FAM20A binds to and regulates FAM20C localization

Mutations in the Family with sequence similarity (FAM) 20 gene family are associated with mineralized tissue phenotypes in humans. Among these genes, FAM20A mutations are associated with Amelogenesis Imperfecta (AI) with gingival hyperplasia and nephrocalcinosis, while FAM20C mutations cause Raine syndrome, exhibiting bone and craniofacial/dental abnormalities. Although it has been demonstrated that Raine syndrome associated-FAM20C mutants prevented FAM20C kinase activity and secretion, overexpression of the catalytically inactive D478A FAM20C mutant was detected in both cell extracts and the media. This suggests that FAM20C secretion doesn’t require its kinase activity, and that another molecule(s) may control the secretion. In this study, we found that extracellular FAM20C localization was increased when wild-type (WT), but not AI-forms of FAM20A was co-transfected. On the other hand, extracellular FAM20C was absent in the conditioned media of mouse embryonic fibroblasts (MEFs) derived from Fam20a knock-out (KO) mouse, while it was detected in the media from WT MEFs. We also showed that cells with the conditioned media of Fam20a WT MEFs mineralized, but those with the conditioned media of KO MEFs failed to mineralize in vitro. Our data thus demonstrate that FAM20A controls FAM20C localization that may assist in the extracellular function of FAM20C in mineralized tissues.

FAM20A, FAM20B and FAM20C are the three structurally-related members of the "family with sequence similarity 20" (FAM20) proteins 1 . Non-sense or missense mutations in the FAM20A gene were identified that are associated with an autosomal recessive type of Amelogenesis Imperfecta (AI), resulting in premature termination or amino acid changes of the peptides (OMIM#614253). These AI patients with FAM20A mutations have several major dental abnormalities such as generalized hypoplastic enamel, intrapulpal calcification, delayed tooth eruption, and failure of tooth development as well as gingival hyperplasia 2,3 . More recently, a non-dental phenotype of nephrocalcinosis has also been reported in the AI patients with FAM20A mutation, further indicating the potential molecular function of FAM20A in biomineralization [4][5][6][7][8][9] . FAM20B encodes a protein kinase that phosphorylates xylose residue in the tetrasaccharide glycosaminoglycan (GAG)-protein linkage region in the Golgi apparatus 10 , thereby regulating the subsequent GAG chain assembly 11 . Proteoglycans derived from FAM20B TALEN knock-out (KO) cells were chondroitin and heparin sulfate GAGs-deficient 11 , which likely explains why the Fam20b deficient zebrafish exhibited poor cartilage matrix production in vivo 12 . FAM20C was found to be a causative gene for Raine syndrome (OMIM#259775) that manifests pulmonary hypoplasia, osteosclerosis, craniofacial dysmorphism, dental anomalies and gingival hyperplasia [13][14][15] . The mice deficient for Fam20c exhibit a systemic hypophosphatemic condition likely by increasing serum levels of the crucial phosphorus regulating hormone FGF23 16 . Dental phenotypes in these mice have also been reported [17][18][19] . Recent reports demonstrated the multi-functionality of FAM20C located in the Golgi and extracellular space. It has been reported that FAM20C acts as the Golgi Casein Kinase (GCK) that phosphorylates casein and the Small Integrin-Binding Ligand, N-linked Glycoproteins (SIBLINGs) which are critical for biomineralization 20,21 . Addition of recombinant FAM20C protein into cell cultures promotes osteoblast differentiation 16 , indicating the potential role of extracellular FAM20C as a growth and differentiation factor-like protein.
