FAM20C directly binds to and phosphorylates Periostin

It is widely accepted that FAM20C functions as a Golgi casein kinase and has large numbers of kinase substrates within the secretory pathway. It has been previously reported that FAM20C is required for maintenance of healthy periodontal tissues. However, there has been no report that any extracellular matrix molecules expressed in periodontal tissues are indeed substrates of FAM20C. In this study, we sought to identify the binding partner(s) of FAM20C. FAM20C wild-type (WT) and its kinase inactive form D478A proteins were generated. These proteins were electrophoresed and the Coomassie Brilliant Blue (CBB)-positive bands were analyzed to identify FAM20C-binding protein(s) by Mass Spectrometry (MS) analysis. Periostin was found by the analysis and the binding between FAM20C and Periostin was investigated in cell cultures and in vitro. We further determined the binding region(s) within Periostin responsible for FAM20C-binding. Immunolocalization of FAM20C and Periostin was examined using mouse periodontium tissues by immunohistochemical analysis. In vitro kinase assay was performed using Periostin and FAM20C proteins to see whether FAM20C phosphorylates Periostin in vitro. We identified Periostin as one of FAM20C-binding proteins by MS analysis. Periostin interacted with FAM20C in a kinase-activity independent manner and the binding was direct in vitro. We further identified the binding domain of FAM20C in Periostin, which was mapped within Fasciclin (Fas) I domain 1–4 of Periostin. Immunolocalization of FAM20C was observed in periodontal ligament (PDL) extracellular matrix where that of Periostin was also immunostained in murine periodontal tissues. FAM20C WT, but not D478A, phosphorylated Periostin in vitro. Consistent with the overlapped expression pattern of FAM20C and Periostin, our data demonstrate for the first time that Periostin is a direct FAM20C-binding partner and that FAM20C phosphorylates Periostin in vitro.

www.nature.com/scientificreports/ while Tenascin-C 7 , CCN3 11 and BMP1 12 through Fas I domain. These findings suggest that Periostin is a multifunctional protein possibly through orchestrating these binding proteins inside and outside of the cell 13 . FAM20C is a member of "Family with sequence similarity 20", consisting of three members; FAM20A, FAM20B and FAM20C. FAM20C, also known as DMP4 in mice 14 , is highly expressed in chondrocytes, osteoblasts, osteocytes, odontoblasts, ameloblasts and cementoblasts as well as dentin, enamel and bone matrices 15 . The expression pattern of FAM20C suggests that it has an important role in the formation of these mineralized tissues and subsequent mineralization process. Recently, FAM20C was identified as an intracellular kinase, namely golgi casein kinase (GCK), which was first described in lactating mammary glands that GCK enzymatically phosphorylates endogenous casein 16 . GCK/FAM20C phosphorylates secretory pathway proteins within Ser-X-Glu/phospho-Ser (SXE/pS) motif, which many of small integrin-binding ligand, N-linked glycoprotein (SIBLING) family members possess 17 . SIBLING proteins are known to function as nucleaters or inhibitors of biomineralization. It has been previously reported that acidic serine-and aspartate-rich motif (ASARM) peptide derived from osteopontin inhibits mineralization by binding to hydroxyapatite in a phosphorylationdependent manner 18 . FAM20C is also known as a causative gene for Raine syndrome (OMIM#259775) 19 . Raine syndrome is an autosomal recessive rare disorder characterized by generalized osteosclerosis with periosteal bone formation, manifesting distinctive facial phenotype 20 . Oral/craniofacial phenotypes have been reported that the non-lethal type of Raine syndrome patients have small teeth with enamel dysplasia 20 , enlarged gingiva 21 and amelogenesis imperfecta with significant gingivitis 22 . A previous report suggested that FAM20C is required for maintenance of healthy periodontal tissues 23 . As FAM20C functions as an intracellular protein kinase in the secretory pathway, it is reasonable to speculate that there might be some secretory proteins working together with FAM20C in PDL tissues.
Here we show that Periostin was identified during the course of FAM20C protein purification by Mass Spectrometry analysis. FAM20C interacted with Periostin in cell cultures. Using recombinant proteins, FAM20C directly bound to Periostin in vitro. We further narrowed down the binding domain of FAM20C and found that Fas I domain in Periostin is necessary for the binding to FAM20C. Immunohistochemical analysis demonstrated that immunolocalization of FAM20C was observed in murine PDL, which was overlapped with that of Periostin. Periostin was phosphorylated by FAM20C in vitro.

