Critical role for αvβ6 integrin in enamel biomineralization

Enamel is the hardest mineralized tissue in the body and the only calcified tissue that is produced by epithelium-derived cells, namely ameloblasts. Amelogenesis consists of secretory, transition and maturation stages (Simmer et al., 2010). During the secretory stage, ameloblasts secrete enamel proteins such as amelogenin (the most abundant enamel matrix protein; Eastoe, 1979), ameloblastin (Krebsbach et al., 1996) and enamelin (Hu et al., 1997) into the enamel matrix. This extracellular matrix undergoes enzymatic modification by enamelysin (MMP-20) and kallikrein 4 (KLK4) in the transition and maturation stages, which results in the formation of a mature enamel that is mainly composed of hydroxyapatite crystallites and a minor amount of residual proteins (Bartlett et al., 1996; Nanci and Smith, 2000). Mutations in amelogenin, enamelin, MMP-20 and KLK4 genes all cause human hereditary amelogenesis imperfecta (AI), in which both enamel formation and its mineralization are affected (Hu et al., 2007). This leads to extensive wear and decay in both the primary and permanent dentition, which may result in tooth loss at a young age or require extensive restorative procedures (Crawford et al., 2007).

During amelogenesis, the apical ameloblast plasma membrane directly abuts against the matrix and the developing enamel crystals (Nanci and Smith, 2000). Although integrins mediate cell-matrix adhesion and signaling in most cell types (Hynes, 2004), the receptors of ameloblasts that mediate ameloblast–matrix adhesion, matrix organization, and signaling are not well characterized. The rat incisor enamel organ has been reported to express the β6 integrin transcript, but its role in enamel formation is not known (Moffatt et al., 2006). Integrin αvβ6 is an epithelial cell-specific integrin that binds to the arginine-glycine-aspartic acid (RGD) amino acid motif in its ligands (Breuss et al., 1993), which include fibronectin, tenascin-C, vitronectin and the latency-associated peptide (LAP) of transforming growth factor-β1 (TGF-β1) and TGF-β3. Its binding to the LAP activates the cytokine (Munger et al., 1999). This integrin-mediated activation plays an important anti-inflammatory role in vivo (Yang et al., 2007), as it regulates experimental TGF-β1-dependent fibrosis in various organs (Hahm et al., 2007; Horan et al., 2008; Patsenker et al., 2008). Furthermore, αvβ6 integrin is expressed in junctional epithelial cells that mediate gingival soft tissue adhesion to tooth enamel (Ghannad et al., 2008). Interestingly, loss of β6 integrin in mice is associated with characteristics of human periodontal disease, suggesting that αvβ6 integrin plays a role in protecting periodontal tissues from inflammatory changes leading to periodontal disease (Ghannad et al., 2008). Junctional epithelium is derived from cells arising from the reduced enamel epithelium during tooth eruption (Schroeder and Listgarten, 1977). In the present study, therefore, we hypothesized that αvβ6 integrin is also expressed in ameloblasts and that lack of its expression is associated with enamel defects. We show that mice deficient in αvβ6 integrin (Itgb6^−/−) have hypomineralized enamel and show excessive accumulation of amelogenin in the mineralizing enamel matrix.

Compared to wild-type (WT) mice, Itgb6^−/− mice had no obvious differences in tooth development or eruption, and there was no malocclusion. However, when incisors of Itgb6^−/− mice were examined more closely, it became obvious that they differed from those of WT mice. The maxillary incisors of WT mice had a smooth and yellowish surface, whereas Itgb6^−/− mouse maxillary incisors lacked the yellow pigmentation and were abnormally white. In addition, mandibular incisors of Itgb6^−/− mice appeared chalky and the tips were noticeably more rounded than those of their WT counterparts (Fig. 1A,B).

Examination of defleshed mandibles from 12-month-old mice using a dissecting microscope or after taking high-resolution radiographs showed that whereas WT mouse molars had prominent and pointed cusps on all three molars (Fig. 1C,E,G), the molar cusps of the Itgb6^−/− mice had extensive wear at their occlusal surfaces (Fig. 1D,F,H) similar to their blunted incisors. This severe attrition of molar cusps was highly evident in the scanning electron micrographs, displaying flattened and heavily worn Itgb6^−/− mouse molar cusps (Fig. 1H). While the WT enamel was smooth, the enamel of Itgb6^−/− mice showed extensive pit formation and roughness (Fig. 1I,J). The rate and amount of attrition in the molar teeth was further investigated in different age groups. As expected, as the mice aged the attrition rate increased in both the WT and Itgb6^−/− mice; however, the rate of attrition was much faster in the Itgb6^−/− group (Fig. 1K). The degree of attrition in the Itgb6^−/− mice reached the maximum level already by 6 months when the cusps were almost completely worn down (Fig. 1K).

Next, we investigated whether the enamel phenotype in the Itgb6^−/− mice could be rescued by re-expressing β6 integrin. For this purpose, the Itgb6^−/− mice were bred with mice overexpressing human β6 integrin under the cytokeratin 14 (K14) promoter in addition to their own endogenous mouse β6 integrin (Häkkinen et al., 2004). Offspring carrying only human β6 integrin on the knockout background were selected to represent the β6 integrin rescue mice (Fig. 1L; integrin expression in the ameloblast layer is presented in Fig. 6A). The incisors of the ‘overexpressing’ and the ‘rescue’ mice appeared macroscopically normal (data not shown). The average attrition rate of 6-month-old Itgb6^−/− mice was significantly higher when compared to the ‘rescue’ and ‘overexpressing’ mice of the same age, indicating that the enamel attrition is effectively rescued by human β6 integrin gene in the Itgb6^−/− mice (Fig. 1M). Moreover, overexpression of β6 integrin did not seem to affect enamel formation.