Currently, there is a debate as to whether endogenous FAM20C is secreted or secretion is an artifact of overexpression in cell culture system 22,23 , thus it is important to determine whether endogenous FAM20C is present in the extracellular space and also to delineate the mechanism how FAM20C secretion is controlled. It has been demonstrated that secretion of FAM20C (by overexpression) occurs regardless of its kinase activity, because a catalytically inactive FAM20C (D478A) mutant, but not most of other Raine syndrome-associated missense FAM20C mutants, is still detected in the conditioned media 23 . This suggests the presence of another molecule(s) which likely determines the FAM20C localization. Therefore, the purpose of our current study was to identify and investigate the molecule(s) that can regulate FAM20C secretion. Here we show that FAM20A binds to FAM20C in cell cultures and in vitro. When FAM20C is transiently overexpressed together with wild-type (WT) FAM20A in 293 cells, FAM20C secretion is increased as compared to cells in which FAM20C is expressed alone. FAM20A forms mimicking AI-mutations fail to enhance the FAM20C secretion. Similarly, extracellular FAM20C is absent in mouse embryonic fibroblast (MEF) cultures derived from Fam20a KO mice, but present in cultures derived from WT MEFs, demonstrating that FAM20A is required for FAM20C extracellular localization. To assess the biological importance of FAM20A-controled FAM20C secretion, in vitro mineralization assay was performed using MC3T3-E1 osteoblastic cells treated with the conditioned media derived from Fam20a KO or WT MEF cells. Our results showed that the extent of mineralization of MC3T3-E1 cells treated with the conditioned media from Fam20a KO MEF cells was markedly lesser than that of MC3T3-E1 cells with the conditioned media from Fam20a WT MEF cells. To the best of our knowledge, our data demonstrate for the first time that FAM20A regulates FAM20C localization and assists in the extracellular function of FAM20C.

Results
Expression of FAM20 family members in various tissues. To investigate the gene expression of all FAM20 family members in various mouse tissues, real-time PCR was performed (Fig. 1A, numerical data in S1 Table). The results demonstrated that the patterns of gene expression in heart, lung and kidney were similar among Fam20 genes, i.e. the expression of Fam20b (red) and Fam20c (green) was higher than that of Fam20a (blue), while the patterns of gene expression in calvaria and tooth were similar, i.e. the expression of Fam20a and Fam20c was higher than that of Fam20b.

Localization of FAM20A and FAM20C proteins in odontogenic cells. Since mutations in FAM20A
as well as FAM20C manifest dental phenotypes, our results led to investigate whether FAM20A and FAM20C proteins are expressed in odontogenic tissues. Immunohistochemical analysis of adult mouse incisors using anti-FAM20A and anti-FAM20C antibodies was performed (Fig. 1B). The results demonstrated that both FAM20A (left panels) and FAM20C (middle panels) proteins were localized in odontogenic cells/tissues including pulp, odontoblasts, ameloblasts and stratum intermedium with granular patterns of the immunoreactivity. Both FAM20A and FAM20C proteins appeared to be accumulated at pre-dentin, dentinal tubules within odontoblasts and Tome's processes of ameloblasts. No immunoreactivity was observed when non-immune goat serum was used (right panels). Similar localization of FAM20A to FAM20C led us to investigate the potential interaction between these proteins. FAM20A binds to FAM20C. Since the localization of FAM20A was similar to that of FAM20C in tooth, and clinical phenotypes of AI patients with FAM20A mutation partially overlap with those of Raine syndrome patients, we investigated the association of FAM20A with FAM20C. The FAM20C expression vector with C-terminal HA tag (pc3-FAM20C WT-HA) was transfected together with pc3-FAM20A WT-Flag into 293 cells, and the interaction was assessed. The results clearly showed that FAM20A interacted with FAM20C ( Fig. 2A, upper panel, lane 2). Furthermore, two types of FAM20C mutant forms (D478A and P328S) were tested to determine that this interaction requires FAM20C kinase activity. As shown in Fig. 2B, the interaction between FAM20A and FAM20C was again confirmed in a dose dependent manner with a different combination of the tag used from Fig. 2A (Fig. 2B, upper panel, lanes 3 and 4). We also found that two of FAM20C mutant forms still bound to FAM20A, demonstrating that FAM20A-FAM20C interaction occurs in a FAM20C-kinase activity independent manner (Fig. 2B, upper panel, lanes 5 and 6).
