X-linked hypophosphatemia (XLH) is the most common genetic disorder of phosphate homeostasis. It is caused by inactivating mutations in the PHEX gene,1–4 which encodes a phosphate-regulating endopeptidase homolog.4, 5 Biochemical hallmarks of this disorder include hypophosphatemia and inappropriately low or normal serum 1,25-dihydroxyvitamin D [1,25(OH)2D, calcitriol] concentrations, resulting in rickets and osteomalacia.
The primary cause of the hypophosphatemia in XLH is elevated levels of a circulating phosphaturic hormone, fibroblast growth factor 23 (FGF23), which inhibits renal tubular phosphate reabsorption and 1,25(OH)2D production.6 Although the majority of patients with XLH have increased circulating FGF23 concentrations, some patients have normal FGF23 levels, that are inappropriate in light of the hypophosphatemia.7–9 Both PHEX and FGF23 are expressed in osteoblasts, osteocytes, and odontoblasts;10–14 however, the mechanism by which PHEX deficiency results in overproduction of FGF23 is currently unclear.
In contrast, persistently low levels of intact FGF23 causes familial tumoral calcinosis characterized by hyperphosphatemia and ectopic calcifications in soft tissues.15–19 Although inactivating mutations in FGF2315, 19–21 and Klotho (KL)22 can cause tumoral calcinosis, most of the patients have mutations in GALNT3,17 which encodes a Golgi-associated glycosyltransferase, UDP-N-acetyl-alpha-D-galactosamine:polypeptide N-acetylgalactosaminyltransferase 3 (ppGalNAc-T3). This enzyme adds O-glycosylation in the subtilisin-like proprotein convertase recognition site in FGF23 protein, thereby protecting FGF23 from proteolytic cleavage and allowing intact FGF23 to be secreted.23, 24 Thus, in the absence of this enzyme, most FGF23 is processed into inactive fragments before secretion, leading to low or undetectable levels of biologically active intact FGF23 in circulation. The low intact Fgf23 and subsequent hyperphosphatemia make tumoral calcinosis a biochemical mirror image of XLH.
The current standard therapy for XLH is high-dose phosphate and calcitriol.25 Although the therapy is effective in treating osteomalacia, it does not ameliorate the main cause of hypophosphatemia (ie, increased FGF23 concentrations) because phosphate and calcitriol are known stimulators of FGF23 expression.6, 26–29 Furthermore, recent studies indicate that standard therapy increases FGF23 concentrations in patients with XLH.30, 31 These observations suggest that FGF23-producing cells (mainly osteoblasts and osteocytes) in XLH sense the phosphate level to be too high even when it nears normal, and this abnormal responsiveness to phosphate may be the underlying cause of increased FGF23 production in XLH. In this study, we tested this hypothesis using murine models of XLH and tumoral calcinosis.
Materials and Methods
Generation and maintenance of experimental animals
A new murine model of XLH carries a nonsense mutation (K496X) in exon 14 of the Phex gene. The PhexK496X mouse line was created and kindly provided by Drs. Jane E Aubin and Frieda Chen at the University of Toronto, Canada, and Dr. Ann M Flenniken, Celeste Owen, and other members of the Centre for Modeling Human Disease, Samuel Lunenfeld Research Institute, Mount Sinai Hospital, Toronto, Canada. A detailed phenotype of this animal will be described elsewhere.32 Creation and initial characterization of Galnt3 knockout mice, the murine model of familial tumoral calcinosis, was described previously.33 To eliminate the effect of strain background, both Phex and Galnt3 mutant mice were backcrossed to C57BL/6J at least 10 generations. These congenic lines were used for all experiments.
Galnt3/Phex double-mutant mice and single-mutant mice were generated by crossing the two congenic lines. Because Galnt3 knockout males (−/−) were infertile,33 homozygous females were crossed to affected Phex males (−/Y) to generate F1 generations. F1 offspring were subsequently intercrossed to generate F2 experimental groups. Animals were fed Teklad Global 18% Protein Extruded Rodent Diet containing 1.01% calcium, 0.65% phosphorus, and 2.05 IU/g vitamin D3 (2018SX, Harlan, Indianapolis, IN, USA) and tap water ad libitum. All animal studies were approved by the Indiana University School of Medicine Institutional Animal Care and Use Committee.