To determine the specific nature of the enamel defects in Itgb6^−/− mice, we first observed enamel microstructure. Electron micrographs of surface acid-etched incisors demonstrated a significant difference in prism structure and organization between the WT and Itgb6^−/− enamel (Fig. 2A,B). In WT mouse incisors, a parallel and well-organized enamel prism structure in alternating rows was present (Fig. 2A). Interestingly, in the Itgb6^−/− incisors, the enamel prism organization was completely lost; the enamel was somewhat layered but showed no regular repeating pattern of prism structure (Fig. 2B).

Next, we compared enamel mineral density in WT and Itgb6^−/− mice. Micro-CT images were taken from 14-day-old Itgb6^−/− and WT mouse mandibular molars. While the mutant mice teeth appeared generally anatomically normal, the X-ray density of the enamel was greatly reduced in both the first and second molars of the Itgb6^−/− mice compared to the WT (Fig. 2C,D). To quantify the difference in mineralization during various stages of enamel formation, we measured the mineral-to-protein ratio of the maxillary and mandibular incisors in both mouse lines by an ashing method. Near the apical loop of the incisors, corresponding to the secretory stage, the mineral-to-protein ratio was similar in WT and Itgb6^−/− mice (Fig. 2E,F). However, as the distance from the apical loop increased (maturation stage starts about 3 mm from the apex), the mineral-to-protein ratio of Itgb6^−/− mouse incisors became drastically decreased compared to that of the WT mice. In fact, from the regions that corresponded to the maturation stage and onwards, the WT incisors were so highly mineralized that it was not possible to cut away incisor enamel strips for sampling, whereas the Itgb6^−/− incisor enamel strips were easily removed along the entire length of the incisors (strip 9 on maxilla and strip 11 on the mandible). In additional mineral analyses using energy-dispersive X-ray spectroscopy, we studied a number of molar teeth. There was no statistical difference in Ca/P ratios between Itgb6^−/− and WT molars with the ratios obtained being typical for hydroxyapatite (data not shown).

To demonstrate that the β6 integrin mRNA and protein are expressed in mouse teeth, in situ hybridization and immunohistochemical studies were performed. In situ hybridization of β6 integrin mRNA with the anti-sense and sense probes was performed in the secretory stage of an upper incisor from a 10-day-old WT mouse. The anti-sense probe demonstrated a strong specific signal that was localized only to the cytoplasm of ameloblasts (Fig. 3A), whereas no signal was detected with the control sense probe (Fig. 3B).

Next, we performed immunostaining of β6 integrin protein in frozen sections from the secretory stage of an upper incisor as well as from developing upper molar cusps from 3-day-old WT mice. In both the incisor and molar teeth, prominent β6 integrin immunoreactivity was localized to the ameloblast layer (Fig. 3C,D). Hematoxylin- and Eosin-stained sections confirmed that the β6 integrin immunoreactivity was indeed localized to the ameloblast cell layer (Fig. 3E,F).

In order to investigate whether expression of β6 integrin in ameloblasts is similar in humans, we immunostained maturation-stage ameloblasts from surgically removed, unerupted human wisdom teeth. Ameloblasts were identified in decalcified sections by morphology as well as by a positive signal for K14 and amelogenin (Fig. 3G–I). The ameloblast layer was strongly labeled with the β6 integrin antibody (Fig. 3J), indicating that expression of αvβ6 integrin is conserved in human ameloblasts.

In order to identify the gene expression profile associated with the disturbed mineralization of Itgb6^−/− enamel, we removed incisor enamel organs of adult Itgb6^−/− and WT mice. A heat map of the differentially expressed genes was generated (Fig. 4A). These data demonstrated significant differences in gene expression profiles in enamel organs between the WT and Itgb6^−/− mice. Using the GOrilla Gene Ontology analysis tool, two of the biological processes that show significant differences were identified as ‘regulation of biomineralization’ and ‘osteoblast differentiation’ (Fig. 4B). The microarray gene expression results for ‘cellular components’ and ‘biological processes’ were further analyzed using DAVID (Database for Annotation, Visualization and Integrated Discovery). In both categories, those genes with significant differences in their expression, including genes relevant to biomineralization, were graphed in a pie chart (Fig. 4C,D, respectively). The rest of the genes were grouped in the ‘other’ category. For the ‘cellular component’ category, differentially regulated genes included genes associated with extracellular matrix or cell surface. For the ‘biological process’ category, gene expression for defense, immune response, wounding, cell-cell signaling, endocytosis, tissue morphogenesis and biomineralization differed significantly between the WT and Itgb6^−/− mice enamel organs.

