Identification and Characterization of a Novel Protein, Periostin, with Restricted Expression to Periosteum and Periodontal Ligament and Increased Expression by Transforming Growth Factor β

Authors


Abstract

We had previously identified the cDNA for a novel protein called osteoblast-specific factor 2 (OSF-2) from an MC3T3-E1 cDNA library using subtraction hybridization and differential screening techniques. Here we describe the localization, regulation, and potential function of this protein. Immunohistochemistry using specific antiserum revealed that in adult mice, the protein is preferentially expressed in periosteum and periodontal ligament, indicating its tissue specificity and a potential role in bone and tooth formation and maintenance of structure. Based on this observation and the fact that other proteins have been called OSF-2, the protein was renamed “periostin.” Western blot analysis showed that periostin is a disulfide linked 90 kDa protein secreted by osteoblasts and osteoblast-like cell lines. Nucleotide sequence revealed four periostin transcripts that differ in the length of the C-terminal domain, possibly caused by alternative splicing events. Reverse transcription- polymerase chain reaction analysis revealed that these isoforms are not expressed uniformly but are differentially expressed in various cell lines. Both purified periostin protein and the periostin-Fc recombinant protein supported attachment and spreading of MC3T3-E1 cells, and this effect was impaired by antiperiostin antiserum, suggesting that periostin is involved in cell adhesion. The protein is highly homologous to βig-h3, a molecule induced by transforming growth factor β (TGF-β) that promotes the adhesion and spreading of fibroblasts. Because TGF-β has dramatic effects on periosteal expansion and the recruitment of osteoblast precursors, this factor was tested for its effects on periostin expression. By Western blot analysis, TGF-β increased periostin expression in primary osteoblast cells. Together, these data suggest that periostin may play a role in the recruitment and attachment of osteoblast precursors in the periosteum.

INTRODUCTION

Bone is formed by two processes: intramembranous ossification and endochondral ossification. The former contributes to the formation and growth of the flat bones of the skull and addition of bone to the periosteum of long bones, while the latter process is responsible for formation of the rest of the bones in the body.(1,2) At the histologic level, both of these processes have been well studied, but the mechanisms involved at the molecular level are not understood. It is known that the periosteum and the periosteal collar are responsible for bone formation and that this tissue contains mesenchymal stem cells and preosteoblasts.(3) This tissue also responds dramatically to bone growth factors such as transforming growth factor β (TGF-β), fibroblast growth factors, and bone morphogenetic protein with a resulting increase in new bone formation.(4,5)

The periostin/OSF-2 gene was isolated in previous studies aimed at identifying novel genes expressed specifically in the osteoblast by using the techniques of subtraction hybridization and differential screening between cDNA libraries of MC3T3-E1 and NIH3T3 cells.(6) Mouse periostin comprises 811 amino acids, and computer analysis of the deduced amino acid sequence revealed a complex protein structure with four repeats of a characteristic domain. A similar structure had been reported for fasciclin I, a homophilic cell–cell adhesion molecule expressed in the central nervous system of insects(7) and βig-h3, a molecule induced by TGF-β that promotes fibroblast attachment and spreading.(8)

In recent years, several novel proteins with a similar structure to fasciclin I have been identified in a variety of species. These include βig-h3 (human and mouse),(9) Algal-CAM (plants),(10) MPB70 (mycobacteria),(11,12) and midline fasciclin (insects).(13) These genes have a 130–150 amino acid long homologous domain, which is characterized by two conserved stretches of 13 and 14 amino acids, respectively. Fasciclin I, βig-h3, midline fasciclin, and periostin contain four homologous domains, while Algal-CAM contains two and MPB-70 consists of one. Outside the two well conserved stretches, the overall sequence homology is generally low. Although their biological functions and the underlying molecular mechanism of their effects are still poorly characterized, several reports suggest that these proteins act as adhesion molecules. βig-h3 was originally cloned as a molecule induced by TGF-β and was subsequently found to be associated with microfibrils(14) and to promote adhesion and spreading of fibroblasts.(8)

In this report, we describe a potential function for periostin as a cell adhesion molecule for preosteoblasts based on its localization in tissues and capacity to induce osteoblast attachment and spreading. This novel gene was previously called “osteoblast-specific factor 2” or OSF-2,(6) but a subsequent study also used this acronym to describe a transcription factor also known as Cbfa1.(15) Therefore, based on its tissue specificity, we have renamed this protein “periostin.” Herein we describe the characterization of this protein and regulation of its expression by TGF-β.

