Cloning and characterization of FGF23 as a causative
factor of tumor-induced osteomalacia
Takashi Shimada*, Satoru Mizutani†, Takanori Muto*, Takashi Yoneya*, Rieko Hino*, Shu Takeda‡§,
Yasuhiro Takeuchi‡, Toshiro Fujita‡, Seiji Fukumoto¶, and Takeyoshi Yamashita*?
*Pharmaceutical Research Laboratory, Nephrology, Kirin Brewery Co. Ltd., 3 Miyahara, Takasaki, Gunma 370-1295, Japan;†Central Laboratories for Key
Technology, Kirin Brewery Co. Ltd., 1-13-5 Fukuura, Kanazawa-ku, Yokohama, Kanagawa 236-0004, Japan; and‡Division of Endocrinology,
Department of Medicine, University of Tokyo School of Medicine, and¶Department of Laboratory Medicine, University of Tokyo
Branch Hospital, 3-28-6 Mejirodai, Bunkyo-ku, Tokyo 112-8688, Japan
Edited by Maurice B. Burg, National Institutes of Health, Bethesda, MD, and approved March 6, 2001 (received for review November 16, 2000)
Tumor-induced osteomalacia (TIO) is one of the paraneoplastic
diseases characterized by hypophosphatemia caused by renal
phosphate wasting. Because removal of responsible tumors nor-
malizes phosphate metabolism, an unidentified humoral phospha-
turic factor is believed to be responsible for this syndrome. To
identify the causative factor of TIO, we obtained cDNA clones that
were abundantly expressed only in a tumor causing TIO and
one major contig, we cloned 2,270-bp cDNA, which turned out to
encode fibroblast growth factor 23 (FGF23). Administration of
recombinant FGF23 decreased serum phosphate in mice within
12 h. When Chinese hamster ovary cells stably expressing FGF23
were s.c. implanted into nude mice, hypophosphatemia with
increased renal phosphate clearance was observed. In addition, a
high level of serum alkaline phosphatase, low 1,25-dihydroxyvi-
tamin D, deformity of bone, and impairment of body weight gain
became evident. Histological examination showed marked in-
crease of osteoid and widening of growth plate. Thus, continuous
production of FGF23 reproduced clinical, biochemical, and histo-
logical features of TIO in vivo. Analyses for recombinant FGF23
products produced by Chinese hamster ovary cells indicated pro-
teolytic cleavage of FGF23 at the RXXR motif. Recent genetic study
indicates that missense mutations in this RXXR motif of FGF23 are
responsible for autosomal dominant hypophosphatemic rickets,
another hypophosphatemic disease with similar features to TIO.
We conclude that overproduction of FGF23 causes TIO, whereas
mutations in the FGF23 gene result in autosomal dominant hy-
pophosphatemic rickets possibly by preventing proteolytic cleav-
age and enhancing biological activity of FGF23.
Because removal of responsible tumors normalizes phosphate
metabolism, an unknown phosphaturic factor sometimes called
phosphatonin is believed to be responsible for this paraneoplas-
tic syndrome (1, 2). Although several groups have reported
inhibitory activity of renal phosphate transport in conditioned
media of tumor cells causing TIO (3–6), the responsible factor
for TIO has not been identified. Similar biochemical findings to
TIO also are observed in X-linked hypophosphatemic rickets?
osteomalacia (XLH), its murine homologue, Hyp, and autoso-
mal dominant hypophosphatemic rickets (ADHR) (7). In addi-
tion, several lines of evidence indicate that XLH and Hyp are
caused by a humoral mechanism (7–10). Therefore, it is possible
that TIO and XLH derive from a common or at least very similar
humoral factor(s). Thus, identification of this phosphaturic
factor causing TIO is indispensable for understanding normal
phosphate metabolism and pathogenesis of several hypophos-
phatemic diseases. In this report, we describe the cloning of a
humoral factor from a TIO tumor and show that this factor has
the ability to rapidly induce hypophosphatemia and reproduce
clinical, biochemical, and histological features of TIO in vivo.
umor-induced osteomalacia (TIO) is one of the hypophos-
phatemic diseases characterized by renal phosphate wasting.
Differential cDNA Screening of TIO and Adjacent Normal Bone Tissue.
