Enamel formation is a unique biomineralization process involving a highly organized matrix protein scaffold deposition and degradation, leading to hydroxyapatite crystal nucleation generating a dense and tightly aligned network of hydroxyapatite crystals. Mature enamel is the body's hardest tissue. Ameloblasts are cells of ectodermal origin responsible for enamel development. These cells secrete enamel proteins, which are required for correct mineralization and structural maturation. Enamel proteins self-assemble and provide a matrix organization that aligns the thin ribbons of calcium phosphate deposited during enamel appositional growth. Enamel formation is characterized by inductive, secretory, and maturation stages. During the inductive stage, the inner enamel epithelium begins to differentiate. Then at the secretory stage, polarized differentiated ameloblasts release enamel proteins, contributing to the enamel matrix (Bei, 2009). Finally at the maturation stage, ameloblasts absorb water and organic matrix. This dehydration allows dense crystal deposition. Mature enamel is thus extremely strong, because of the density and fine organization of its crystal layers (Bei, 2009; Simmer et al., 2010).

Amelogenesis imperfecta (AI) refers to a group of rare genetic diseases presenting with defects in enamel formation either as isolated trait, or in association with other symptoms. Patient cases of AI are classified into hypoplastic, hypomineralized, or hypomaturation categories based on enamel quantitative or qualitative defects, i.e., thickness, hardness, and/or color. To date, mutations in over 30 genes are associated with non-syndromic or syndromic AI (Bloch-Zupan et al., 2012). These diseases can be recapitulated in several mouse models. For example, hypoplastic or aplastic enamel deficiencies are produced by amelogenin (Amel), ameloblastin (Ambn), and enamelin (Enam) mutations, recapitulating defects in patients (Gibson et al., 2001; Fukumoto et al., 2004; Masuya et al., 2005; Hu et al., 2008, 2014). Studies in mouse models have contributed to understanding how defective ameloblast protein secretion contributes to these diseases. For a given genetic defect, interfamilial, intrafamilial, and individual intraoral variations in phenotype severity are often seen, suggesting that environmental factors come into play. Indeed, ameloblasts are highly sensitive to their environment (Simmer et al., 2010). Body stressors including high fever (Ryynänen et al., 2014), excess fluoride (Yang et al., 2015), and endocrine disrupters such as bisphenol A disrupt ameloblast function (Jedeon et al., 2014). It is likely that a variety of factors contribute to clinical heterogeneity found in AI and to acquired enamel defects such as molar incisor hypomineralization (MIH) or hypomineralized second primary molars (HSPM; Alaluusua, 2010; Jeremias et al., 2013). Our objective is to discover novel factors initiating and influencing enamel regulatory networks, with the aim of designing new strategies to alleviate dental defects. To further characterize nutritional and environmental factors regulating enamel formation, we have focused on the role of retinoic acid (RA), the main active form of vitamin A that plays key roles during vertebrate development (Niederreither and Dollé, 2008). RA is the ligand for nuclear receptors (RARα, β, and γ), which bind as heterodimers with RXRs to DNA regulatory elements termed RA-response elements (RAREs, Rochette-Egly and Germain, 2009). Through this mechanism, RA regulates expression of various target genes (Balmer and Blomhoff, 2002). RA distribution within embryonic tissues is tightly controlled through an interplay between enzymes involved in its synthesis (mainly retinol dehydrogenase 10 and retinaldehyde dehydrogenases [Raldh]1, 2, and 3) and catabolism (Cyp26A1, B1, and C1). As both RA deficiency and excess results in diverse developmental defects, the distribution of active retinoid signaling requires tight regulation to limit potent adverse effects (Rhinn and Dollé, 2012).

Expression of RA receptors, synthesizing, and catabolizing enzymes has been detected in the developing teeth (Bloch-Zupan et al., 1994; Berkovitz et al., 2001; Cammas et al., 2007). Severe dietary vitamin A deficiency in rats leads to hypoplastic enamel, enamel organ metaplasia, dentine dysplasia, and delayed tooth eruption (Mellanby, 1941; Punyasingh et al., 1984; McDowell et al., 1987). Hypervitaminosis A during rodent pregnancy induces exencephaly and craniofacial malformations, along with tooth fusions and supernumerary incisors (Knudsen, 1967 and references therein). These effects may reveal an evolutionary role of RA signaling in the posterior pharyngeal region controlling tooth number (Seritrakul et al., 2012). In culture, RA excess can retard molar growth (Mark et al., 1992), reduce ameloblast differentiation (Kronmiller et al., 1992), and diminish tooth alkaline phosphatase production (Jones et al., 2008). Retinoids may also regulate incisor cervical loop maturation, increasing mitosis and laminin gene expression (Bloch-Zupan et al., 1994).