The gene profiling analysis showed that 67 genes were downregulated and 171 genes upregulated in Itgb6^−/− enamel organs compared to WT enamel organs (absolute fold change >1.6; data not shown). Interestingly, the two most upregulated genes in Itgb6^−/− enamel organs were amelogenin (Amelx; 21-fold) and enamelin (Enam; 7.6-fold; Fig. 5A). Other highly upregulated genes were progressive ankylosis gene (Ank; 5-fold) and RNA motif binding proteins (RBMs; data not shown). Expression of Klk4 was also increased, whereas ameloblastin (Ambn), MMP-20 (Mmp20) and amelotin (Amtn) remained unchanged (Fig. 5A, and data not shown). The expression of β1 and β4 integrin genes was not changed, suggesting that there was no compensatory change in other integrins in the absence of β6 integrin (Fig. 5A). Also, the level of ameloblast marker gene K14 was unchanged. Among the genes that showed the strongest downregulation in the Itgb6^−/− enamel organ, were arrestin domain containing 3 protein (Arrdc3; −8-fold), photocadherin-21 (Pcdh21; −5.9-fold) and Fam20B (family with sequence homology 20, member B; −2.9-fold; data not shown). Regulation of these genes in the Itgb6^−/− enamel organ may contribute to the hypomineralized enamel phenotype in these animals but their role needs to be further investigated.

Next, we verified the expression levels of amelogenin, enamelin and other key molecules associated with enamel mineralization using quantitative RT-PCR (Fig. 5B). Among the tested mRNAs, only the expression of Amelx, Enam and Klk4 showed significant increases both in gene arrays and in RT-PCR (Fig. 5A,B).

We then verified the integrin expression profiles of the incisor enamel organs of 3–6-month-old WT, Itgb6^−/− and β6 integrin rescue mice by western blotting. As expected, Itgb6^−/− incisors were negative for β6 integrin, whereas it was expressed in the WT animals (Fig. 6A). In ‘rescue’ and ‘overexpressing’ mice, β6 integrin expression was elevated due to the constitutive K14 promoter activity in the enamel organs. Confirming the qRT-PCR results, the expression of β1 integrin was similar in all mouse lines. However, the amount of αv integrin protein, which partners with β6 integrin and others, was reduced in the Itgb6^−/− enamel (−2.5-fold; Fig. 6A) likely because of intracellular degradation due to lack of receptor dimerization.

Analysis of protein expression in incisor enamel organs of 3- to 6-month-old WT and Itgb6^−/− mice by western blotting revealed strong abnormal accumulation of amelogenin in the incisors of Itgb6^−/− mice (on average 10-fold; Fig. 6A,C). The main amelogenin bands were between 20 and 25 kDa, which is consistent with major cleaved amelogenins. Consistent with the molar attrition data, the amelogenin accumulation in the incisors of the Itgb6^−/− mice was rescued by human β6 integrin expression (Fig. 6A), showing that the accumulation of amelogenin was specifically attributable to β6 integrin deficiency. The expression of MMP-20 and KLK4 was also slightly increased in the Itgb6^−/− mice (Fig. 6D). The protein levels of enamelin were not tested because of the lack of availability of a specific antibody for mouse enamelin.

Because one of the main functions of αvβ6 integrin is to activate TGF-β1, we compared the phosphorylation and expression levels of Smad3 and Smad2, downstream targets of TGF-β1, in enamel organs. While TGF-β1 effectively induced pSmad3 phosphorylation in cultured ameloblast-like cells (not shown), there was no significant difference in the expression or phosphorylation of Smad3 or Smad2 between the WT and Itgb6^−/− enamel organs (Fig. 6B), suggesting that the mineralization defect and accumulation of amelogenin in the Itgb6^−/− enamel is not attributable to impaired TGF-β1 signaling.

To visualize the organization of the ameloblast layer, undecalcified histological sections from the late secretory stage and early maturation stage incisors of 14-day-old Itgb6^−/− and WT mice were compared. In WT mice, the ameloblast layer was well organized against mineralizing enamel (Fig. 7A,C). In the Itgb6^−/− mice, the ameloblast layer was similarly organized but showed abnormal accumulation of amorphous unmineralized matrix between the cells as well as between the ameloblasts and the forming enamel surface (Fig. 7B,D). In order to identify this material, we performed transmission electron microscopy (TEM) and immunogold labeling for amelogenin localization. Ameloblasts in the WT mouse showed typical organization and secretion of enamel matrix (Fig. 7E) that subsequently became highly mineralized in the maturation stage (Fig. 7I). In the Itgb6^−/− ameloblast layer, amorphous protein pools appeared in cyst-like structures between ameloblasts and towards the enamel surface (Fig. 7F,G). By immunolabeling, this material was shown to contain abundant amelogenin (Fig. 7H). Abnormal accumulation of amelogenin was observed both between the cells and within the cells. Electron micrographs of the surface of WT mouse molar enamel revealed a smooth, well organized, and parallel crystal structure at the enamel-ameloblast border (Fig. 7I) in the early maturation stage, whereas corresponding regions in the Itgb6^−/− mouse showed an enamel surface that was irregular and layered (Fig. 7J).