To determine whether or not the binding observed was direct (i.e. without the presence of other molecules), we initially examined the in vitro binding by incubating rhFAM20C WT-V5/His protein purified from conditioned media of stably transfected 293 cells with recombinant GST-FAM20A fusion protein produced in E. coli and purified, followed by a GST pull down assay. However, the binding between FAM20A and FAM20C was not clearly observed. We reasoned that a protein kinase didn't usually associate firmly with substrate-like proteins 24 . It has been previously reported that the mutation of the aspartate residue at 478 in FAM20C abolished its intracellular kinase activity 20 , consistent with the requirement of a divalent cation binding to this residue for the catalysis 25 . We thus investigated the in vitro binding assay using rhFAM20C D478A-V5/His protein and tested this possibility (  The results indicate that wild type FAM20A is necessary for FAM20C extracellular accumulation and that AI-associated FAM20A forms appear to fail to enhance FAM20C secretion. In order to determine whether the data obtained in the 293 cell model reflects what occurs in primary cells, we further examined whether or not extracellular FAM20C is present in cultures of Fam20a KO MEF cells. Conditioned media and cell lysates of MEF cells derived from WT and KO embryos were collected and the levels of extracellular FAM20C, and intracellular FAM20C and FAM20A were evaluated using anti-FAM20C and anti-FAM20A antibodies. Our results demonstrated that extracellular FAM20C was found in the conditioned media from WT MEF cells (  Quantitative real-time PCR analysis was performed using 5-week-old mouse tissues (brain, heart, lung, kidney, calvaria, tooth). The mean fold change in the expression of each Fam20 member was calculated based on the normalization to that of glyceraldehyde-3-phosphate dehydrogenase (Gapdh) using the value of Fam20a in brain as a calibrator. The values are shown as the mean + S.D. based on triplicate assays. The expression of Fam20a is indicated by blue bars, Fam20b by red and Fam20c by green. (B) Immunohistochemical analysis of FAM20A and FAM20C in adult mouse incisors. FAM20A (a,d) and FAM20C (b,e) were both localized in the same cell/ tissue types. No immunoreactivities were observed when non-immune goat serum was used as a negative control (c,f). Scale bar, 50 μ m. Images of odontoblasts and ameloblasts were shown at a higher magnification on the right corner of each image (scale bar, 10 μ m). Pu; pulp, Od; odontoblasts, PD; pre-dentin, D; dentin, E; enamel, Am; ameloblasts, Si; stratum intermedium.
WT or KO MEF cells, and mineralized nodules were stained with Alizarin Red S dye. The results showed that MC3T3-E1 cells cultured with CM from Fam20a KO MEF cells failed to mineralize, while cells cultured with CM from Fam20a WT fully mineralized. The Alizarin Red S dye was extracted from the mineralized nodules deposited in cell cultures and the concentration of the dye was quantified. The values extracted from MC3T3-E1 cells with CM from Fam20a WT were significantly greater than those with CM from Fam20a KO (Fig. 4B), indicating that Fam20a is required for FAM20C-mediated mineralized nodule formation in osteoblasts.

Discussion
FAM20A was originally identified as an upregulated gene in murine bone marrow cells with a retrovirus expressing a dominant negative retinoic acid receptor α (RARα ) protein, and public database searches also identified FAM20B and FAM20C by similarity 1 . FAM20B is evolutionally different from the other 2 members, as this gene has been conserved from Caenorhabditis elegans to human, while FAM20A and FAM20C have been identified from Fugu rubripes (pufferfish) to human 1 . Gene knockout study also suggests that the phenotype of Fam20b is distinct and different from the other 2 members; Fam20b KO mice show embryonic lethality at E12.5, while Fam20a and Fam20c KO mice do not 26 . Our data in Fig. 1 demonstrate that the expression patterns of Fam20a and Fam20c are similar and well correlated with the tissues where clinical phenotypes in AI with FAM20A mutation and Raine syndrome were found, while that of Fam20b is different. The evolutionary conservation, gene KO phenotype and differential expression among FAM20 members may suggest that the molecular functionality of FAM20A is associated with FAM20C, but not FAM20B. This notion is also confirmed by the recent report that FAM20A forms a protein complex with FAM20C and regulates FAM20C kinase activity 27 .
There are currently conflicting results reported whether FAM20A, FAM20B and FAM20C are secretory proteins or intracellular Golgi-proteins or both 1,20,21,28,29 . Overexpression of FAM20 proteins have been previously performed in several studies. The expression of FAM20A 1,20 and FAM20B 20 was identified in both cell lysate and conditioned media, while it was detected only in the cell lysate in another report 21 . FAM20C secretion by overexpression or at the endogenous level was observed in all studies tested 1,20,21,27,30 . Although less quantitative, endogenous FAM20 protein immunostaining has been shown in skeletal/dental tissues. The immuno-localization of FAM20A 2,6 and FAM20B 2 were observed only within cells, while that of FAM20C detected in both extracellular matrices and cells 2,28 . Our immunohistochemical data demonstrated that the immunoreactivities to anti-FAM20C antibody were mainly observed intracellularly, in the cytoplasmic areas of odontogenic cells including odontoblasts, ameloblasts, stratum intermedium and pulp with granular patterns. Our immunohistochemical analysis also demonstrated the presence of FAM20C in pre-dentin, which may suggest that secretion of FAM20C occurs in a tissue-specific manner.