Serum biochemistry measurements
Four- and twelve-week-old mice were anesthetized with a ketamine/xylazine mix, and blood was drawn by cardiac puncture and stored at −80°C until analysis. Serum alkaline phosphatase, blood urine nitrogen (BUN), calcium, creatinine, and phosphorus were measured, using Roche COBAS Mira S (Roche Diagnostics, Indianapolis, IN, USA). 1,25(OH)2D concentrations were measured, using 1,25-Dihydroxy Vitamin D EIA (Immunodiagnostic Systems Ltd., Fountain Hills, AZ, USA). Intact Fgf23 levels were measured using FGF-23 ELISA Kit, which detects only intact Fgf23 (Kainos Laboratories, Inc., Tokyo, Japan).9 Fgf23 levels were also measured using mouse FGF-23 (C-Term) ELISA Kit, which detects both intact and C-terminal fragments of Fgf23 (Immutopics International, San Clemente, CA, USA). Fgf23 concentrations in Galnt3 knockout mice, Phex mutant mice, and Galnt3/Phex double-mutant mice were measured in serums diluted at 1:10 with diluents suggested by the manufacturers.
Fgf23 mRNA quantification
After the blood draw, femurs were collected and stored in RNAlater RNA Stabilization Reagent (Qiagen Inc., Valencia, CA, USA). Total RNA was extracted from the whole femurs, using TRIzol Plus RNA Purification System (Invitrogen, Carlsbad, CA, USA) and used for first-strand cDNA synthesis, using High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA, USA). The cDNA was subsequently used for quantification of Fgf23 expression by probe-based quantitative PCR, using TaqMan Gene Expression Master Mix in the 7900HT Fast Real-Time PCR System (Applied Biosystems). To identify genes stably expressed across various mice, 10 different “housekeeping” genes (18s, Actb, B2m, Gapdh, Gusb, Hmbs, Hprt, Sdha, Tbp, and Tfrc) were amplified in eight cDNA samples each of male and female wild-type, Phex mutant, Galnt3 knockout, and Galnt3/Phex double-mutant mice (both 4 and 12 weeks old). The most stable gene, Tbp (TATA box binding protein), determined by NormFinder,34 was used to normalize the expression of Fgf23. Relative gene expression was determined by analyzing the data using the relative standard curve method. Probe-based quantitative real-time PCR assays were designed using Integrated DNA Technologies SciTools (Coralville, IA, USA).
The forward primer, probe, and reverse sequences were: GGTGATAACAGGAGCCATGAC, TCAGCCCAGAGAATTGCAAGTTCCG, and TGCTTCTGCGACAAGTAGAC for Fgf23 (NCBI Reference Sequence: NM_022657) and AAGAAAGGGAGAATCATGGACC, CCTGAGCATAAGGTGGAAGGCTGTT, and GAGTAAGTCCTGTGCCGTAAG for Tbp (NCBI Reference Sequence: NM_013684). Sequences for other genes are available upon request.
Means, standard deviations, and standard errors were calculated for all outcome measures by genotype. Differences between the 12 genotypic groups were tested for significance using analysis of variance (ANOVA). When the ANOVA p values were significant, differences between two groups were tested for significance using unpaired Student's t test. Because of significant differences between sexes and ages, the between-genotype comparisons were made for each sex and age separately. P values < 0.05 were considered significant for all analyses.
Galnt3/Phex double-mutant mice have an intermediate phenotype between wild-type and Phex mutant mice
To produce Galnt3/Phex double-mutant mice, Phex mutant mice carrying a nonsense mutation, K496X, were crossed to Galnt3 knockout mice, a murine model of hyperphosphatemic tumoral calcinosis.33 The double-mutant mice were born at the expected ratio (χ2 test p = 0.52), and the presence of both Galnt3 and Phex mutations had no effect on the survival of these mice up to 12 weeks. The presence of Galnt3 heterozygosity by itself had no major effects on the phenotype of the mice, and the mean values for various measurements were comparable between wild-type littermates and Galnt3 heterozygotes (Supplemental Figs. S1–S3). Therefore, phenotypic comparisons were limited to four main groups within each sex: wild-type, Galnt3 knockout (Galnt3−/−), Phex mutant (Phex−/Y and +/−), and double-mutant (Galnt3−/−, Phex−/Y, or Galnt3−/−, Phex+/−). Phex mutant mice had characteristic short, stubby tails (Fig. 1A). The tails of double-mutant mice were less stubby than Phex mutant littermates but shorter and thicker than wild-type and Galnt3 knockout littermates. Growth of Galnt3 knockout mice was comparable to that of wild-type littermates; however, Phex mutant mice (in Galnt3 wild-type background) had significant growth retardation from before weaning at 3 weeks old (Fig. 1B). Both male and female Galnt3/Phex double-mutant mice were also smaller than their respective wild-type controls. Sizes of double-mutant and Phex mutant mice were comparable at weaning but diverged after weaning, resulting in double mutants having an intermediate body weight between wild-type and Phex mutant mice. Interestingly, body weight of female double mutants was closer to that of wild-type, whereas male double-mutant mice were more similar to Phex mutant mice.