MATERIALS AND METHODS

Cells and cell culture

Primary osteoblasts were prepared from calvaria of newborn ddY mice as previously described.(16) Primary osteoblasts, MC3T3-E1 (mouse calvaria-derived osteoblast-like cell line) and ST2 (mouse bone marrow derived-stromal cell line) were grown in alpha modified Eagle's medium (ICN Biochemicals, Costa Mesa, CA, U.S.A.). Ltk (mouse fibroblastic cell line), C3H10T1/2 (mouse embryonal cell line), NIH3T3 (mouse fibroblastic cell line), and 3T3-L1 (mouse embryo-derived preadipose-like cell line) were grown in Dulbecco's modified Eagle's medium (DMEM; Nissui Pharmaceutical Co., Tokyo, Japan). Each media was supplemented with 10% fetal calf serum (FCS; Dainippon Pharmaceutical Co., Tokyo, Japan), penicillin G (100 U/ml), and streptomycin (100 μg/ml). ATDC5 (mouse embryonal carcinoma-derived chondrogenic cell line) was grown in a 1:1 mixture of DMEM and F12 medium (Nissui Pharmaceutical Co.) containing 5% FCS, 10 μg of human transferrin (Boehringer Mannheim GmbH, Mannheim, Germany) and 3 × 10–8 M sodium selenite as described previously.(17) Cells were incubated at 37°C in a humidified atmosphere of 5% CO2 in air.

Treatment with TGF-β

Primary osteoblasts isolated from mouse calvaria were incubated with 5–20 ng/ml of human TGF-β1 for 48 h. The cells were then harvested and the cell lysate was prepared for SDS-PAGE. Periostin levels were determined by Western blot analysis as described below. Western blots for glyceraldehyde-3-phoshpate dehydrogenase (GAPDH) was used as a control. Human TGF-β1 was purchased from Immuno-Biological Laboratories (Gunma, Japan), and anti-GAPDH antibody was purchased from Biogenesis (Sandown, NH, U.S.A.).

Production of mouse periostin antisera

A peptide (NKRVQGPRRRSREGRSQ) homologous to a hydrophilic region of mouse periostin was synthesized using the Model 430 A peptide synthesizer (Applied Biosystems, Foster City, CA, U.S.A.). For primary immunization of rabbits, the peptide was conjugated to bovine serum albumin (BSA) using the glutaraldehyde method. For primary immunization, 1.0 mg of conjugated peptide in phosphate-buffered saline (PBS) containing 10 mM sodium phosphate (pH 7.3) and 150 mM NaCl, was diluted 1:2 in Freund's complete adjuvant (Difco Laboratories, Detroit, MI, U.S.A.). Three successive immunizations were performed with 200 μg of nonconjugated peptide in Freund's incomplete adjuvant at monthly intervals. The polyclonal antiserum was affinity purified using a CNBr-sepharose 4B (Amersham Pharmacia Biotech, Buckinghamshire, U.K.) conjugated with the synthetic peptide. To completely remove the remaining anti-BSA antibodies, the antiserum was passed through the BSA-conjugated CNBr-sepharose column twice.

Western blot analysis

The conditioned medium of MC3T3-E1 was concentrated 8-fold using a Centriflo ultrafiltration membrane cone CF25 (Millipore Corp., Ann Arbor, MI, U.S.A.) and then dissolved in SDS electrophoresis buffer (100 mM Tris-HCl, 4% SDS, 0.2% bromophenol blue, 20% glycerol) either with or without the addition of dithiothreitol (0.1 M) for separation by SDS-PAGE. For the samples of cell lysate, primary osteoblasts and cell lines described above were used. Cultured cells were harvested with a cell scraper and dissolved in lysis buffer (2% NP-40, 20 mM Tris-HCl, pH 8.0, 150 mM NaCl, 5 mM MgCl2, 5 mM EDTA, 2 mM NaN3 and 1 mM phenylmethylsulfonyl fluoride). The supernatant was recovered and dissolved in SDS electrophoresis buffer.

Samples were separated by SDS-PAGE (7.5%). Prestained molecular weight markers were used as standards. Proteins were blotted onto nitrocellulose membranes (Schleicher & Schuell, Dassel, Germany) using a semidry electroblotter. The membranes were blocked with PBS containing 5% nonfat dry milk overnight at 4°C. Subsequently the membranes were incubated at room temperature for 1 h with antiperiostin antisera, and bound antibodies were detected with a horseradish peroxidase–conjugated goat anti-rabbit immunoglobulin G (IgG) antibody (Southern Biotechnology Associates, Birmingham, AL, U.S.A.). The membrane was exposed to X-ray film (Kodak, Rochester, NY, U.S.A.) using the enhanced chemiluminescence system (Amersham Pharmacia Biotech).

