Periostin is expressed within the developing teeth at the sites of epithelial–mesenchymal interaction


  • Agnieszka Kruzynska-Frejtag,

    1. Institute of Molecular Medicine and Genetics, Medical College of Georgia, Augusta, Georgia
    Search for more papers by this author
  • Jian Wang,

    1. Cardiovascular Development Group, Herman B. Wells Center for Pediatric Research, Indiana University School of Medicine, Indianapolis, Indiana
    Search for more papers by this author
  • Manabu Maeda,

    1. Cardiovascular Development Group, Herman B. Wells Center for Pediatric Research, Indiana University School of Medicine, Indianapolis, Indiana
    Search for more papers by this author
  • Rhonda Rogers,

    1. Cardiovascular Development Group, Herman B. Wells Center for Pediatric Research, Indiana University School of Medicine, Indianapolis, Indiana
    Search for more papers by this author
  • Edward Krug,

    1. Department of Cell Biology and Anatomy, The Medical University of South Carolina, Charleston, South Carolina
    Search for more papers by this author
  • Stanley Hoffman,

    1. Department of Cell Biology and Anatomy, The Medical University of South Carolina, Charleston, South Carolina
    Search for more papers by this author
  • Roger R. Markwald,

    1. Department of Cell Biology and Anatomy, The Medical University of South Carolina, Charleston, South Carolina
    Search for more papers by this author
  • Simon J. Conway

    Corresponding author
    1. Institute of Molecular Medicine and Genetics, Medical College of Georgia, Augusta, Georgia
    2. Cardiovascular Development Group, Herman B. Wells Center for Pediatric Research, Indiana University School of Medicine, Indianapolis, Indiana
    • Riley Hospital for Children, 702 Barnhill Drive, Room 2614, Indiana University School of Medicine, Indianapolis, IN 46202
    Search for more papers by this author


Periostin was originally isolated as an osteoblast-specific factor that functions as a cell adhesion molecule for preosteoblasts and is thought to be involved in osteoblast recruitment, attachment, and spreading. The protein was renamed “periostin” because of its expression in the periosteum and periodontal ligament, indicating a potential role in bone and maintenance of tooth structure. Periostin has structural similarity to insect fasciclin-I and can be induced by TGF-β and Bmp2. Because tooth and periodontium development is a well-described genetic model for organogenesis governed by a reciprocal set of epithelial–mesenchymal interactions, thought to be controlled by various TGF-β superfamily members, we investigated whether periostin is present during tooth morphogenesis. Both periostin mRNA and protein expression were analyzed throughout normal tooth development (embryonic day [E] 9.5–newborn) and within both Bmp4- and Msx2-null embryos. Periostin mRNA is initially present within the E9.5 first branchial arch epithelium and then shifts to underlying ectomesenchyme. Both mRNA and protein are asymmetrically localized to the lingual/palatal and buccal side during the early epithelial–mesenchymal interactions. Periostin is also present in dental papilla cells and within the trans-differentiating odontoblasts during the bell and hard tissue formation stages of tooth development. We suggest that periostin plays multiple roles as a primary responder molecule during tooth development and may be linked to deposition and organization of other extracellular matrix adhesion molecules during maintenance of the adult tooth, particularly at the sites of hard-soft tissue interface. Developmental Dynamics 229:857–868, 2004. © 2004 Wiley-Liss, Inc.


Organogenesis requires the establishment of multiple signaling pathways and physical interactions between adjacent tissues for effective morphogenesis to occur. The initial morphologic development of all epithelial organs (including whiskers, hair follicles, mammary glands, and teeth) is similar and it is governed by epithelial–mesenchymal interactions (Dassule and McMahon, 1998). Recently, tooth morphogenesis is beginning to be understood at the gene level (Thesleff, 2000). Since the developmental regulatory genes have been conserved to a high degree during evolution and similar gene networks regulate the development of teeth as of other vertebrate organs, the developing mouse tooth germ provides a powerful tool for elucidation of the molecular mechanisms controlling the development of many organs (Thesleff, 2000; Miletich and Sharpe, 2003).