To determine whether ameloblasts use αvβ6 integrin as a cell adhesion receptor for the enamel matrix, we seeded WT mouse ameloblasts in wells coated with a porcine enamel matrix extract (EMD; over 90% amelogenin). Ameloblasts spread on EMD only in the absence of the protein synthesis blocker cycloheximide (Fig. 8A), suggesting that their adhesion on EMD was dependent on endogenous protein synthesis. Stimulating the cells with 200 nM 12-O-tetradecanoylphorbol-13-acetate (TPA), an integrin activator, further increased their adhesion and spreading by about 50% (Fig. 8A). To determine whether this adhesion was mediated by αvβ6 integrin, cell spreading was analyzed in the presence of function-blocking anti-αvβ6 integrin antibodies. Three different antibodies blocked 35–50% of the cell spreading in a statistically significant manner (Fig. 8B), indicating that ameloblasts deposit endogenous, amelogenin-binding extracellular matrix proteins, some of which are αvβ6 integrin ligands. As expected, these antibodies did not affect the spreading of Itgb6^−/− ameloblasts (data not shown).

We next tested whether αvβ6 integrin is involved in the regulation of amelogenin endocytosis in mouse ameloblasts. Amelogenin endocytosis occurred in a similar fashion in both WT and Itgb6^−/− ameloblasts (Fig. 8C–H), suggesting that αvβ6 integrin does not play a significant role in this internalization process.

Epithelial cell integrins regulate a variety of cell functions during development and tissue repair (Larjava et al., 2011). However, little information is available regarding integrin function in ameloblasts during enamel formation. In the present report, we show that αvβ6 integrin plays a crucial role in enamel biomineralization via regulation of amelogenin and enamelin gene expression.

Integrins likely play key roles in tooth development since several integrins, including α6, αv, β1, β4 and β5 integrin subunits, are expressed in the dental epithelium (Salmivirta et al., 1996). Integrin αvβ5 may regulate tooth morphogenesis, as its expression oscillates between dental mesenchyme and epithelium (Yamada et al., 1994). Ameloblasts express α2β1 integrin when they assume their columnar shape (Wu and Santoro, 1994), but no tooth phenotype was observed in α2 integrin knockout mice (unpublished data from our laboratory). Loss-of-function mutations in either α6 or β4 integrin cause human junctional epidermolysis bullosa, in which the epithelium detaches from the basement membrane causing skin or mucosal blistering (Wright, 2010). In these patients, ameloblast adhesion to developing enamel is also reduced, leading to hypoplastic enamel (Wright, 2010).

Only a few potential integrin-binding ligands have been identified in the enamel matrix. Amelogenin (Snead et al., 1983) and amelotin (Iwasaki et al., 2005; Moffatt et al., 2006) do not possess RGD motifs or other known integrin recognition sequences. The RGD motif in enamelin it is not evolutionarily conserved (Hu et al., 1997; Nawfal et al., 2007). Ameloblastin contains binding sites for the RGD motif-binding integrins (Černý et al., 1996), and it has been shown to bind to ameloblasts, but the receptor has not yet been identified (Fukumoto et al., 2004). Ameloblasts also express bone sialoprotein that has binding sites for both hydroxyapatite and RGD-binding integrins (Ganss et al., 1999; Harris et al., 2000; Stubbs et al., 1997). Dentin sialoprotein, which also contains the RGD motif, is only transiently expressed in presecretory ameloblasts and may contribute to formation of the dentino-enamel junction rather than to enamel maturation (Bègue-Kirn et al., 1998; Paine et al., 2005).

EMD enhances the adhesion, proliferation and matrix production of fibroblast-like but not epithelial cells (Cattaneo et al., 2003; Gestrelius et al., 1997; Haase and Bartold, 2001; Hoang et al., 2000; Kawase et al., 2000). Attachment of human periodontal ligament cells to EMD is possibly mediated by bone sialoprotein as it can be inhibited by RGD-containing peptides or with an anti-αvβ3 integrin antibody (Suzuki et al., 2001). Also β1 integrins seem important to cell adhesion to EMD (van der Pauw et al., 2002). More recently, it was demonstrated that both β1 and αv integrins mediate periodontal ligament fibroblast adhesion to EMD (Narani et al., 2007). Because amelogenins do not contain any known recognition sites for integrins, it is possible that self-aggregation of amelogenins exposes cryptic integrin recognition sites. Consistently with this hypothesis, amelogenins promote binding of fibroblasts and endothelial cells via multiple integrins, including αvβ3, αvβ5 and αvβ1 (Almqvist et al., 2010; Almqvist et al., 2011). In the present study, ameloblast αvβ6 integrin did not directly bind to EMD but the cells attached via endogenously deposited matrix. Some of these matrix molecules appeared to be αvβ6 integrin ligands, such as fibronectin that readily binds to amelogenin (Narani et al., 2007). However, also Itgb6^−/− ameloblasts were able to attach and spread on this matrix, suggesting collaboration by other integrin receptors in ameloblast adhesion. It appears unlikely that αvβ6 integrin is essential for ameloblast cell adhesion in vivo either, as the ameloblasts retained their columnar shape and adhesion to the developing enamel in the Itgb6^−/− mice.