We constructed a cDNA library with the Lambda ZAP II vector
from the frozen tumor tissue responsible for TIO of this patient
and performed a differential cDNA screening using two kinds of
probes specific for the tumor and the adjacent normal bone
tissue. Total RNAs (0.7 ?g each) from both tissues were used to
synthesize cDNA by the SMART cDNA synthesis kit (CLON-
TECH). We prepared tumor-specific and nontumor probes by
subtraction between both cDNAs by using the PCR-Select
cDNA subtraction system (CLONTECH). The probes were
labeled with alkaline phosphatase by the AlkPhos Direct system
(Amersham Pharmacia). Plaques generated from the TIO
cDNA library were lifted on two nylon membranes (Hybond N?;
Amersham Pharmacia). One membrane was hybridized with a
tumor-specific probe and the other with a probe specific for the
normal bone. Clones specifically hybridized only with a tumor-
specific probe were subjected to DNA sequencing.
Expression of Fibroblast Growth Factor 23 (FGF23). Full-length hu-
man FGF23 cDNA was amplified with F1EcoRI primer (5?-
taining EcoRI site at the 5? terminus and LhisNot primer
ATGATGGATGAACTTGGCGAA-3?) containing the se-
quence encoding the His6 tag and NotI site. The PCR product
was digested with EcoRI and NotI and ligated to pcDNA3.1zeo
(Invitrogen). A Chinese hamster ovary (CHO) cell line stably
producing FGF23 (CHO-FGF23) was cloned by limited dilution.
Northern Blot Analysis. Three kinds of human multiple tissue blots
(CLONTECH) derived from 23 tissues were used. Those tissues
were heart, brain, placenta, lung, liver, skeletal muscle, kidney,
pancreas, spleen, thymus, prostate, testis, ovary, small intestine,
colon, peripheral blood leukocyte, stomach, thyroid, spinal cord,
lymph node, trachea, adrenal gland, and bone marrow. The
amplified FGF23 cDNA fragment using F1 (5?-AGCCACTCA-
GAGCAGGGCAC-3?) and L1 (5?-CACGTTCAAGGGGTC-
This paper was submitted directly (Track II) to the PNAS office.
Abbreviations: TIO, tumor-induced osteomalacia; FGF, fibroblast growth factor; XLH,
X-linked hypophosphatemic rickets?osteomalacia; ADHR, autosomal dominant hypophos-
phatemic rickets; CHO, Chinese hamster ovary; 1?-OHase, 1?-hydroxylase; 1,25(OH)2D,
1,25-dihydroxyvitamin D; OK, opossum kidney; DMP-1, dentine matrix protein-1; MEPE,
matrix extracellular phosphoglycoprotein.
Data deposition: The sequence reported in this paper has been deposited in the GenBank
database (accession no. AB047858).
See commentary on page 5945.
§Present address: Department of Molecular and Human Genetics, Baylor College of Med-
icine, Houston, TX 77030.
?To whom reprint requests should be addressed. E-mail: firstname.lastname@example.org.
The publication costs of this article were defrayed in part by page charge payment. This
article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C.
§1734 solely to indicate this fact.
May 22, 2001 ?
vol. 98 ?
no. 11 www.pnas.org?cgi?doi?10.1073?pnas.101545198
CCGCT-3?) primers was labeled and used for hybridization.
Total RNAs were isolated from whole kidney with Isogen
(Nippon Gene, Toyama, Japan) and subjected to Northern
blotting for renal 25-hydroxyvitamin D 1?-hydroxylase (1?-
OHase). A 1,090-bp cDNA fragment for mouse 1?-OHase was
amplified by using 1AF (5?-CAGACAGAGACATCCGTG-
TAG-3?) and 1AR (5?-CCACATGGTCCAGGTTCAGTC-3?)
primers, labeled and used for hybridization. The filters were
washed to the stringency of 0.1? SSC with 0.1% of SDS at 50°C
for 40 min. A BAS imaging system (Fuji) was used for identifi-
cation of signals.
and L1 primers for FGF23. The PCR products were analyzed by
Southern blotting using the32P-labeled FGF23 cDNA fragment.
The filter was washed to the stringency of 0.1? SSC with 0.5%
of SDS at 65°C for 30 min.
Western Blot Analysis. Twenty microliters of conditioned medium
from CHO-FGF23 cells was resolved by 10–20% gradient SDS?