Mammalian teeth form by a series of reciprocal epithelial–mesenchymal interactions during the odontogenic developmental program, from the early patterning of the future dental axis to the initiation of tooth development at specific sites within the ectoderm (Cobourne and Sharpe, 2003). Tooth development is initiated when the mandibular epithelium instructs cranial neural crest-derived ectomesenchymal cells to aggregate at specific sites (Chai et al., 2000; Cobourne and Sharpe, 2003). Members of the Bmp (Tucker et al., 1998), FGF (St Amand et al., 2000; Kettunen et al., 2000), Hedgehog (Hardcastle et al., 1998; Dassule et al., 2000), and Wnt (Sarkar et al., 2000) families of signaling molecules induce early regionally restricted expression of downstream target genes in the odontogenic ectomesenchyme. Bmp4 is initially expressed in the embryonic day (E) 9.0 epithelium of the facial processes but later shifts to the mesenchyme (Vainio et al., 1993). This is followed by down-regulation of Bmp4 mRNA expression in the E10.5 epithelium overlying the maxillary and mandibular processes, and only a very weak signal in the underlying mesenchyme of both processes (Bennett et al., 1995). However, Bmp4 expression appears again at E11 within epithelium and stays there for a short time during the bud stage (E12–E13), and subsequently shifts completely to the condensed dental mesenchyme around the epithelial bud (Vainio et al., 1993; Aberg et al., 1997). The shift in odontogenic potential is associated with a condensation of the mesenchyme and synthesis of several extracellular matrix (ECM) molecules, including syndecan and tenascin (Thesleff et al., 1990). Bmp4 induces the mesenchymal expression of Msx1, which is in turn required for mesenchymal Bmp4 expression. In Msx1-nulls, Bmp4 expression is significantly reduced in molar mesenchyme and tooth bud development is arrested (Chen et al., 1996). However, Msx2 plays an important role in later stages of tooth development. The earliest defects in Msx2-null tooth germs are detectable at E16.5, and the Msx2-null molars undergo severe degeneration, whereas the incisors are brittle and misaligned (Satokata et al., 2000). Numerous in vitro studies using dental papilla cells suggest a regulatory role for TGF-β1, Bmp2, 4, and 7 in differentiation of the pulp cells into preodontoblasts, influencing genes expression (Nakashima et al., 1994) and stimulation of matrix secretion (Begue-Kirn et al., 1992, 1994). Continued expression of Msx2 and Dlx2 (Lezot et al., 2000) and dramatic increases in tenascin and alkaline phosphatase expression within the mesenchyme coincides with dental mineralization (Mackie et al., 1987). After cell–cell interaction, the neural crest-derived mesenchyme differentiates into odontoblasts and cementoblasts, laying down dentin and cementum, respectively. The neural crest also gives rise to the highly organized periodontal ligament (PDL) tissue between the cementum and the alveolar bone (Chai et al., 2000). The PDL has a remarkable capacity for renewal and repair, playing a pivotal role in periodontal regeneration (Beertsen et al., 1997). Additionally, the fibroblastic PDL cells are thought to be multipotent cells (McCulloch and Bordin, 1991) that may be a source of osteoblasts for continued remodeling of alveolar bone of the jaws. PDL cells have osteoblast-like properties, such as alkaline phosphatase activity (Yamashita et al., 1987), being responsive to parathyroid hormone and producing bone sialoprotein in response to 1,25-dihydroxyvitamin D3 (Nojima et al., 1990).

Periostin, originally called osteoblasts-specific factor-2 (Osf-2), was first isolated from the mouse MC3T3-E1 osteoblastic cell line (Takeshita et al., 1993) and is reported to support MC3T3-E1 cell attachment and spreading (Horiuchi et al., 1999). MC3T3-E1 cells readily undergo osteoblastic differentiation in response to both vitamin D and parathyroid hormone (Krishnan et al., 1995). Periostin has structural similarity to insect fasciclin-I, a homophilic adhesion protein involved in the neuron growth cone guidance (Takeshita et al., 1993), and can be induced by TGF-β (Horiuchi et al., 1999) and Bmp2 (Ji et al., 2000). Purified recombinant periostin has been shown to be a ligand for αvβ3 and αvβ5 integrins and proved to promote αvβ3 and αvβ5 integrin-dependent cell adhesion and enhance cell motility (Gillan et al., 2002). The original cDNA clone was 3,187 bp and contained an open reading frame of 2,436 bp, corresponding to 811 amino acids. Human OSF-2/periostin, originally cloned from both placenta and osteosarcoma tissue, has an 89.2% identity to the mouse protein. There are five human periostin isoforms, with the differences occurring in the C-terminal domain and constitute in-frame deletions or insertions, suggesting alternative splicing events (Takeshita et al., 1993). The secreted disulfide-linked 90-kDa protein, purified from MC3T3-E1–conditioned medium was renamed “periostin,” because of its localization in both the periosteum and periodontal ligament, indicating its potential role in bone and tooth formation and maintenance of structure (Horiuchi et al., 1999). Recently, Wilde et al. (2003) demonstrated that periostin mRNA is up-regulated in the pressure sites in bone and periodontal tissue remodeling after mechanical stress during experimental tooth movement. Significantly, periostin is also reported to play a role in various pathologic conditions. Multiple reports have demonstrated elevated periostin levels in tumor samples from neuroblastoma (Sasaki et al., 2002), within epithelial ovarian cancer (Gillan et al., 2002) and in patients with non-small cell lung carcinoma (Sasaki et al., 2001) that had undergone epithelial–mesenchymal transformation and metastasized. It is also suggested to be responsible for ECM deposition after myocardial infarction (Stanton et al., 2000). Given that the cellular interactions are not always direct and the extracellular environment is involved in signal transduction (Lonai, 2003), we investigated the expression pattern and the localization of periostin, a protein secreted to the ECM during tooth and periodontal ligament development.


Localization of periostin mRNA During the Early Epithelial–Mesenchymal Induction of Tooth Development

Periostin mRNA is initially expressed by the E9.5 first arch epithelium (Fig. 1A–C), which is thought to contain the major odontogenesis signaling center responsible for the formation of later epithelial ingrowths into the ectomesenchyme. Subsequently, periostin expression disappears from the epithelium, and strong expression is now observed in the E10.5 caudal mesenchyme of the first branchial arch area facing the pharyngeal region of foregut (Fig. 1D,E). Only a very weak signal was detected within a few cells in the E10.5 arch mesenchyme underlying the presumptive odontogenic epithelium, but periostin expression is markedly up-regulated at E12.0, when the inductive signals shift to the mesenchyme (Fig. 2A,B). Periostin expression remains mesenchymal during the later stages of the tooth development (Figs. 2, 3).

Figure 1.