It is not clear, at this point, why FAM20A directly binds to a kinase-inactive form, but not WT of FAM20C (Fig. 2B) although FAM20A binds to FAM20C WT in cell cultures ( Fig. 2A,B). It is generally accepted that the kinase doesn't always stably associate with its corresponding substrate because the biochemical phosphorylation reaction is transient 31 and this transient contact typically prevents or at least strongly reduces the efficacy of the substrate detection by the commonly used protein-protein interaction methods 24 , suggesting the elaboration of biochemical techniques is required to identify kinase targets. This concept may be the basis for the direct binding between wild type FAM20A and FAM20C was not detected, while FAM20C D478A (inactive kinase) did co-immunopreciptate with FAM20A. Very recently, it has been demonstrated that FAM20A functions as a pseudokinase in a complex with FAM20C 27 . Molecular function of some pseudokinases has been studied and reported 32 . STE20-related pseudokinase, STRADα , for example, forms a protein complex with a protein kinase LKB1 33 , resulting in the activation of LKB1 kinase activity. The kinase activity and localization of LKB1 was further enhanced with MO25, the scaffold molecule with STRADα 34 . We showed that the mutated forms of FAM20A failed to enhance extracellular FAM20C accumulation, which is consistent with the recent report that FAM20A binds to ATP, which may stabilize FAM20A conformation 27 . Although the biochemical function of FAM20A in conjunction with FAM20C GCK activity was demonstrated, questions especially in vivo phenotypes still remain. Fam20c knockout mice exhibit hypophosphatemia due to the inability to properly control FGF23 secretion, however the levels of Ca and phosphate in Fam20a knockout mice were normal 26 . This raises a possibility that FAM20A may function in FAM20C-independent manner. Consistent with this notion, there are broad and distinct clinical features in AI patients with FAM20A mutation (OMIM#614253) and Fam20a KO models 26,35 . Previous studies suggested that the mineralization inhibitor(s) may be FAM20A's potential substrate(s), because FAM20A mutations in human or Fam20a deficiency in mouse KO models are associated with ectopic calcification phenotypes (Loss of FAM20A kinase activity, if any, may result in loss of phosphorylation in these molecules and thus loss of inhibition) 4-6,26,36,37 . Among these inhibitors, it has been reported that alpha-2-HS-glycoprotein (also known as fetuin) deficiency resulted in the ectopic calcification phenotype 38 , resembling to Fam20a KO mice 26 . However, the finding that FAM20A is a pseudokinase conflicts with this idea. Thus, it is an interesting future study to delineate molecular mechanism(s) by which the ectopic calcification phenotype occurs in Fam20a KO mice.
In the present study, we show, for the first time, that FAM20A regulates FAM20C secretion into the extracellular space and propose a model in Fig. 5. It is of particular interest that DMP4, a mouse homolog of FAM20C was originally identified as an extracellular calcium-binding protein which promotes HAT-7 ameloblastic cell mineralization 30 . Addition of FAM20C into MC3T3-E1 osteoblast cell cultures up-regulates the expression of several bone differentiation markers such as Bone sialoprotein (BSP), Dentin matrix protein (DMP)1 and Osteocalcin 16 , indicating that extracellular FAM20C functions as a growth and differentiation factor in mineralizing cells. Therefore, the regulation of FAM20C secretion by FAM20A is crucial for mineralized tissue biology. In support with this notion, our results showed that addition of the conditioned media from Fam20a KO MEF cell cultures (which had under detection levels of FAM20C) failed to form mineralized nodules in MC3T3-E1 osteoblastic cultures in vitro (Fig. 4B).
As the emerging FAM20 kinase family functions become clear, the effect of FAM20A on FAM20C localization may define the novel mechanisms of mineralized tissue biology at a molecular level and is currently under investigation in our laboratory.

RNA extraction and quantitative real-time PCR.