Results
Identification of Periostin during FAM20C protein purification process by Mass Spectrometry (MS) analysis. We intended to obtain FAM20C proteins, thus FAM20C-stablely transfected clones were first generated by transfecting FAM20C-WT-V5/His or FAM20C-D478A-V5/His expression vector into HEK 293 cells. HEK 293 cells were selected for their higher transfection efficiency.
The conditioned media from the FAM20C-WT-V5/His-transfected cell clone or the FAM20C-D478A-V5/ His-transfected cell clone were collected, and FAM20C proteins were purified by Ni-NTA purification system. The expression of FAM20C-WT-V5/His and -D478A-V5/His were confirmed by Western blotting with both anti-V5 and anti-FAM20C antibodies after purification (data not shown). The purified FAM20C-WT-V5/His and FAM20C-D478A-V5/His proteins were prepared, electrophoresed, and the gel was stained with CBB. A single CBB-positive band was then found at a molecular weight corresponding to the size of FAM20C-WT-V5/His or -D478A-V5/His protein (~ 75 kDa band in Fig. 1). The purified FAM20C-V5/His fusion protein was detected at a slightly higher molecular weight position than previously reported 17 likely due to the presence of V5/His tag (~ 5 kDa). Interestingly another band appeared at a higher molecular weight than the FAM20C-WT-V5/ His or FAM20C-D478A-V5/His protein when the amount of FAM20C-WT-V5/His (Fig. 1A, lane 3, indicated by an arrow) or -D478A-V5/His (Fig. 1A, lane 5, indicated by an arrowhead) protein applied was increased, suggesting the presence of FAM20C-binding protein(s). These CBB positive bands with broader appearance were cut (Fig. 1A, indicated by an arrow and an arrowhead), further treated with trypsin and the digested peptides were subjected to protein identification analysis. To exclude the false-positive peptides, the data were analyzed with the filter setting condition (Xcorr2; 2.2, Xcorr3; 3.5) as previously reported 24 . This analysis revealed that the peptides of FAM20C-WT and -D478A were identified with a protein coverage of 58.73% (56 unique peptides and 197 total peptides found) and 59.25% (49 unique peptides and 179 total peptides found) (  www.nature.com/scientificreports/ To determine whether the binding between POSTN and FAM20C was direct without the presence of other molecules, we examined the binding between recombinant POSTN protein produced by Sf21 baculovirus system and FAM20C-WT-V5/His or -D478A-V5/His protein purified from the conditioned media in a dose dependent manner. Proteins were incubated and the immunoprecipitation (IP)-Western blot analysis was performed. The results showed that the binding between POSTN and FAM20C-WT (    www.nature.com/scientificreports/ show higher magnification views of the open boxed areas in (a), (c), (e) or (g), respectively. Immunolocalization for FAM20C (brown color) was present in PDL (a, b), with preferential distribution at the bone surface region (b; arrows) and bone embedded Sharpey's fibres (b; arrow heads). Periostin was present (brown color) in PDL (c, d) with strong positive signals along the thick PDL collagen, both at the bone surface and cementum surface regions of PDL (d; arrows), and bone embedded Sharpey's fibres (d; arrow heads). No immunoreactivities were detected when non-immune IgG was used (e-h).