PAGE under reduced condition and electroblotted onto a
poly(vinylidene difluoride) membrane. The membrane was in-
cubated with anti-His (C-term)-horseradish peroxidase antibody
(Invitrogen), and signals were detected by the ECL system
Amino Acid Sequence. The recombinant proteins separated by
SDS?PAGE were electroblotted to a poly(vinylidene difluoride)
membrane and stained with Coomassie brilliant blue. The band
corresponding to each recombinant protein was dissected and
subjected to automated Edman degradation in gas phase protein
sequencer model 492A (Perkin–Elmer).
Recombinant Protein Preparation. Ten liters of conditioned me-
dium from CHO-FGF23 cells was filtrated through 0.2-?m
membrane (SuporCap, Pall Gelman Laboratory, Ann Arbor,
MI) at 4°C. The filtrate was applied to SP-Sepharose FF
(Amersham Pharmacia), and the column was washed with 50
mM sodium phosphate buffer, pH 6.7. The retained proteins
were eluted with a linear gradient of NaCl ranging from 0 to 0.7
M in a buffer of 50 mM sodium phosphate, pH 6.7. The mature
full-length polypeptide was collected at ?0.3 M NaCl. Purified
mature recombinant FGF23 was concentrated into a buffer
consisting of 5 mM Hepes and 0.1 M NaCl, pH 6.9.
Animals and Experimental Designs. Approximately 4 ?g of recom-
mice (SLC, Hamamatsu, Japan). To examine the long-term
in 0.1 ml of PBS were s.c. implanted into both sides of backs of
6-week-old male BALB?c athymic nude mice (SLC). The con-
trol group was injected with 0.1 ml of PBS as well. Tumor volume
was calculated according to the formula of 1?2 ? (long diam-
eter) ? (short diameter)2. The calculated result by this formula
was proved to correlate well with the volume measured by water
bred in metabolic cages for 24 h. Blood samples were taken
under anesthesia with diethyl ether. All animals received a
commercial rodent diet (CE-2; CLEA Japan, Osaka) containing
1.1% phosphate and 1.0% calcium. Diets and tap water were
provided ad libitum throughout the experiments. All experi-
ments were reviewed and approved by the institutional animal
care and use committee at the Pharmaceutical Research Labo-
ratory, Kirin Brewery and were conducted in the specific patho-
gen-free area where no infectious problem has been reported.
Measurement of 1,25-Dihydroxyvitamin D [1,25(OH)2D] in Serum.
Serum 1,25(OH)2D was measured by RIA (Immunodiagnostic
Systems, Boldon, U.K.). Because 500 ?l of serum was required
for one measurement, 250 ?l of serum from individual animals
in the same groups (n ? 5) was pooled and subjected to duplicate
measurements. The detection limit was 5 pmol?liter.
imaging system (?FX-1000; Fuji). Exposure of imaging plates
was conducted at 25 kV, 0.1 mA. Exposure period was 10 sec for
the whole body and 5 sec for isolated femurs. Exposed plates
were analyzed by a BAS imaging system.
Histological Examination. Freshly isolated tibiae were fixed and
subjected to the Villanueva bone stain. The tissues were em-
bedded in methyl methacrylate resin (Wako Pure Chemical,
Osaka), and undecalcified sections with 4-?m thickness were
prepared with the microtome (Supercut RM2065; Leica, Solms,
Germany). Villanueva-Goldner counterstain for these sections
In Vitro Transport Assay and Measurement of cAMP Production.
Opossum kidney (OK) cells were obtained from the American
Type Culture Collection (CRL-1840). Phosphate, glucose, and
alanine transport were measured separately at 0.1 mM solute
concentration with [32P] dibasic potassium phosphate, [14C]
D-glucose, or [3H] L-alanine in the uptake solution containing
137 mM NaCl, 5.4 mM KCl, 2.8 mM CaCl2, 1.2 mM MgSO4, and
10 mM Hepes, pH 7.4 (13). Na-independent uptake was mea-
sured with Na-free uptake solution in which NaCl was replaced
by choline chloride. To examine cAMP production, OK cells
were incubated with FGF23 for 10 min in the presence of 1 mM
3-isobuthyl-1-methylxanthine. cAMP in cells was extracted and
measured by EIA (Amersham Pharmacia).
Statistical Analyses. All measurements were expressed as mean ?
SEM. Statistical significance was evaluated by either Student’s t
test or one-way ANOVA followed by Dunnett’s test for com-
parison of multiple means. A P value less than 0.05 was consid-
ered to be significant.