Purification of the mouse periostin

Approximately 2 l of conditioned media of MC3T3-E1 cells was diluted with 1 l of 20 mM PBS, pH 6.4 and filtered using a 0.22 μm membrane filter (Millipore Corp.). The diluted conditioned medium was applied to a HiLoad SP Sepharose FF column (16 × 100 mm) equilibrated with 20 mM PBS, 0.1 M NaCl, pH 6.4. After the column was washed with the same buffer, proteins were eluted with a step-wise gradient from 50–1000 mM NaCl. The eluate was applied to a Supredex 200 pg column (16 × 600 mm) equilibrated with 20 mM PBS, 0.1 M NaCl, pH 7.8, and then to a Hi-trap Q column (1 ml) equilibrated with 20 mM PBS, 50 mM NaCl, pH 7.8. The column was washed with the same buffer and then eluted with a linear gradient from 0.05–0.25 M NaCl. The purification was monitored by SDS-PAGE followed by silver stain and Western blot analysis of each fraction. The positive fractions were pooled. The protein concentration was determined with the DC protein assay kit (Bio-Rad Laboratories, Hercules, CA, U.S.A.) using albumin (Pierce, Rockford, IL, U.S.A.) as a standard. All columns used were purchased from Amersham Pharmacia Biotech.

Construction and expression of the recombinant periostin protein

The coding residue 1552–2451 of periostin was amplified by polymerase chain reaction (PCR) using the following primers: 5′-AGAGGAAGCAAGCAGGGAA-AGGA-3′ (forward direction) and 5′-TTCTGCAGGAATTCGGATCCTGAGAACGGCCTTCTCTTGA-3′ (reverse direction). The PCR product was digested with PstI and replaced with the periostin coding residue 1582–3077, resulting in the deletion of residue 2452–3077 and the addition of an artificial BamHI, EcoRI, and PstI adapter. The expression vector pFC362(18) (provided by Dr. Satoshi Nakamura, Tokyo Institute of Technology, Japan), which contains a cDNA of human IgG Fc, was digested with BamHI, and then the IgG Fc fragment (686 bp) was inserted through the artificial BamHI adapter. The periostin-Fc chimeric sequences were cloned into the expression vector pCXN2 (provided by Dr. Jun-ichi Miyazaki, Osaka University, Japan) which contains chicken β-actin promoter. Expression vectors were transfected into Ltk cells using Lipofectamine (GIBCO Laboratory, Rockville, MD, U.S.A.) and transfectants were selected in 400 μg/ml G418 sulfate (Calbiochem-Novabiochem Corp., La Jolla, CA, U.S.A.). The conditioned medium containing the periostin-Fc recombinant protein was purified by affinity chromatography using Protein A Sepharose 4 Fast Flow (Amersham Pharmacia Biotech) as described previously.(19)

Reverse transcription PCR and sequence analysis

Total RNA was extracted from cultured cells using the guanidinium thyocyanate procedure.(20) Reverse transcription was performed using the first-strand cDNA synthesis kit as instructed by the manufacturer (Life Sciences, St. Petersburg, FL, U.S.A.). Each RNA (1 μg) sample was reverse-transcribed with 10 U of reverse transcriptase of avian myoloblastosis virus. Transcribed cDNAs were used to amplify periostin genes and the GAPDH gene by PCR. Primers of periostin and GAPDH(21) and expected sizes of each PCR product are shown in Fig. 3A. cDNA samples were amplified using Taq DNA polymerase (Takara Biomedicals, Shiga, Japan). The PCR conditions were 95°C for 1 minute, 60°C for 1 minute, and 72°C for 1 minute for 25 cycles. PCR products were run on 2% agarose gels and stained with ethidium bromide along with a molecular size marker.

Figure FIG. 3..

Diagram of periostin cDNA and the positions of the primers used in RT-PCR analysis (A). Expression of periostin mRNA in cell lines analyzed by RT-PCR (B). The repeat domain 1 (RD1) and the C-terminal domain were amplified as described in the Materials and Methods. GAPDH is the positive control. Higher levels of expression are observed in primary calvarial osteoblasts (COB), ATDC5, and MC3T3-E1. Amplification products of the C-terminal domain consist of several bands indicating heterogeneity in this domain. (C) Expression of periostin protein in cell lines by Western blot analysis. Cell extracts were separated by SDS-PAGE under reducing conditions and detected using antiperiostin antiserum. High levels of expression are observed in COB, ATDC5, and MC3T3-E1 cells. These experiments show correlation between mRNA levels and protein expression.

PCR products of the C-terminal domain were cloned, and the nucleotide sequences were determined using the Li-Cor 4000 l automated DNA sequencer (Aloka, Tokyo, Japan).

