Exome sequencing reveals FAM20c mutations associated with fibroblast growth factor 23–related hypophosphatemia, dental anomalies, and ectopic calcification

Authors

  • Silje Hjorth Rafaelsen MD,

    Corresponding author
    1. Center for Medical Genetics and Molecular Medicine, Haukeland University Hospital, Bergen, Norway
    • Department of Clinical Medicine, University of Bergen, Bergen, Norway
    Search for more papers by this author
  • Helge Ræder,

    1. Department of Clinical Medicine, University of Bergen, Bergen, Norway
    2. Department of Pediatrics, Haukeland University Hospital, Bergen, Norway
    Search for more papers by this author
  • Anne Kristine Fagerheim,

    1. Medical Department, Nordland County Hospital, Bodø, Norway
    Search for more papers by this author
  • Per Knappskog,

    1. Department of Clinical Medicine, University of Bergen, Bergen, Norway
    2. Center for Medical Genetics and Molecular Medicine, Haukeland University Hospital, Bergen, Norway
    Search for more papers by this author
  • Thomas O Carpenter,

    1. Departments of Pediatrics (Endocrinology) and Orthopaedics and Rehabilitation, Yale University School of Medicine, New Haven, CT, USA
    Search for more papers by this author
  • Stefan Johansson,

    1. Center for Medical Genetics and Molecular Medicine, Haukeland University Hospital, Bergen, Norway
    2. Department of Biomedicine, University of Bergen, Bergen, Norway
    Search for more papers by this author
  • Robert Bjerknes

    1. Department of Clinical Medicine, University of Bergen, Bergen, Norway
    2. Department of Pediatrics, Haukeland University Hospital, Bergen, Norway
    Search for more papers by this author

Address correspondence to: Department of Clinical Medicine, Section for Pediatrics, University of Bergen, Haukeland University Hospital, N-5021 Bergen, Norway. E-mail: silje.rafaelsen@pedi.uib.no

ABSTRACT

Fibroblast growth factor 23 (FGF23) plays a crucial role in renal phosphate regulation, exemplified by the causal role of PHEX and DMP1 mutations in X-linked hypophosphatemic rickets and autosomal recessive rickets type 1, respectively. Using whole exome sequencing we identified compound heterozygous mutations in family with sequence similarity 20, member C (FAM20C) in two siblings referred for hypophosphatemia and severe dental demineralization disease. FAM20C mutations were not found in other undiagnosed probands of a national Norwegian population of familial hypophosphatemia. Our results demonstrate that mutations in FAM20C provide a putative new mechanism in human subjects leading to dysregulated FGF23 levels, hypophosphatemia, hyperphosphaturia, dental anomalies, intracerebral calcifications and osteosclerosis of the long bones in the absence of rickets.

Introduction

Over the last decade, our understanding of human phosphate metabolism, including genes and mechanisms involved, has greatly increased. Of particular importance is the characterization of a bone-kidney signaling pathway that includes a new class of phosphate regulating molecules; i.e., the phosphatonins.[1] Fibroblast growth factor 23 (FGF23) has been recognized as a key phosphate regulating hormone in this pathway, inhibiting the renal reabsorption of phosphate through regulation of sodium-phosphate co-transporters in proximal tubule cells, and downregulating the 1-alpha-hydroxylase responsible for the conversion of 25-hydroxy vitamin D (25(OH)D) to 1,25-dihydroxyvitamin D (1,25(OH)2D).[2] The net effect is a reduction of body phosphate in response to increasing levels of serum phosphate, 1,25(OH)2D or parathyroid hormone (PTH).[3]

Elevated levels of FGF23 result in phosphaturia, hypophosphatemia, and low or inappropriately normal levels of 1,25(OH)2D as occurs in various monogenic renal wasting conditions such as X-linked hypophosphatemic rickets (XLH) caused by PHEX mutations,[4] autosomal dominant hypophosphatemic rickets (ADHR) caused by FGF23 mutations,[5] and autosomal recessive hypophosphatemic rickets (ARHR) type 1 caused by DMP1 mutations.[6] A recent work described a second form of ARHR, ARHR type 2, caused by ENPP1 mutations, associated with arterial calcification of infancy and elevated FGF23 levels.[7] The clinical spectrum of FGF23-dependent disorders and our understanding of their genetic basis is continuously expanding,[8, 9] allowing for increased insight into the pathophysiological mechanisms involved in hypophosphatemic conditions.