To investigate the gene expression of FAM20 family members in various tissues, total RNA was extracted with TRIzol reagent (Life Technologies) using mouse tissues, i.e. brain, heart, lung, kidney, calvaria, and tooth (C57BL/6, 5-week-old, male). Two μ g of the total RNA extract was used for RT using the Omniscript RT Kit (Qiagen). Real-time PCR was performed in triplicate using the specific primers-probe for mouse Fam20a (Applied Biosystems, ABI assay number; Mm00464089_m1), mouse Fam20b (Mm01286165_m1), mouse Fam20c (Mm00504210_m1), or rodent glyceraldehyde-3-phosphate dehydrogenase (Gapdh; 04308313), and the expression levels were analyzed by the ABI 7300 real-time PCR system. The mean fold changes in the expression of each Fam20 member relative to that of Gapdh were calculated using the values obtained from the brain cDNA with Fam20a expression as a calibrator by means of 2 −ΔΔ CT method as we previously reported 39 . Three independent experiments were performed and the results were essentially identical.

Immunohistochemistry.
To determine the localization of endogenous FAM20A protein in vivo, formalin fixed paraffin embedded sections of non-diseased mouse heads (C57BL/6, 4-week-old, male) were used to perform immunohistochemical staining as we reported 39,40 . Sections were incubated with anti-FAM20A (goat polyclonal, Santa Cruz Biotech, catalog number: sc-164308, diluted in 1:800), anti-FAM20C antibody (goat polyclonal, Santa Cruz Biotech, catalog number: sc-160322, diluted in 1:200), or non-immune goat serum as a negative control. The sections counterstained with hematoxylin were photographed under a light microscope.
Generation of FAM20A and FAM20C forms in mammalian expression vectors. All vector construction was performed by polymerase chain reaction (PCR) using HotStarTaq DNA polymerase (Qiagen) and site-directed mutagenesis with the megaprimer PCR procedure 41 . The plasmids containing the full length sequence of human FAM20A and FAM20C were purchased from OriGene and used as PCR templates. The primer sequences used in this study were described in Table 1. The coding region of wild type (WT) FAM20A without stop codon was amplified by PCR (forward-reverse primers; F1-R1, Table 1) and the amplified fragment was subcloned into pcDNA3-3′ -Flag-tag (pc3-Flag) and pcDNA3-3′ -HA-tag (pc3-HA) mammalian expression vectors. The coding region of WT FAM20C was amplified by the primers (F5-R6), and it was subcloned into pc3-HA and pcDNA3.1-3′ -V5/His-tag (pc3.1-V5/His) vectors. The plasmids harboring pc3-FAM20A WT-Flag, pc3-FAM20A WT-HA, pc3-FAM20C WT-HA and pc3.1-FAM20C WT-V5/His were then obtained. To generate AI form of FAM20A with c.1175_1179delGGCTC mutation 3 in the expression vector, which resulted in the frame shift mutation at arginine residue 392 to proline followed by 21 extra amino acids and the premature stop codon (p.R392PfsX22), the forward and reverse primers with this mutation (F2 and R2) and the reverse primer corresponding to the extra amino acids without premature stop codon (R3) were designed, thus generating FAM20A R392Pfs21. The PCR fragment of FAM20A R392Pfs21 was amplified by the primers (F1-R3) using (F1-R2) and (F2-R3) fragments by means of megaprimer PCR method, and the resulted F1-R3 fragment was subcloned into pc3-Flag. To create another AI form of FAM20A with c.992G > A mutation 4 , which resulted in the missense mutation at glycine residue 331 to aspartic acid (p.G331D), the forward and reverse primers with this missense mutation were designed (F3 and R4). The PCR fragment of FAM20A G331D was amplified by the primers (F1-R1) using (F1-R4) and (F3-R1) fragments by means of megaprimer PCR method, and the resulted F1-R1 fragment was subcloned into pc3-Flag. The plasmids harboring pc3-FAM20A R392Pfs21-Flag and pc3-FAM20A G331D-Flag were then obtained. We also generated the kinase inactive form of FAM20C (FAM20C D478A) and one of the Raine syndrome-mutant form of FAM20C (FAM20C P328S) in the expression vectors (pc3.1-FAM20C D478A-V5 and pc3.1-FAM20C P328S-V5) according to the previous report 20 (F6 and R7 for FAM20C D478A, and F7 and R8 for FAM20C P328S). All plasmids generated were sequenced, and their DNA sequences were 100% identical to the respective reference DNA sequences available at NCBI or the respective mutations.