Immunohistochemistry

For adult mouse specimens, 5-week-old ddY mice were used. After intracardiac perfusion with 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4), the mandible and the lower limbs were dissected and immersed in the same fixative for 5 h. The specimens were decalcified in 10% EDTA for 7 days at 4°C and embedded in paraffin wax. Sections were cut at about 5 μm and mounted onto glass slides. The sections were preincubated with 1% BSA in PBS for 30 minutes, then incubated for 1 h with either antiperiostin antisera or anti-tissue nonspecific alkaline phosphatase (ALP) antisera (provided by Dr. Kimimitsu Oda, Niigata University, Niigata, Japan) diluted 1:100 in 1% BSA in PBS. The sections were washed three times with PBS and incubated for 30 minutes with an horseradish peroxidase–conjugated goat anti-rabbit IgG antibody diluted 1:100, as a secondary antibody. The sections were washed three times and incubated with DAB solution (0.05% diaminobenzidine and 0.02% H2O2 in 0.05 M Tris-HCl buffer). After development of the color reaction, sections were washed with PBS three times and counterstained with methyl green.

RNA in situ hybridization

Paraffin sections were collected on poly L-lysine coated glass slides, deparaffinized with xylene, and rehydrated in decreasing concentrations of ethanol prior to rinsing with PBS. Sections were then treated with 4% paraformaldehyde in 0.1 M phosphate buffer for 15 minutes and then with proteinase K (10 μg/ml) in 10 mM Tris-HCl (pH 8.0) at 37°C for 15 minutes followed by incubation with 0.2 M HCl for 10–15 minutes. Acetylation of the sections was performed by incubation for 10 minutes with 0.25% acetic anhydride in 0.1 M triethanolamine (pH 8.0). For RNA in situ hybridization, periostin cRNA was used as described previously.(22) RNA probes of the antisense periostin were prepared with digoxigenin-labeled UTP. A hybridization mixture (50% formamide, 10 mM Tris-HCl, pH 7.6, 100 μg tRNA, 1× Denhardt's solution, 10% dextran sulfate, 600 mM NaCl, 0.25% SDS, and 1 mM EDTA) was preheated for 10 minutes at 90°C. Each RNA probe was adjusted to a concentration of 0.1–1.0 μg/ml. The mixture was denatured by heating at 90°C for 2–3 minutes and then applied to the sections. Hybridization was performed overnight at 50°C. After hybridization, sections were washed with 50% formaldehyde in 2× SSC at 50–55°C for 30 minutes and treated at 37°C with a solution of 10 mM Tris-HCl (pH 8.0), 0.5 M NaCl, and 1 mM EDTA (TNA). Nonspecific binding of probes was reduced by RNAse A treatment (20 μg/ml in TNE solution) at 37°C for 30 minutes. Visualization was performed using nitroblue tetrazolium salt and 5-bromo-4-chloro-3-indolylphosphate.

Solid-phase binding assay

Solid-phase binding assay was performed essentially as described previously(23) with some modifications. Tissue culture plastic dishes with 96 wells were coated overnight at 4°C with proteins diluted in PBS. Along with the purified periostin protein and the periostin-Fc recombinant protein, a bovine fibronectin (Nacalai Tesque, Tokyo, Japan) was used as a positive control. Following several washes with PBS, the wells were blocked with 3% BSA in PBS for 1 h at room temperature. Wells coated with BSA alone served as negative controls. Following the incubation with BSA, wells were washed twice with PBS. Cultured cells were harvested by treatment with 0.05% trypsin and 0.02% EDTA in PBS and rinsed twice with 2% FCS in PBS and once with FCS-free medium. The cells were then resuspended in FCS-free medium and incubated in the protein coated wells for 90–120 minutes at 37°C. Wells were examined by phase contrast microscopy and photographed. Spread cells (defined as having at least one process extension of at least one cell diameter) and round cells were enumerated and the percentage of spread cells was calculated. At least 100 cells were counted for each determination. To determine the effect of antibody on cell spreading, we used antiperiostin antisera and rabbit antimaltose binding protein (anti-MBP) antisera (New England BioLabs, Beverly, MA, U.S.A.) as a negative control. Antisera were diluted 1:10 in PBS and incubated in the periostin (30 μg/ml) or fibronectin (5 μg/ml)-coated wells blocked with 3% BSA, for 1 h at room temperature.

RESULTS

Periostin is a 90 kDa secreted protein

Our previous study revealed that periostin transcripts are abundantly expressed in MC3T3-E1 cells, but no study was performed examining protein expression. If one examines the diagram in Fig. 1, the structural homology between periostin and insect fasciclin I is striking. Periostin contains two highly conserved sequences maintained in fasciclin I and in other members of this family (Fig. 1B). To determine if this protein may play a role in cell adhesion, it was necessary to first develop tools such as specific antibody and recombinant protein.

Figure FIG. 1..

A diagram of the structure of proteins in the fasciclin I family. Two highly conserved sequences in each repeated domain (RD) are indicated as black boxes (a) and hatched boxes (b), respectively, in (A). Comparison of the highly conserved sequences among the members of the fasciclin I family and the consensus sequence (B). The proteins shown are mouse periostin, Drosophila fasciclin I, human βig-h3, Volvox Algal-CAM, and mycobacterium MPB70. The conserved consensus sequence is shown at the bottom.