In vivo models to substantiate RA actions in tooth development are lacking. Interestingly, though, Cyp26b1 knockout mice show a defect in maxillary bone compaction around upper incisors (Maclean et al., 2009). Observing that the fetal tooth has robust expression of Cyp26b1 and Cyp26c1 RA-degrading enzymes, we hypothesized that in vivo, conditions of RA excess may have adverse effects on osteoblast and ameloblast growth regulatory networks. We report that mice born from dams exposed to RA during mid-late pregnancy using a food-based supplementation suffer adult-stage enamel hypoplasia. Effects were strongest when treatments began at the tooth dental lamina-placode stage (E12.5), and continued until early bell developmental stages. Reductions of Enam, Ambn, and Odam mRNA expression in E16.5 lower incisors were observed. High throughput RNA sequencing (RNA-seq) analysis of lower incisors revealed that RA excess perturbs neural crest lineage determinants and pro-ossification growth factor and transcription networks. Combinatorial changes in collagen, extracellular matrix, and calcium homeostasis genes occur at E14.5, followed by a decrease at E16.5 of transcripts encoding pre-ameloblast secretory-stage proteins. These alterations in gene expression are observed several days before the ameloblast lineage begins to differentiate. Retinoid excess targets fetal odontoblasts, along with a range of epithelial enamel protein targets. Our data provides potential avenues through which environmental and nutritional changes may alter the penetrance and expressivity of human enamel defects such as AI or MIH.

To analyze RA-dependent tooth alterations, we employed a protocol where RA added to the food supply was administered to pregnant CD1 mice (Niederreither et al., 2002), at a concentration of 0.4 mg/g food beginning at E12.5 or later. The treatment period began after craniofacial neural crest migration into the head was complete (Minoux and Rijli, 2010), avoiding earlier stage lethality due to exencephaly. In another study, HPLC analysis carried out after similar RA treatment at 0.1 mg/g food showed that serum RA levels were moderately increased (~20%) compared with untreated controls (Mic et al., 2003). Treated dams bore litters with 50% lethality, typically with 5–7 pups. Incisors in both groups erupted at the same age. Once the pups reached adulthood (4–7 weeks-old), we compared 100 randomized control and RA-treated groups macroscopically for dental defects. The labial side of rodent incisor (analog of the crown and covered with enamel) normally has a yellow/orange pigmentation, due to iron present at a net weight of about 0.03% in enamel (Pindborg, 1953). It gives mouse teeth a characteristic color, which is a dark orange in the upper incisors (Figures 1A,E). When RA treatments were performed from E12.5 to 16.5, they were found to produce a chalky lightening and length reduction of incisors, changes more pronounced in lower incisors (Figures 1B,F). The whiter color and less shiny surface may reflect reduced enamel thickness typical of mouse enamel hypoplasia models (Gibson et al., 2001; Masuya et al., 2005; Hu et al., 2014). Treatments performed at later stages (E13.5–16.5: Figures 1C,G, or E14.5–17.5: Figures 1D,H) had less severe effects on enamel, suggesting RA has early roles in the oral epithelium starting at the placode stage of tooth initiation.

Samples shown in Figures 1A,E (WT) and B,F (RA-treated) were analyzed by X-ray micro-computed tomography (micro-CT). This analysis confirmed lower incisor shortening (Figures 2A,B; Figure S1). Optical sections revealed a reduction in enamel mineral density and thickness (molars in Figures 2C,E; incisors in Figures 2D,F). Alveolar bone at the level of the molars showed greater porosity in the RA-treated animals (boxes in Figures 2C,E; Figure S2). Although retinoid gradients shape pharyngeal tooth evolution (Seritrakul et al., 2012), our micro-CT analysis revealed normal molar eruption and cusp morphology (Figure S3), as predicted because RA treatments are initiated at E12.5, after neural crest has completed its migration into the jaw. This analysis also revealed no evident signs of dental attrition from both groups. Figure 2G shows scanning electron micrographs (SEM) of enamel prisms of control lower incisor, exhibiting a well-organized decussating pattern. This was disrupted following RA treatment. The most outer enamel was less mineralized when compared with the control tooth, as outer enamel appeared darker in RA-treated animals. Enamel rods of RA-supplemented animals were less densely packed, and as a consequence, areas normally filled with interprismatic enamel seemed empty and showed holes-like pattern (Figure 2H), similar to Enam haploinsufficent mice (Hu et al., 2014).