In this study, we investigated a family with an autosomal recessive pattern of hypophosphatemia, hyperphosphaturia, dental anomalies, intracerebral calcifications, and osteosclerosis of the long bones. Using whole-exome sequencing, we aimed to find the genetic explanation and assess the relationship of the findings to the serum levels of FGF23.

Patients and Methods

Subjects

The family was recruited from a national population-based cohort of familial hypophosphatemia in Norway and included two affected brothers (Fig. 1). A detailed description of their clinical history is given in the supplementary material (Clinical report). Written informed consent was obtained from all study participants. The study was approved by the Regional Committee for Medical and Health Research Ethics, Region West, Norway (REK number 2009/1140).

Figure 1.

(A) Family pedigree. (B) Picture of the face and hands of sibling 1 and 2 and their unaffected father. Note the high forehead, flat, broad nasal bridge, hypotelorism, flat malar area, and short distal phalanges in the brothers, as compared to their father.

Biochemical parameters

Blood samples were collected after an overnight fast. Circulating levels of Ca, albumin, phosphate, and alkaline phosphatase (ALP) activity in serum were analyzed using the Modular P-system from Roche Diagnostics (Basel, Switzerland). PTH was measured with a two-site chemiluminescent enzyme-labeled immunometric assay for intact PTH (Immulite 2000; Siemens Healthcare Diagnostics, Deerfield, IL, USA). Measurement of 25(OH)D levels was performed using an in-house–developed liquid chromatography double mass spectrometry (LC-MS/MS) method.[10] A spot sample of urine collected at the time of blood sampling was analyzed for calcium, phosphorus, and creatinine. The maximal tubular reabsorption of phosphate per glomerular filtration rate (TmP/GFR) was calculated according to the algorithm based on the nomogram of Walton and Bijvoet.[11, 12] Plasma intact FGF23 was measured with the FGF23-ELISA kit (Kainos Laboratories, Japan), with a lower detection limit of 3 pg/mL and a intraassay coefficient of variation (CV) of 5.8%, and inter assay CV of 7.5%.

Genetic analysis

Genomic DNA was purified from blood using the QiaSymphony system (Qiagen, Hilden, Germany). Whole-genome single-nucleotide polymorphism (SNP) genotyping was performed with Genome-Wide Human SNP 6.0 array (Affymetrix, Santa Clara, CA, USA). Whole-exome capture using Roche-Nimblegen's SeqCap EZ Exome v2 and sequencing on the Illumina HiSeq was performed at the HudsonAlpha Institute for Biotechnology (Huntsville, AL, USA) to a median coverage of 154× according to the manufacturer's protocol.[13] The paired-end 100-nt reads were analyzed with CASAVA v1.8 (Illumina Inc., San Diego, CA, USA) followed by alignment with Burrows-Wheeler transformation.[14] PCR duplicates were removed with PICARD (http://picard.sourceforge.net) and the Genome analysis toolkit (GATK) was used for base quality recalibration.[15] SNPs and indels were called using the GATK unified genotyper using a minimum threshold of 8× sequencing depth and quality score ≥30. ANNOVAR[16] and in-house scripts were used for variant annotation.

Sanger sequencing of FAM20C

Exon 3 and 4 of family with sequence similarity 20, member C (FAM20C), including the intron-exon boundaries, were sequenced in all five members of the family.

