establishes an association between fgf23 expression and Phex deficiency than has been observed in cross-sectional evaluations of subject with XLH (28,31,32), a setting in which variable circulating levels of FGF23 could be explained by additional factors regulating serum phosphorus and FGF23 expression/metabolism. Moreover, the fact that both fgf23 and Phex are expressed in osteoblasts, along with the observation that osteoblasts derived from Hyp mice have intrinsic abnormalities of mineralization (38) and secrete phosphaturic factors (23), suggests that osteoblasts may play a central role in the pathogenesis of both defective mineralization and phosphaturia. To this end, it has been proposed that increments in mepe and/or the impaired metabolism of this extracellular matrix protein might contribute to the pathogenesis of abnormal mineralization in Hyp mice (44,46). The current findings that mepe expression is highly correlated with increased fgf23 in Hyp bone (Fig. 7) suggest that fgf23 stimulates mepe expression or that common pathways downstream of Phex may be responsible for increasing both mepe and fgf23 expression in osteoblasts.
The novel finding that fgf23 message levels are increased in Hyp skeleton was unexpected and is contrary to the recently proposed single enzyme/substrate hypothesis that PHEX cleaves FGF23 at either the RXXR motif or a nearby site (33,34). We believe that the initial report of Phex-dependent metabolism of FGF23 may have been because of other proteolytic enzymes contaminating the reticulolysate preparation (34), be-cause we observed degradation of recombinant FGF23 after incubation with reticulolysates alone, as well as with reticulolysates expressing either the inactive 3Ј truncated Phex or the full-length wild-type Phex (Fig. 5). Others have recently reported the failure to document Phex-dependent cleavage at the RXXR site in FGF23 (35,43). The results in COS-7 cells and reticulolysate systems are consistent with the presence of widely expressed proteolytic enzymes capable of hydrolyzing FGF23 (27,30).
MEPE and FGF23 (43), as well as parathyroid hormonerelated peptide residues 107-139 (47), have amino acid sequences that would be predicted to be cleaved by Phex. Indeed, quenched fluorescence peptides derived from FGF23, as well as peptides derived from MEPE, are hydrolyzed by rPhex in vitro (43). In the current study, the failure to demonstrate that full-length FGF23 containing putative cleavage sites are hydrolyzed by Phex, and a prior study (46) that also failed to show Phex-dependent hydrolysis of full-length MEPE, suggest that the mere presence of consensus cleavage sites for an enzyme in a protein does not necessarily indicate that these site are available to the enzyme. Indeed, the inability of Phex to hydrolyze intact FGF23 or MEPE may indicate a constraint placed by incompatible three-dimensional structures that limit enzyme substrate interactions (48,49). Consequently, large proteins such as FGF23 and MEPE might require processing into smaller fragments before being cleaved by Phex, or other proteins such as Dentin matrix protein 1 may have conformations necessary for Phex-dependent hydrolysis (49). The latter possibility seems more likely, because neither N-nor C-terminal FGF23 fragments are hydrolyzed by Phex (see Figs. 1 and 3). Moreover, the processing of FGF23 leads to its inactivation (26), and to date, biological activities have not been ascribed to the N-and C-terminal FGF23 fragments. Thus, even if Phex could be shown to hydrolyze the N-and C-terminal FGF23 fragments under different conditions, the biological significance of further degradation of these inactive fragments is uncertain.
Finally, the in vitro finding that a furin-like enzyme present in COS-7 cell membranes cleaves FGF23 into N-and C-terminal fragments (see Figs. 1 and 2) is of uncertain physiological significance. Similar degradation by cell surface convertases in vivo might be expected to limit the biological activity of FGF23. It is possible that other factors may be present in the circulation that bind to and prevent FGF23 cleavage that are not present in the in vitro culture systems. Although only fulllength FGF23 has been shown to have phosphaturic actions (50), additional studies are needed that investigate its in vivo metabolism and pharmacokinetics.
In conclusion, the increased expression of fgf23 in the skeleton of Hyp mice suggests that increased synthesis may be an important mechanism of elevated circulating FGF23 levels in XLH. In addition, the failure to confirm fgf23 hydrolysis by Phex leaves open the possibility that physiologically relevant Phex substrates remain to be discovered. Taken together, these data support a more complex model, with intermediate steps linking PHEX to FGF23, to explain the pathogenesis of XLH. One possibility is that increased fgf23 expression in the Hyp skeleton results from the actions of yet to be identified Phex substrate or from other downstream consequences of inactivating mutations of Phex. Furthermore, additional direct or indirect effects of fgf23 and Phex to modify mepe expression and metabolism may also explain the intrinsic defect in osteoblastmediated mineralization that is independent of hypophosphatemia. Whatever the exact mechanism, the interactions between Phex and fgf23 likely occur within the skeleton, where both are expressed. This implicates bone as an endocrine organ FIG. 8. Expression of fgf23 mRNA in normal and Hyp-derived osteoblasts. Immortalized osteoblasts from normal wild-type (TMOb-Nl) and Hyp (TMOb-Hyp) mice were cultured for up to 14 days in the presence of ascorbic acid and ␤-glycerol phosphate to induce differentiation as described previously (35). Fgf23 expression was assessed by real-time reverse transcriptase PCR as described under "Experimental Procedures." TMOb-Hyp osteoblasts expressed higher levels of fgf23 at 7, 10, and 12 days of culture. Values represent the mean Ϯ S.E. of three wells per group. Values with different superscript letters are significantly different at p Ͻ 0.0001. that produces a hormone, fgf23, to regulate renal phosphate conservation to meet the need for bone mineralization.