Cloning of cDNAs Specifically Expressed in a Tumor Causing TIO. To
isolate a putative phosphaturic factor, differential cDNA screen-
ing was carried out. Of ?320,000 clones, 456 clones dominantly
expressed in the tumor were selected and sequenced. All se-
quences obtained were compared with each other and analyzed
against the GenBank database. Consequently, each clone was
classified into contigs, which were aligned by the number of
consisting clones to reflect the frequency of appearance in the
screening. The most frequently detected contig was dentine
matrix protein-1 (DMP-1) followed by heat shock protein-90,
osteopontin, and a novel sequence contig named OST311 that
was composed of nine different cDNA clones (Table 1). We also
obtained a contig termed OST190 that has been recently iden-
tified as a novel molecule, matrix extracellular phosphoglyco-
protein (MEPE), produced by TIO tumors (14). The longest
cDNA clone for OST311 contained 2,270-bp nucleic acids and
had an ORF encoding 251-aa residues with putative molecular
mass of 28 kDa (Fig. 1A). The sequence coding OST311 protein
derived from three exons spanning more than 10 kbp on 12p13.3
where the responsible gene for ADHR exists (15). Deduced
amino acid sequence shows that OST311 protein has a signal
peptide sequence at its N terminus and has a homology to FGF
family members, especially in the middle portion of the protein.
OST311 protein also has a unique C-terminal portion charac-
terized by the presence of several potential phosphorylation sites
Shimada et al.PNAS ?
May 22, 2001 ?
vol. 98 ?
no. 11 ?
by casein kinase. Recent reports indicated that OST311 is
identical with FGF23 (16, 17).
Expression of FGF23. Expression of FGF23 could not be detected
in all 23 human tissues by Northern blot analysis using about 2
?g of poly(A)?RNA even after 24 h exposure to the imaging
plate (data not shown). Abundant expression of FGF23 was
observed in the tumor tissue responsible for TIO by reverse
transcriptase–PCR followed by Southern blot analysis (Fig. 1B).
Weak expression of FGF23 also was detected in liver, lymph
node, thymus, and heart. The expression of FGF23 could not be
observed in other tissues, including kidney, even by this method.
These results indicate that a few normal tissues can express a
small amount of FGF23, but a tumor responsible for TIO
abundantly expresses FGF23.
Preparation and Analysis of Recombinant FGF23. To investigate the
biological activity of FGF23, we prepared recombinant CHO-
FGF23 cells. Western blotting detecting the C terminus of
recombinant proteins revealed that conditioned media con-
tained two recombinant products with molecular masses of ?30
and 10 kDa (Fig. 1C). Amino acid sequencing showed the larger
product was a mature protein lacking signal sequence, which is
composed of the first 24 aa residues of FGF23. The smaller
product had180Ser residue at its N terminus. The preceding
amino acid sequence of
agrees with the consensus proteolytic cleavage sequence of
RXXR (18). The proteolytic processing also was supported by
the detection of about a 16-kDa N-terminal fragment using a
polyclonal antibody raised against the synthetic peptide corre-
sponding to amino acids48Arg to67Gln (data not shown).
Biological Activity of FGF23. Recombinant mature FGF23 was
injected into mice three times with intervals of 5 h, and blood
samples were collected at 12 and 24 h after the initial injection.
Administration of mature FGF23 reduced serum phosphate
significantly at both times without affecting serum calcium (Fig.
urine samples were collected. Serum was obtained just after the
urine collection. In this experiment, treatment with FGF23
reproduced hypophosphatemia and increased renal phosphate
excretion (P ? 0.001) (Table 2). These hypophosphatemic and
phosphaturic effects were reproduced in two other experiments
(data not shown). These data indicate that reduced renal phos-
phate reabsorption by FGF23 at least in part contributed to the
development of hypophosphatemia in those animals. Serum and
urinary levels of calcium and other solutes did not change with
FGF23 (Table 2). Urinary excretion of measurable 11 kinds of
amino acids was determined. Because the absolute value of each
amino acid was variable, we calculated the ratio of amino acid
excretion in the FGF23 group to that in the control group. As
shown in Table 2, FGF23 did not increase renal excretion of
amino acids. Furthermore, we did not observe increased urinary
cAMP levels in FGF23-treated mice. Thus, FGF23 specifically
reduced renal reabsorption and decreased the serum level of
To further investigate the biological function of FGF23 and its
Table 1. The tumor-specific contigs by frequency of appearance
frequency Gene name
Heat shock protein-90
Translationally controlled tumor protein
The most frequently cloned 10 contigs in the differential screening are
The gene names were identified by comparing with the sequence data in
did not hit any data in Genbank by the BLAST homology search at that time
and MEPE, respectively.