In situ hybridization analysis of the early periostin mRNA expression in the first branchial arch. A,B: Low-power, sagittal and transverse sections through the embryonic day (E) 9.5 stage mouse embryos. C: A high-power image of the first branchial arch in B. D: Low-power, sagittal section through the E10.5 embryo. E: A high-power image of the first branchial arch in D. A–C: Arrows indicate periostin expression in odontogenic epithelium, and arrowheads indicate expression in aboral epithelium facing the pharyngeal region of foregut. E: Arrow indicates low-level periostin expression within mesenchyme underlying odontogenic epithelium, and asterisk indicates strong expression in aboral mesenchyme. fb, hb, fore- and hindbrain, respectively; h, heart; 1st, first branchial arch; nt, neural tube.

Figure 2.

In situ hybridization analysis of periostin mRNA in early tooth morphogenesis. A,C,E: Low-power, transverse sections through mid-face region within embryonic day (E) 12.0 (A), E13.5 (C), and E16.5 (E) embryos. B,D,F: High-power images of developing lower molar area of A, C, and E, respectively. A,B: An E12.0 developing tooth at the stage of epithelial ingrowth of oral ectoderm into the ectomesenchyme. C,D: An E13.5 developing molars at the bud stage. Stronger expression on the lingual side of the developing tooth is indicated by arrows, whereas lower levels on the lateral side by arrowheads in B and D. E,F: E16.5 lower molar at the early bell stage. Arrows indicate strong expression in cells adjacent to the future cervical loop (E). Asterisks, teeth; t, tongue; Mc, Meckel's cartilage; pp, lateral palatine process; p, palate; mn, developing mandible; mx, developing maxilla; ns, nasal septum.

Figure 3.

In situ hybridization analysis of periostin mRNA in lower molars. A,C: Low-power, transverse sections through lower molars of a newborn mouse during the hard tissue formation crown stage. B,D: Higher magnifications of cervical loop region in A and the apex of the dental epithelial fold in C. The arrows in B and C indicate the expression in dental follicle cells, and the arrows in D preodontoblasts. A,B: The arrowheads and the asterisks indicate the ectomesenchymal expression around the cervical loop and in the region where the future apical foramen will form. d, dentin; a, ameloblasts; e, enamel; dp, dental papilla; o, odontoblasts; cl, cervical loop; em, ectomesenchyme.

Asymmetric Localization of periostin mRNA During the Epithelial Thickening, Bud, and Cap Stages of Tooth Development

During the early thickening and the bud stage, periostin mRNA was restricted to the ectomesenchymal cells, closely packed around the epithelial bud and was absent from the overlying epithelial cells (Fig. 2A–D). It has been proposed that, around E12.5, the inductive potential shifts to the dental mesenchyme, which condenses around the early tooth bud (Thesleff and Sharpe, 1997; Cobourne and Sharpe, 2003). Periostin mRNA continues to be exclusively expressed in the ectomesenchyme during the cap and early bell stage, when the cells encapsulate the proliferating dental organ and form the dental follicle, which will give rise to the supporting tissues of the tooth. Significantly, starting from E12.0 (Fig. 2B), periostin mRNA is present within the whole condensing mesenchyme, but it is expressed at a higher level on the lingual/palatal side of the epithelial ingrowth when compared with the buccal side. The asymmetrical expression is particularly visible around lower first molar bud (Fig. 2D), but it is also present in the upper tooth (Fig. 2C). Later, the expression is present during the formation of dental papilla on both sides of the dental lamina, connecting the tooth germ with the oral epithelium, where it is also stronger medially (Fig. 2E,F). Periostin expression remains asymmetric in the late bell stage around the epithelial islands formed from the dental lamina at E18 (data not shown).

Localization of periostin mRNA During the Bell Stage and in Trans-differentiating Odontoblasts During the Hard Tissue Formation

Additional to the asymmetric pattern around dental lamina, periostin mRNA is also strongly expressed in the ectomesenchyme of E16.5 dental papilla (Fig. 2E,F). Significantly, high-level expression is present in cells located close to the proliferating cervical loop of the dental organ (Fig. 2F).

Periostin expression during hard tissue formation is limited to three sites in the tooth: preodontoblasts located in the cuspids formation sites, ectomesenchymal cells adjacent to the Hertwig's epithelial root sheath and dental follicle cells (Fig. 3). Neither the ectomesenchymal cells of the dental papilla, nor the mature, dentin-producing odontoblasts express periostin mRNA (Fig. 3D). Thus, only those cells changing fate and migrating from the dental papilla toward the dentin express high levels of periostin mRNA (arrows in Fig. 3D). We noticed only a few ectomesenchymal cells located close to the elongating internal dental epithelium (elongation is the first step toward future amelogenesis) that express periostin at E18 (data not shown). However, in newborns, expression becomes stronger and the number of the periostin-positive preodontoblasts significantly increases (Fig. 3D). In newborns, periostin continues to be expressed by the ectomesenchyme surrounding the root forming Hertwig's epithelial sheath, which proliferates to form the cervical loop (Fig. 3B), including the cells of the dental follicle (Fig. 3B). Periostin is also expressed by the dental follicle cells, surrounding the developing tooth crown before the eruption (Fig. 3C) and robustly in adult periodontal tissues (Figs. 6, 8).