FAM20C phosphorylates Periostin in vitro.
We examined S-X-E FAM20C phosphorylation consensus sites in mouse Periostin and there were several potential phosphorylation sites (Fig. 5B). We then investigated whether Periostin protein could be phosphorylated by FAM20C in vitro. In vitro kinase assay was performed according to the previous reports 17, 25 and phosphorylated substances were enriched with Phos-tag agarose. Isolated phosphorylated proteins were separated by SDS-PAGE and detected by anti-Periostin antibody. Our results demonstrated that Periostin was detected when FAM20C WT (Fig. 5A, lane 4), but not D478A (Fig. 5A, lane 5), was incubated with Periostin. When Periostin alone was used in the absence of FAM20C protein, immunoreactivity to anti-Periostin antibody was not observed (Fig. 5A, lane 6). The data thus indicate that FAM20C phosphorylates Periostin in vitro.

Discussion
There has been increasing evidence that Periostin plays an important role in periodontal tissue development.
Periostin knock-out (KO) mice were generated and the homozygous mice showed periodontal tissue phenotypes [26][27][28] . In Periostin KO mice, PDL fibroblasts were irregularly distributed among collagen fibrils and collagen fibrils were disorganized 28 . While periodontium of WT and KO mice appeared intact when teeth were unerupted, after tooth eruption, dramatic periodontal defects were observed including enlarged gingival tissue, attachment loss, irregular PDL width, alveolar bone loss and external root resorption. Removal of masticatory forces partially rescued the PDL phenotype, suggesting the potential role of Periostin via transcription level 26 . It has also been reported that Periostin deficiency leads to decrease in collagen fibril diameter, which may explain why KO mice's PDL collagen fibrils were susceptible to occlusal force and damaged 8 . Besides the PDL defect, Periostin KO mice exhibited dwarfism and enamel defects, suggesting that Periostin plays crucial roles in bone and tooth development 27 . Despite its possible critical role of Periostin reported in certain tissues including PDL, there has been so far no gene mutations found in humans. Among the FAM20 family, it has been reported that periodontitis may be a part of clinical phenotypic spectrum in FAM20A mutations 29 . Some reports showed the presence of periodontitis in non-lethal type of Raine syndrome, where FAM20C is mutated, but genotype-phenotype correlation was not established. Although FAM20A is considered to be an allosteric activator of FAM20C kinase 30 and some clinical phenotypes in patients with FAM20A mutations are overlapped with those who with FAM20C mutations, it has not been ruled out that periodontitis in Raine syndrome patients is just coincidence. On the contrary, in mice, the involvement of FAM20C in periodontal tissues has been more extensively studied and established. The expression of Fam20C mRNA is detected in PDL fibroblasts at 5 and 7 weeks and FAM20C protein is expressed in PDL matrix in the incisor at 7 weeks 15 . Consistent with this, our data demonstrated that FAM20C is expressed in PDL tissues by immunohistochemistry (Fig. 4). More recently, conditional Fam20C KO mouse in cells expressing type I collagen (Fam20C cKO) was generated and showed periodontal disease phenotype. Since collagen fibers appeared thinner and unevenly distributed in these mice, this disrupts PDL structure, which likely allows direct infiltration of bacteria from periodontal pocket leading to periodontal disease. Additionally, in Fam20C cKO mice, the expression level of Periostin in PDL was dramatically reduced 23 . Fam20C global KO mouse (Fam20C −/− mouse) was generated by another group and demonstrated approximately 20% of mortality rate. Surviving Fam20C −/− mice showed decreased body weight and length, bone abnormalities and lack of tooth enamel 31 , manifesting similar clinical phenotypes to Raine syndrome patients. Therefore, these findings in mouse model studies and observations in human case reports highly suggest the possible relationship between FAM20C and Periostin in PDL, bone and enamel, and these two genes are genetically associated.
It has been known for almost 30 years that non-collagenous proteins in bone, dentin and enamel contain phosphoproteins with O-phosphoserine 32 . Despite the critical biological importance of phosphorylation in noncollagenous matrix proteins, kinase(s) responsible for phosphorylation have not been identified until recently. FAM20C kinase favors phosphorylation of an S-X-E/phosphorylated-S motif 17,25,33 . As illustrated in Fig. 5B, Periostin has several S-X-E motifs, showing one of potential candidate substrates for FAM20C kinase. Our data indicate that Periostin is phosphorylated by FAM20C (Fig. 5A), however, we could not identify the precise phosphorylation site(s) in Periostin using MS analysis due to lack of availability of the peptides containing these motifs (Fig. 1B). Four S-X-E motifs are found in mouse Periostin (S 140 -N-E, S 548 -E-E, S 611 -K-E, S 805 -R-E; based on NCBI Reference Sequence; NP_056599.1), of which three of them are conserved between mouse and human Periostin (S 138 -N-E, S 546 -E-E, S 609 -K-E; based on NCBI Reference Sequence; NP_006466.2). The conserved three S-X-E motifs are located within Fas I domains, i.e. S 140 -N-E in mouse RD1, S 548 -E-E and S 611 -K-E in mouse RD4. A possible reason why we could not retrieve any S-X-E motif peptides from our MS analysis is likely due to the detection limitation. As trypsin was used for peptide digestion and due to the presence of K or R residues near S-X-E motif, the length of peptide is too short to detect by the analysis. For example, S 546 -E-E containing peptide is supposed to be digested as "GMTSEEK", and S 609 -K-E containing peptide as "SK". At this point, it is still unclear why we could not detect any S 138 -N-E containing motif, however the peptide detection setting is normally 7-20 peptides and this S 138 -N-E containing peptide may be too long to detect. Further studies are required to verify the location of phosphorylation site(s) in Periostin.