sequence is underlined. The consensus proteolytic cleavage site is boxed. (B)
Expression profile of FGF23 in adult normal tissues analyzed by reverse tran-
scriptase–PCR followed by Southern blotting. Template cDNAs are as follows:
lanes 1, bone marrow; 2, brain; 3, colon; 4, heart; 5, kidney; 6, leukocyte; 7,
liver; 8, lung; 9, lymph nude; 10, muscle; 11, ovary; 12, pancreas; 13, placenta;
TIO tumor. (C) Western blotting with anti-His (c-term) antibody that recog-
nizes the carboxyl-terminal tag sequence of recombinant FGF23 proteins
secreted into media by CHO-FGF23 cells.
(A) Amino acid sequence of human FGF23. The signal peptide
and calcium (B) concentrations in mice treated with purified recombinant
protein (n ? 6; filled columns) or vehicle (n ? 6; open columns). FGF23 was i.p.
injected three times with intervals of 5 h. Blood samples were collected at 12
and 24 h after the first injection. Each column represents the mean ? SEM.*,
P ? 0.05;**, P ? 0.001 by Student’s t test.
Effects of recombinant FGF23. The changes of serum phosphate (A)
www.pnas.org?cgi?doi?10.1073?pnas.101545198Shimada et al.
role in the development of TIO, we used the tumor-bearing nude
mouse system. To examine the time course of biochemical and
phenotypic changes, six mice in each group were killed sequen-
tially at indicated time points. Mice with CHO-FGF23 cells
showed hypophosphatemia as expected from the experiments of
recombinant protein injection. This hypophosphatemia was ac-
companied by inappropriately increased renal phosphate clear-
ance after 6 days of implantation (Fig. 3 A and B). In addition,
serum alkaline phosphatase activity increased in CHO-FGF23
mice as in patients with TIO (Fig. 3C). Furthermore, inappro-
priately reduced level of 1,25(OH)2D in the presence of hy-
pophosphatemia is reported in TIO patients (1, 2). In agreement
with these observations, the serum level of 1,25(OH)2D in
CHO-FGF23 mice was reduced after 6 days of tumor cell
implantation compared with that in control and wild-type CHO
14.2 pmol?liter, respectively). To confirm the abnormal vitamin
D metabolism, we examined the expression level of renal 1?-
OHase by Northern blot analysis using kidneys isolated on day
3 after implantation. The decreased expression of 1?-OHase in
CHO-FGF23 mice was observed (Fig. 4). Mice with CHO-
FGF23 cells showed growth retardation after 2 weeks, although
the amount of food intake of CHO-FGF23 mice on day 45
(5.85 ? 0.62 g) was comparable to those of vehicle-treated
Fig. 5A shows the typical appearance of CHO-FGF23 mice,
which is characterized by leanness and a round back. All mice in
the CHO-FGF23 group, but no mice in other groups, showed
such a phenotype at 45 days after the implantation. Soft x-ray
images showed that bone mineral content of a mouse bearing
CHO-FGF23 cells was markedly reduced compared with a
mouse bearing CHO cells (Fig. 5B). Distortion of rib cage also
was observed in a mouse bearing CHO-FGF23 cells. Bone
bearing CHO-FGF23 was remarkably reduced compared with
that of control mice (Fig. 5C). The ratio of ash weight to dry
weight of femurs in CHO-FGF23 mice was significantly de-
creased compared with that of wild-type CHO mice (0.534 ?