Amelogenin or its fragments are endocytosed by ameloblasts via a receptor-mediated mechanism involving LAMP1 and CD63 proteins (Shapiro et al., 2007). This uptake may serve as a feedback loop to upregulate amelogenin expression through stabilization of its mRNA in the cytoplasm (Xu et al., 2006). We did not find evidence for the direct involvement of αvβ6 integrin in amelogenin endocytosis. However, we found that expression of several RBMs (Rbm26, Rbm45, Rbm8a) was increased (4.2- to 5-fold) in the enamel organs of Itgb6^−/− mice compared to WT. RBMs have been shown to critically regulate post-transcriptional RNA metabolism (Janga and Mittal, 2011). Therefore, these proteins could regulate amelogenin and enamelin mRNA stability. It is possible that the lack of αvβ6 integrin signaling in the Itgb6^−/− ameloblasts may disrupt the feedback system for amelogenin expression and lead to its dysregulated accumulation in the tooth enamel by an as yet unknown mechanism.

The best-known function of αvβ6 integrin is in the activation of latent TGF-β1 (Yang et al., 2007). TGF-β and its receptors are important regulators of early tooth development (Chai et al., 1994; Chai et al., 1999; Pelton et al., 1991). In addition, TGF-β1 is expressed in ameloblasts of developing enamel (Gao et al., 2009). Interestingly, the tooth development in TGF-β1 mutant mice is unaffected, which could be partially due to rescue by maternal TGF-β1 (D'Souza and Litz, 1995). TGF-βs binding to type I and II TGF-β receptors (TGFβRI and TGFβRII) induces activation of Smad2/3 (Moustakas and Heldin, 2008). Human mutations in TβRI and II cause Loeys-Dietz syndrome due to abnormal receptor signaling, but no enamel phenotypes have been reported for these patients (Loeys et al., 2005). However, there is evidence that dysregulated TGF-β1 signaling may be detrimental to amelogenesis. Ameloblasts in mice overexpressing TGF-β1 under the dentin sialoprotein promoter show enamel defects, including ameloblast detachment with a pitted and hypoplastic enamel (Haruyama et al., 2006). In K14-Smad2-overexpressing mice, the ameloblast layer is disorganized and amelogenin-containing matrix is found between ameloblasts (Ito et al., 2001). No tooth abnormalities have been reported in Smad7^−/− (an inhibitor of TGF-β signaling) mice. Mice overexpressing Smad7 under the K5 promoter, however, show suppressed Smad3-mediated signaling and a failure to produce proper enamel (Helder et al., 1998; Klopcic et al., 2007). In the present study, we did not find evidence for the enamel defects in Itgb6^−/− mice being caused by altered TGF-β1 activation. Neither phosphorylation nor the expression level of Smad2/3 was altered in the Itgb6^−/− enamel organ. We also compared histology of enamel from Smad3^−/− mice to that of Itgb6^−/−, and they lacked similarity (unpublished results). Previous studies have also shown that ameloblast morphology remains unaltered in Smad3^−/− mice although the enamel is poorly mineralized due to defective protein removal at the maturation stage (Yokozeki et al., 2003). Recently, it was reported that Smad3 and FoxO1 (transcriptional co-factor for Smads) collaboratively regulate genes involved in enamel formation (Poché et al., 2012). Interestingly, in both Smad3 and FoxO1 mutant teeth, the expression of Ambn, Amelx, Enam, Mmp20 and Klk4 were all significantly downregulated. The fact that Amelx and Enam are highly upregulated in β6 integrin mutant teeth also strongly suggests that the observed enamel phenotype is not caused by defective TGF-β1 activation. Clearly, TGF-β signaling regulates enamel formation but significant involvement of αvβ6 integrin in the activation of ameloblast TGF-βs seems unlikely.

We explored the gene expression profiles in the WT and Itgb6^−/− mouse enamel organs to identify additional changes that could contribute to disturbed mineralization of Itgb6^−/− mouse enamel. Interestingly, expression of Ank was increased in the Itgb6^−/− enamel organ. ANK functions in transporting inorganic pyrophosphate to the extracellular space, where pyrophosphate serves as a potent inhibitor of extracellular matrix mineralization in calcified tissues (Harmey et al., 2004; Wang et al., 2005; Zaka and Williams, 2006). Future studies should therefore investigate whether inorganic pyrophosphate levels are increased in Itgb6^−/− enamel organs, thereby contributing to the mineralization defect.

Other significant changes in gene expression observed in the Itgb6^−/− enamel organ included reduced expression of arrestin Arrdc3, photocadherin Pcdh21 and Fam20B. ARRDC3 functions in cell signaling and molecular trafficking (Rajagopal et al., 2010). It remains to be shown whether it is involved in the regulation of Amelx expression. Mutations in PCDH21 cause autosomal recessive cone-rod dystrophy in the eye (Ostergaard et al., 2010), which is associated with AI in some patients, thus demonstrating possible links between retinal functions and enamel mineralization (Parry et al., 2009; Polok et al., 2009). Since cell-cell adhesion of ameloblasts via cadherins has been demonstrated to be critical for enamel mineralization (Bartlett et al., 2010), it is possible that downregulation of Pcdh21 contributes to weakened ameloblast cell-cell adhesion and subsequent accumulation of amelogenins between the cells. FAM20B is a kinase implicated in phosphorylation and control of proteoglycans (Koike et al., 2009). It and other members of the FAM20 protein family have multiple roles in mineralized tissues. Mutations in FAM20A cause hypoplastic AI and gingival overgrowth (Cho et al., 2012; O'Sullivan et al., 2011), whereas FAM20C that is expressed in mineralized tissues is essential for normal bone development (Hao et al., 2007; Simpson et al., 2007; Wang et al., 2010; Ishikawa et al., 2012). FAM20B is expressed during the maturation stage of amelogenesis (O'Sullivan et al., 2011), but its function in enamel remains unknown.