Scientific Reports
| (2020) 10:17155 | https://doi.org/10.1038/s41598-020-74400-6 www.nature.com/scientificreports/ The potential molecular function of Periostin has been characterized. It has been reported that Periostin is associated with cell proliferation, migration and activation of the survival signaling pathway PI3K/AKT/ mTOR 34 . More recently, functional domain of this Periostin's biological effect has been identified and reported by two independent groups; one is that a peptide sequence (amino acids 142-151) of Periostin stimulates chemotactic migration, adhesion, proliferation and endothelial tube formation of human endothelial colony forming cells in vitro 35 and another that monoclonal antibodies that recognize amino acids 136-151 of Periostin inhibit Periostin-induced migration of human endothelial colony forming cells 36 . Interestingly, the functional domain reported in both studies is located within Fas I domain of Periostin, more specifically in RD1 which contains S 138 -N-E motif (Fig. 5B). Therefore, this suggests that phosphorylation by FAM20C may regulate Periostin-mediated cell functions and it is of our particular future interest to investigate the biological role of phosphorylated Periostin and its molecular function in both healthy periodontal tissues and periodontal disease.

Materials and methods
Ethics statement. The use of animals and all animal procedures in this study were approved by the Institutional Animal Care and Use Committee (IACUC) at Boston University Medical Campus (approved protocol number: AN-15050), and all efforts were made to minimize suffering animals. This study was performed in accordance with the NIH Guide for the Care and Use of Laboratory Animals.
Cell culture. The human embryonic kidney (HEK) 293 cells were maintained as previously described 37,38 and used in this study.
Reagents and antibodies. X-tremeGENE 9 DNA transfection reagent was obtained from Roche Life Science. Recombinant mouse Periostin protein (2955-F2) was obtained from R&D Systems. The antibodies used in this study were as follows; anti-V5 (Life Technologies), anti-HA (clone 12CA5, Roche Life Science), anti-HA high affinity (clone 3F10, Roche Life Science), goat polyclonal anti-FAM20C (Santa Cruz Biotechnology), and rabbit polyclonal anti-Periostin (ab14041, Abcam) antibody. Rabbit polyclonal anti-Periostin antibody against RD1 domain of Periostin previously generated 39 was used in the immunohistochemical analysis.