In short, DNA targets were first amplified by PCR (list of primers available upon request) using the AmpliTaq Gold DNA polymerase system (Applied Biosystems, Life Biosystems, Carlsbad, CA, USA). PCR amplicons were purified with 2 µL of ExoSapIT. Using the Big Dye Terminator, chemistry sequencing was performed on the 3730 DNA analyzer (Applied Biosystems) and analyzed using the SeqScape software (Applied Biosystems).

All exons and intron/exon boundaries of FAM20C were sequenced in an additional five Norwegian subjects with familial hypophosphatemia, for which mutational analysis of the PHEX, DMP1, FGF23, ENPP1, and Klotho genes had not revealed the underlying genetic cause. In addition, we sequenced FAM20C in five PHEX-negative subjects with hypophosphatemia and variable degrees of osteosclerosis, from the Yale Center for XLH.

Results

Clinical studies

We investigated a family in which the affected individuals displayed a pattern of autosomal recessive hypophosphatemia and hyperphosphaturia (Fig. 1, Table 1), in whom tooth decay was evident by approximately 18 months of age, and in whom osteosclerosis rather than rickets was later found in radiographs of long bones. After puberty, the degree of renal phosphate wasting gradually diminished such that by 18 to 20 years of age, the affected brothers had circulating phosphate levels in the lower normal range and urine phosphate excretion in the higher normal range (Fig. 2, Table S1). In addition, the subjects displayed ectopic calcifications in the brain, as well as mild facial and acral dysmorphic features (see Supplementary Fig. S1 and the Supplementary Material/Clinical report for a detailed clinical description). Previous genetic studies had revealed no chromosomal disorders or mutations in the PHEX, FGF23, or DMP1. The father, the mother, and the half-brother had normal serum and urine phosphate levels (Table 1), normal teeth, and no dysmorphic features.

Table 1. Biochemical Parameters at Inclusion
ParameterNormal rangeUnitSibling 1 (18 years old)Sibling 2 (16 years old)Mother (48 years old)Father (46 years old)
  1. s = serum; ALP = alkaline phosphatase activity; iPTH = intact parathyroid hormone; FGF23 = fibroblast growth factor 23; TmP/GFR = tubular maximum reabsorption of phosphate/glomerular filtration rate.
  2. aReference values for s-phosphorus are age- and sex-dependent: male and female 1–17 years: 1.2–1.8 mmol/L; female >17 years: 0.85–1.65 mmol/L; male 18–49 years: 0.75–1.65 mmol/L; male >49: 0.75–1.35 mmol/L.
  3. bReference values for s-ALP are age-dependent: <18 years: <400 U/L; >18 years: 35–105 U/L.
  4. cConversion factor: pmol/L × 9.497 = pg/mL.
  5. dReference value from the laboratory in which our samples were measured.[13]
s-PhosphateAge dependentammol/L0.880.991.021.35
s-Calcium2.2–2.55mmol/L2.442.382.262.35
s-Magnesium0.71–0.94mmol/L0.980.960.930.86
s-ALPAge dependentbU/L1754436049
s-iPTH1.3–6.8cpmol/L5.91211.45.2
s-1,25 (OH)2 vitamin D50–150nmol/LNA196NA146
FGF23<50dpg/mL90894940
TmP/GFR>0.85mmol/L0.710.80.891.14
Figure 2.

Development of (A) s-phosphate levels and (B) TmP/GFR with age for the two affected siblings. The symbol represents sibling 1 and the symbol × represents sibling 2. The broken lines represent the upper and lower limits of the reference interval.

Genetic studies

Whole-genome SNP genotyping in all family members did not reveal any large region of shared homozygosity between the two affected siblings and we therefore decided to perform whole-exome sequencing on sibling 1. Median exome coverage was 154 ×. We found a total of 18,036 variants in the exome, out of which 8560 were within coding regions (excluding synonymous variants). After filtering against an in-house database of variants found in other whole-exome sequenced Norwegian samples and the 1000 Genomes database (http://www.1000genomes.org) (>0.5% allele frequency), we were left with 207 very rare variants in sibling 1 (Table 2). Using a recessive disease model, only 12 genes remained. Next, using the whole-genome genotyping SNP data we excluded eight of the genes not shared two-identical-by-descent by the affected siblings (for autosomes) and two gene-variants on the X-chromosome carried identically by the non-affected half-brother. Of the two remaining candidate genes; FAM20C and ANKRD36B, only the FAM20C c.C803T/c.C915A mutations could be verified in the proband. The putative homozygous ANKRD36B c.G1627A/p.D543N mutation showed only a borderline quality score and suboptimal mapping quality in the exome sequencing data and could not be validated by Sanger Sequencing. We consider this an erroneous call.