Multiple Isoforms of periostin mRNA and Protein Are Expressed During Tooth Development

The published full-length periostin cDNA sequence cloned from the MC3T3-E1 osteoblast cell line (Takeshita et al., 1993) contains 23 exons; however, after screening of an E12.0 mouse embryo cDNA library, we noted that embryonic mouse periostin contains 24 exons (data not shown). Of interest, both Horiuchi et al. (1999) and EST BC007141 report finding the identical 81-bp insert (Fig. 4C). Thus, we designed reverse transcriptase-polymerase chain reaction (RT-PCR) primers that amplify this alternatively exon to determine what isoform/s are present during craniofacial development and whether there is developmentally regulated differential splicing. Only periostin isoform 1 (to use the nomenclature described by Horiuchi et al., 1999) is present in E12.0 mouse embryo head samples (Fig. 4A). However, during morphogenesis, isoform 2 appears such that within the newborn head both isoforms are equally represented and in the adult head samples isoform 2 predominates. The isolated adult incisor, molar, and periodontal ligament tissues all predominantly express isoform 2 (isoform 1 is detectible when additional PCR cycles are used). Given that periostin is highly expressed within the fibroblastic-like periodontal ligament cells (Fig. 4A) and was originally cloned using subtractive hybridization of MC3T3-E1 osteoblastic cell line and NIH3T3 fibroblastic cell line, we used our “alternatively exon” RT-PCR primers to determine which isoforms (if any) were expressed in these cell lines. Surprisingly, both NIH3T3 and MC3T3-E1 cell lines express periostin, but NIH3T3 cells only express isoform 1, whereas the MC3T3-E1 cells express both isoforms (Fig. 4B).

Figure 4.

Differential expression of periostin isoforms and Western blotting analysis. A: Reverse transcriptase-polymerase chain reaction (RT-PCR) analysis of periostin expression throughout craniofacial development and in isolated tissues. The housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH) is used as a control for loading. B: RT-PCR analysis of periostin expression in MC3T3-E1 and NIH3T3 cell lines. C: Representative sequences obtained from PCR shown in A and B. Note that isoform 2 is identical to isoform 1, except that a 81-bp exon is spliced out (underlined). D: Western blot analysis of periostin protein expression. Specificity of binding is demonstrated using either the specific immunogenic peptide or a nonspecific control peptide as a competitor of antibody binding. E14, embryonic day 14. E: Periostin protein levels in the developing craniofacial region. The migration of standard proteins is indicated.

An anti-periostin antibody was prepared against the N-terminal region that enabled us to detect all known isoforms as described in the Experimental Procedures section for use in Western blotting. To verify its specificity, we show that the immunogenic peptide blocked the ability of the antibody to detect several polypeptides of appropriate molecular weight (∼90 kDa), which were detected when the antibody was incubated with a control peptide (Fig. 4D). We then determined the levels and molecular forms of periostin present in the craniofacial region of mice at several stages of development using actin as a loading control (Fig. 4E). Periostin is hardly detectable at E9/E10 (see Fig. 4E, but periostin is detected when films are exposed 10-times longer); intermediate levels are present at E12; while high levels are present in the E16–newborn craniofacial region. In adult mice, considerably less periostin is present. At least four, closely spaced in molecular weight, forms of periostin were detected and are noteworthy for their heterogeneity and for the fact that different forms predominate at different stages of development. Interestingly, the highest molecular weight form is predominant at E12 and E16, the lowest molecular weight form is predominant at E18, and an intermediate molecular weight form is predominant in adult tissue.

Periostin Protein Expression During Tooth Development

Our anti-periostin antibody produced very striking positive staining and was specific, as negative controls did not result in any nonspecific staining (data not shown). At E9.5, periostin protein is predominantly present in the ECM between the epithelium and the ectomesenchyme of the first branchial arch, but is also detectable within the ectomesenchyme itself (Fig. 5A,B). During the early thickening stage, both mRNA and protein are colocalized within the condensing ectomesenchyme, as periostin mRNA is expressed by the ectomesenchymal cells and protein is released into the immediately adjacent ECM. Significantly, high levels of expression is evident in the ECM between the proliferating tooth bud and the overlying ectomesenchyme (Fig. 5C,D).

Figure 5.

Immunohistochemical detection of periostin protein during critical stages of tooth development in mouse. A,B: Transverse section through the mandibular component of the first branchial arch of a embryonic day (E) 9.5 embryo, low and high power, respectively. Arrows indicate the expression in extracellular matrix between the epithelium and the ectomesenchyme of the first branchial arch; arrowheads indicates periostin in the ectomesenchyme. C,D: Transverse section through E13.5 molars at the bud stage. D: A higher power image of C, in which small arrows indicate high levels of expression in the matrix adjacent to the invaginating epithelium. Stronger, lingual expression is indicated by arrow, weaker buccal expression by arrowhead. E,F: Transverse section through the incisor and cervical loop in newborn mouse. F is a higher power image of E, in which arrows indicate periostin expression in area located close to the cervical loop (cl). G,H: Transverse section through newborn lower molar. Arrow indicates expression. Asterisks, teeth; nt, neural tube; 1st, first branchial arch; oph, oropharyngeal region; ph, pharyngeal region of the gut; mx, maxilla; mn, mandible; dp, dental papilla; d, dentin; em, ectomesenchymal cells; o, odontoblasts.

Of interest, similar to the asymmetric localization of periostin mRNA during the bud and cap stages, periostin protein is also present at higher levels on the lingual/ palatal side compared with the buccal side. Both mRNA and protein are colocalized throughout the cap and bell stage (data not shown) as well as during hard tissue formation of the tooth. In newborn mouse, periostin protein is present in the dental follicle and tissues surrounding the tooth (Fig. 5E,G), in the ECM around the Hertwig's epithelial root sheath proliferating from the cervical loop (Fig. 5F), and in the preodontoblasts that undergo trans-differentiation from the ectomesenchyme to give rise to the dentin-producing odontoblasts (Fig. 5H).