0.007 vs. 0.608 ? 0.004, n ? 5, P ? 0.001). Fig. 5D shows the
histological appearance of tibial proximal metaphysis. There was
a prominent increase of osteoid and widening of growth plate in
a mouse with CHO-FGF23 cells. Osteomalacia and muscle
weakness are typical phenotypes of TIO. In addition, growth
retardation and skeletal deformities are reported in younger
patients (1). Thus, we could reproduce biochemical, clinical, and
histological features of TIO by implanting CHO-FGF23 cells
into nude mice. These biochemical and phenotypic changes by
Table 2. Effects of recombinant FGF23 on serum and urinary
Ca (? 10?2dl?mg)
Glucose (? 10?3dl?mg)
Na (? 10?2dl?mg)
Cl (? 10?2dl?mg)
Amino acid (T?C ratio)
9.55 ? 0.21
9.86 ? 0.14
178.8 ? 5.4
152.0 ? 0.9
5.33 ? 0.18
111.9 ? 0.6
7.85 ? 0.26
9.68 ? 0.10
181.4 ? 8.5
151.2 ? 0.6
5.30 ? 0.18
112.7 ? 0.4
P ? 0.001
1.38 ? 0.06
1.50 ? 0.05
2.45 ? 0.14
4.61 ? 0.30
2.37 ? 0.21
9.17 ? 0.39
1.0 ? 0
83.3 ? 10.7
1.77 ? 0.05
1.55 ? 0.07
2.85 ? 0.24
4.81 ? 0.16
2.23 ? 0.10
9.42 ? 0.31
1.04 ? 0.04
60.2 ? 7.1
P ? 0.001
Purified recombinant FGF23 (n ? 12) or vehicle (n ? 12) was injected three
times as shown in Methods. All biochemical markers except for amino acids
were measured in each mouse. Renal excretion was evaluated by the follow-
ing formula, urinary concentration divided by serum concentration and uri-
(Thr, Asn, Glu, Gln, Pro, Gly, Ala, Phe, His, Lys, and Arg) were determined by
HPLC. The ratio of each amino acid excretion in FGF23-treated mice to that of
control mice (T?C ratio) was calculated. The values for Gly, Ala, and Lys, which
are commonly excreted in aminoacidurias, were 1.08, 0.99 and 0.99, respec-
tively. The average of the ratios from 11 amino acids is shown. Results
represent the mean ? SEM. P ? 0.001 by Student’s t test.
difference from control group by one-way ANOVA followed by Dunnett’s multiple comparison test;*, P ? 0.05;**, P ? 0.01;***, P ? 0.001.
Biochemical changes in mice transplanted with CHO cells producing FGF23. Time course of changes of serum phosphate level (A), renal phosphate
CHO-FGF23 cells. Vehicle-treated mice (n ? 3), control CHO-implanted mice,
RNAs (20 ?g each) were used for the determination of expression level of
1?-OHase by Northern blot analysis. Expression of glyceraldehyde-3-
phosphate dehydrogenase (G3PDH) also was determined by using the same
Expression of renal 1?- OHase in mice implanted with CHO cells or
Shimada et al.PNAS ?
May 22, 2001 ?
vol. 98 ?
no. 11 ?
CHO-FGF23 cells were confirmed in other independent
In Vitro Activity of FGF23. Because several reports showed the
inhibitory activity on phosphate transport in conditioned media
of tumor cells causing TIO (3–6), we evaluated in vitro activity
of recombinant FGF23 in OK cells. However, although phos-
phonoformic acid, a specific inhibitor for Na-dependent phos-
phate transport, clearly inhibited phosphate transport, FGF23
did not inhibit transport of phosphate (Table 3). Transport of
glucose and alanine did not change by FGF23, either. In
addition, FGF23 did not increase cAMP production in OK cells
Causative factors of TIO have been sought for a long time. Using
OST311, which turned out to be identical to FGF23, were picked
factor of TIO, we produced recombinant cells that secrete these
candidate gene products. Then we examined the in vivo effects
of these gene products so that we could identify the causative
factor for TIO even if it does not work on renal tubular cells
directly. We first evaluated the most frequently cloned DMP-1
by the tumor-bearing system shown here. However, mice with
CHO cells secreting DMP-1 showed neither hypophosphatemia
nor increased renal phosphate clearance. Mice with CHO cells
secreting MEPE did not show hypophosphatemia either (data
not shown). The abundant expression of MEPE was demon-
strated in TIO tumors (14), and DMP-1 is a candidate gene for
dentinogenesis imperfecta (19). Both DMP-1 and MEPE are
members of matrix proteins that may have some role in calcifi-
cation. However, our results show that these proteins are not
causative factors of TIO. We could observe hypophosphatemia
only with CHO cells secreting FGF23.