Serum biochemistries were measured in 4- and 12-week-old mice. Phex mutant mice, which were significantly smaller than wild-type and Galnt3 knockout mice, had lower creatinine and BUN levels, particularly at 12 weeks (Fig. 2). Similar to other murine models of XLH,35, 36Phex mutant mice used in this study had hypocalcemia, hypophosphatemia, and high alkaline phosphatase levels (Fig. 2). 1,25(OH)2D level was only elevated in male Phex mutant mice. As previously described,33Galnt3 knockout mice had an opposite biochemical phenotype including hyperphosphatemia, mild hypercalcemia, and low alkaline phosphatase concentrations. Although Galnt3 knockout mice had higher BUN and creatinine levels than Phex mutant mice, these levels were mostly normal compared with wild-type littermates (Fig. 2). Similar to Phex mutant mice, Galnt3/Phex double-mutant mice were hypophosphatemic with increased alkaline phosphatase levels; however, the phenotype of the double mutant was less severe than that of the Phex mutant (Fig. 2). Serum phosphorus in 12-week-old double-mutant mice were only 16% less than wild-type mice, compared with 43% in Phex mutants. At 4 weeks, the degree of the reduction was similar with 49% in Phex mutant and 40% in double-mutant mice; however, the difference between the two groups was still significant. Double-mutant mice also exhibited hypocalcemia observed in Phex mutant mice but only in 4-week-olds.
Although same-age males and females of the comparable genotypes had statistically significant differences in some biochemical parameters (most notably alkaline phosphatase for 12-week-old mice), overall trends between genotypes were essentially the same for male or female groups.
Galnt3/Phex double-mutant mice have higher Fgf23 gene expression than Phex mutant mice, despite less severe hypophosphatemia
To test for the presence of altered phosphate responsiveness in XLH, we compared serum biochemistries and femoral Fgf23 mRNA expression between wild-type, Phex mutant, Galnt3 knockout, as well as Galnt3/Phex double-mutant mice. Phex mutant mice had increased femoral Fgf23 expression, resulting in elevated serum intact Fgf23 levels (Fig. 3) and consequent hypophosphatemia. Consistent with our previous study,33Galnt3 knockout mice had significantly increased Fgf23 expression but lower circulating concentrations of intact Fgf23 than wild-type littermates (Fig. 3). As a result of the low intact Fgf23 concentration, Galnt3 knockout mice had hyperphosphatemia (Fig. 2).
Although the Galnt3/Phex double-mutant mice were still hypophosphatemic, their serum phosphorus level was significantly higher than Phex mutant mice and closer to normal because the double mutant had lower intact Fgf23 concentrations compared with the Phex mutant (Figs. 2 and 3). In the absence of the altered phosphate responsiveness, Fgf23 mRNA expression in the double mutant should remain similar to Phex mutant mice. Instead, despite the presence of the improved but still low phosphorus levels, the double-mutant mice upregulated Fgf23 gene expression even higher than Phex mutant mice (on average, 5- and 24-fold higher than 4- and 12-week-old Phex mutant mice, respectively). Of note, these levels were 57- or 113-fold higher than wild-type controls (Fig. 3).
Post-translational processing of Fgf23 is increased in Galnt3 knockout mice but reduced in Phex mutant mice
Circulating Fgf23 levels were also measured using a newly developed sandwich ELISA. Because this assay measures both biologically active intact Fgf23 protein and inactive C-terminal Fgf23 fragments, this measurement represents total amounts of Fgf23 produced in the bone. Thus, the observed patterns between the four genotypic groups were similar to those found in the Fgf23 expression in the bone (Fig. 3). Galnt3 knockout mice had significantly higher levels of total Fgf23; total Fgf23 levels were about 9- and 19-fold higher than 4- and 12-week-old normal mice, respectively. However, when the proportion of intact Fgf23 in circulation was determined in these mice, <2% of Fgf23 were intact (Fig. 3).