Table 2. Variant Filtration of Exome Sequencing Data From Individual 1 Compared With Whole-Genome Genotyping Data From the Two Affected Siblings, Their Parents, and Their Half-Brother Revealed Two Novel Compound Heterozygous Mutations in the FAM20C Gene as the Most Likely Cause of the Disease in the Family
FilterCount
  1. aSegregation analysis was performed using whole-genome genotyping data from the two affected siblings, their parents, and their half-brother. Eight genes on the autosomes were excluded (not shared 2-IBD by the affected brother), and two gene variants located on the X-chromosome could be excluded as they reside on the identical haplotype carried by the unaffected half-brother.
  2. bOnly the FAM20C c.C803T/c.C915A mutations were confirmed in both affected patients. The putative homozygous ANKRD36B c.G1627A/p.D543N mutation had low mapping quality and borderline quality score and could not be validated by Sanger Sequencing. We consider this an erroneous call.
Exomic variants18,036
Excluding synonymous8,560
Not in in-house database240
Not in 1000 Genomes (0.5% MAF)207
Putative recessive12 genes
In regions segregating with disease in the familya2 genes
Validated by Sanger sequencing in the familyb1 gene

Thus, the FAM20C gene appeared to be the only likely candidate gene, and on sequencing we identified compound heterozygous mutations in the FAM20C gene in both affected subjects. There was a missense mutation in exon 3 (c.803 C > T, p.T268M) inherited from the mother, located in the ATP-binding P-loop, and a nonsense mutation in exon 4 (c.915 C > A, p.Y305X) inherited from the father (Fig. 3A). The latter is predicted to cause a premature truncation of the protein, and is unlikely to lead to a viable protein. In addition, the missense mutation in exon 3 was found in heterozygous state in the half-brother. None of the two mutations were found in 192 healthy Norwegian blood donors.

Figure 3.

(A) Overview of the FAM20C gene and mutations identified so far.[23][25][31][32] Green brackets signify survivors and red brackets signify lethal cases. In the gene/protein model, blue color represents the signal peptide at amino acid position 1–26, whereas green color represents the kinase domain (Pfam 06702; DUF 1193) at amino acid position 353–573. In our family we found a missense mutation in exon 3 (c.803 C > T, p.T268M) inherited from the mother, affecting the ATP-binding P-loop, and a nonsense mutation in exon 4 (c.915 C > A, p.Y305X) inherited from the father. (B) Scatter diagram of the FGF23 values in the presented family.

FGF23 analysis

We assessed FGF23 levels in the studied family members and found that the affected subjects had elevated FGF23 levels (Fig. 3B, Table 1). The unaffected mother and father had normal FGF23 levels. The half-brother was not available for FGF23 testing.

Discussion

Using whole-exome sequencing in a Norwegian family, we identified mutations in FAMC20 associated with autosomal recessive hypophosphatemia, suggesting that this condition may represent a third form of ARHR. Recent murine data supports the notion that inactivation of FAMC20 leads to hypophosphatemia.[17] Moreover, FGF23 levels were elevated in our affected family members, as well as in bone cells and serum of FAMC20 knockout mice,[17] indicating that our patients may suffer from a new monogenic form of FGF23-related hypophosphatemia and hyperphosphaturia. We screened DMP1- and PHEX-negative subjects from a national Norwegian population-based material of familial hypophosphatemia and selected U.S. patients with familial hypophosphatemia and osteosclerosis for FAM20C mutations, but were unable to identify other families. Hence, recessive FAM20C mutations do not appear to be a common cause of classical hypophosphatemia.