To prove that FGF23 induces hypophosphatemia, we purified
recombinant FGF23 protein and administered it into normal
mice (Fig. 2). Furthermore, we examined the time course of
changes of biochemical markers in tumor-bearing nude mice
(Fig. 3). Administration of recombinant protein induced a
reproducible and significant decrease of serum phosphate with
increased renal phosphate excretion. Long-term exposure to
FGF23 induced severer hypophosphatemia, osteomalacia, and
inappropriately low 1,25(OH)2D level with decreased 1?-OHase
expression. Although the magnitude of the decrease of serum
phosphate by recombinant FGF23 was smaller than that in nude
mice bearing CHO-FGF23 cells, this disparity may be caused by
the difference in serum concentration of FGF23 or the exposed
period to FGF23 in these systems.
It is possible that FGF23 has a stimulatory effect on tumor
growth and large tumor burden indirectly modified phosphate
and bone metabolism in mice with CHO-FGF23. However, the
tumor burden was not statistically different between CHO-
FGF23 mice and wild-type CHO mice at 45 days after the
implantation (0.74 ? 0.15 vs. 0.99 ? 0.41). In addition, neither
CHO cells by recombinant FGF23 was observed (data not
shown). Furthermore, serum phosphate was clearly decreased in
mice with CHO-FGF23 cells at 6 days after the implantation
Whole skeletal soft roentgenogram of a mouse bearing wild-type CHO cell
tumors (Left) and a mouse bearing CHO-FGF23 cell tumors (Right) on day 45
after implantation. (C) Soft roentgenogram of femurs isolated from control
mice (Left) and mice bearing CHO-FGF23 cell tumors (Right). (D) Histological
a mouse bearing CHO-FGF23 (Lower) on day 45. The undecalcified tissue
sections were subjected to the Villanueva bone stain and the Villanueva-
Goldner stain to discriminate mineralized bone tissues (green) from osteoid
Table 3. Effects of FGF23 on Na-dependent transport of phosphate, glucose, and alanine, and
cAMP production in OK cell cultures
Na-dependent transport, nmol?mg per 6 min
cAMP, nmol?mg PhosphateGlucose Alanine
PFA (10 mM)
FGF23 (4 ng?ml)
FGF23 (40 ng?ml)
FGF23 (400 ng?ml)
2.09 ? 0.08
0.62 ? 0.05**
2.08 ? 0.06
2.05 ? 0.07
2.04 ? 0.04
0.81 ? 0.03
0.83 ? 0.05
0.79 ? 0.03
0.78 ? 0.03
3.33 ? 0.08
3.40 ? 0.05
3.33 ? 0.09
3.32 ? 0.10
12.3 ? 1.0
13.5 ? 3.1
OK cells were exposed to FGF23 for 20 h and then subjected to the transport assay (n ? 4). Phosphonoformic
acid (PFA), a specific inhibitor for Na-dependent phosphate transport, was added to an assay buffer. Cytosolic
www.pnas.org?cgi?doi?10.1073?pnas.101545198 Shimada et al.
when tumors were barely detectable. These results indicate that Download full-text
tumor burden cannot account for the abnormal phosphate and
bone metabolism in mice with CHO-FGF23.
To clarify the mechanism of actions of FGF23 on phosphate
metabolism, we examined the effects of recombinant FGF23 on
sodium-dependent phosphate transport in OK cells. In contrast
to previous reports (3–6), we did not observe a clear inhibitory
activity of FGF23. There are several possibilities that can explain
these results. For example, FGF23 may not act on renal cells
directly, but affects renal phosphate reabsorption indirectly in
vivo. It is also possible that FGF23 needs some modification in
vivo to exert its effect on phosphate metabolism. Because actual
target cells of FGF23 are unknown at the moment, these
possibilities should be examined in future experiments.
It is interesting to note that the responsible gene for ADHR
has been recently identified and named FGF23 (17). The
missense mutations at176Arg and179Arg are responsible for
ADHR. These amino acids are in the consensus proteolytic
cleavage sequence of RXXR. Lack of expression of FGF23 in
kidney (Fig. 1B) indicates that ADHR also is caused by a
humoral mechanism. Because recombinant FGF23 produced by
CHO cells was partially cleaved between179Arg and180Ser (Fig.