Amelogenesis imperfecta represents a collection of genetic disorders that affect enamel formation both in the primary and permanent dentition in the absence of systemic manifestations (Hu et al., 2007). Mutations in several genes are associated with human AI (Hu et al., 2007; Stephanopoulos et al., 2005). In general, mutations that affect enamel matrix production (AMELX, AMBN, ENAM) tend to result in hypoplastic enamel while those that affect matrix removal (MMP20, KLK4) tend to result in hypomaturation of the enamel (Hu et al., 2007). Interestingly, previous studies with overexpression of normal amelogenin or its various fragments have not shown major alterations in enamel structure (Chen et al., 2003; Gibson et al., 2007; Paine et al., 2004). However, the present study clearly demonstrates that the enamel defect resulting from αvβ6 integrin deficiency and the subsequent overexpression of amelogenin and enamelin is detrimental to enamel formation and leads to a condition that mimics the hypomaturation type of AI. Therefore, ITGB6 should be considered a candidate gene for human AI with normal thickness but altered prism structure and reduced mineralization. As discussed above, there are several putative mechanisms for how αvβ6 integrin regulates enamel biomineralization that need to be explored further.

The Animal Care Committee of the University of British Columbia approved all animal procedures used in this study. The mouse lines used were: WT (FVB background), Itgb6^−/− (a generous gift from Dr Dean Sheppard, University of California, San Francisco, CA, USA), K14β6 that overexpresses human β6 integrin under the K14 promoter (Häkkinen et al., 2004) and β6 rescue mice, which were the progeny of Itgb6^−/− mice bred with the K14β6 mice (F1; also from Dr Sheppard). The mice were maintained in a conventional animal care facility and had free access to standard mouse chow (Purina 5001) and water. Animals were sacrificed by CO₂ inhalation.

The incisors of the maxillas and mandibles of 3- to 9-month-old mice (three animals, 8–10 incisors per group) were removed, and their enamel organs scraped off with a microsurgical blade and pooled for western blotting. The antibodies used were: ameloblastin (ab72776; Abcam, Cambridge, MA, USA), amelogenin (pc-062; Kamiya Biomedical Company, Seattle, WA, USA), KLK4 (ab3636; Abcam), MMP-20 (ab76109; Abcam), Smad3 (ab28379; Abcam), pSmad3 (ab52903; Abcam), Smad2 (#3103; Cell Signaling Technology, Danvers, MA, USA), pSmad2 (#3101; Cell Signaling), β6 integrin (AF2389; R&D Systems, Inc., Minneapolis, MN, USA), αv integrin (sc-6618, Santa Cruz Biotechnology, Santa Cruz, CA, USA), β1 integrin (4080; Larjava et al., 1990), β-actin (ab8227; Abcam) and β-tubulin (mAb3408, Millipore, Temencula, CA, USA). Appropriate peroxidase-conjugated IgGs (Santa Cruz) were used as secondary antibodies. After washing, the protein bands were detected using the ECL Western Blotting Detection Kit (GE Healthcare, Baie d'Urfe, QC, Canada), and the digitized images were quantified using the ImageJ software (available at http://rsb.info.nih.gov/ij; developed by Wayne Rasband, National Institutes of Health, Bethesda, MD, USA). Alternatively, IRDye-conjugated secondary antibodies (LI-COR Biosciences, Lincoln, NE, USA) were used and the blots analyzed and quantified with an Odyssey infrared reader (LI-COR).

The enamel organs were collected from the incisors of 6-month-old WT and Itgb6^−/− mice, pooled (three animals, 8–10 incisors per group), placed into RNAlater (Ambion, Life Technologies, Inc., Burlington, ON, Canada) and stored at −80°C. Total RNA was extracted using either NucleoSpin RNA II or XS kit (Macherey-Nagel, Bethlehem, PA, USA) and treated with DNaseI digestion. RNA samples were analyzed using Illumina Mouse WG-6 v.2.0 Expression BeadChip at the Finnish Microarray and Sequencing Centre (Turku Centre for Biotechnology, Turku, Finland) followed by data analysis with R (R Development Core Team, 2008) and Bioconductor (Gentleman et al., 2004) softwares. The data were quantile normalized, and statistical analysis for detecting the global differences in gene expression between the groups was carried out using Bioconductor's Limma package. The chosen thresholds used in filtering the differentially expressed genes were FDR P-value <0.05 and absolute fold change >1.6 as the comparison cutoff.