Although 4- and 12-week-old Phex mutant mice had on average 27- and 10-fold higher intact Fgf23 than wild-type mice, the increase was less pronounced in total Fgf23 measured by the new assay—only 7- and 6-fold higher than normal littermates (Fig. 3). To corroborate this finding, the proportion of intact Fgf23 in total Fgf23 was >40% higher in Phex mutant mice than wild-type mice. In 4-week-old Phex mutants, 63% of Fgf23 exist as intact, whereas almost all Fgf23 (97%) in 12-week-old Phex mutant mice were intact.
Consistent with the Fgf23 expression data, double-mutant mice had significantly higher total Fgf23 levels than Phex mutant and Galnt3 knockout mice. Double-mutant mice at 4 and 12 weeks increased overall production of Fgf23 to 32- and 55-fold higher than sex-matched, wild-type controls. Furthermore, the intact-to-total Fgf23 ratio revealed that compared with Galnt3 knockout mice, there was a small but significant increase in the proportion of intact Fgf23 in double-mutant mice at 4 weeks.
In light of data suggesting that the current therapy for XLH (ie, high-dose phosphate and calcitriol25) may actually aggravate FGF23 production,30, 31 we investigated whether bone cells in XLH have an abnormal response to circulating phosphate concentrations, using two murine models of disorders of phosphate homeostasis. Compared with Phex mutant mice, Galnt3/Phex double-mutant mice had phosphate levels that were closer to normal levels, although still hypophosphatemic. However, despite the fact that double-mutant mice had improved serum phosphate levels, they attempted to bring phosphate concentrations down to the level of Phex mutant mice by increasing Fgf23 expression and, to some extent, reducing proteolytic cleavage of intact Fgf23. This suggests that the mutation in the Phex gene alters the responsiveness of bone cells to extracellular phosphate concentrations. Although the improved phosphorus levels were likely the major stimulus for the increased Fgf23 expression, it is possible that the upregulation of Fgf23 expression was in part the result of inappropriately normal 1,25(OH)2D in double-mutant mice. Furthermore, the increased Fgf23 expression could result from the abundant C-terminal Fgf23 fragments in these mice, which have been implicated as an inhibitor of intact FGF23 that increases circulating phosphate levels.37 Another possibility for the increased Fgf23 expression in double-mutant mice is that hypophosphatemia acts as a “brake” on the system such that as the phosphate level increases, the “brake” suppressing Fgf23 expression is released. However, given the magnitude of the observed increase in Fgf23 expression, this is less likely.
The altered phosphate responsiveness in Phex mutant mice can be explained by at least three potential mechanisms. First, Fgf23-producing bone cells have a phosphate sensor (or sensing mechanism), which is analogous to the calcium-sensing receptor that controls production of parathyroid hormone.38Phex mutations may directly affect the phosphate-sensing ability of this sensor and lower the set point for extracellular phosphate concentrations. Thus, normalization of phosphate levels in Phex mutant mice would lead to enhanced Fgf23 production. Second, Fgf23 synthesis may be a secondary gene response to chronic changes in intracellular phosphate concentrations. Bone cells in Phex mutant mice may maintain intracellular levels of phosphate higher than normal even when extracellular concentration is low, leading to increased Fgf23 expression. Third, Phex itself may have a direct role in detecting changes in extracellular levels of phosphate. In this scenario, bone cells lacking normal Phex function may become more sensitive to extracellular phosphate and produce more Fgf23. Determining the exact mechanism by which Fgf23 is regulated in bone cells is important for our understanding of the pathogenesis of XLH, as well as other FGF23-mediated disorders.
The increased FGF23 levels in XLH (and Phex mutant mice) were believed to be a result of FGF23 upregulation. However, our data indicate that Phex mutant mice not only overexpressed the Fgf23 gene but also secreted more intact Fgf23 protein, likely because of reduced proteolytic cleavage of Fgf23. In agreement with our data, inactive Fgf23 fragments were also undetectable by western blot in another murine model of XLH (Hyp).39 These data indicate that decreased cleavage of Fgf23 is not dependent on the type of Phex mutation and is part of the pathophysiology in Phex mutant mice and, most likely, XLH patients as well. Furthermore, assays that detect both intact FGF23 and inactive fragments may underestimate the severity of FGF23 elevations in XLH patients because they are making a greater percentage of biologically active FGF23 compared with normal individuals.