Binding of FAM20A to FAM20C and the effect of FAM20A on FAM20C localization. The 293 cells were plated onto 6 well culture plates at a density of 3 × 10 5 cells/well and transfected with pc3-FAM20A WT-Flag and pc3-FAM20C-HA. After 24 hours, cells were lysed and then immunoprecipitated with anti-Flag antibody and Western blot (WB) analysis was performed with anti-HA antibody (clone 3F10, Roche Diagnostics) to investigate the interaction between FAM20A and FAM20C. The expression of FAM20C-HA or FAM20A-Flag was also verified using the same cell lysates by immunoprecipitation (IP)-WB analysis with anti-HA or anti-Flag antibody, respectively. To determine the interaction between FAM20A and mutant forms of FAM20C, 293 cells were transfected with pc3-FAM20A WT-HA and pc3.1-FAM20C WT-V5, pc3.1-FAM20C D478A-V5, or pc3.1-FAM20C P328S-V5. Cell lysates were used for immunoprecipitation with anti-V5 antibody and WB analysis was performed to detect the interaction with anti-HA antibody. The expression of FAM20A WT-HA, or FAM20C WT-V5 FAM20C D478A-V5, or FAM20C P328S-V5 was verified using the same lysates by WB analysis with anti-HA or IP-WB analysis with anti-V5 antibody. Chemiluminescent detection of bound antibodies was determined using the ECL Western blotting detection reagents (Amersham ECL Prime, GE Healthcare Life Sciences).
To investigate the effect of FAM20A co-transfection on the localization of FAM20C, 293 cells were plated in the same manner as described and transfected with pc3-FAM20A WT-Flag, pc3-FAM20A R392Pfs21-Flag, or pc3-FAM20A G331D-Flag in the presence of pc3-FAM20C-HA. The conditioned media and cell lysates were collected and analyzed by IP-WB analysis in the same manner as described above.
Generation of recombinant human Glutathione-S-Transferase (GST)-FAM20A fusion protein. The primer sequences of FAM20A cDNA corresponding to its mature protein (residues 22-541; Asp 22 -Ser 541 ) were described ( Table 1, F4-R5). The PCR products were amplified, cloned into pGEX6P-1 vector (GE Healthcare Life Sciences), sequenced, and the vector that expresses the mature FAM20A protein (pGEX-FAM20A) was obtained. The pGEX-FAM20A or pGEX empty vector was transformed into the BL21-CodonPlus bacterial strain (Agilent Technologies), cultured, and the synthesis of GST-FAM20A fusion protein or GST protein alone was induced by isopropyl-D-1-thiogalactopyranoside (IPTG). The bacteria were further cultured overnight at 20 °C. The cultures were then centrifuged and lysed in PBS containing 1% Triton X-100. After the centrifugation, the supernatants were incubated with glutathione-sepharose beads (GE Healthcare Life Sciences) overnight and the beads were extensively washed with PBS. The beads were then treated with the elution buffer (10mM glutathione, 50 mM Tris-HCl pH8.0) to release the conjugated proteins. The eluted proteins were pooled, dialyzed against distilled water, lyophilized and resuspended in distilled water. The protein concentration of each GST protein was measured using a Detergent Compatible (DC) protein assay kit (Bio-Rad) and a plate reader. Aliquots of each GST protein were mixed with SDS sample buffer, applied to 4-12% SDS-PAGE, and stained with Coomassie Brilliant Blue R-250 (CBB) (Bio-Rad). An aliquot of the GST-FAM20A protein generated was subjected to WB analysis with anti-FAM20A or anti-GST (goat polyclonal, GE Healthcare Life Sciences, product code: 27-4577-01) antibody and the immunoreactivity of the protein was confirmed.