Antiserum was raised against a peptide homologous to a sequence only found in mouse periostin and was found to be useful for Western blot analysis. From the cDNA sequence, which encodes a typical N-terminal signal sequence and lacks coding for a transmembrane domain, it would be predicted that periostin would be a secreted protein. This was verified by Western blot analysis. The conditioned medium of MC3T3-E1 cells was concentrated 8-fold and separated by SDS-PAGE. Figure 2 shows a thick doublet is present suggesting that a large amount of periostin protein is present in the conditioned medium. Under reducing conditions, the apparent molecular weight of the periostin protein was ∼90 kDa, and was in good accordance with the calculated molecular weight of 90,254, indicating that the glycosylation of periostin protein is, if it occurs at all, not significant. The mobility of the protein was slightly retarded upon reduction, suggesting the presence of internal disulfide bond(s). On SDS-PAGE, the stained bands appeared rather broad and could be resolved into two or three close bands, suggestive of the existence of closely related isoforms. In a previous study, we showed that in human cells, several variations were present in the C-terminal domain of periostin, supposedly caused by alternative splicing events.(6) The present results indicate that such diversity also exists in mouse.

Figure FIG. 2..

Western blot analysis of the conditioned medium of MC3T3-E1 using antiperiostin antibody. Negative control, secondary antibody only (A) and the antiperiostin antiserum (B). Conditioned medium was concentrated 8-fold and separated by SDS-PAGE under either nonreducing (nr) or reducing conditions (r). The immunoreactive bands, which consist of two closely placed bands, were observed in the lanes with the antiperiostin antiserum. The apparent molecular weight is about 90 kDa under reducing conditions. Under reducing conditions, the migration of the bands is slightly retarded indicating the presence of internal disulfide bond(s).

Periostin mRNA is strongly expressed in osteoblastic and precursor cell lines

To investigate whether mesenchymal-derived cell lines express periostin transcripts, RT-PCR was performed using mouse calvaria-derived osteoblast-like cell line, MC3T3-E1, mouse embryonal cell line, C3H10T1/2, mouse muscle-derived myogenic cell line, C2C12, mouse fibroblastic cell line, NIH3T3, mouse bone marrow-derived stromal cell line, ST2, mouse embryonal carcinoma-derived chondrogenic cell line, ATDC5, mouse embryo-derived preadipose-like cell line, 3T3-L1, and primary calvarial osteoblasts from a newborn mouse (COB). For PCR, we designed two sets of primers to examine possible variations in the C-terminal domain (see Fig. 3A). One set of primers amplifies repeat domain 1 (RD1), and the other the C-terminal domain, as shown in Fig. 3A. High levels of expression were observed in MC3T3-E1, ATDC5, and COB cells and to a lesser extent in C3H10T1/2 and 3T3-L1 cells (Fig. 3B). In C2C12, NIH-3T3, and ST2 cells, very little expression was detected even when the number of PCR cycles was increased to 33 (data not shown). Each PCR product of the C-terminal domain consisted of three or four closely migrating bands, suggesting possible diversity or truncation in this domain. The expected length of the PCR product from the previously published periostin cDNA is 547 bp, so it is likely that both longer and shorter isoforms exist.

We also performed Western blot analysis to confirm these PCR findings at the protein level. Extracts of cells were separated under reducing conditions by SDS-PAGE and the protein was detected as described in the Materials and Methods. As shown in Fig. 3C, the levels of protein was almost identical to levels of mRNA as detected by RT-PCR of RD1 in Fig. 3B, demonstrating a high level of expression only in MC3T3-E1, ATDC5, and COB cells. Minor bands, supposedly the isoforms, appeared when the membrane was exposed for a longer period of time.

The structure and sequence analysis of isoforms

The PCR products of C-terminal domain were cloned, and we determined their nucleotide sequences and found four different sequences. Figure 4A represents the nucleotide sequence of the C-terminal domain of the longest isoform (isoform 1 in Fig. 4B). The sequence could be divided into at least six sections (designated as sections a, b, c, d, e, and f in Fig. 4A), and as shown in Fig. 4B, each isoform consists of four to six of these sections, suggestive of alternative splicing events. Because the number of nucleotide residues of each fragment is a multiple of three, deletion or insertion of any of these will not lead to a frame-shift in translation. One of the isoforms of human periostin (hOSF-2 pl)(6) corresponds to a deletion of section b and a part of section c in Fig. 4A. Thus, section c may be further divided into two fragments. The calculated molecular weight of these isoforms ranges from 93,159 to 87,181 and this may explain the broad bands of Western blots (Figs. 2, 3C, and 8).

Figure FIG. 4..