Total RNA (0.5 µg) was reverse-transcribed with SuperScript VILO cDNA Synthesis Kit (Invitrogen, Life Technologies). Real-time PCR amplification was performed on the MiniOpticon Real-Time System (Bio-Rad, Mississauga, ON, Canada) using 5 µl of RT products mixed with 10 µl of 2× iQ SYBR Green I Supermix (Bio-Rad) and 5 pmoles of primers, in a final volume of 20 µl. An amplification reaction was conducted for target genes with Actb, Gapdh and Hprt1 as reference genes and replicated nine times for each sample. The data were analyzed based on the comparative C_T program of Gene Expression Analysis for iCycler iQ Real-Time PCR Detection System (Bio-Rad). Primer sequences (5′–3′) were: Actb CTTCCTTCTTGGGTATGGAATC, TAGAGGTCTTTACGGATGTCAAC; Ambn CTTCCTTCTTGGGTATGGAATC, TAGAGGTCTTTACGGATGTCAAC; Amelx GCTTTTGCTATGCCCCTA, CTCATAGCTTAAGTTGATATAACCA; Itgb6 AATCACCAACCCTTGCAGTAG, AATGTGCTTGAATCCAAATGTAG; Enam TCTCTGCTGCCATGCCATTC, TTGATTATATCGCATCATCTCTTCAC; Klk4 CATCCCTGTGGCTACCCAA, GGGCAGTTTCCCATTCTTTA; Mmp20 TGCTGTGGAACTGAATGGCTA, ACACTAACCACGTCTTCCTTC; Hprt1 TGTTGGATTTGAAATTCCAGACAAG, CTTTTCCAGTTTCACTAATGACACAA; Krt14 ATCCTCTCAATTCTCCTCTGGCTC, ACCTTGCCATCGTGCACATC; Itgb1 GCTGGTTCTATTTCACCTATTCA, CAACCACGCCTGCTACAA; Itgb4 CCAGCTGAGACCAATGGCGA, GAGCACCTTCTTCATAGGTCCA; Gapdh CTTTGTCAAGCTCATTTCCTGGTA, GGCCATGAGGTCCACCA.

Maxillas and mandibles of 3- to 24-month-old WT and Itgb6^−/− mice were defleshed mechanically and with 2% KOH (EM Science, Merck, Darmstadt, Germany). Attrition of the molars was scored based on cusp heights where 0 = <10% attrition of the cusps, 1 = attrition reaching up to 50% of the cusp height, and 2 = more than 50% attrition. In a cross-sectional study, we also compared the attrition rate of molars in 6-month-old WT, Itgb6^−/−, K14-β6 and rescue mice.

WT and Itgb6^−/− mice (11- to 14-months old) were sacrificed, decapitated and their heads were fixed in 4% formaldehyde in phosphate-buffered saline (PBSp; pH 7.2). The mandibles were dissected and further fixed in 2.5% glutaraldehyde for 1 h at 4°C. The samples were post-fixed with 1 M OsO₄ for 1 h at RT. After three rinses, the specimens were dehydrated, critical-point dried, gold-coated and viewed by scanning electron microscopy (SEM). Three standardized areas (mesial, middle and distal) on the mid-lingual surface of the third molar were used to measure calcium-to-phosphorus ratios using energy dispersive X-ray spectroscopy (Hitachi S-3000N SEM with light element EDX).

Defleshed and dried incisors from WT and Itgb6^−/− mice were frozen by immersion in liquid nitrogen and then fractured (along the line from root to crown) with a pre-cooled scalpel blade. The fractured portion of the incisor was treated at RT in 35% phosphoric acid for 30 s, rinsed and air-dried (Sánchez-Quevedo et al., 2006). The incisors were then mounted and coated with a 10 nm layer of gold/palladium. Each sample was imaged by SEM with identical imaging conditions at 4 kV.

Immunostaining of β6 integrin was performed on frozen sections (6 µm) of the secretory stage of an upper incisor and a developing upper molar from a 3-day-old WT mouse using the β6 integrin antibody β6B1 (a generous gift from Dr Dean Sheppard).

Impacted human third molars containing the dental follicle were collected anonymously from patients requiring extractions of these teeth as a part of their treatment (approved by the Clinical Research Ethics Board, University of British Columbia). Teeth with soft tissue remaining were fixed with 2% formaldehyde in PBS and then decalcified in the same fixative containing 0.4 M EDTA. Frozen sections (6 µm) containing patches of cells resembling ameloblasts were identified and used for immunolocalization of K14 (MCA890; AbD Serotec, Oxford, UK), amelogenin and β6 integrin as described above using a fluorescently-labeled secondary antibody (Alexa Fluor 488; Invitrogen).

Hemi-maxillas and -mandibles from 7-week-old male and female WT and Itgb6^−/− mice were removed and cleaned of adhering soft tissues. Procedures employed for isolating and removing sequential 1-mm-long strips of developing enamel from maxillary and mandibular incisors have been described in detail previously (Smith et al., 2005; Smith et al., 2009; Smith et al., 2011). Directly measured ‘before’ and ‘after’ heating weights were used to calculate various parameters including mineral-to-volatiles ratio, which represents the after-heating weight (mineral content of sample) divided by the difference between the initial dry weight minus the after-heating weight (amount of volatiles in sample; mostly protein). Weight data for enamel strips were collected from a minimum of nine maxillary and nine mandibular samples per genotype.

Jaws from 10-day-old mice were fixed in 4% paraformaldehyde, decalcified for 24 h using Morse's solution (Nakatomi et al., 2006) and then embedded in paraffin. Sections (5 µm) were treated with proteinase K at 7 µg/ml in PBS for 25 min, postfixed with paraformaldehyde and then treated with 0.2 M HCl for 10 min prior to the hybridization using digoxigenin–UTP-labeled riboprobes as previously described (Yoshida et al., 2010). A plasmid for integrin β6 integrin riboprobe was generated by inserting PCR products from mouse cDNA into pCRII-TOPO vector (Invitrogen). Primers used for the PCR were 5′: AGGGGTGACTGCTATTGTGG and 3′: GGCACCAATGCTTTACACT.