Generation of recombinant human FAM20C WT-and D478A-V5/His proteins. Recombinant
human (rh) FAM20C WT-and D478A-V5/His proteins were generated and purified in the same manner as we previously reported 40 . Briefly, 293 cells were transfected with the pc3.1-FAM20C WT-or D478A-V5/His vector, and the clones derived from G418-resistant cells were isolated. Conditioned media were collected from each clone (out of 10 clones), and IP-WB analysis was performed with anti-V5 antibody. The clone which synthesized the highest level of rhFAM20C-V5/His protein was further cultured, the conditioned media were collected, and rhFAM20C WT-or D478A-V5/His protein was purified using a Ni-NTA agarose resin (Qiagen). The purified each rhFAM20C protein was pooled, dialyzed against distilled water, lyophilized and dissolved in distilled water. The protein concentration was measured using a DC protein assay kit and it was kept at − 20 °C until use. The purity of the rhFAM20C WT-or D478A-V5/His protein was assessed by 4-12% SDS-PAGE stained with CBB and by WB analyses with anti-V5 and anti-FAM20C antibodies. The presence of hFAM20C was also confirmed by Mass Spectrometry analysis (data not shown).
GST pull-down assay. To investigate the direct interaction between FAM20A and FAM20C, GST protein or GST-FAM20A protein was incubated with rhFAM20C WT-or D478A-V5/His overnight at 4 °C. After addition of glutathione-sepharose beads, the samples were washed with lysis buffer twice, mixed with SDS sample buffer, applied to 4-12% SDS-PAGE and WB analysis was performed with anti-V5 antibody. The expression of GST proteins (GST and GST-FAM20A) used for GST pull-down as input was analyzed by Western blotting with anti-GST antibody.

Isolation and primary cell culture of mouse embryonic fibroblasts (MEFs). Heterozygous Fam20a
KO mice previously generated 35 were intercrossed and the embryos at 14.5 days were dissected. Genomic DNA was extracted to determine the genotypes of embryos by PCR as previously established 35 and the embryos were further minced. Cells were released by treating the minced samples with trypsin for 1 hour at 37 °C and plated onto the culture plates. Only 2-5 passages of MEF cells were used for the experiments.

FAM20C expression in Fam20a KO MEF cells. MEF cells from Fam20a
WT and KO embryos were plated and cultured for 48 hrs. Conditioned media were collected and the expression of extracellular FAM20C was then analyzed by IP-WB analysis with anti-FAM20C antibody. Cells were lysed and cell lysates were analyzed by Western blotting using anti-FAM20C and anti-FAM20A antibodies to verify the expression levels of intracellular FAM20C and FAM20A in the cell lysates. The expression level of β -TUBULIN as a loading control in the cell lysates was measured by WB analysis with anti-β -TUBULIN antibody (mouse monoclonal, BD Biosciences, catalog number: 556321). Films with non-saturated exposure were used and densitometric analysis was carried out using Image-J software (NIH). The mean gray value intensity was quantified by measuring the expression of intracellular FAM20C and β -TUBULIN, and the value intensity of intracellular FAM20C was normalized to that of β -TUBULIN in each cell type using the same cell lysate. Three independent experiments were performed and data are represented as the mean + standard deviation (SD) with statistical analysis.
Scientific RepoRts | 6:27784 | DOI: 10.1038/srep27784 In vitro mineralization assay and Alizarin Res S staining quantification. To collect the conditioned media from Fam20a KO and WT MEFs, cells were plated onto 15 cm culture dishes. After cells reached confluence, the serum was starved for 24 hours and the conditioned media were collected. The media were centrifuged at 1,500 rpm for 5 min to remove any dead cells/cellular debris, and kept at 4 °C until use. MC3T3-E1 cells were plated onto 35 mm culture dishes in triplicates at a density of 3 × 10 5 cells/dish. After cells reached confluence, the media were refreshed twice a week as follows; 50% v/v of the conditioned media from Fam20a WT or KO MEFs, and 50% v/v of α -MEM containing 20% of FBS and antibiotics described above in the presence of 50 μ g/ml of ascorbic acid and 4 mM of β -glycerophosphate. To visualize the mineralized nodules in vitro, Alizarin Red S staining was performed as previously reported 42 . Alizarin Red S dye deposited onto the cell matrices was extracted with 10% cetyl pyridinium chloride buffer in 10 mM sodium phosphate. Absorbance was measured at 560 nm, Alizarin Red S concentration in each sample was calculated using an Alizarin Red S standard curve, and the values are expressed as the mean + SD (n = 3) with statistical analysis.

Statistical analysis.
Student T-test was used to assess the statistical significance between groups. P value with less than 0.01 was considered significant.