Nucleotide sequence analysis of C-terminal domain of mouse periostin. The nucleotide sequence of C-terminal domain of isoform 1 (1918–2532 bp; the longest isoform so far found) is shown (A). The sequence structure of C-terminal domain of four isoforms (B). Isoform 2 represents the structure of previously published periostin/OSF-2.(6) The length of each C-terminal domain and calculated molecular weight of each isoform are also shown.

Figure FIG. 8..

TGF-β increases the expression of periostin by osteoblast cells as determined by Western blot analysis. Primary osteoblasts were incubated with TGF-β1 for 48 h as described in the Materials and Methods. TGF-β dose dependently (5–20 ng/ml) increases the proposed isoforms of periostin, especially the lower band that is difficult to detect except in the presence of TGF-β.

High levels of expression of periostin are observed in the periosteum but not in the endosteum and bone matrix

We had previously investigated the tissue specific expression of periostin by RNA dot-blot analysis and showed that periostin is only expressed in primary calvarial osteoblasts and in MC3T3-E1 cells.(6) Although a weak signal was observed in the lung, in no other tissues including the brain, heart, kidney, liver, muscle, placenta, spleen, testis, and thymus, could the transcripts for periostin be detected. To evaluate the expression pattern of periostin in bone tissue in detail, immunohistochemistry was performed using antiperiostin antiserum.

Sections of the tibia of a 5-week-old mouse revealed strong positive staining in the periosteum but not in endosteum nor bone matrix (Fig. 5A). At high magnification, a positive reaction was observed in the cambial layer and to a lesser extent in the fibrous layer of the periosteum (Fig. 5B). The staining pattern indicates that the periostin protein is not anchored to the surface of the cell but secreted into the surrounding extracellular matrix possibly by osteoblasts or osteoblast precursors. Although ALP positive cells, possibly preosteoblasts and mature osteoblasts, were present both in the periosteum and on the surface of trabecular bone (Fig. 5C), expression of periostin was all but restricted to the ALP positive cells in the periosteum. No expression was observed in cartilaginous tissue, including the articular surface, growth plate, and calcified cartilage. Essentially identical findings were observed using in situ hybridization with the periostin RNA antisense probe (Fig. 5D). The periosteal osteoblasts were positive for periostin mRNA.

Figure FIG. 5..

Femur of a 5-week-old mouse immunostained for periostin (A, ×100; B, ×400) or ALP (C, ×100), and hybridized with periostin antisense probe (D, ×400). At low magnification, strong positive reaction is observed in periosteum (arrowheads) but not in endosteum or in the bone matrix. No staining is found in cartilage tissue (arrows) (A). High magnification shows high levels of expression of periostin in the cambial layer and to a lesser extent in the fibrous layer (B). A very weak reaction is noted in the cortical bone. The contour of the cells is not sharply outlined. ALP positive cells, probably preosteoblasts and mature osteoblasts, are observed in periosteum (arrowheads) and the surface of trabecular bone (arrows) (C). Periostin transcripts are observed in the cells of periosteum, consistent with the findings of immunohistochemical staining for the protein (D). po, periosteum; cb, cortical bone; tb, trabecular bone.

Expression of periostin in tooth

Immunohistochemistry of sections of a 5-week-old mouse mandible revealed high levels of periostin expression in the periodontal ligament (Fig. 6A). The contour of the cells is not as sharply outlined as in periosteum, indicating that the periostin protein is present in the extracellular matrix and secreted possibly from periodontal fibroblasts (Fig. 6B). No positive reaction was observed in enamel, dentin, cementum, dental pulp, and alveolar bone. The cells surrounding the blood vessels, possibly endothelial cells, showed little or no signal.

Figure FIG. 6..

Mandibular bone of a 5-week-old mouse immunostained for periostin (A, ×100; B, ×400). At low magnification a positive reaction is clearly noted in periodontium but not in either alveolar bone or tooth (A). A positive reaction is also observed in periosteum (arrowheads) of alveolar bone as observed in periosteum of the femur. High magnification of periodontium showing high levels of periostin expression in periodontal ligament (B). Endothelial cells of blood vessel show little staining (arrows). to, tooth; pl, periodontal ligament; ab, alveolar bone.

Periostin supports cell spreading and attachment in vitro

Solid-phase binding assays were performed using MC3T3-E1 cells, and the purified periostin protein from the conditioned media of MC3T3-E1 cells and the periostin-Fc recombinant protein. MC3T3-E1 cells were trypsinized and incubated in wells coated with either purified periostin (a, 20 μg/ml), recombinant periostin-Fc protein (b, 30 μg/ml), fibronectin (c, 5 μg/ml), or BSA alone (d) as described in the Materials and Methods. The cells spread on the periostin, periostin-Fc- and fibronectin-coated wells, but not on the BSA alone coated wells (Fig. 7A).

Figure FIG. 7..