The affected family members had hypophosphatemia secondary to renal phosphate wasting and the hallmarks of an FGF23-mediated disorder; i.e., inappropriate normal levels of 1,25(OH)2D and PTH within or slightly above the normal range. Interestingly, the renal phosphate wasting decreased after puberty and at the age of 18 to 20 years, both brothers had normal circulating phosphate levels and no significant hyperphosphaturia. A similar pattern has been described in some families with ADHR,[18-20] and it has been speculated whether this phenomenon may be caused by adaptation to the increased levels or altered metabolism of FGF23 over time.[18-20] The FGF23 levels were significantly elevated in the two affected siblings,[21, 22] even after the renal phosphate wasting defect was gone. The mechanisms by which FAM20C mutations are linked to FGF23-mediated renal phosphate wasting and hypophosphatemia are at the moment unclear. FAM20C mutations have previously been associated with Raine syndrome, a rare disorder originally described as a lethal syndrome of microcephaly, hypoplastic nose, exophthalmos, gum hyperplasia, cleft palate, low-set ears, and osteosclerosis.[23-25] Recently, four cases of non-lethal Raine syndrome have been described, and of the three in which s-phosphate was measured, two had hypophosphatemia.[23, 25] We demonstrate for the first time that FAM20C mutations are also associated with nonlethal, monogenic, FGF23-dependent hypophosphatemia.

FAM20C is composed of 10 exons spanning 605 bp, and encodes the phosphorylation enzyme Golgi-enriched fraction casein kinase (GEF-CK). There is a short N-terminal signal peptide (amino acid position 1–22) and a highly conserved domain, the Pfam 06702 (DUF 1193) in the C-terminal end, at amino acid position 354–573. Within the Pfam-domain there are highly conserved residues necessary for the kinase activity, as well as a Greek key calcium binding domain, and an RDG (Arg-Gly-Asp) motif.[26] A recent publication suggests the SIBLING-proteins to be substrates for FAM20C27 and shows that osteopontin (OPN), DMP1 and matrix extracellular phosphoglycoprotein (MEPE) are phosphorylated by FAM20C in vitro.[27] Actually, the recently reported FAM20c knockout model shows a phenotype resembling the DMP1 knockout mouse phenotype. Moreover, in these mice, loss of FAM20c resulted in a significant downregulation of DMP1.[17] Dentin sialophosphoprotein (DSPP) is also included in the SIBLING protein family, and contain the S-x-E motif recognized by FAM20C.[28] The overlapping clinical features of dentinogenesis imperfecta type II[29] and Raine syndrome may suggest that DSPP may also be a substrate for FAM20C. Moreover, it has earlier been reported that a phosphorylation defect of osteopontin in Hyp-mice is due to defective activity of skeletal casein kinase activity in these mice.[30] Taken together, defective phosphorylation seems to be a mechanism that is involved in the pathways that generate FGF23-dependent hypophosphatemia.

When comparing the genetically characterized cases of Raine syndrome in literature, we found that the lethal cases had mutations that affected the highly conserved Pfam domain in the C-terminal part of FAMC20.[31, 32] In contrast, most of the surviving cases have had mutations in the N-terminal part of FAMC20 (see Fig. 3A).[23, 25] There is, however, one exception in which a mild phenotype was associated with the mutation c.1351 G > A that leads to the substitution of aspartic acid by asparagine at position 451 in the Pfam domain.[23] It is possible that the structural similarities between these two amino acids may explain why this mutation may be less damaging. Although one may speculate that FAM20C mutations affecting the C-terminal part of the GEF-CK may result in a more severe phenotype, a confirmation of this will have to await further studies.

