Amelogenins are expressed by ameloblasts during enamel formation. A heterogeneous mixture of amelogenins has been extracted from the forming enamel with large variations in MW from 28kDa to 5kDa. These multiple forms have been attributed to alternative splicing [9] and/or proteolytic cleavages [2] and [15]. Amelogenins are expressed by seven exons, and seven alternatively spliced amelogenin mRNAs have been isolated from the cDNA library, coding for amelogenins formed by 194, 180, 156, 141, 74, 59 and 44 amino acids residues in mouse [27]. In addition, the 25–28 kDa nascent proteins rapidly loose the telopeptides and the resultant amelogenin (23 kDa) then accumulates in the matrix after further cleavage (20–21 kDa). This is a crucial step in the mineralisation process, as the 21–20-kDa amelogenin can then bind minerals and form a solid-state aggregate. This follows by cleavage by either MMP20 or MMP-9; both metalloproteinases have been shown to degrade amelogernin in vitro ([2] and [8] and our unpublished data). A degradation cascade has been described [25], which results in small molecular species of 13 kDa and 11k Da, and then 5-, 4.5-, 4.3-kDa fragments and finally residual 3-kDa peptides, which are gradually eliminated from the forming tissue [8], [18] and [25].

Three domains have been identified in the amelogenin molecule. The first is a highly conserved amino-terminal ‘TRAP’ (tyrosine-rich amelogenin peptide) sequence of 44/45 residues. This sequence is phosphorylated in serine-16, but not glycosylated. The TRAP segment has lectin-like properties and is similar to wheat germ agglutinin as it binds N-acetyl-d-glucosamine [23]. The second domain is a central hydrophobic core of 100–130 residues, extremely rich in proline (one residue on four). Finally, the carboxyl-terminal domain is formed by a small acidic hydrophilic sequence of 13 amino acids [8].

Although amelogenins do not possess in vitro the specific crystal-modulating properties characteristic of certain acidic mineralised tissue proteins [17], examination of the forming enamel at the ultrastructural level supports the existence of tubules and stippled (granular) materials where the first crystal may be formed (see [11] for a review). The significance of the stippled materials has been disputed as it was suggested to result from defective fixation methods, or to constitute residual biological material before intracellular re-internalisation of excesses or/and degraded peptides during enamel formation [12] and [24]. However, rapid freezing and freeze substitution that preserve the actual native structure of the forming enamel, and radioautography, suggest that this stippled material is a real structure and represents an enamel precursor. These granules result from self-association properties between the N- and C- termini of amelogenin, which lead to the formation of nanospheres where initial mineralisation may be triggered [7]. The role of amelogenin in enamel formation and mineralisation is substantiated by the following in vivo findings: •

the supra-molecular control of HAP crystal growth is under the control of amelogenin [5];

The importance of amelogenins in enamel formation has been clearly demonstrated. The molecules are already expressed by presecretory ameloblasts, but no enamel is formed at the early stages, as the secreted enamel proteins presumably diffuse through the porous predentin and the unmineralized mantle dentin. At least some of these proteins, and in particular the different low-molecular-weight species of amelogenin are thought to be internalised into odontoblasts, thus preventing their accumulation in the dentin. The internalised amelogenins are then presumably destroyed in the lysosomes of the odontoblasts. Amelogenin diffusion is gradually diminished by dentin mineralisation, causing amelogenins and the other enamel proteins, to be retained at the dentin–enamel junction (DEJ), initiating enamel formation.

Two series of data have demonstrated that the situation is in fact more complex and suggest that the amelogenin peptides may actually serve as signalling molecule in the secretory odontoblasts. Firstly, the demonstration of amelogenin splice products in the odontoblast gene library [21] and [30], and secondly, the identification of amelogenin degrading enzyme, enamelysin or MMP-20, in odontoblasts.