Solid-phase binding assays showing adhesion of MC3T3-E1 cells to immobilized periostin protein. MC3T3-E1 cells were trypsinized and incubated in wells coated with either purified periostin (a, 20 μg/ml), recombinant periostin-Fc protein (b, 30 μg/ml), fibronectin (c, 5 μg/ml), or BSA alone (d) as described in the Materials and Methods (A). The cells spread on the periostin-, periostin-Fc–, and fibronectin-coated wells, but not on the BSA alone coated wells. MC3T3-E1 cells attach to periostin-Fc–coated wells in a concentration-dependent manner (B). Trypsinized MC3T3-E1 cells were incubated in wells coated with periostin-Fc recombinant protein at the indicated concentrations. Spread cells and rounded cells were enumerated, and the percentage of cell spreading was calculated. Antiperiostin antiserum interfered with cell spreading on periostin-Fc (30 μg/ml)–coated wells but not with cell-spreading on fibronectin (5 μg/ml)-coated wells (C). Protein-coated wells were incubated with either antiperiostin antiserum (1:10) or rabbit anti-MBP antiserum (1:10) for 1 h at 37°C prior to the plating of MC3T3-E1 cells. The negative control, anti-MBP antiserum had no effect on either periostin-Fc– or fibronectin-coated wells. Data represent the average and standard error of the mean for three wells. At least two independent experiments were performed with similar results obtained.

As shown in Fig. 7B, on periostin-Fc coated plates, cell spreading occurred in a concentration-dependent manner. Periostin also supported cell attachment under reducing conditions, so secondary structure is not essential for this effect (data not shown). An addition of 5 mM EDTA completely inhibited cell adhesion on both periostin- and fibronectin-coated plates, while GRGDS peptide (1 mM) had no effect (data not shown).

Antiperiostin antiserum impaired cell spreading on the periostin-Fc–coated plates, while on fibronectin-coated plates no significant effect was observed (Fig. 7C). The negative control, rabbit anti-MBP antiserum did not interfere with the cell spreading on either periostin-Fc– or fibronectin-coated plates. Similar results were observed using ST2 cells, which express few periostin transcripts as described above (data not shown).

TGF-β increases expression of periostin in osteoblast cells

Because TGF-β has been shown to increase matrix formation and stimulate new bone formation (for review see Bonewald, 1996(24)) and stimulates periosteal expansion and osteoblast recruitment,(4) it was tested on the expression of periostin by primary murine calvarial osteoblast cells. As can be seen in Fig. 8, TGF-β increased all three bands recognized by the antiserum, but especially the lower band. The significance of an increase in this proposed isoform of periostin remains to be determined.

DISCUSSION

Here we describe the localization, regulation, and potential function of a novel protein, periostin, the first protein to be described with tissue specificity to the periosteum and the periodontal ligament. The localization of periostin is unique and no other glycoprotein including osteocalcin, vitronectin, osteopontin, fibronectin, nor thrombospondin has a similar distribution pattern.(25) The expression of this protein is also regulated by TGF-β and this protein has the potential to be a tissue-specific mediator of the effects of TGF-β on new bone formation.

Immunohistochemical analysis showed that periostin is preferentially expressed in the extracellular matrix of the periosteum and the periodontal ligament in the adult mouse. Periosteum and periodontal ligament are both highly vascular and cellular connective tissues that play critical roles in mineralized tissue generation and support. Periosteum covers the outer surface of bone and contributes to the growth in diameter of bone and in membranous bone formation. Mesenchymal cells are present in the periosteum and can differentiate directly into osteoblasts, but also have the potential to differentiate along the chondrogenic pathway in the process of fracture healing.(26,27) The periodontal ligament is situated between tooth and alveolar bone and supports the attachment of teeth to the alveolar bone. Like the mesenchymal cells in the periosteum, periodontal fibroblasts are able to differentiate into multiple types of cells and appear to play an essential role in response to mechanical forces of the tooth and in repair of damaged matrix.(28) Localization of periostin to these tissues suggests a role in mineralized tissue formation.

Markers described for preosteoblasts such as ALP are also found on mature osteoblasts.(29) However, periostin appears to be a specific marker for the preosteoblast. Not all of the ALP positive cells in bone express periostin, only those ALP positive cells in the periosteum. Periostin also appears to be secreted into the surrounding extracellular matrix. Because strong expression of periostin is observed in periosteal osteoblasts and not in the osteoblasts lining the trabecular bone where bone remodeling is actively taking place, it is assumed that periostin is not directly but indirectly involved in osteogenesis or in the process of bone formation or repair. Periostin expression in osteocytes was nearly undetectable, suggesting that osteoblasts cease production of periostin during the process of differentiation. A recently described osteocyte cell line, MLO-Y4, lacks detectable periostin transcripts,(30) which further supports this observation.