FAM20C and Periostin expression vectors. Human FAM20C expression vector constructs, including
wild-type (WT), Raine syndrome-mutant form (P328S), and a mutant form lacking its kinase activity (D478A) were generated by PCR methods as previously reported 40 . The plasmids harboring FAM20C-WT, FAM20C-D478A, and FAM20C-P328S cDNAs followed by V5-6XHis-tag (pcDNA3.1-FAM20C-WT-V5/His, Purification of FAM20C proteins. The stable transfected HEK 293 cell clones that overexpress FAM20C-WT-V5/His or FAM20C-D478A-V5/His were generated and FAM20C-V5/His proteins were purified in the same manner as previously described 41 . Briefly, cells were transiently transfected using X-tremeGENE 9 DNA transfection reagent with pcDNA3.1-FAM20C-WT-V5/His and pcDNA3.1-FAM20C-D478A-V5/His according to the manufacturer's protocol. The transfected cells were treated with 400 μg/ml of G418 neomycin analogue and further cultured. Ten of single colony-derived clones transfected with either FAM20C-WT or FAM20C-D478A were isolated, further cultured and the expression of FAM20C proteins was verified by Western blotting with anti-V5 antibody. The cell clone that expressed the strongest band intensity for FAM20C-WT-V5/His or FAM20C-D478A-V5/His was chosen, further cultured in a larger scale for protein production, and the conditioned media of FAM20C-WT-V5/His and FAM20C-D478A-V5/His were collected. The conditioned media were centrifuged at 1500 rpm for 5 min, the supernatant was incubated with Ni-NTA agarose beads (Qiagen), and FAM20C-WT-V5/His and FAM20C-D478A-V5/His proteins were purified as previously described 38,41 . The purified proteins were dialyzed against distilled water, lyophilized and resuspended in distilled water. The protein concentration was measured and the purified proteins were kept at -20 °C until use.  www.nature.com/scientificreports/ (b-)-WT-HA; POSTN-WT-HA) and FAM20C-WT-V5/His, FAM20C-D478A-V5/His, or FAM20C-P328S-V5/ His. Total amount of cDNA was kept constant by supplementation with empty vector. After 24 h of transfection, cell lysates were collected, immunoprecipitated with anti-V5 antibody and subjected to Western blotting (WB) analysis using anti-HA antibody to identify the binding. The same membrane was stripped by stripping buffer, and WB with anti-V5 antibody was performed to verify the expression of FAM20C-V5/His proteins. An aliquot of the same cell lysates was subjected to WB analysis with anti-HA antibody to verify the expression of POSTN-WT-HA.

Protein identification by mass spectrometry (MS) analysis. Various amounts of purified
In vitro binding assay. The mouse recombinant Periostin and FAM20C-WT-V5/His or FAM20C-D478A-V5/His proteins were prepared in FAM20C protein's dose dependent manner and incubated in PBS where the total amount of proteins per sample was kept constant by adding bovine serum albumin (BSA). Samples were then immunoprecipitated with anti-V5 antibody followed by Western blot (WB) analysis with anti-Periostin (Abcam) antibody. The same membrane was stripped, and WB with anti-V5 antibody was performed to verify the expression of FAM20C-V5/His proteins. Immunohistochemistry. The maxillary periodontal tissue including molars were dissected from mouse neonates (C57BL/6 strain, male, postnatal day 28), fixed, decalcified and embedded in paraffin. Immunohistochemical staining was performed as previously described 37 with anti-FAM20C antibody (1:100 dilution), anti-Periostin antibody (1:400 dilution) 39 or a non-immune goat or rabbit immunoglobulin (IgG) (at the same concentration as primary antibodies) as negative controls. The immuno-reactivity was amplified using VECTASTAIN Elite ABC HRP kit (Vector Laboratories Inc.). The sections were counterstained with hematoxylin (Sigma). www.nature.com/scientificreports/