Histological analysis of 1-week old mice after hematoxylin-eosin staining confirmed that RA treatments produced a shorter, smaller, and disorganized layer of ameloblasts in both molars (Figures 3A,B) and incisors (Figures 3C–H). This was clearly seen in the secretory zone, where ameloblasts displayed disrupted epithelial sheet organization (Figure 3H). In all treated samples, ameloblast adhesion to enamel was impaired (Figures 3B,D,G,H). This was first seen histologically as a detachment of the basement membrane, likely causing preameloblast separation from forming odontoblasts (Figure 3G, red arrowhead). Non-cellular organic material was present between layers (Figure 3B, black arrowhead). The lower incisor secretory zone enamel layer was slightly thinner (Figure 3H), suggesting RA treatment delayed enamel maturation and/or reduced overall mineralization.

Enamel matrix proteins are secreted by ameloblasts and form a matrix directing enamel mineral deposition. Among these proteins, amelogenin is the most abundant (Eastoe, 1979). Enamel fails to form or is hypoplastic in amelogenin-deficient mice (Gibson et al., 2001), and in Enam- or Ambn-null mutant mice (Fukumoto et al., 2004; Masuya et al., 2005; Hu et al., 2014). Our real-time RT-PCR analysis of E16.5 lower incisors following E12.5–16.5 RA treatment revealed up to 20-fold reductions in Enam, Ambn, and Odam mRNAs (Figure 4). Notably, no RA-induced alterations in Enam, Ambn, or Odam were observed at E14.5 (data not shown).

To assess if reduced enamel protein expression was linked to ectopic activation of RA signaling, we used RARE-hsp68-lacZ transgenic mice as a reporter for RA activity (Rossant et al., 1991). No expression was seen in the tooth germ areas at E13.5 (data not shown), although activity was found in the mandible next to the tooth germs at E14.5 (Figures 5A,B). This retinoid reporter transgene exhibited low level of expression in both mandible and maxilla adjacent to the growing incisors and molars in E16.5 untreated fetuses (Figure 5C and data not shown). In E16.5 fetuses following RA treatment from E12.5 to 16.5, domains of RARE-lacZ activity broadly extended into alveolar bone and surrounding mesenchyme, but appeared absent from enamel organ and dental ectomesenchyme (Figure 5D and data not shown). Note that very low levels of RA signaling may not be detected by such a reporter transgene. Proper control of RA levels is necessary for early neural crest patterning, but RA deficiency does not alter first branchial arch formation (Niederreither et al., 2000). To explore why RA activity was absent from most tissues of the tooth buds, we examined expression of CYP26 family cytochrome P450 enzymes specifically involved in RA catabolism. At E13.5, Cyp26b1 expression (Figure 5E) was found to be prominent in mesenchyme surrounding the forming incisors. At E14.5, Cyp26c1 (Figure 5F) was prominently expressed in the enamel organ, whereas a low amount of Cyp26a1 transcripts was seen in the dental papilla (Figure 5F, insert).

Excesses of vitamin A or RA lead to skeletal fragility by reducing bone formation and by stimulating bone-resorbing osteoclasts (Henning et al., 2015). In the Cyp26b1 mutant, impaired RA catabolism causes long-bone fusions and induces premature osteoblast differentiation into mineralizing osteocytes, truncating bone development in the craniofacial region (Maclean et al., 2009). Alizarin red/alcian blue staining of E14.5 skulls 2 days into RA regime revealed shortened mandibles, and truncated regions of ossified maxilla and premaxilla in the RA-treated animals (Figures 6A,B). The skull frontal plate was also smaller in treated animals, with less ossification. Malformed Meckel's cartilage and truncated mandibles could lead to incisor shortenings. Skeletal staining performed at E15.5 confirmed RA-reduced mineralization (Figures 6C,D). RT-PCR analysis showed reductions in Runx2 mRNA in both lower incisor and adjacent alveolar bone at E14.5 in RA-treated animals (Figures 6E,F). A 3-fold reduction in the expression of this key target might account for reduced bone mineral deposition (Ducy and Karsenty, 1998), and its mesenchymal localization (Figure 6G) implies indirect enamel effects. To exclude systemic non-specific RA effects, we cultured isolated lower incisors or lower incisors with adjacent alveolar bone from E13.5 embryos (Figure S4). When 10⁻⁸ M RA was added to culture medium, Enam levels were dramatically reduced in isolated incisor cultures, indicating RA acted directly on tooth.