Expression of periostin was found in ATDC5 cells, which undergo chondrogenic differentiation with the addition of insulin(17) and also were found in the primary culture of rat growth plate of the rib (data not shown). However, there was no detectable expression of periostin in either cartilage tissue or sclerotome during embryonic development (data not shown). This opposing observation suggests that in vitro, expression of periostin may be deregulated in some cells in a similar fashion to that observed with osteonectin (also designated as “culture shock protein”), resulting in expression by cells in culture that would not be expressed in vivo.(2)

The amino acid sequence of periostin has homology with that of fasciclin I, a protein expressed on the surface of a subset of axon pathways in the embryonic central nervous system in insects.(7,31) In recent years, novel proteins with a similar structure have been identified one after another in a wide range of species, such as βig-h3 (human and mouse),(9) Algal-CAM (alga Volvox),(10) MPB70 (mycobacteria),(11) and midline fasciclin (Drosophila).(13) All of these appear to have some role in cell adhesion. Fasciclin I supports cell aggregation and mediates cell sorting,(32) and disruption of fasciclin I causes defects in axonogenesis.(33) Null mutation of midline fasciclin, a recently cloned gene in Drosophila, leads to subtle defects in axonogenesis as seen in disruption of fasciclin I.(13) Algal-CAM is localized in cell–cell contacts in embryo, and specific antibodies interfere with cell contact formation.(10) The function of MPB70, a secreted protein of mycobacteria, is unknown.(12) Among the genes in the fasciclin I family, the homology between periostin and βig-h3 is relatively high. The comparison of amino acid sequence reveals identity in each domain ranging from 59% to 36%. βig-h3, which was originally cloned as a molecule induced by TGF-β, promotes adhesion and spreading of fibroblasts in vitro,(8) and may be associated with the microfibrils in vivo.(14) Homology of periostin with members of this family suggest a function as a cell adhesion molecule.

In this study, we showed that both the purified periostin and the periostin-Fc recombinant protein supported cell adhesion and spreading in vitro. Although the members of the fasciclin I family are basically implicated in cell adhesion, the mechanism of this adhesion itself and the existence of a possible signaling pathway are poorly understood except for fasciclin I in which a signal transduction pathway has been described involving the Abelson tryosine kinase.(33) It is likely that periostin is, like the other members of the fasciclin I family, not only involved in cell adhesion but in signal transduction.

Periostin has one possible N-glycosylation site, but the apparent molecular weight revealed by SDS-PAGE is in good accordance with the calculated molecular weight of 90,254, suggesting that glycosylation is not a major component of the mature protein. Periostin lacks a transmembrane domain, as do the other members of the fasciclin I family, and is secreted into medium by cultured cells. Although no member of the fasciclin I family contains a typical transmembrane domain, fasciclin I, Algal-CAM, and midline fasciclin are thought to be anchored to the cell surface.(10,13,34) Fasciclin I has a phosphatidylinositol lipid membrane anchor and is tethered to the cell surface, but during Drosophila development a significant amount of fasciclin I protein is also present in a free, soluble form.(34) The biological functions of this secreted form as well as its regulation are not well understood. The secreted form of fasciclin I could be the result of a cleavage of phosphatidylinositol lipid linkage as a way of down-regulating cell adhesion. It is not known at this time whether periostin is inserted into preosteoblast membranes in a similar manner. Periostin also contains a Leu-Arg-Glu (LRE) motif shown to be an adhesive site in S-laminin, specifically for motor neurons and the peptide inhibits neurite outgrowth.(35,36) The significance of this motif in periostin remains to be determined.

In a previous study, we showed that there are variations in the C-terminal domain in human periostin possibly caused by alternative splicing events.(6) The results in this study revealed that multiple isoforms also exist in mouse. We found four isoforms and determined their nucleotide sequence. The C-terminal domain of mouse periostin could be divided into at least six sections. Insertions and deletions of these sections all occurred in-frame, therefore no frame shifts in the transcripts were observed. The patterns of alternatively spliced periostin mRNA differed among the cell lines, indicating that the alternative splicing of periostin transcripts are under developmental control and that these isoforms may have different functions. Multiple isoforms (supposedly caused by alternative splicing events) are also found in fasciclin I,(37) Algal-CAM,(10) and βig-h3.(14,38) Since these variant forms are commonly seen among the members of the fasciclin I family, this could be another feature of this family. In the present study, TGF-β increased all three isoforms detected by Western blot analysis. A more dramatic increase was observed in the lower molecular weight isoform from almost undetectable to the same intensity as the two upper bands. The characterization of these isoforms and their regulation in vitro and in vivo by TGF-β are the focus of present and future studies.

Acknowledgements

We thank Dr. Kiyotoshi Sekiguchi (Osaka University, Japan) for advice with the cell adhesion study. This work was supported by grants from the Ministry of Education, Science and Culture of Japan and Mitsubishi Foundation.

Ancillary