WT and Itgb6^−/− mouse mandibles were immersion-fixed overnight in sodium-cacodylate-buffered aldehyde solution and cut into segments containing the molars, underlying incisor and surrounding alveolar bone. The samples were dehydrated through a graded ethanol series and infiltrated in acrylic resin (LR White; London Resin Company, Berkshire, UK) followed by polymerization of at 55°C for 2 days. 1-µm sections of hemi-mandibles were cut with a diamond knife using an ultramicrotome, and glass slide-mounted sections were stained with 1% Toluidine Blue and von Kossa reagent to detect mineralization. Immunogold labeling for amelogenin coupled with TEM was performed as described previously (McKee and Nanci, 1995; McKee et al., 1996) using an anti-amelogenin antibody kindly provided by Dr Takashi Uchida of Hiroshima University, Japan (Uchida et al., 1991).

Micro-computed tomography (Micro-CT, model 1072; Skyscan, Kontich, Belgium) of undecalcified hemi-mandibles was performed at the level of the first molar from three samples of each genotype. The X-ray source was operated at maximum power (80 KeV) and at 100 µA. Images were captured using a 12-bit, cooled, charge-coupled device camera (1024×1024 pixels) coupled by a fiber optic taper to the scintillator. Using a rotation step of 0.9°, total scanning time was 35 min for each sample with a scan resolution of 5 µm, after which ∼300 sections (slice-to-slice distance of 16.5 µm) were reconstructed using Skyscan tomography software. Appropriate imaging planes were selected to show three-dimensional longitudinal and cross-sectional ‘sections’ (segments) of the first molar and underlying incisor.

Maxillary and mandibular incisors of 5- to 6-month-old WT and Itgb6^−/− mice were prepared under aseptic conditions and placed in cell culture wells pre-coated with bovine fibronectin (10 µg/ml; Millipore) and collagen (30 µg/ml; PureCol®; Advanced BioMatrix, Inc., San Diego, CA, USA) in keratinocyte growth medium (KCM) (Häkkinen et al., 2001). The cells growing out from the explants were differentially trypsinized to selectively remove the more easily released fibroblastic cells. The remaining epitheloid cells were confirmed to express ameloblast marker K14 as well as to be positive (WT) or negative (Itgb6^−/−) for β6 integrin expression by PCR. The ameloblasts were routinely grown in KCM.

For cell spreading assays, WT ameloblasts were seeded in plates coated with EMD (10,000 µg/ml in 10 mM acetic acid; Emdogain®, Straumann, Basel, Switzerland) in triplicates in the presence or absence of 50 µM cycloheximide (a protein synthesis blocker) and 200 nM TPA (an integrin activator). Cell spreading was quantified as described previously (Narani et al., 2007).

To explore the role of αvβ6 integrin in cell spreading on EMD, WT or Itgb6^−/− ameloblasts were pre-incubated on ice with anti-integrin antibodies and then allowed to spread on EMD in the presence of TPA and in the absence of cycloheximide. The anti-αvβ6 integrin antibodies used were: MAB2077Z (50 µg/ml; Millipore), 6.8G6 (10 µg/ml; Biogen Idec, Cambridge, MA, USA) (Weinreb et al., 2004) and 6.3G9 (10 µg/ml; Biogen).

WT and Itgb6^−/− ameloblasts were seeded onto glass coverslips (10,000 cells per ml) for 24 h in KCM. The cells were treated with KCM only, with 100 µg/ml of EMD or pre-treated for 10 min with 20 µM E64d [(2S,3S)-trans-epoxysuccinyl-L-leucylamido-3-methylbutane ethyl ester; inhibits lysosomal degradation of endocytosed proteins (McGowan et al., 1989)], followed by addition of 100 µg/ml of EMD. The cells were then incubated for 24 h and processed for immunofluorescence staining with anti-amelogenin antibody (PC-062).

The experiments were repeated at least three times. The difference between WT and Itgb6^−/− mice was calculated using unpaired, two-tailed Student's t-test using GraphPad InStat 3 software. Multiple comparison tests were performed using one-way ANOVA with Tukey's post-test. Statistical significance was set at P<0.05.

The authors thank Mr Cristian Sperantia for his technical help, Mrs Ingrid Ellis for her editorial help, Dr Dean Sheppard for providing the Itgb6^−/− and rescue mice, Dr Shelia Violette for the β6 integrin antibodies and Dr Takashi Uchida for the amelogenin antibody.

L.M. collected and analyzed most of the data and wrote a large part of the manuscript; L. Koivisto assisted in western blotting and cell spreading experiments and manuscript writing; G.J. performed real-time PCR analyses; L. Kytömäki performed and analyzed the gene arrays; M.H. and C.W. assisted in collecting samples for immunohistochemistry; G.R.O. assisted in endocytosis experiments; Y.X. assisted in immunolocalization studies; K.H. provided samples and assisted in manuscript writing; T.Y. performed in situ hybridization experiments; C.E.S. analyzed the mineral and protein contents of enamel; J.H. assisted in gene profiling and manuscript writing; L.H. participated in experiment planning and manuscript writing; M.D.M. performed hard tissue sectioning and EM studies and assisted in manuscript writing; H.L. supervised all experiments and participated in manuscript writing.