2A), it is possible that mutations at176Arg and179Arg prevent
proteolytic cleavage. Taking into account our present data,
mutations of the FGF23 gene in ADHR are considered to result
in the enhancement of biological activities of FGF23 as a
The involvement of an unknown humoral factor(s) also is
proposed for XLH and its murine homologue, Hyp (4, 5). Even
though XLH?Hyp is caused by mutations in the putative endo-
peptidase, a phosphate-regulating gene with homologies to
endopeptidases on the X-chromosome (PHEX?Phex) (20), the
mechanism by which mutations of PHEX?Phex cause phospha-
turia is unclear. It has been postulated that PHEX?Phex protein
inactivates a putative phosphaturic factor, phosphatonin, by its
proteolytic activity and maintains serum phosphate level (2). It
is necessary to investigate whether FGF23 interacts with PHEX
protein. Further studies are also necessary to clarify the rela-
tionship between FGF23 and a putative humoral factor causing
XLH?Hyp. Because both the overproduction and the missense
mutations of FGF23 cause hypophosphatemia with renal phos-
phate wasting, we conclude that FGF23 is at least one of the
causative factors of TIO and is an important regulator of
phosphate and bone metabolism. A recent report showing
overexpression of FGF23 in several tumors causing TIO also
supports this notion (21). Elucidation of the detailed actions of
FGF23 is indispensable for understanding normal phosphate
metabolism and will lead to the development of novel diagnostic
and therapeutic approaches for several diseases with abnormal
We thank Prof. T. J. Martin for valuable comments on the manuscript
and Dr. Shinichiro Kato for helpful discussions. This work was supported
in part by grants from the Ministry of Education, Culture, Sports,
Science, and Technology, the Ministry of Health, Labor, and Welfare,
1. Drezner, M. K. (1999) in Primer on Metabolic Bone Diseases and Disorders of
Mineral Metabolism, ed. Favus, M. J. (Lippincott, Philadelphia), pp. 331–337.
2. Kumar, R. (2000) Bone 27, 333–338.
3. Cai, Q., Hodgson, S. F., Kao, P. C., Lennon, V. A., Klee, G. G., Zinsmiester,
A. R. & Kumar, R. (1994) N. Engl. J. Med. 330, 1645–1649.
4. Wilkins, G. E., Granleese, S., Hegele, R. G., Holden, J., Anderson, D. W. &
Bondy, G. P. (1995) J. Clin. Endocrinol. Metab. 80, 1628–1634.
5. Nelson, A. E., Namkung, H. J., Patava, J., Wilkinson, M. R., Chang, A. C.,
Reddel, R. R., Robinson, B. G. & Mason, R. S. (1996) Mol. Cell. Endocrinol.
6. Rowe, P. S., Ong, A. C., Cockerill, F. J., Goulding, J. N. & Hewison, M. (1996)
Bone 18, 159–169.
7. Econs, M. J. (1999) Bone 25, 131–135.
8. Morgan, J. M., Hawley, W. L., Chenoweth, A. I., Retan, W. J. & Diethelm,
A. G. (1974) Arch. Intern. Med. 134, 549–552.
9. Meyer, R. A., Jr., Meyer, M. H. & Gray, R. W. (1989) J. Bone Miner. Res. 4,
11. Fukumoto, S., Takeuchi, Y., Nagano, A. & Fujita, T. (1999) Bone 25,
13. Biber, J., Malmstrom, K., Reshkin, S. & Murer, H. (1991) Methods Enzymol.
14. Rowe, P. S. N., De Zoysa, P. A., Dong, R., Wang, H. R., White, K. E., Econs,
M. J. & Oudet, C. L. (2000) Genomics 67, 54–68.
15. Econs, M. J., McEnery, P. T., Lennon, F. & Speer, M. C. (1997) J. Clin. Invest.
16. Yamashita, T., Yoshioka, M. & Itoh, N. (2000) Biochem. Biophys. Res.
Commun. 277, 494–498.
17. White, K. E., Evans, W. E., O’Riordan, J. L. H., Speer, M. C., Econs, M. J.,
Lorenz, B., Grabowski, M., Meitinger, T. & Strom, T. M. (2000) Nat. Genet.
18. van de Loo, J. W., Creemers, J. W., Bright, N. A., Young, B. D., Roebroek, A. J.
& Van de Ven, W. J. (1997) J. Biol. Chem. 272, 27116–27123.
19. MacDougall, M., Gu, T. T. & Simmons, D. (1996) Connect. Tissue Res. 35,
20. The Hyp Consortium (1995) Nat. Genet. 11, 130–136.
J. Clin. Endocrinol. Metab. 86, 497–500.
Shimada et al. PNAS ?
May 22, 2001 ?
vol. 98 ?
no. 11 ?