The periodontium surrounding both the adult mouse molars (Fig. 6) and incisors (data not shown) is strongly positive for periostin protein. Using our antibody, we can confirm that periostin protein is present in all periodontal ligament fibers connecting the cementum with the alveolar bone from the apex of the root (Fig. 6D) to the fiber bundles attached to the alveolar crest (Fig. 6E). Alveolar bone, as well as the cementum are both negative for periostin protein. However, contrary to previous published reports, we did detect periostin protein within the dental pulp. The expression is limited only to the matrix surrounding small groups of cells but is present within the pulp chamber (Fig. 6B) and within the root canal (Fig. 6C). In both cases, periostin protein is detected in cells that exhibit a different morphology to that of the odontoblasts and which are adjacent to the unmineralized dentin matrix, called predentin, and is thought to be responsible for maintaining the integrity of dentin.

Figure 6.

Immunohistochemical detection of periostin protein in young adult mouse teeth. A: Low-power image of sagittal section of upper and lower molars. Note periostin protein is visible throughout the periodontal ligament and at lower levels within the dental pulp. B,C: High-power image of pulp chamber (B) and root canal (C) of the second and first lower molar, as boxed in A. D: Apex of the second lower molar front root. E: Alveolar crest (ac) between the first and second lower molar. Arrows indicate periostin expression. 1st, 2nd, 3rd, first, second, and third molars, respectively; mx, maxilla; mn, mandible; d, dentin; dp, dental pulp; c, cementum; ab, alveolar bone; PDL, periodontal ligament.

Bmp4 Is Not Required for the Onset of periostin Expression in the Epithelium, but Is Required for the Shift in Expression to the Mesenchyme

Given the central role that Bmp4 signaling plays throughout tooth development (Tucker et al., 1998; Yamashiro et al., 2003), we performed both RT-PCR and in situ hybridization analysis on the Bmp4-null embryos to determine whether or not periostin functions within the same signaling pathway. We recently reported that, when the Bmp4lacZ-null mutants (Lawson et al., 1999) are out-bred to a CBA background, approximately 50% survive until E11.0, despite the lack of posterior structures and an allantoic connection (Conway et al., 2003). This finding enabled us to assess of the spatiotemporal expression pattern of periostin within the Bmp4-nulls (n = 5 “normal” E10.5 Bmp4-null embryos). RT-PCR revealed that periostin was down-regulated within the E9.5 Bmp4-null whole embryo (data not shown). Periostin mRNA was detected by in situ hybridization exclusively within first branchial arch epithelium at E9.5, at levels similar to the wild-type embryos (Fig. 7A). However, by E10.5 when periostin expression in normal littermate embryos is shifted to the mesenchyme, we detected continued periostin epithelial expression and no shift into the adjacent mesenchyme (Fig. 7C–F). Additionally, mesenchymal expression was absent in aboral region of the Bmp4-null mandibular primordial suggestive of a generalized developmental delay. However, periostin expression within the umbilical vessels is unaltered in the Bmp4-nulls.

Figure 7.

Periostin expression in wild-type and Bmp4-null embryos, detected by in situ hybridization. A,B: Sagittal section of embryonic day (E) 9.5 wild-type (+/+) embryo. Periostin expression in the first branchial arch epithelium indicated by arrow and arrowhead in B. C.D: Sagittal section of E10.5 wild-type embryo. Arrows in D indicate the expression in the mesenchyme. E,F: Sagittal section of E10.5 Bmp4-null mutant embryo. Note that periostin remains in the epithelium and is not expressed within the underlying mesenchyme (indicated by arrowhead in F). 1st, first branchial arch; h, heart; u, umbilical vessels.

Periostin Is Normally Expressed in the Msx2 and Msx1/2 Null Embryos but Is Overexpressed Within the Msx2 Null Adult

Similarly, given the central roles that Msx1 and Msx2 and their interactions with various Bmp signaling molecules play throughout tooth development (Yamashiro et al., 2003), we performed both RT-PCR and in situ hybridization analysis on the Msx2 and Msx1/2 double-nulls. Throughout early development, periostin mRNA is normally expressed in the Msx2-null embryo craniofacial region (data not shown). Periostin mRNA expression around the E14 tooth bud is identical to that previously described in normal embryos (Fig. 8A). The first defects in the Msx2-null are observed at E16.5, resulting in later progressive ameloblast degeneration and reduced amount of enamel within adult teeth causing the inability to chew solid food (Satokata et al., 2000). After in situ hybridization analysis, it is evident that the Msx2-null adult PDL cells expresses significantly higher levels of periostin compared with age-matched control littermates (Fig. 8C,D). Finally, we used in situ hybridization to analyze periostin mRNA expression in Msx1/2 double-null embryos to determine whether the lack of both family members would lead to altered periostin spatiotemporal expression. In the Msx1/2 double-null embryos, tooth development is arrested at either the lamina or early bud stage (Bei and Maas, 1998). At E12.5, periostin mRNA expression does not differ from that of normal embryos (Fig. 8E,F) and is stronger medially than laterally. Thus, lack of both Msx genes does not affect the onset and localization of periostin expression, but the lack of Msx2 results in an inability to regulate periostin expression in the adult PDL.

Figure 8.