In the family we report, a severe dental phenotype was associated with mutations in the FAM20C gene. Although the dental anomalies were of clinical and histological similarity to those in XLHR and ARHR type 1, both cases had dental anomalies starting at a very early age and affecting both deciduous and permanent teeth, rendering them both edentulous before the age of 18 years. Thus the dental phenotype appears to be more severe in FAM20C mutations than in most cases of XLH. Earlier surviving patients with Raine syndrome have also shown dental effects23,25 and, recently, two different groups have demonstrated that FAM20C knockout mice have significantly defective formation and mineralization of dentin, cementum, and enamel[17] or show an amelogenesis imperfecta (AI) phenotype.[33] In addition, the dental phenotype of our patients closely matched the findings in dentinogenesis imperfecta type II, a disorder caused by mutations in the putative FAMC20-encoded protein substrate DSPP.[29] Taken together, these findings suggest that phosphorylation of SIBLING proteins by FAM20C-encoded protein may be important for proper amelo- and dentinogenesis.

The FGF23-mediated hypophosphatemic disorders may be associated with a variety of skeletal affections. Typically, XLHR and ADHR are characterized by rickets and osteomalacia,[34] but adults with XLHR often demonstrate higher than average bone mineral density by dual-energy X-ray absorptiometry (DXA),[35] and radiographs may have a sclerotic appearance,[36] whereas patients with ARHR type 1 show osteosclerosis of the skull bones and rickets of the long bones.[37] The affected brothers of our family with FAM20C mutations had osteosclerosis of the long bones, but no rickets, a picture also observed earlier in the few surviving patients with Raine syndrome.[23, 25] In contrast, the mice in the FAM20C inactivation model were characterized by rickets and no osteosclerosis,[17] a discrepancy that may suggest species-specific differences in gene expression.[38]

The reported patients show intracerebral calcifications, especially in the periventricular white matter and basal ganglia, a feature in common with Raine syndrome patients.[23, 25, 31, 32, 39-48] Recently, Wang and colleagues[49] identified mutations in SLC20A2, encoding the type III sodium-dependent phosphate transporter 2 (PiT2), in families affected by idiopathic basal ganglia calcification. Ectopic calcification in other tissues, e.g., enthesopathies, is an important feature of the different hereditary hypophosphatemic disorders, including XLHR,[50] ARHR-I,[37] and ARHR-II.[51] The exact mechanism behind these ectopic calcifications needs to be studied further.[37] In the reported siblings, the intracerebral calcifications have remained stable during the last years.

An important characteristic of Raine syndrome, also found in the two brothers in this study, is dysmorphic features of the midface. A flat nasal bridge and a high arched or cleft palate, as well as dental involvement, have been described in most cases. Associations between FAM20C and both the canonical WNT and the transforming growth factor β (TGFβ) signaling pathways have been reported,[17, 52] both important in the development of the palate.[53] Taken together this may point to a role for FAM20C-encoded protein phosphorylation function in the embryonic development of structures in the head region.

In conclusion, we have identified the nonlethal variant of Raine syndrome as a new form of FGF23-related hypophosphatemia and hyperphosphaturia associated with FAMC20 mutations, and we propose this disease entity to be the third autosomal recessive form of hypophosphatemia. Additional clinical features of this hypophosphatemic disorder include a severe dental phenotype, intracerebral calcifications, osteosclerosis, as well as facial and acral dysmorphic features.

Disclosures

All authors state that they have no conflicts of interest.

Acknowledgments

The family members are thanked for making this study possible, by being available for clinical examination and blood tests, as well as giving access to medical information. We also thank Ketil Mevold, MD at Nordland County Hospital, for the initial referral of the family and the staff at the Centre of Medical Genetics and Molecular Medicine, Haukeland University Hospital, Bergen Norway for technical support.

Authors' roles: SHR, HR, SJ, and RB designed the study; SHR, HR, AKF, TOC, and SJ collected the data; SHR, HR, PK, TOC, SJ, and RB contributed to data analysis and interpretation. The manuscript was drafted by SHR, HR, SJ, and RB. All authors contributed to the revision and approved the final version of the manuscript.

Ancillary