The levels of mRNA expression and “total” Fgf23 protein had similar overall trends. However, the increase in Fgf23 mRNA in double-mutant mice was markedly higher than that of total Fgf23 protein measured by the new ELISA. This observation suggests that Fgf23 mRNA expression is so high that not all mRNA can be translated into protein. Alternatively, Fgf23 fragments may be degraded quickly such that these cannot be detected by the ELISA used in this study. However, this finding may simply reflect an inherent detection limit and sensitivity of quantitative PCR and ELISA. It should also be noted that the levels of both intact and total Fgf23, as well as the proportion of intact Fgf23 in circulation, change between the two ages studied as Fgf23 was generally higher in the 12-week-old mice.
The finding that Phex mutations alter responsiveness of bone cells to extracellular phosphate has major clinical implications. Although the current therapy of high-dose phosphate and calcitriol is effective in treatment of osteomalacia, our data suggest that in XLH the bone is actively trying to reduce serum phosphate levels by producing more FGF23. Thus, the current therapy may increase FGF23 production in XLH not only because both phosphate and calcitriol can stimulate FGF23 production in normal and pathological conditions,6, 26–29 but also because PHEX mutations cause the intrinsic defect in phosphate responsiveness of bone cells. Therefore, long-term use of the current therapy can lead to marked increases in FGF23 concentrations as suggested in recent studies30, 31 and potentially reduce the effectiveness of this therapy.
A few observations on the growth and serum biochemistries are worth noting. First, our initial characterization of Galnt3 knockout mice showed that heterozygous and homozygous males were smaller than wild-type littermates.33 However, in this study the presence of Galnt3 mutant alleles did not affect growth of the mice up to 12 weeks. This discrepancy may be attributable to the strain difference used in these two studies. The original study used F2 generations in mixed 129SvEv and C57BL/6J background, whereas the current study used F2 offspring from two congenic C57BL/6J strains. Second, Galnt3/Phex double-mutant mice had an intermediate phenotype between wild-type and Phex mutant mice. However, because O-glycosylation of Fgf23 by ppGalNAc-T3 plays a critical role in stability of Fgf23, it was conceivable that lack of Galnt3 would prevent secretion of most intact Fgf23 and cause hyperphosphatemia in double-mutant mice. Instead, the double-mutant mice had an opposite biochemical phenotype with increased intact Fgf23 concentration (although less than Phex mutant mice). This finding suggests that if Galnt3 knockout mice produce large amounts of Fgf23, they can override Galnt3 deficiency and produce more intact Fgf23 protein.
In conclusion, Phex mutations lead to intrinsic defects in the responsiveness of Fgf23-producing cells (osteoblasts and osteocytes) to phosphate levels. The impaired responsiveness in turn results in overexpression of the Fgf23 gene, as well as reduction in proteolytic cleavage of intact Fgf23. Therefore, elucidation of the mechanism for the abnormal phosphate responsiveness in these bone cells may open up a new therapeutic approach to control elevated FGF23 concentrations in XLH.
MJE receives royalties from and is a consultant for Kyowa Hakko Kirin Co. Ltd. All the other authors state that they have no conflicts of interest.
We thank Dr Erik A Imel at Indiana University (Indianapolis, IN, USA) for helpful suggestions on intact Fgf23 measurements and Jeffrey Lavigne at Immutopics International (San Clemente, CA, USA) for providing some mouse FGF-23 (C-Term) ELISA Kits.
This study was supported in part by Indiana University School of Medicine Biomedical Research Grant (SI), Showalter Research Trust Fund Grant (SI), and National Institutes of Health Grant R01 AR042228 (MJE). The preparation of this article was partially supported by a career development award to SI from the Indiana Clinical and Translational Sciences Institute funded in part by National Institutes of Health Grant RR025760.
Authors' roles: SI and MJE designed the study; SI, AMA, and AKG collected data; SI, AMA, and AKG analyzed data; SI and MJE interpreted data; SI drafted the manuscript; AMA, AKG, and MJE revised the manuscript content; SI, AMA, AKG, and MJE approved the final version of the manuscript. SI and MJE take responsibility for the integrity of the data analysis.