Periostin expression in Msx-nulls, detected by in situ hybridization. A,B: Transverse section (low and high power, respectively) of embryonic day (E) 14 Msx2-null mid-face region. C,D: Sagittal section through adult upper incisors of wild-type (C) and Msx2-null mouse (D). E,F: transverse section through Msx1/2 double-null E12.5 embryo. Arrow and arrowhead indicate the asymmetric lingual vs. buccal expression. p, palate; t, tongue; e, epithelium; m, mesenchyme; d, dentin; PDL, periodontal ligament.


This is the first study to show expression of periostin mRNA in the developing mouse dental tissues. In this study, we have successfully described the spatial and temporal expression of periostin during the normal developmental process and how expression is altered in both the Bmp4- and Msx-null mutants. Significantly, periostin is present throughout all stages of mouse tooth development, particularly within the embryonic sites of epithelial–mesenchymal interaction and by later newborn cells that trans-differentiate from one phenotype into another. Our data show that periostin is expressed early in epithelial and mesenchymal cells destined to form the mineralized tissues of the tooth and periodontum. Expression of periostin was also observed in the adult PDL and dental pulp at the sites of hard-soft tissue interfaces and was noticeably absent from terminally differentiated mineralized tissues.

Our results demonstrate that periostin mRNA is expressed early within the E9.5 first branchial arch epithelium overlying the mesenchyme, before any overt morphologic changes associated with tooth induction and early morphogenesis (Bennett et al., 1995; Bachler and Neubuser, 2001) and the protein is secreted into the ECM adjacent to the epithelium.

During the budding, cap, and bell stages, the highest levels of periostin mRNA and protein expression are localized immediately surrounding the dental epithelium, of expanding tooth bud and cells proliferating from the cervical loop epithelial sheath. Similar to tenascin, an ECM multimodular glycoprotein that possesses neurite outgrowth-stimulating properties (Sahlberg et al., 2001), periostin expression is predominantly in the ECM surrounding migratory cells and tissues. Significantly, tenascin has an almost identical spatiotemporal expression pattern during tooth development to that of periostin and is also regulated by TGF-β1 and FGFs. Given the association of periostin with various integrins (Gillan et al., 2002) and the colocalization with several ECM molecules, the presence of periostin within the ECM may create the appropriate microenvironment enabling the epithelial tooth bud and cervical loop to invade the dental mesenchyme. Because periostin is responsive to several well-known developmentally important signaling molecules such as TGF-β (Horiuchi et al., 1999), Bmp2 (Ji et al., 2000), Wnt3 (Haertel-Wiesmann et al., 2000), and Twist (Oshima et al., 2002), we suggest that periostin may act as one of the epithelial–mesenchymal responder genes that enables cellular migration and transformation.

One striking feature of both the periostin mRNA and protein localization is that both are asymmetrically expressed and are present at a much higher intensity within the condensing mesenchyme on the lingual/palatal side of the epithelial ingrowth rather than the buccal side. Teeth, and particularly molars, are not symmetrical structures. Crown patterning is thought to be controlled by the enamel knot, and although periostin is not expressed within this structure, it is expressed in the adjacent mesenchyme. After Bmp4-mediated apoptosis of the enamel knot, the mesenchyme assumes the inductive role (Jernvall et al., 1998). Several other genes are also known to be asymmetrically expressed by the ectomesenchymal cells surrounding the tooth germ, indicating possible influence of this structure on the future shape of the tooth. In contrast to periostin, Bmp4 is predominantly expressed on the buccal side of the tooth germ (Vaahtokari et al., 1996; Aberg et al., 1997). However, similar to periostin, both Fgf10 (Kettunen et al., 2000) and αv-integrin (Salmivirta et al., 1996) are predominantly expressed on the lingual side. The asymmetric coexpression of both periostin and αv-integrin is intriguing, given that human periostin secreted by epithelial ovarian carcinoma has been shown recently to be a ligand for αv-β3 and αv-β5 integrins and promote cell motility in vitro (Gillan et al., 2002). These data indicate that periostin may function as a ligand for integrins to support adhesion and migration specifically of the lingual elements and may stabilize the future shape of the teeth.

Numerous reports and reviews have described the many interactions and central roles played by Bmps (particularly Bmp4) and Msxs during tooth development (Chen et al., 1996). However, neither Bmp4, Msx2, or Msx1/2 are required for the onset and normal initial expression of periostin. Given that Bmp4 has been shown to play a regulatory role in both the dental epithelium and mesenchyme of several different genes at this stage (Vaahtokari et al., 1996; Aberg et al., 1997), our data suggest that epithelial periostin expression is not dependent upon the epithelial Bmp4 signaling pathway. Because Bmp4 mutants are fairly severely affected, growth retarded, and that the 50% that survive gastrulation all die around E11.0, we are unable to determine whether the switch in periostin expression from the epithelium to the mesenchyme is Bmp4-dependent. Of interest, Msx1 is expressed in dental mesenchyme and excluded from epithelium in early stages and is required for mesenchymal expression of Bmp4 (Chen et al., 1996; Bei et al., 2000). Msx2 itself is thought to play a major role during in later stages of tooth development, as Msx2-null mice molars undergo severe degeneration and the incisors are brittle and misaligned (Satokata et al., 2000). Msx2 is not expressed by ameloblasts, but it is expressed in ovoid preodontoblasts and subjacent papilla cells and may suppress osteocalcin expression immediately preceding odontoblast terminal differentiation (Bidder et al., 1998). Significantly, periostin is normally expressed within Msx1/2 double-nulls and is only overexpressed in the adult Msx2-null PDL. As PDL cells are known to have osteoblast-like properties and that there is cross-talk between Msx/Dlx homeobox genes and vitamin D during tooth mineralization (Lezot et al., 2002), these data may indicate that the continued elevated levels of periostin are preventing normal mineralization. However, given that the viable Msx2-null adult mice are maintained on powdered food, the difference in periostin mRNA expression may be caused by a nonspecific lower mechanical load upon the teeth/PDL as the mice do not have to chew solid food.