To compare global gene expression changes in control vs. RA-treated samples, E14.5 and E16.5 lower incisors were analyzed by high throughput RNA sequencing (RNA-seq). Principal component analysis (PCA plot, Figure S5) and scatter plots (Figure S6) revealed the overall changes observed at E14.5 were tightly clustered, statistically significant, yet often seen as net reductions of ~20–50% (Table 1). RNA-seq analysis confirmed broad effects of RA treatment affecting mineralization-inducing pathways, extracellular collagens, and calcium networks. Genes reduced in expression at E14.5 included Runx, Dlx, Bmp, and Shh, all known inducers of inner enamel epithelium differentiation (Bei, 2009). Net reduction in Runx2 and Dlx5 (Table 1) may act combinatorially to reduce the bone biomarker BGLAP (osteocalcin, (Hassan et al., 2004)), a target reduced by ~2-fold at E16.5 in RA-treated samples (Table 2). A novel RA-inhibited target is the matricellular protein Smoc2, whose mutation in human produces oligodontia, microdontia, abnormal root development, dentin dyplasia, and reduced alveolar bone growth (Bloch-Zupan et al., 2011). Retinoids are known to drive uncommitted progenitor cells into neuronal lineages (Maden, 2007). Consistently, DAVID (Database for Annotation, Visualization, and Integrated Discovery) analysis revealed a functional enrichment score of 1.4 E⁻⁶ for neuronal differentiation pathways. Increased Neurogenin1 (Neurog1), Neurogenin2 (Neurog2), and the notch target Hes5, are typical of an RA-induced neuronal differentiation profile (Table S2). At E16.5, Enam was the most reduced target (Table 2), also markedly reduced by in situ hybridization analysis (Figure S7). Reductions in ameloblastin, X- linked amelogenin, amelotin, and kallikrein-related peptidase 4 (Klk4), all encoding enamel proteins (Núñez et al., 2015), were observed (Table 2). Notably, these reductions at E16.5 occur much earlier than the characterized times of action of the corresponding proteins in inductive, secretory, or maturation stages of rodent enamel development. Odontoblast-originating signals control ameloblast induction (Bei, 2009). Reductions in mineralization targets (alkaline phosphatase), odontoblast structural proteins (dentin matrix protein 1, dentin sialophosphoprotein), ossification biomarkers (Bglap1/osteocalcin, Bglap2), and calcium homeostasis pathways (calcitonin, vitamin D receptor) were all observed in E16.5 RA-treated samples. Table S3 summarizes how RA excess at E16.5 increases the expression of genes involved in retinoid signaling (including Cyp26b1), Wnt signaling, and neuronal differentiation.

Study design: SM, AB, and KN. Data collection: SM, VL, EM, BS, and KN. Data analysis: SM, VL, JH, AB, and KN. Drafting manuscript: SM, PD, AB, and KN. Revising manuscript content: SM, JH, PD, AB, and KN. Approving final version of manuscript: SM, VL, JH, EM, BS, PD, AB, and KN. KN takes responsibility for the integrity of the data analysis.

This work was supported by a grant from the University Hospital of Strasbourg (API, 2009–2012, “Development of the oral cavity: from gene to clinical phenotype in Human”), the EU-funded projects (ERDF) A27 “Oro-dental manifestations of rare diseases” in the framework of the RMT-TMO Offensive Sciences initiative INTERREG IV and INTERREG V No. 1.7 RARENET, the Institut d'Etudes Avancées (Institute of Advanced Studies) de l'Université de Strasbourg (USIAS Fellows 2015), and the grant ANR-10-LABX-0030-INRT managed by the Agence Nationale de la Recherche under the frame program Investissements d'Avenir ANR-10-IDEX-0002-02. Sequencing was performed by the IGBMC Microarray and Sequencing platform, supported by the France Genomics National Infrastructure, funded as part of the Investissements d'Avenir program (ANR-10-INBS-0009).