The group of Veis [1] identified in dentin extract a chondrogenic inducing agent, which was later revealed to be an amelogenin [21]. Veis et al. [30] have demonstrated that it was a spliced-gene product. Two specific cDNAs were found in the odontoblast gene library, the first one being expressed by exons 2 to 7, and the second one by exons 2, 3, 5, 6d and 7, in other words, all exons except 4. The first corresponding recombinant molecule was designated as A+4 and was shown to have a molecular weight of 8.1 kDa. The second, named A–4, was smaller (6.9 kDa). They both enhanced the in vitro incorporation of sulphate into proteoglycans and increased type-II collagen synthesis. The two molecules also stimulated Sox9 and Cbfa1 mRNA expression. In vivo, the expression of BSP and BAG-75 (DMP-1) was enhanced in implant assays. Along this line, it was also recently demonstrated that amelogenins are transiently expressed by odontoblasts at early stages of dentin formation [22]. Furthermore, results from our laboratory, using agarose beads soaked in A+4 or A–4 solutions and implanted in the pulp of rat's molar, have shown that amelogenin molecules stimulated the recruitment of stem cells or the dedifferentiation of resident adult cells. The recruited cells were shown to divide both in the area of implantation and in the central part of the pulp. The daughter cells developed two phenotypes: an osteoblast-like phenotype for the cells regrouped around agarose beads, with osteopontin labelling, and an odontoblast-like cells, as suggested by the positive DSP immunostaining, for the cells located at the surface of the wounded pulp and involved in the construction of a reparative dentinal bridge. However, final conclusions for these experiments await confirmation of the phenotype of these cells by more specific markers.

Taken together, these data suggest that amelogenins are not only involved in enamel formation, but may also function as signalling molecules. This property of amelogenin is currently exploited therapeutically in periodontal and bone regeneration (Emdogain). Our experiments open gates on the healing of pulp damages.

During crown formation, the enamel organ plays a role in regulating enamel formation and consequently in the secretion and diffusion of amelogenins in dentin. During root formation, the epithelial Hertwig's root sheath is instrumental in the recruitment and differentiation of root odontoblasts. For a long time, differences between the root and crown dentin have been recognized. It is also the case for the coronal and root pulp [28], shown to respond in a different way to BMP-7. It was suggested that the environment was different during the formation of the tooth, and this may be related to the presence or absence of amelogenin. Bosshardt and Nanci [3] denied the presence of amelogenin within Hertwig's sheath. As a conflicting view, two alternate splice forms of amelogenins (M180 and LRAP) were identified in the periodontal region (bone, ligament and cementum) of mouse tooth roots [16]. Altogether, and taking into account the different methods used, the different stages of formation and the tissues that were examined, it may be stated that the absence of amelogenin in the Hertwig's root sheath seems to interfere with the root dentin formation, but is not crucial, may be because some other molecule(s) are over-expressed and rescue partially the deficiency. It comes out clearly from the study that was carried out on the amelogenin null mice [16] that the lack of M180 and LRAP expression is correlated with cementum defects. The defects are associated with an increase expression of nuclear factor kB ligant (RANKL), a regulator of osteoclastogenesis. This provides another evidence for the signalling functions of amelogenins.

Our recent immunohistochemical and immunogold electron microscopic results confirmed the presence of MMP-20 protein in both ameloblasts and odontoblasts of rat incisor and showed that MMP-20 protein colocalizes with amelogenin, suggesting that in odontoblasts too, amelogenin may be the target substrate. This co-localization of amelogenin and MMP-20 would argue for local synthesis of both enzyme and substrate.

Unlike ameloblasts, where MMP-20 is found to be expressed in the cells and secreted into the matrix, suggesting a role in the processing of enamel proteins leading to enamel maturation, MMP-20 expressed in odontoblasts was not detected in the corresponding predentin/dentin matrix. This suggests that the local degradation of amelogenin by MMP-20 in odontoblasts would produce low-molecular-mass peptides, which may serve as signalling molecules. A schematic representation of the possible various pathways and hypothesis is shown in Fig. 3.