Periostin is also highly expressed in the PDL during its morphogenesis and development, and in the adult animal. The PDL is a soft connective tissue interposed between the roots of teeth and alveolar bone, and it is characterized by rapid turnover and a high remodeling capacity, which give it adaptability, maintaining a constant width despite being exposed to rapidly changing physical forces such as mastication, speech, and orthodontic tooth movement (Kirkham et al., 1993; Beertsen et al., 1997). Periostin protein continues to be highly expressed in the developing dental follicle and in the adult mouse periodontium, particularly at the sites of hard–soft tissue interfaces. This finding is significant as recent studies have demonstrated that cells in the circumference of the dental follicle migrate in an apical direction, suggesting cells migrating through the dental follicle connective tissue may contribute to the formation of the periodontium (Diekwisch, 2002). Additionally, active movement of PDL fibroblasts is thought to be an important component of tooth eruption (Weinreb et al., 1997) and that periodontal fiber disruption (which occurs during increased stretching) can provide a mechanism to stabilize the eruption site (Katona and Qian, 2001). Given that periostin is also highly expressed in the dental follicle and PDL ECM during and after tooth eruption, periostin's role may be to create an environment appropriate for cell migration within the dental follicle during periodontium formation, for fibroblast movement during tooth eruption or for osteoclast recruitment and activation. Given that fibroblasts are the predominant cell type within the PDL (Beertsen et al., 1997) and are capable of producing and digesting matrix components, fibroblastic cells have been assumed to be responsible for self-renewal of the PDL and regeneration of periodontal tissue (Beertsen et al., 1997). Fibroblastic cells of the PDL have also been suggested to be a source of osteoblasts for continued remodeling of alveolar bone of the mandible as studies have demonstrated that cells isolated from the PDL have osteoblast-like properties, such as alkaline phosphatase (ALPase) activity (Yamashita et al., 1987), are responsive to PTH (Nojima et al., 1990) and produce bone sialoprotein in response to 1,25-dihydroxyvitamin D3 (Nojima et al., 1990). Significantly, periostin-expressing MC3T3-E1 cells have osteoblast-like properties, such as ALPase activity, are responsive to PTH and produce bone sialoprotein in response to 1,25-dihydroxyvitamin D3 (Takeshita et al., 1993). The mechanism of induction of PDL cells to become either osteoblast or cementoblast has been suggested to be dependent upon the microenvironment (Saito et al., 2002). Thus, the continued high expression of both periostin mRNA and protein within the adult PDL ECM may indicate that periostin maintains the PDL, as a nonmineralized tissue and as a renewable source of cells required for the regeneration of the aveolar bone and root cementum. Further studies may determine the potential role of periostin in the alveolar bone regeneration process and whether it is affected during the pathogenesis of periodontal diseases.


Mice and Genotyping

Bmp4(lacZneo) heterozygous mice were provided by Dr. Brigid Hogan (Duke, NC), crossed onto a mixed CBA genetic background that allows Bmp4-null mutants to survive until ∼11.0 days post coitum (as reported in Conway et al., 2003). Bmp4 mutants were genotyped as previously described (Lawson et al., 1999). Viable Msx2-null knockout mice were provided by Dr. Robert Maxson (USC, CA) and genotyped as previously described (Satokata et al., 2000). Additionally, Msx1/2 double-null embryos were generously provided by Dr. Maxson. Wild-type samples used in this study came from our breeding colony CBA mice.

RT-PCR Reactions

Isolated mouse E13.5, newborn and adult heads and teeth of normal mouse and adult teeth of Msx2-/- mutants were dissected in ice-cold phosphate buffered saline (PBS), and the periodontal ligament cells were removed by scrapping (so that no periodontal ligament was visible under the dissecting microscope). RNA was also isolated from MC3T3-E1 and NIH3T3 cells cultured by using established methods (Horiuchi et al., 1999). RNA isolation and RT-PCR were carried out as previously described (Kruzynska-Fejtag et al., 2001). Primers used for periostin amplification were designed around the site of alternative splicing based on the GenBank sequence NM015784.1. This strategy results in the amplification of two different isoforms (5′-TCCAGCAGATATTCCAGTTG and 3′-TTTCGCCTTCTTTAATCAGC). PCR amplification was performed for 24 cycles (housekeeping gene GAPDH) and 30 cycles (periostin), such that products were still within linear range, and analyzed on a 1.5% agarose gel. Both isoforms were excised from the gel purified using Qiagen Gel Extraction Kit and sequenced.

Isolation of Mouse periostin cDNAs

Two periostin cDNA probes were used, one has been described previously (Kruzynska-Fejtag et al., 2001) and is 574 bp and spans the signal peptide, cysteine-rich region and into the first Fasciclin repeat. A second 623-bp probe was also PCR generated by using primers within the second and fourth Fasciclin repeats (TTC CTG ATT CTG CCA AAC AAG T and CAG GTG ATA AAG AAT GAT GAT GTT T). The 623-bp probe was cloned into the pCRII-TOPO vector (Invitrogen) and sequenced. In situ hybridization with both probes gave identical spatiotemporal patterns of expression within the craniofacial region (data not shown).

In Situ Hybridization

Both sense and antisense [35S]UTP-radiolabeled RNA probes were transcribed from both the 574-bp and 623-bp periostin fragments and used for in situ hybridization on wax sections as previously described (Kruzynska-Frejtag et al., 2001). Adult heads were initially de-calcified in 20% ethylenediaminetetraacetic acid (EDTA; pH 7.4) before wax embedding and sectioning. In situ hybridization was carried out on both sagittal and transverse sections of at least eight normal mouse embryos at E9.5–E18, sections of three newborn and two adult mice heads. Also, three Bmp4-null embryos at E9.5–E10.5; eight Msx2-null E9.5–E18, newborn and adult samples and E12.5 Msx1/2 double-null embryos (double-nulls are not viable at later stages) were analyzed by using the techniques described previously (Conway, 1996). Specific signal was only observed when sections were hybridized with the antisense probe, and there was no nonspecific signal when hybridized with the sense probe (data not shown). All sections were counterstained with 0.02% toluidine blue for 30 sec to visualize the tissue and photographed by using Zeiss Photomicroscope.

Preparation and Testing of Affinity-Purified Anti-Periostin Antibodies

The peptide KLREEIEGKGSYTYFAPSN representing amino acids 123–141 of periostin was synthesized with an additional C residue at the N-terminus to facilitate its attachment to Sulfolink Coupling Gel (Pierce, Rockford, IL). The conjugation of the peptide to KLH, the injection of rabbits, the purification of total IgG, and the affinity purification of IgG by using the peptide attached to Sulfolink Couplin Gel were all performed as previously described (Zanin et al., 1999). In Western blotting experiments, the affinity-purified anti-periostin antibody was used at a final concentration of 0.4 μg/ml in antibody buffer. In peptide competition experiments, the antibody (0.4 μg/ml in antibody buffer) was preincubated with 50 μg/ml of either the immunogenic peptide or a control peptide (CNKRVQGPRRRSREGRSQ, the 17 amino acids at the C-terminus of periostin with an additional C residue at its N-terminus) for 1 hr at room temperature before use on Western blots.

Western Blotting of Tissue Extracts

Mouse tissues were isolated and microdissected in cold PBS (pH 7.4), and a limb bud was removed for PCR genotyping. E14 embryo, whole embryos minus extraembryonic membranes and placenta; Nb heart, newborn ventricles, atria, and outflow tract vessels; E9 head and E10 head, whole heads; E12 head, E16 head, and E18 head, microdissected upper and lower jaws and associated craniofacial tissues including tongue; Nb head and adult head, dissected upper and lower jaws. Samples were placed in 20 volumes of sample buffer (2% sodium dodecyl sulfate [SDS], 50 mM Tris pH 7.0, 5% glycerol and 0.5% bromophenol blue dye), protease inhibitors added (10 mM N-ethyl maleimide, 5 mM benzamidine, 50 μg/ml leupeptin, 5 μg/ml pepstatin-A, and 2 mM phenylmethyl sulfonyl fluoride), and homogenized on ice for 5 min. For reducing conditions, 1/20th volume of β-mercaptoethanol (final conc. of 5%) was added to the sample buffer. The multiple bands seen in reduced samples demonstrate heterogeneity in monomers. As preliminary Western analysis using the affinity-purified anti-periostin antibody gave similar results under both reducing and nonreducing conditions, all data presented have β-mercaptoethanol present. Protein samples were boiled for 3 min and allowed to cool to room temperature, and an additional aliquot of protease inhibitors was added. Samples were first normalized on test gels by using both amido black staining and probing of blots with polyclonal rabbit anti-actin antibody (Sigma). Protein isolates were run on a 6% SDS-polyacrylamide gel electrophoresis gels and transferred to nitrocellulose. Immunopositive polypeptides were detected by enhanced chemiluminescence (ECL) using standard methods.


Mouse embryo heads and fetal, newborn, and adult craniofacial tissues were collected at various developmental stages, rinsed briefly in cold PBS, and fixed in 4% paraformaldehyde. Embryo, fetal, and newborn tissues were processed for wax embedding and sectioning by using standard methods. Adult heads were initially decalcified in 20% EDTA (pH 7.4), then wax embedded. Serial 10-μm sections were de-waxed, washed, and incubated in Avidin/Biotin Block (Vector Laboratories, Inc.) and goat serum for 15 min. Sections were incubated with primary affinity-purified anti-periostin antibody diluted 1:6,000 in 0.1 M Tris (pH 7.6)/1% BSA for at least 1 hr. Slides were washed in 0.1 M Tris (pH 7.6)/1% Tween and incubated for at least 30 min with goat α-rabbit-biotin–conjugated (Vector Laboratories, Inc.) secondary antibody diluted 1:1,000 in 0.1 M Tris (pH 7.6)/1% BSA. Slides were washed in 0.1 M Tris (pH 7.6)/1% Tween and incubated for 10 min with Biotin Tyramide Amplification Reagent and processed as described by the manufactured for the Vectastain “Elite” ABC Kit (Vector Laboratories, Inc.). After diaminobenzidine (DAB) staining, sections were counterstained with methyl green. Negative controls that lacked either the primary or secondary or both did not produce any DAB staining (not shown).


We thank Dr. Brigid Hogan (Duke) and Dr. Robert Maxson (USC) for providing us with mice and embryos, and Dr. Yuji Amagai (Fukushima, Japan) for the MC3T3-E1 cell line.